MRI t1 semi-quantitative imaging method and system based on non-immunostimulant polysaccharide cross-linked colloidal nanoparticles and formulation-normalized potentiation index (f-NEM)

Abstract

The present invention relates to a method and a system for MRI T1 semi-quantitative imaging based on non-immunostimulating polysaccharide cross-linked colloidal nanoparticles satisfying PRR / BCR non-operability and a formulation-normalized potentiation index (F-NEM).

Description

MRI T1 Semi-Quantitative Imaging Method and System Based on Non-Immunostimulating Polysaccharide Cross-linked Colloidal Nanoparticles and Formulation-Normalized Enhancement Index (F-NEM)

[0001] The present invention relates to a non-immunostimulating polysaccharide cross-linked colloidal nanoparticle satisfying PRR / BCR non-operativity and an MRI T1 semi-quantitative imaging method and system based on a formulation-normalized enhancement index (F-NEM).

[0002] Magnetic resonance imaging (MRI) has been established as a standard tool for evaluating the vascular and circulatory systems and diagnosing cardiovascular diseases due to its ability to visualize deep structures at high resolution and provide excellent soft tissue contrast. Recently, MR lymphangiography (MRL) has been gaining attention for the diagnosis and pathological analysis of lymphatic circulation disorders, including lymphedema. As the lymphatic system is a crucial part of the circulatory and immune systems, precisely imaging the microstructures of lymphatic vessels and nodes distributed throughout the body is key to establishing diagnostic and treatment plans. However, achieving near-real-time high-resolution visualization is technically challenging due to the small diameter of lymphatic vessels and the complex intertwining of superficial and deep pathways.

[0003] Lymphedema is a chronic, progressive disease caused by various factors such as congenital malformation of lymphatic vessels or lymph nodes, secondary damage occurring after tumors, surgery, or radiation therapy, and parasitic or bacterial infections, characterized by the accumulation of fluid with a high protein content in peripheral tissues, accompanied by persistent inflammation and fibrosis. Pathophysiologically, increased mechanical resistance in lymphatic drainage pathways, an imbalance between capillary filtration and lymphatic reabsorption, the activation of macrophages and fibroblasts, and the pathological proliferation of adipose tissue form a reciprocal amplification loop, creating a vicious cycle leading to inflammation, fibrosis, and fatification. In clinical practice, indirect indicators such as limb circumference measurements or water displacement volume measurements are widely used; however, they face significant anatomical limitations and measurement variability in quantitatively evaluating structural and functional abnormalities of deep lymphatic vessels. Furthermore, standardized imaging and molecular indicators capable of consistently tracking treatment responses over the long term have not been sufficiently established. Typical imaging findings of lymphedema include delayed lymphatic transport, cutaneous reflux, and reduced lymph node visualization. While non-contrast MRI clearly presents increased subcutaneous tissue thickness, honeycomb structures, and dilated lymphatic vessels, it has limitations in providing real-time functional indicators of lymphatic flow.

[0004] In the preclinical field, the mainstream approach involved blocking lymphatic outflow in rodent tail or hind limb models through surgery, ligation, or irradiation, followed by verifying the blockage using near-infrared fluorescence lymphangiography (NIRF-ICG) or lymphoscintigraphy. However, small animals differ significantly from humans in terms of anatomical scale, lymphatic vessel diameter, valve structure, and carrying capacity, which limits the potential for translation into clinical techniques such as microsurgery, reconstructive surgery, medical devices, and cell therapies. Reproducibility is low because the rate and extent of collateral reperfusion formation vary among individuals even when the same surgical scope is applied, and tail models fail to adequately reflect the biomechanics and lymphomechanics of limb-centered pathologies. To address these issues, large animal models such as pigs and dogs were proposed; however, standardization has been insufficient due to complications associated with extensive resection, short model maintenance periods, and a lack of cross-validation at the imaging and molecular levels. In particular, integrated protocols that non-invasively confirm lymphatic blockade in large animals, quantify volume changes by tracking the same individual over a long period, and simultaneously verify mechanistic changes at the tissue and molecular levels have been reported only to a limited extent.

[0005] Large animal models possess strengths in terms of clinical translatability. Because limb diameters and soft tissue thicknesses are similar to those of humans, they can simulate actual clinical procedures with high fidelity, including surgical approaches, the anatomical locations of lymph nodes and collecting ducts, suturing and reconstruction techniques, and compression and physical therapy. MRI-based 3D volumetric analysis enables quantitative evaluation with higher repeatability than indirect indicators such as circumference and hydro-displacement, and when combined with contrast-enhanced MRL, it allows for the cross-validation of structural blockades and functional drainage impairments within the same workflow. Furthermore, the correlation between imaging and molecular indicators can be established by confirming changes in lymphatic marker expression (podoplanin, LYVE-1, PROX1, VEGFR-3, etc.), valve structural abnormalities, and diameter increases using immunofluorescence staining, and by capturing the activation of inflammatory and fibrotic pathways such as NF-κB and TGF-β / SMAD using RNA sequencing. This multi-layered validation provides evidentiary value regarding not only sensitivity and specificity but also clinical translatability in the preclinical evaluation of therapeutic candidates and reconstruction procedures.

[0006] Each lymphatic visualization technique has its own inherent limitations. Fluorescent lymphangiography is advantageous for real-time observation of superficial lymphatic vessels, but it has limitations regarding deep lymphatic vessels and lymph nodes due to signal loss caused by attenuation and scattering, as well as low quantification. Lymphoscintigraphy presents the contours of systemic drainage, but its low spatial resolution makes it difficult to interpret and quantify local lesions. Conventional MR lymphangiography (MRL) can visualize deep structures in high resolution, but gadolinium chelate contrast agents with small molecular weights cause rapid venous leakage, leading to frequent venous contamination. Furthermore, protocol variations between institutions and equipment result in significant signal variability, hindering consistency. To enhance clinical metastasis potential, a lymph-specific contrast platform is required that visualizes the deep lymphatic system in high resolution, avoids venous contamination, and standardizes signal interpretation even in multi-institutional settings.

[0007] Polysaccharides can activate innate immunity by binding to pattern recognition receptors (PRRs; e.g., TLRs, CLRs) on the surfaces of macrophages and dendritic cells. Through these interactions, some polysaccharides, such as dextran, can induce increased phagocytosis, enhanced dendritic MHC-II expression and antigen presentation, and the promotion of cytokine (IL-6, TNF-α, IL-12) secretion. Apart from this immune activation, particulate formulations with surfaces modified by polysaccharides can induce osmotic gradients and local immune responses at the injection site. When high concentrations of high-molecular-weight polysaccharides accumulate locally, the osmotic pressure relative to surrounding tissues increases, leading to water movement and fluid redistribution, which may result in transient edema. Simultaneously, if certain polysaccharide surface patterns interact with innate immune receptors, inflammatory responses such as vasodilation, increased permeability, and leukocyte influx are initiated, which may lead to lymphatic compression or impaired endothelial function. When lymphatic contractility and valve function are impaired, reflux increases and absorption decreases, exacerbating edema. These changes can exaggerate the local absorption and retention of contrast agents, creating image artifacts that resemble actual structural abnormalities. In particular, in environments rich in immune cells, non-specific capture and excessive local enhancement of contrast agents become prominent, making it difficult to distinguish between pathological lymphoid findings and inflammatory artifacts. Furthermore, variables of contrast agent origin (receptor occupancy, internalization rate, and changes in the inflammatory microenvironment) distort the time-signal curve, increasing dataset heterogeneity and distribution shifts between training and validation sets even within the same protocol. This undermines the calibration reliability and threshold portability of imaging AI models.

[0008] Therefore, to achieve reproducible quantitative interpretation throughout the entire clinical and preclinical cycle, a new contrast imaging platform is required that simultaneously satisfies a non-perturbative contrast mechanism that minimizes interaction with the biological system and lymph-selective pharmacokinetics.

[0009] The present invention aims to provide a foundation for the accurate pathological evaluation of lymphatic diseases, including lymphedema, standardized monitoring of treatment responses, and the clinical application of imaging artificial intelligence by combining a non-immunostimulatory, lymphoselective contrast imaging platform satisfying PRR / BCR non-operability with comparative quantitative analysis based on formulation-normalized enhancement markers (F-NEM).

[0010] To this end, the present invention proposes a polysaccharide cross-linked colloid-based contrast agent designed to suppress exogenous disturbances in target density and distribution during imaging windows through immunoreceptor non-operational design, and to minimize venous capillary leakage to follow the interstitial-lymphatic priority pathway. Furthermore, based on the premise that the absolute value of the contrast signal varies depending on the equipment, formulation, and environment, a standardized framework is adopted to interpret time-signal (PSE, WIR, TTP, AUC) and morphology (ES, NF, LV) indicators in a semi-quantitative and trend-based manner, using a formulation-normalized enhancement index (F-NEM) corrected by formulation metadata such as dosage, infusion-scan interval, and lot-specific r1 characteristics as a common metric. ΔR1 can be used as an auxiliary indicator under operating range, masking, and quality control (QC) only when a T1 map is obtained, and the reporting and AI pipeline of the present invention can be configured to operate centered on F-NEM even without ΔR1. This strategy can ensure reproducibility between training, validation, and actual deployment by progressively eliminating sources of error such as venous contamination, infusion leakage, T2* effects, SNR degradation, and B1 heterogeneity through QC and masking, and by increasing the consistency of input distribution in a multi-institution environment.

[0011] A first aspect of the present invention is a non-immunostimulating polysaccharide crosslinked colloidal particle composition that is dispersible in an aqueous medium and satisfies pattern recognition receptor (PRR) and / or B cell receptor (BCR) non-operational properties, wherein (a) the hydration diameter in an aqueous solvent is 2 to 10 nm (preferably 2 to 8 nm), (b) the ζ-potential is -20 mV to 0 mV, (c) the -OH functional groups of the monosaccharides serving as repeating units of the polysaccharides to be crosslinked are selectively modified with an epoxide-based first crosslinking agent to form a compact spherical three-dimensional network structure that is intramolecularly and / or intermolecularly crosslinked, (d) at least 60% (preferably 70% or more, more preferably 90% or more, even more preferably 95% or more) of the total number of monosaccharides are modified by the crosslinking agent, and (e) some or all of the amine functional groups derived from the crosslinking agent exposed on the surface are converted to -COOH Post-modification to achieve the ζ-potential of (b), and (f) optionally, iron ions (Fe) coordinately bonded to crosslinker-derived amine functional groups or -COOH functional groups exposed on the surface of the polysaccharide crosslinked colloid particles. 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ The present invention provides a particle composition characterized by including ) or iron oxide nanoparticles.

[0012] A second aspect of the present invention provides an MRI T1 semi-quantitative imaging method using a particle composition of the first aspect, comprising: (a) administering the particle composition via intradermal, subcutaneous, or other peripheral routes; (b) calculating a formulation-normalized enhancement metric (F-NEM) using formulation metadata (administration amount, injection-scan interval (τ), r1 reference value per formulation, etc.) from T1-weighted images or T1 map-based data at different time points in the same anatomical region; (c) calculating one or more of time-signal indices (PSE, WIR, TTP, AUC) and / or morphological indices (ES, NF, LV) with the F-NEM as an axis; and (d) providing the calculated indices as a standardized numerical report.

[0013] A third aspect of the present invention provides a semi-quantitative processing system configured to receive input data (image time series, formulation metadata, QC log) and (i) calculate a formulation-normalized enhancement index (F-NEM), (ii) calculate a percentage signal enhancement (PSE), a wash-in rate (WIR), a time-to-peak (TTP), an area under the curve (AUC), edge sharpness (ES), a necrotic fraction (NF), and / or a lesion volume (LV), (iii) calculate a tissue damage index (TDI) (weighted combination), (iv) determine QC pass / mask status, (v) generate a DICOM-SR / HL7-FHIR structured report, and (vi) record and store audit logs such as model, version, hyperparameter, random seed, and processing time.

[0014] A fourth aspect of the present invention provides a computer-readable recording medium that stores an instruction instructing a method step of the second aspect to be performed when executed by a processor.

[0015] A fifth aspect of the present invention provides a use of the particle composition of the first aspect for MR lymphangiography to visualize superficial and / or deep lymphatic networks while avoiding venous contamination.

[0016] A sixth aspect of the present invention provides an imaging kit comprising the particle composition of the first aspect and further comprising instructions for use (administration site, dosage, infusion-scan interval), a formulation metadata record sheet, and a standard report form for calculating the formulation-normalized enhancement index (F-NEM) and / or tissue damage index (TDI).

[0017] The seventh aspect of the present invention provides an artificial intelligence (AI) learning method that, in the method of the second aspect, performs labeling of superficial and / or deep lymph network segmentation, nodal filling delay, reflux presence / grade, bypass asymmetry, lesion activity (active / inactive), and / or fibrosis co-occurrence on a training dataset, and trains a classification, segmentation, and time-series prediction model using a formulation-normalized enhancement index (F-NEM) and a group of derived indicators as inputs.

[0018] The eighth aspect of the present invention provides a use for the particle composition of the first aspect for tracking the rate of change in vascular permeability, interstitial retention, and lymphatic drainage impairment (PSE / WIR / TTP / AUC) and evaluating morphological changes (ES / NF / LV) in neoplastic lesions.

[0019] A ninth aspect of the present invention provides a particle composition of the first aspect for evaluating activity (acute / inactive), extent (focal / multifocal), and / or tissue damage based on baseline-referenced change and / or side-to-side comparison (ipsilateral vs. contralateral) in inflammatory and fibrotic diseases.

[0020] The tenth aspect of the present invention provides a use of the particle composition of the first aspect for the continuous visualization of superficial / deep networks in lymphatic circulation disorders and for the quantification of functional indicators based on formulation-normalized enhancement indicators (F-NEM).

[0021]

[0022] The present invention will be described below.

[0023] In magnetic resonance imaging (MRI), the contrast signal is nonlinear with respect to contrast agent concentration (low-concentration under-response, high-concentration saturation) and strongly dependent on the tissue environment (blood flow, interstitial pH, protein binding, micro-magnetic homogeneity, etc.), so there are inherent limitations in interpreting the absolute signal value at a single time point as an absolute indicator of pathophysiology. Even with the same formulation, the signal scale and dynamic range vary depending on operating conditions such as scanner magnetic field strength, sequence parameters, coil environment, lipid inhibition, and B1 heterogeneity, so comparisons based on "absolute quantification" between equipment and institutions have low reliability.

[0024] Considering these physical and biological limitations, the present invention does not aim for absolute values ​​themselves, but adopts a standard frame based on semi-quantitative and trend-based readings. The key is to use the formulation-normalized enhancement metric (F-NEM) as the central axis to provide consistent numerical values ​​for relative comparisons, such as same-patient time series, left-right contrasts, and rates of change from baseline. F-NEM is designed to normalize signals by reflecting formulation metadata (dosage, infusion-scan interval, r1 pre-representation values ​​by formulation, etc.), thereby primarily offsetting variations caused by formulation and protocol, and allowing the remaining changes to reflect the temporal trends of pathophysiology (vascular permeability, interstitial retention, lymphatic drainage impairment, etc.). For example, left-right comparison includes normalizing and calculating F-NEM and its derived indices (PSE, WIR, TTP, AUC, ES, NF, LV, TDI) between the ipsilateral (lesion side) and contralateral (health side) of paired structural organs such as limbs, breasts, and kidneys.

[0025] The present invention also uses non-immunostimulatory polysaccharide cross-linked colloidal particles to ensure input stability. The particles are designed to satisfy PRR / BCR non-activation and to exhibit predictable pharmacokinetics of lymph-selective distribution and renal excretion dominance while suppressing venous contamination through a hydration diameter of 2-8 nm, a weak negative charge of -20 to 0 mV, and a high-substitution cross-linked structure. This creates a non-perturbative environment in which the contrast agent does not significantly induce receptor occupancy, internalization, or cytokine / complement activation during the imaging window, thereby reducing variability in time-series input.

[0026] Based on the above premise, the present invention can calculate time-signal indicator groups (PSE, percentage signal enhancement; WIR, wash-in rate; TTP, time-to-peak; AUC, area under the curve) and shape indicator groups (ES, edge sharpness; NF, necrotic fraction; LV, lesion volume) semi-quantitatively, centered on F-NEM. Semi-quantitative means comparing and tracking time-series change rates, left-right differences, and baseline variation as standardized values ​​on a formulation-normalized scale, without requiring absolute signal-concentration mapping. Furthermore, through the Tissue Damage Index (TDI), a synthetic indicator defined as a weighted combination of F-NEM and ES / NF / LV, complex pathological conditions can be summarized into a single score, and a progression flag can be assigned upon a threshold increase (e.g., ≥20% relative to baseline) to support clinical decision-making.

[0027] r1 (longitudinal relaxation enhancement capability), which varies depending on the device, magnetic field strength, and environment, is not directly compared as a performance indicator of the present invention, and harmonization between scanners or phantom-based absolute calibration is not a requirement of the present invention. Instead, consistency of the input distribution is ensured through F-NEM normalization using formulation metadata and quality control (QC) rules. QC rules include operating range constraints such as automatic exclusion of frames in which venous contamination is detected in time-resolved images (frames at the point in time where early linear high signal appears along the anatomical venous pathway and wash-out is synchronized with the venous system) and frames in which infusion leakage (extravasation) is detected (frames determined by extensive blurring around the injection site and the absence of temporal distal progression), satisfaction of predefined SNR hurdles (e.g., frame average SNR ≥ 15), flip angle and B1 correction suitability, and / or masking of high concentration saturation and T2* influence intervals.

[0028] For example, a venous-contaminated frame may be a frame within a time-resolved image (time series) at a point in time where linear or tubular high signals appear early along anatomical venous pathways rather than lymphatic pathways, and a rapid wash-out pattern synchronized with the deep venous system is observed, indicating venous inflow. In other words, it refers to a specific frame among multiple time points where findings of venous contamination are detected. Injection-site extravasation is a pattern characterized by diffuse spread and the absence of temporal progression or distal displacement around planned intradermal / subcutaneous injection sites, signifying a state where the contrast agent has permeated local tissues but failed to enter the lymphatic pathway. The SNR hurdle is a predefined lower limit for the signal-to-noise ratio used to determine analysis eligibility. For instance, a lesion / background-based frame average SNR ≥ 15 or a voxel-based SNR ≥ 10 may be used, but the specific threshold follows the values ​​defined in the operating manuals for each formulation, sequence, and instrument. Frames / voxels that do not meet the hurdles are excluded as QC-fail.

[0029] ΔR1(=1 / T1 post - 1 / T1 pre ) is treated as an auxiliary indicator rather than an essential indicator in the present invention. ΔR1 may be selectively referenced only when the predefined operating range, masking, and QC are satisfied, and key judgment and reporting are performed by F-NEM and its derivative indicators (PSE, WIR, TTP, AUC, ES, NF, LV, TDI). This configuration provides “reproducible comparative quantification” required in actual clinical practice, while bypassing the weaknesses of absolute quantification arising from contrast signal-concentration nonlinearity and environmental dependence.

[0030] In summary, the present invention (i) acknowledges the limitations of interpreting absolute contrast signal values, (ii) ensures input stability using non-immunostimulatory and non-agonisting contrast agents, (iii) semi-quantitatively calculates time-signal and morphology indicators using the formulation-normalization-based F-NEM axis, and (iv) guarantees data distribution consistency through QC and operating range management rather than scanner harmonization. As a result, standardized numerical reports suitable for time-series within the same patient, left-right comparisons, and inter-institutional comparisons are generated, enabling reliable trend-based readings in diagnosis, treatment response evaluation, and clinical trial design.

[0031] In this specification, side-to-side comparison refers to a method of comparing and normalizing indices by using the ipsilateral (lesion side) and contralateral (health side) within the same individual as internal controls. It can be applied to paired structural organs such as the upper and lower limbs, breasts, kidneys, and salivary glands, or to left-right symmetric anatomy. It can compensate for differences in background signal, physique, and imaging conditions between individuals and increase sensitivity by interpreting changes in indices such as F-NEM, PSE, WIR, TTP, AUC, ES, NF, LV, and TDI as relative values ​​within the same individual.

[0032] When one side is not appropriate due to bilateral disease, surgical history, etc., a proximal-distal comparison, a region-of-interest internal control, or a reference atlas may be used as alternatives.

[0033] According to the present invention, non-immunostimulatory polysaccharide cross-linked colloidal particles designed to withstand enzymatic degradation in the body, disperse in lymph nodes without gelation or aggregation, and satisfy pattern recognition receptor (PRR) and / or B cell receptor (BCR) non-operational properties can not only overcome the limitations of immunogenicity (immune cell binding / activation, increased side effects upon repeated administration) caused by polysaccharide-based drug delivery and contrast agents, but also ensure signal input stability in MR lymphangiography (MRL). For example, non-immunostimulating polysaccharide cross-linked colloidal particles are formed by selectively modifying the -OH functional groups of repeating units (monosaccharides) of various polysaccharides, including dextran, with an epoxide-based first crosslinker, and then inducing (i) direct crosslinking between the modified -OH pairs and / or (ii) intramolecular and / or intermolecular crosslinking mediated by a polyvalent amine (second crosslinker) to form a compact spherical three-dimensional network with significantly restricted rotational degrees of freedom. Subsequently, the surface-exposed primary amine is post-modified with -COOH to adjust the ζ-potential to the range of -20 mV to 0 mV and control the hydration diameter to 2 to 8 nm. This physicochemical design (highly substituted crosslinking, small hydration diameter, weak negative charge) simultaneously achieves the following immunoevasion and stability mechanisms.

[0034] First, PRR / BCR non-agonistic. By geometrically restricting the exposure of the repeat sequence and maintaining the surface charge at a weak negative charge, it significantly reduces binding affinity to pattern recognition receptors (PRRs; e.g., Toll-like / C-type lectin receptors) on macrophages and dendritic cells, as well as B cell receptors (BCRs). As a result, receptor clustering and internalization are not induced, so NF-κB pathway activation, cytokine (TNF-α, IL-6, IL-1β) release, and complement (C3a / C5a) activation remain within the detection limit or tolerance range. In particular, by structurally inhibiting frequent BCR cross-linking in T-independent multivalent antigens, it lowers the potential for memory B cell formation and anti-drug antibody (ADA) induction, thereby increasing the tolerance for repeated administration.

[0035] Second, enzyme resistance and predictable elimination. The high-substitution cross-linking network reduces sensitivity to glycosidic hydrolases, thereby delaying hydrolysis in vivo, and thanks to its size of 2–8 nm and weak negative charge, non-specific binding to serum proteins and opsonization are inhibited. As a result, it follows a planned movement sequence from the interstitial to the lymphatic to the venous without gelation or aggregation at the injection site, and the fraction that enters systemic circulation is predominantly excreted by renal filtration, minimizing the risk of organ retention and metal accumulation (confirmed by bladder signal in the examples).

[0036] Third, suppression of immune-induced image distortion and input stability. Due to the above non-operational design, target density and distribution during the imaging window are not disturbed by the contrast agent itself, so temporal signal changes mainly reflect variations in target physiology (vascular permeability, interstitial retention, lymphatic drainage). Accordingly, ΔR1 (=1 / T1post - 1 / T1pre), calculated from the T1 change before and after contrast, is treated only as an auxiliary indicator, but can be consistently converted into a formulation-normalized enhancement metric (F-NEM) by applying formulation metadata (dosage, infusion-scan interval τ, pre-correction r1 by lot and magnetic field strength), thereby ensuring reproducibility for multi-institutional and multi-equipment comparison and time-series tracking.

[0037] Fourth, the engineering significance of the cross-linking substitution rate. The particles of the present invention must be modified with a cross-linking agent comprising at least 60% (preferably ≥70%, more preferably ≥90~95%) relative to the total number of monosaccharides. This leads to (i) a reduction in the spatial accessibility of repeat sugar epitopes, (ii) a disruption of hydrogen bonding and hydrophobic interaction patterns with BCR / antibody, and (iii) suppression of exposure of PRR binding motifs, thereby simultaneously achieving immune recognition evasion and enzyme resistance. Above the above threshold, a decrease in PRR / BCR binding affinity and inactivation of the immune switch were experimentally confirmed (Figs. 2, 4, and 6).

[0038] Fifth, safety and clinical suitability. In preclinical studies (repeated-dose toxicity, hematology, histopathology) and early clinical studies (phase 1 / 2a), no significant signs of systemic inflammation, hypersensitivity reactions, or ADA induction were observed, and continuous filling images from peripheral lymphatic vessels to lymph nodes to central lymphatic vessels were realized without residue (see Fig. 18 and Examples 5-7). This supports the fact that the polysaccharide cross-linked colloid particles of the present invention do not induce capture or activation even in immune cell-dense environments such as lymph nodes.

[0039] According to the present invention, non-immunostimulating polysaccharide cross-linked colloid particles satisfying PRR / BCR non-operativity do not cause inflammation, edema, allergic and / or hypersensitivity reactions when administered in the body. It has been experimentally confirmed that these particles are not recognized or bound by Pattern Recognition Receptors (PRRs), or at least do not cause significant cytokine release and NF-κB signaling activation in imaging and observation windows (Figs. 2, 4, and 6). Representative PRRs include Toll-like receptors (TLRs), C-type lectin receptors (CLRs), Retinoic acid-inducible gene-I-like receptors (RRRs), Nucleotide oligomerization domain-like receptors (NLRs), and Absent-in-melanoma-like receptors (ALRs). For reference, clinical symptoms of hypersensitivity reactions include urticaria (a localized skin reaction accompanied by itching and swelling), angioedema (swelling of the deep layers of skin around the eyes and lips), and anaphylaxis (a severe systemic allergic reaction that may include symptoms such as difficulty breathing, hypotension, and shock).

[0040] As shown in Fig. 2, unlike linear dextran (Dextran T10) which is the target of crosslinking, NEMO-103, which uses crosslinked dextran (Cdex), did not induce an acute edema reaction in vivo. Furthermore, according to Examples 3-12 and Fig. 10, the polysaccharide crosslinked colloid particle-based T1 MRI contrast agent of the present invention maintained its contrast performance until it was excreted in urine through the kidneys via vascular circulation after administration into the body. When combined with the results of hematological analysis from a 4-week repeated (1 time / 2 weeks, total 3 times) intradermal toxicity test in Beagle dogs INV-001 (Examples 5-8, Table 7) and a human Phase 1 / 2a shoulder MR arthrography test (NEMO-103, Examples 4-5), this suggests that the composition is not recognized as an exogenous antigen or significantly causes an inflammatory response to eliminate it.

[0041] Consistent results were also confirmed at the cellular level. In Figure 4, Cdex did not stimulate the NF-κB activation pathway when treating macrophages, and in Figure 6, it did not have a significant effect on cell viability even when treated at high concentrations. From this, it can be inferred that Cdex has a non-agonistic profile, which does not induce cell proliferation or cytokine release through signaling within macrophages. Furthermore, in the hematological tests of the repeated-dose toxicity test in beagle dogs (Examples 5-8, Table 7) and the human Phase 1 / 2a study (Examples 4-5), no clinically significant abnormal changes were observed in any of the major indicators, including complete blood count, neutrophils, eosinophils, basophils, lymphocytes, and monocytes. The convergence of these preclinical and clinical data supports the fact that the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention are not significantly recognized or bound by PRRs, or at least do not cause immune activation accompanied by cytokine release, and consequently have a low likelihood of inducing inflammation and hypersensitivity reactions.

[0042] The non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention secure PRR / BCR non-operability through a triple design of “highly substituted cross-linked network based on epoxide selective modification + weak negative charge of -20 to 0 mV + hydration diameter of 2 to 8 nm” and realize a predictable pharmacokinetic signature characterized by enzyme resistance, lymph selectivity, and renal excretion. Accordingly, stable input with minimized contrast agent origin variables is secured across the imaging window, and a semi-quantitative pipeline operates to consistently calculate time-signal indicators (PSE, WIR, TTP, AUC), conformation indicators (ES, NF, LV), and synthesis indicators (TDI) based on F-NEM (formulation-normalized enhancement metric) normalized by formulation metadata. These immunological and physicochemical compatibility support industrial applications as a safe contrast and delivery platform capable of repeated administration in the early diagnosis of inflammatory diseases, evaluation of lymphatic function via MR lymphangiography (MRL), and staging and monitoring of treatment response in lymphedema, tumors, and fibrotic lesions.

[0043] The MRI T1 imaging procedure using the above contrast agent is based on lymphoselective and non-perturbative contrast that avoids venous contamination, and acquires 3D T1-weighted images or T1 maps at least two time points within an observation window of 5 to 120 minutes after administration. ΔR1 is calculated selectively as an auxiliary indicator within the scope of quality control, but final interpretation and reporting are performed based on F-NEM and its derivative indicators (PSE, WIR, TTP, AUC, ES, NF, LV, TDI). Through this, microvascular permeability, interstitial retention, lymphatic drainage, and tissue damage dynamics in lymphedema, neoplastic lesions, fibrotic lesions, and ischemic / reperfusion injury can be standardized and presented as time-series rates of change, left-right comparisons, and baseline contrast changes, rather than absolute concentration values.

[0044]

[0045] 1. Mechanisms of Immune Response and Non-Immunostimulatory Design Rationale

[0046] According to the present invention, a non-immunostimulating polysaccharide cross-linked colloidal particle composition that is disperseable in aqueous media such as lymph nodes without gelation / aggregation and satisfies pattern recognition receptor (PRR) and / or B cell receptor (BCR) non-operationality is,

[0047] (a) In an aqueous solvent, the hydration diameter is 2 to 10 nm (preferably 2 to 8 nm), and

[0048] (b) The ζ-potential is -20 mV to 0 mV, and

[0049] (c) having a compact spherical three-dimensional network structure that is intramolecularly and / or intermolecularly crosslinked by selectively modifying the -OH functional groups of the monosaccharide repeating units of the crosslinking target polysaccharide with an epoxide-based first crosslinking agent,

[0050] (d) At least 60% of the total number of the monosaccharides (preferably 70% or more, more preferably 90% or more, even more preferably 95% or more) are modified by a crosslinking agent, and

[0051] (e) post-modifying some or all of the crosslinking agent-derived amine functional groups exposed on the surface with -COOH to achieve the ζ-potential of (b), and

[0052] (f) Optionally, iron ions (Fe) coordinately bonded to amine functional groups or -COOH functional groups derived from the crosslinking agent exposed on the surface of polysaccharide crosslinked colloidal particles 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ It is designed to include ) or iron oxide nanoparticles.

[0053] Here, PRR / BCR non-operativity is defined in human PBMC or monocyte / macrophage-based in vitro tests as not showing an increase in TNF-α, IL-6, IL-1β secretion relative to the vehicle exceeding a pre-defined threshold, and / or signal amplification in TLR reporters being below the threshold, or C3a / C5a complement activation being within the acceptable range.

[0054] The particles of the present invention can be designed so that, under conditions where there is no targeting by cell surface ligands, the binding affinity with PRR or BCR is lower than a predetermined Kd upper limit (e.g., above the μM range) so that endocytosis does not significantly occur.

[0055] In this specification, “vehicle” refers to a reference formulation containing only the solvent, buffer salt, stabilizer, isotonic agent, and pH regulator in the same composition, excluding the active ingredients, which are non-immunostimulating polysaccharide cross-linked colloidal particles and metal coordination agents (Fe / Gd / Mn or iron oxide). Even in the case of metal coordination compositions, the vehicle does not contain free metal ions (or is below an unavoidable background level) and is composed such that the ionic strength, osmotic pressure, and pH are equivalent to those of the test formulation. For example, the vehicle may be composed of 10 mM HEPES, 150 mM NaCl, 5% w / v mannitol, and pH 7.4.

[0056] In this specification, the “preset threshold” refers to an upper limit for each biomarker that is not considered a significant immune stimulus relative to the vehicle, and is applied as a conservative double rule combining relative and absolute criteria. Typically, under human PBMC or monocyte / macrophage-based in vitro conditions (e.g., 2–5 × 10⁶ 5In cells / well, 37 °C, 5% CO₂, culture for 24±4 hours, with n≥3, TNF-α, IL-6, and IL-1β should not increase by more than 1.3 times (30%) relative to the vehicle mean, and at the same time, the Δ concentration should not exceed the vehicle mean plus 2 standard deviations (SD). In practice, absolute upper limits are recommended as ΔTNF-α ≤50 pg / mL, ΔIL-6 ≤50 pg / mL, and ΔIL-1β ≤25 pg / mL, but if the statistical criteria are stricter, those values ​​should be applied first. For TLR reporters (NF-κB / SEAP or luciferase), a value is determined to be “below threshold” when the vehicle normalized signal is less than 1.2 times (20%) relative to the vehicle and simultaneously less than 10% of the positive control (LPS 100 ng / mL or Pam3CSK4 1 μg / mL), and a value to be defined as “within acceptable range” when complement activation (C3a / C5a, 10% human serum, 37 °C, 30-60 min) is less than 1.2 times relative to the vehicle and does not exceed the upper limit of the laboratory standard (e.g., C3a 300 ng / mL, C5a 50 ng / mL). The stricter of the two criteria is always applied for the determination, and reproducibility is verified with multiple serum lots. The values ​​are “exemplary thresholds.”

[0057] In this specification, the “predefined Kd upper limit” refers to the upper limit of binding affinity measured against the extracellular domain of the PRR or BCR (purified protein immobilized surface or receptor overexpressing cell surface). Binding evaluation is performed using one or more of surface plasmon resonance (SPR), biolayer interferometry (BLI), or flow cytometry-based equilibrium binding methods, and includes (i) low-density ligand immobilization, (ii) low-flow rate / mass transport correction, and (iii) verification of competition with known ligands to minimize apparent affinity distortion due to the multivalentity of the nanoparticles. Non-binding low-affinity conditions may be defined as an apparent Kd(app) ≥ 10 μM (preferably ≥ 50 μM) or a binding signal of less than 3-5% relative to Rmax (including negative surfaces and buffer injection baselines on the same chip). For example, as a cell-based alternative criterion, a normalized Mean Fluorescence Intensity (MFI) corrected for non-specific binding (homologous isotype / F(ab')2, blocked with excess BSA) at 4 °C and 37 °C is adopted when it is less than 1.2 times that of the vehicle (same buffer / substrate backbone / metal-uncoordinated control) and less than 5% of the positive control (e.g., anti-TLR2 IgG, anti-IgM F(ab')2, LPS-biotin). In C-type lectin series (PRR) assays, Ca 2+To confirm dependence, an additional check is performed to determine if the same signal is lost under EDTA (+ / -) conditions. Under the same conditions, endocytosis indicators are evaluated using pHrodo signaling or LysoTracker co-locating and surface fluorescence quenching techniques, with an additional requirement that the proportion of positive cells be less than 5%. All tests are performed with n ≥ 3 replicates using a minimum 4-arm design consisting of a vehicle, a negative control (untreated or unbound nanoparticles), a positive control (LPS, R848, Poly(I:C), or receptor-specific antibody), and test samples (at least 4-6 concentrations), and are reported with the mean ± standard deviation or a 95% confidence interval. Additionally, endotoxin content (≤ 0.1 EU / mL), complement activation indicators (C3a / C5a) negativity, and cytotoxicity (viability ≥ 90%) are verified for the same batch samples to enhance the specificity of non-immunostimulatory assessment.

[0058] These principles of definition, threshold, and measurement are presented as exemplary standards to ensure reproducible practice while considering background variability between formulations and institutions; in actual application, determination is made based on the stricter value between the mean±SD-based statistical threshold (e.g., mean+2SD) and the presented absolute upper limit, in accordance with each laboratory's validated SOP.

[0059] The innate immune system is composed of cells that non-specifically recognize pathogens and respond rapidly, including macrophages, dendritic cells, neutrophils, and natural killer (NK) cells. These cells perform initial defense against external pathogens by expressing different pattern recognition receptors (PRRs). These receptors recognize sugar structures present on the surface of pathogens and can induce cell activation and inflammatory responses. Representative PRRs include TLR4 (Toll-like receptor 4), which recognizes microbial polysaccharides such as lipopolysaccharides (LPS); the Mannose receptor (MR), which recognizes mannose structures derived from pathogens; and the C-type lectin receptor (CLR), a cell membrane receptor that recognizes polysaccharide structures and plays a central role in the response to fungi and some viruses.

[0060] Antigens are foreign substances capable of inducing an immune response, and polysaccharides are easily targeted by innate and adaptive immunity due to their repetitive sugar arrangements. Typically, when foreign substances enter the body, they are recognized by PRRs or BCRs, initiating a series of reactions such as cytokine secretion, complement activation, and antibody production; this can lead to the rapid elimination of drug carriers and side effects. In particular, since BCRs promote activation, differentiation, and antibody production when efficiently cross-linked by multivalent antigens, there is a risk of hypersensitivity reactions and the formation of memory B cells upon repeated administration. The cross-linking design of the present invention structurally reduces the BCR binding affinity and cross-linking potential, which are established by a complex of hydrogen bonding, van der Waals, and hydrophobic interactions, by lowering the degrees of freedom of the sugar chain and the accessibility of exposed carbohydrate epitopes through a high substitution rate of monosaccharide -OH groups (e.g., ≥60% of total monosaccharides, preferably ≥70%, more preferably ≥90% or more modification). As a result, the particles of the present invention are difficult to recognize as multivalent antigens, so the potential for BCR clustering and antibody induction is significantly low.

[0061] No immune-related adverse reactions were observed in the non-immunostimulating polysaccharide cross-linked colloidal particle-based contrast agent of the present invention in clinical and preclinical studies. In a Phase 1 / 2a clinical trial (Stage 1 Safety Set: n=9; Stage 2 Safety Set: n=23), no significant changes in major indicators, such as leukocyte count or C-reactive protein suggestive of systemic inflammatory response, hypersensitivity, or allergic reaction, were confirmed in hematological analyses within 24 hours of a single intravenous administration of the dextran cross-linked T1 contrast agent NEMO-103 (Examples 4-5). This suggests that the contrast agent components are safely eliminated via physiological excretion pathways without unnecessary interaction with the immune system. Furthermore, in a repeated-dose toxicity study of the iron-based lymphoid contrast agent INV-001 in rats, no significant toxicity was observed in clinical symptoms, hematological indicators, or histopathology even when administered at doses up to 1,500 mg / kg based on dextran (Examples 5-8, Table 7). In Phase 1 clinical trials, INV-001 demonstrated stable, high-resolution visualization of lymph nodes and lymphatic vessels, while failing to reach the Maximum Tolerable Dose (MTD) and showing no occurrence of drug-related adverse events (ADRs) or serious adverse events (SAEs). Furthermore, non-immunogenicity based on low binding affinity to BCR / PRRs was confirmed, thereby demonstrating both lymph-specific imaging efficacy and safety (Example 6). Therefore, the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention are effective as a non-immunostimulating polymer-based drug delivery and contrast imaging platform that effectively suppresses immune adverse reactions such as inflammation, allergies, and hypersensitivity by minimizing immune system disturbances along with precise control of pharmacokinetic behavior. Additionally, iron ions (Fe²⁻) coordinately bound to amine or -COOH functional groups derived from the crosslinking agent exposed on the surface of the polysaccharide cross-linked colloidal particles 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ T1 MRI contrast agents, including those containing iron oxide particles, support the possibility of various biomedical applications.

[0062] Considering the structural background of polysaccharide-derived antigenicity, sugar arrangements and orientations, such as the α-1,6 main chain and α-1,3 branching points of dextran, contribute to BCR binding specificity. By limiting the three-dimensional accessibility and rotational degrees of freedom of these arrangements through epoxide selective modification and high-substitution crosslinking, the present invention weakens non-covalent interactions with the antibody variable region (CDR) and raises the initiation threshold for antigen-mediated BCR cross-linking. Consequently, the likelihood of memory B cell formation and long-term antibody response induction is reduced, making it easier to establish clinically repeated administration strategies.

[0063] As a result, it simultaneously provides a safety profile that enables repeated administration in the early diagnosis of inflammatory diseases, lymphatic diseases, and cancer, lesion target delivery, T1 MRI contrast imaging, and other (semi)quantitative imaging applications, and predictable pharmacokinetic characteristics characterized by avoidance of venous contamination, lymphoselectivity, and renal excretion.

[0064] Accordingly, the present invention is characterized by precisely defining the nanophysical properties (size, charge, and cross-linking substitution rate) and surface chemistry (epoxide selective modification → high-substitution cross-linking → surface -COOH post-modification) of polysaccharide cross-linked colloids so as not to disrupt the innate-adapted immunity cross-pathway (PRR-complement-BCR), thereby consistently realizing a non-immunostimulatory state.

[0065] Major trigger points of the immune response include (i) binding and signal amplification of pattern recognition receptors (PRRs; TLR / CLR, etc.), (ii) a surge in C3a / C5a following the activation of the complement pathway, and (iii) cross-linking and endocytosis of B cell receptors (BCRs) by multivalent binding, and these events converge on the activation of the NF-κB pathway and the secretion of TNF-α, IL-6, and IL-1β.

[0066] The particle composition of the present invention is designed to operate non-agonistically with respect to such signaling networks, the core of which is a compact spherical three-dimensional network composed of (i) a hydration diameter of 2–10 nm (preferably 2–8 nm), (ii) a weak negative charge of a ζ-potential of -20–0 mV, and (iii) highly substituted crosslinking (≥60% relative to monosaccharides, preferably ≥90–95%) based on the epoxide selective modification of monosaccharide -OH groups. Surface-exposed amines are post-modified with -COOH to achieve the above ζ-potential, and if necessary, Fe on carboxylate / amine / hydroxyl / phosphate / catechol donor groups 2+ / Fe 3+ , Gd 3+ , Mn 2+ Alternatively, surface-coordinate iron oxide nanoparticles to impart T1 (optional T2) contrast, while restricting formulation conditions to prevent aggregation and r2* dominance.

[0067] In this specification, the definition of non-operation is explicitly defined by reproducible test criteria. Non-operation is determined if, in human PBMC or monocyte / macrophage-based in vitro tests, TNF-α, IL-6, and IL-1β secretion does not increase beyond a predefined threshold relative to the vehicle, if signal amplification in TLR reporters is below the threshold, or if C3a / C5a in serum tests remains within the acceptable range. If necessary, BCR-related indicators (e.g., Ca 2+ Also check whether flux (psYK / BTK) is maintained at a low level compared to the anti-IgM positive control group.

[0068] If the hydration diameter of polysaccharide cross-linked colloidal particles exceeds 10 to 20 nm or if ζ becomes excessively negatively charged to -30 mV or less, the tendency for activation of the complement alternative pathway may increase, and if the cross-linking substitution rate drops below 60%, the PRR / BCR binding and NF-κB reporter positivity rates may significantly increase. If the -COOH conversion of surface amines is insufficient, electrostatic interactions with the negative charge of the cell membrane increase, which may lead to increased cytokine secretion. The present invention systematically avoids these risks through a “triple design window” of 2 to 8 nm, -20 to 0 mV, and ≥60% (preferably ≥90%).

[0069] The mechanistic basis for achieving non-operometry is explained by the interaction of structure, charge, and size. Highly substitutable cross-linked networks structurally inhibit multivalent binding-based BCR cross-linking by reducing the accessibility and flexibility of epitopes per repeat. Weak negative charges interfere with the formation of multiple electrostatic and hydrogen bond assemblies required for CLR / TLR co-binding. The 2–8 nm size allows for preferential drainage into the interstitial lymph by moving outside the efficient clustering window of macrophage scavenger receptors, thereby reducing the duration of high-concentration exposure at the injection site. Consequently, PRR / BCR occupancy, internalization, and complement activation do not occur significantly during the imaging window, and the imaging signal more directly reflects changes in pathophysiology (vascular permeability, interstitial retention, lymphatic drainage) rather than disturbances of contrast agent origin.

[0070] The manufacturing and feasibility of the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention are ensured by a standardized process. After primary epoxide modification (selective -OH reaction) under alkaline conditions, secondary crosslinking by polyhydric amine spotting and -COOH post-modification by organic acid anhydrides are sequentially performed, and the reaction ratio and time are finely adjusted until the target physical property range is reached. Optionally, metal coordination is performed on the same surface donor group, and the range where aggregation (r2* influence) is possible after coordination can be managed by masking rules in the acquisition / analysis stage. With such an integrated design-verification-manufacturing framework, a person skilled in the art can repeatedly manufacture non-immunostimulating polysaccharide cross-linked colloidal particle compositions with the same performance without excessive trial and error, and the non-operability of each batch can be objectively determined by the above test definition.

[0071] According to the present invention, a non-immunostimulating polysaccharide cross-linked colloidal particle-based T1 MRI contrast agent formed in an aqueous solvent by (i) modifying the -OH functional group of a monosaccharide, which is a building block of a linear polysaccharide, branched polysaccharide, or cyclic polysaccharide, with a first crosslinking agent having an epoxide group, and (a) directly between the functional group modified by the first crosslinking agent and a spatially adjacent -OH functional group, and / or (b) intramolecularly and / or intermolecularly crosslinking between two spatially adjacent functional groups modified by the first crosslinking agent through a second crosslinking agent having two or more amine groups (-NH2), can be inferred that it was not hydrolyzed by enzymes in the body by imaging its pathway and distribution after administration in the body, for example, until it is excreted from the body in urine after circulation in the vascular system (Examples 3-12 and FIG. 10). Therefore, it can be inferred that the polysaccharide cross-linked colloid particle-based T1 MRI contrast agent of the present invention can be excreted from the body without exposure to potential sugar-based immunogenic sites due to hydrolysis, and that the resulting immunogenicity is minimized. Specifically, in the animal test of Example 3 (intravenous administration), the Phase 1 / 2a clinical trial of Examples 4-5 (musculoskeletal contrast agent), and the animal test of Examples 5-8 (lymphatic contrast agent), it was confirmed through MRI imaging that the polysaccharide cross-linked colloid particle-based T1 MRI contrast agent of the present invention did not cause an inflammatory response when distributed in the body, and was not phagocytosed by macrophages or hydrolyzed by enzymes in the body, resulting in excretion through the kidneys in urine.

[0072] It can be seen that the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention, although formed by cross-linking not only linear polysaccharides or cyclic polysaccharides but also complex branched polysaccharides with high immunogenicity, are not recognized or bound by pattern recognition receptors (PRRs) on the surface of immune cells when 60% or more, 70% or more, 90% or more, or 95% or more of the total number of monosaccharides, which are the building blocks of polysaccharides, are modified by a cross-linking agent (Examples 4-2 and 5-7).

[0073] The degree of immunogenicity of the polysaccharide cross-linked colloid particles of the present invention may vary depending on the degree of cross-linking, the size of the polysaccharide molecules, and specific chemical modifications.

[0074] Therefore, the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention can minimize immunogenicity by significantly weakening the binding affinity with pattern recognition receptors (PRRs) on the surface of immune cells in the body in applications such as drug delivery systems, by controlling the degree of cross-linking and the structural / chemical properties of the polysaccharides.

[0075] In summary, the present invention combines a triple design of “small hydration diameter-weak negative charge-highly substitutive crosslinking” with surface coordination chemistry to achieve stable T1 contrast while maintaining the PRR / BCR·complement axis in a non-operotropic state. The resulting signal is suitable for formulation metadata-based comparative quantification (e.g., time-signal·form indicators and synthesis indicators TDI centered on F-NEM) and time-series readings, providing reliable image input for clinical applications requiring repeated-dose safety and reading reproducibility.

[0076]

[0077] 2. Classification of Contrast Agents and Drug Carriers and Comparison of Targeted vs. Non-targeted

[0078] The “stable signal” of the present invention is obtained when the sequence of particle design, pharmacokinetics, and signal generation is aligned to minimize disturbance. First, non-immunostimulating polysaccharide cross-linked colloidal particles form a compact spherical three-dimensional network with low rotational degrees of freedom through epoxide-based selective modification of monosaccharide -OH and a high cross-linking substitution rate, and partially convert surface-terminal amines to -COOH to adjust the ζ-potential to the range of -20 to 0 mV. This combination of properties structurally inhibits agonism with pattern recognition receptors (PRRs) and B cell receptors (BCRs), significantly lowering receptor occupancy, internalization, cytokine release, and complement activation across the imaging window (non-agonistic). In other words, the interaction between the particle and the immune system is reduced to a “negligible level,” thereby securing a non-perturbative state that does not disturb the momentary balance of the biological system.

[0079] These surface chemistry and structures are directly linked to pharmacokinetic signatures. Hydration diameters of 2–8 nm inhibit capillary leakage and venous outflow during intradermal and subcutaneous injection, and cause the substance to preferentially follow the physiological drainage pathway leading from the interstitial space to lymphatic capillaries, peripheral lymphatic vessels, and lymph nodes (lymph selectivity). A weak negative charge (ζ -20–0 mV) reduces non-specific binding to serum proteins and opsonins, thereby mitigating rapid clearance into the hepatic and splenic macrophage systems, and the fraction that enters systemic circulation tends to be excreted renally via glomerular filtration (renal-dominant excretion). As a result, in MR lymphangiography (MRL), superficial and deep lymphatic networks can be visualized continuously while avoiding venous contamination.

[0080] In terms of signal generation, paramagnetic metals coordinated to carboxylate / amine / hydroxyl / phosphate donor sites on the particle surface (e.g., Fe 2+ / Fe 3+ , Gd 3+ , Mn 2+It promotes water exchange between the inner and outer spheres to increase r1 (=1 / T1) while maintaining a low r2 / r1 ratio. The nanometer-scale hydration diameter maximizes the surface-to-volume ratio, increasing the density of active coordination sites per unit mass, and consequently provides stable T1-positive contrast (bright signal) even at low doses. In this case, the signal appears not because the contrast agent "self-induced" changes, but as a result reflecting existing target physiology (vascular permeability, interstitial retention, lymphatic drainage impairment).

[0081] The advantages of this design become even more pronounced when compared to targeted ligand-based formulations. Targeted nanoformulations containing receptor-agonistic ligands or immuno-activating surface patterns can induce a chain of events—such as receptor clustering and internalization, cytokine secretion, and complement activation—immediately after administration, thereby altering the target density and microenvironment of the lesion during the imaging window. This makes the time-signal curve sensitive to formulation origin variables and amplifies data shifts between equipment, institutions, and cohorts, thereby degrading the model's calibration and reproducibility during the training, validation, and field-use phases. Conversely, the non-agonistic lymphoselective T1 formulation of the present invention minimizes immune-vascular interactions, reducing the variability of input signals within time series of the same protocol and serving as the basis for consistently generating "standardized numerical reports" that are comparable even when equipment or institutions differ.

[0082] The present invention operates a group of semi-quantitative indicators centered on the formulation-normalized enhancement metric (F-NEM), which is based on formulation metadata (administrative dose, infusion-scan interval, r1 pre-determination by formulation, etc.), rather than absolute concentration quantification. Here, semi-quantitative refers to a “relative value capable of comparison across identical patient time series, left-right comparisons, and inter-institutional comparisons” by offsetting and normalizing formulation-attributed variations under defined operating ranges and quality control (QC) rules. Time-signal indicators (PSE, WIR, TTP, AUC) and morphological indicators (ES, NF, LV) are calculated using F-NEM as the axis, and if necessary, a weighted Tissue Damage Index (TDI) is calculated to provide a progression flag. In this case, input stability is ensured by combining pharmacokinetic characteristics derived from particle design—such as non-action, lymphoselectivity, avoidance of venous contamination, and renal excretion—with a signal generation mechanism (promotion of water relaxation via surface coordination chemistry).

[0083] In summary, the mechanism of action of the present invention is explained in three stages: ① a surface and structural design featuring high crosslinking, weak negative charge, and hydration diameters in the nanometer range that ensures PRR / BCR non-operability; ② predictable pharmacokinetics regarding lymph selectivity, avoidance of venous contamination, and renal excretion resulting therefrom; and ③ stable generation of T1-positive contrast based on surface coordination chemistry. This closed-loop design minimizes input signal disturbances across the entire imaging window, thereby enhancing the reproducibility of F-NEM-based semi-quantitative indicators and consequently providing a “low-variability input distribution” necessary for clinical decision-making and the learning and inference of medical artificial intelligence.

[0084] Meanwhile, the present invention relates to a paramagnetic metal ion (Fe) on a surface donor group (primary amine derived from the crosslinking agent, its -COOH post-modification residue, carboxylate, hydroxyl, phosphate-based functional group) of a non-immunostimulating polysaccharide crosslinked colloid particle. 2+ / Fe 3+ , Gd 3+ , Mn 2+By coordinating iron oxide nanoparticles or metal complexes, a metal complex-based contrast agent and carrier platform can be realized that selectively provides T1 positive or T2 negative contrast in magnetic resonance imaging (MRI). Particle properties, characterized by a nanometer-scale hydration diameter (2–8 nm), a weak negative charge (ζ -20–0 mV), and a highly substitutive cross-linked network, support PRR / BCR non-operation, keeping the immune and complement axes inactive during the imaging window. Additionally, the metal complex state, with its high density of surface coordination sites, promotes intra-extrasphere water exchange, inducing r1 enhancement (r2 / r1 inhibition). Consequently, it provides bright and stable T1 signal enhancement (or T2 signal degradation upon application of superparamagnetic iron oxide) relative to the same dose, while maintaining predictable pharmacokinetic characteristics characterized by lymph selectivity, avoidance of venous contamination, and renal excretion.

[0085] Non-targeted formulations are designed to minimize non-specific binding to cell surface receptors and intracellular uptake, providing stable semi-quantitative input within an operating range that follows the physiological pathway from interstitial to lymphatic to bloodstream and avoids non-linear and saturation zones in background tissues. If necessary, they can be converted into targeted contrast agents / carriers by attaching proprietary binding modules (antibodies, Fab / ScFv, peptides, small molecule ligands, sugar chain mimics, etc.) to the particle surface via chemical bonding (e.g., carbodiimide, click chemistry, chelator-linker bridge) or biocompatible bonding (e.g., streptavidin-biotin). The design principles at this time are to (i) limit the surface density of the binding module within the PRR / BCR non-operational threshold to suppress disturbances caused by receptor clustering and internalization, (ii) prevent occlusion of metal coordination sites to minimize the reduction in r1 efficiency, and (iii) secure formulation metadata (dosage, infusion-scan interval, pre-formulation r1 expression by formulation) to maintain linkage with F-NEM (formulation-normalized enhancement metric)-based semi-quantitative analysis.

[0086] In the composition of targeted contrast agents, ligand binding affinity (KD), binding module / particle ratio, and linker length and flexibility are optimized to match the expression density, internalization kinetics, and microenvironment (e.g., pH, protein concentration) of the target receptor. As a result, selective accumulation within the lesion, along with localized amplification of T1 relaxation shortening (bright signal) or T2* induction (dark signal), allows for the visualization of changes at the molecular and cellular levels. Representative applications include (a) detection of micrometastasis and residual tumors by targeting markers (HER2, EGFR, etc.) in the tumor microenvironment, (b) lesion mapping by targeting adhesion molecules / immune receptors (e.g., integrins, selective TLR binding modules) in inflammatory or infected lesions, and (c) high-resolution functional and molecular imaging through targeting of amyloid / tau or specific neuroreceptors in the nervous system. Local enrichment based on targeting reduces the potential for side effects by lowering the effective dosage, and provides a signal-to-noise ratio (SNR) suitable for time-series comparisons, left-right comparisons, and analysis of base-to-base change rates centered on the F-NEM axis.

[0087] In the composition of drug delivery systems, therapeutic substances (small molecules, proteins / peptides, nucleic acids, radioisotopes) can be simultaneously loaded onto the same surface platform, or theranostic systems combining diagnosis and treatment can be designed. In this case as well, PRR / BCR non-operability, ζ-potential, hydration diameter, and maintenance of metal coordination sites are key control variables, and distribution, penetration, and efflux can be monitored via T1 / T2 signals on MRI after administration to optimize dosage and scheduling. In particular, lymphoselective pathways suppress unnecessary capture in immune cell-dense regions (lymph nodes and collecting ducts), and in areas without target binding, they are rapidly eliminated through renal excretion, thereby reducing the risk of retention in the body and metal accumulation.

[0088] The operational advantages are summarized as follows. First, F-NEM-centric semi-quantitative reporting is possible through the generation of consistent T1 / T2 contrasts via metal coordination chemistry under identical formulation and quality control (QC) conditions. Second, the non-functional design minimizes anthropogenic variations in target density and the microenvironment during the imaging window, thereby enhancing the stability of time-series data. Third, easy switching between non-target and targeted modules enables customization by indication, and combined with standardized formulation metadata facilitates multi-institutional comparison. Fourth, it seamlessly integrates with MR lymphangiography (MRL), volumetric MRI, and medical AI pipelines to provide standardized numerical reports linked to time-signal indicators such as PSE, WIR, TTP, and AUC, morphological indicators such as ES, NF, and LV, and even the synthetic indicator TDI.

[0089] Accordingly, the metal complex-based non-immunostimulating polysaccharide cross-linked colloid platform of the present invention serves as a targeted and non-targeted contrast agent and drug carrier, providing a high-performance MRI solution that simultaneously satisfies non-agonist signal generation, predictable pharmacokinetics, and F-NEM-centered semi-quantitative analysis across lymphatic diseases, inflammatory / fibrotic diseases, and neoplastic lesions.

[0090] Meanwhile, the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention are designed to minimize the possibility of recognition by pattern recognition receptors (PRRs) expressed by innate immune cells (e.g., macrophages, dendritic cells, neutrophils, etc.), thereby reaching inflammatory lesions without stimulating immune cells and effectively being expelled without remaining in tissues, thereby increasing their potential for use as imaging contrast agents or therapeutic adjuvants (Example 7). The colloidal particles of the present invention can expand the therapeutic window by extending the circulation half-life through immune system evasion. Immune-derived side effects, such as inflammation or hypersensitivity reactions, are suppressed. Based on immune phagocytosis evasion characteristics, targeted drug delivery to specific tissues or cells is possible.

[0091] The “contrast-free carrier mode,” which does not coordinate metal or iron oxide nanoparticles, uses the same non-immunostimulating polysaccharide cross-linked colloidal backbone but does not confer surface coordination metals, and binds drugs (small molecules, peptides / proteins, nucleic acids, immunomodulators, etc.) to -COOH / amine / catechol donors using hydrolytic, reducing, pH-sensitive linkers. The 2-8 nm size and -20-0 mV surface charge suppress venous leakage and preferentially follow the interstitial → lymphatic vessel → lymph node pathway, enabling targeted delivery and local release to superficial / deep lymph nodes. Since there is no metal, it does not generate intrinsic MRI contrast, but it safely achieves aggregation at target sites (e.g., lymph nodes) for therapeutic purposes (e.g., TLR agonists, anticancer / antifibrotic agents, nucleic acid therapeutics) and, if necessary, can be tracked with trace amounts of labeling (fluorescent / radiolabeling, or combined with isolated contrast tracers). As shown in Figs. 4, 6, and 7, even when loaded with an active ligand, the particle backbone maintains PRR / BCR non-operation, thereby reducing acute toxicity and apoptosis (maintaining cell viability), and mitigates systemic exposure and side effects through local concentration and delayed release of the drug. This carrier mode can be operated in combination or alternately with the contrast agent mode, and since it is manufactured and quality-controlled using the same formulation metadata, it offers high standardization and scalability for clinical application.

[0092] Accordingly, the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention can be combined with drugs (including contrast agents) or utilized as drug delivery vehicles, and are useful as a means to maintain high local concentrations of drugs for therapeutic or diagnostic purposes at a target site. The non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention have various biomedical application possibilities as drug delivery vehicles, imaging contrast agents, immune-evasive therapeutic adjuvants, etc. In particular, they are useful as delivery platforms for inflammatory diseases, autoimmune diseases, or vaccines.

[0093]

[0094] 3. Principles of MRI Signal Generation and T1 / T2 Mechanisms (MR Physics & Relaxivity Essentials)

[0095] The present invention adopts a contrast principle that stably secures T1 positive contrast by integrating the “relaxation mechanism of water protons” and “relaxation enhancement via surface coordination chemistry.” The signal observed in magnetic resonance imaging is mainly formed by the difference in recovery rates between transverse magnetization and sessile magnetization of water protons, where T1 (sessile relaxation) and T2 (transverse relaxation) are determined as a function of tissue, environment, and contrast agent interactions. The contrast agent accelerates relaxation by inducing magnetic field fluctuations in adjacent water molecules, and broadly speaking, the signal is enhanced by (i) the inner sphere (when primary hydration number q is directly coordinated to the center of the metal), (ii) the outer sphere (water diffusing through the vicinity of the metal), and (iii) exchange (alternation of coordinated and non-coordinated water). The interaction between the contrast agent, water, and magnetic field can be understood using the classical Solomon-Bloomberg-Morgan (SBM) framework, and the key process variables are rotational correlation time (τR), water residence time (τM), primary hydration water (q), electron spin relaxation, and outer-sphere interaction strength, which is dependent on diffusion and viscosity.

[0096] (1) Principle of T1 positive contrast formation

[0097] T1 contrast accelerates seeding recovery, making the signal “bright” in T1-weighted images. In the endurance sphere contribution, if q is 0, there are no direct coordination sites, reducing the potential for r1 up, while if q is excessively large, τM becomes long, which can degrade T1 efficiency. r1 is maximized at an optimal q (typically 1 to 2) and an intermediate range of τM (too short results in insufficient interaction, too long results in exchange bottlenecks). Since the endurance sphere contribution depends on surface coordination site density and diffusion control, small, spherical nanostructures that provide a larger surface-to-volume ratio (S / V) relative to the same mass are advantageous. The polysaccharide cross-linked colloidal particles of the present invention maintain a large S / V with a hydration diameter of 2 to 8 nm, and amplify extrasphere interactions by forming stable surface coordination with paramagnetic centers such as Fe(III) or Mn(II) through surface carboxylate / amine / hydroxyl / phosphate / catechol donor groups. At the same time, the high-substitution cross-linking (≥60% relative to monosaccharides, preferably ≥90%) suppresses molecular rotational degrees of freedom to increase τR, while suppressing the increase in magnetization sensitivity caused by excessive aggregation / coagulation (unnecessary increase in r2). As a result, r1 is increased while r2 / r1 is kept low, making the T1 positive contrast dominant.

[0098] (2) Control of T2 and T2* Decay

[0099] T2 contrast “darkens” the signal through the acceleration of transverse magnetization decay. In the presence of large magnetic sensitivity (magnetic susceptibility) non-uniformity or a magnetic core (e.g., superparamagnetic iron oxide), the local field gradient increases, leading to prominent T2* decay and potentially eroding T1 contrast. The present invention (i) predominates T1 contribution by surface-coordinating paramagnetic ions such as Fe(III) / Mn(II) around a metal center, (ii) prevents unwanted magnetization clusters with a weak negative charge surface (ζ -20 to 0 mV) that suppresses aggregation between nanoparticles, and (iii) prevents T2-induced distortion from interfering with F-NEM output by automatically detecting and masking T2* patterns (maintaining high signal only in spin echo compared to gradient echo, abnormal loss in long TE) during the quality control (QC) stage. Conversely, for applications requiring T2 / T2* contrast (e.g., SPIO-based T2 negative contrast), only surface modification may be applied to the same platform, but the primary implementation of this specification aims for T1 positive contrast.

[0100] (3) Correlation between physical properties and relaxation ability (r1 / r2) and operating range

[0101] Contrast signals exhibit concentration-signal nonlinearity. At low concentrations, the signal is weak due to insufficient contact probability; at intermediate concentrations, the signal is maximized due to an increase in r1; and at high concentrations, it may decrease again due to the influence of saturation and T2*. The present invention practically expands the "intermediate concentration range" through the structure and physical properties of the contrast agent, and systematically excludes nonlinear regions through operating range and masking rules during the analysis phase. Specifically, a hydration diameter of 2–8 nm increases the density of surface coordination points exposed per unit mass, thereby strengthening extrasphere interactions; high-substitution crosslinking increases τR to enhance r1; and a weak negative charge of -20–0 mV inhibits non-specific binding of serum proteins, thereby avoiding aggregation and cluster-induced r2 increases. This combination maximizes effective r1 exposure at the same dosage while keeping the r2 / r1 ratio low, thereby ensuring stability relative to T1. During the acquisition and analysis phases, the percentage signal enhancement (PSE) is limited to a reasonable range (e.g., greater than 0% and less than or equal to 300%), voxels with rapid signal changes (changes of ≥4 times compared to the previous time point) are automatically masked, and suspected T2* regions are excluded. This operational range concept simultaneously reflects the advantages of contrast agent properties and the constraints of signal physics to ensure that F-NEM maintains a monotonic relationship (consistent change over dose and time).

[0102] (4) The role of surface coordination chemistry (creation of non-active contrast)

[0103] The present invention promotes the exchange of the inner and outer spheres by densely distributing coordination-capable donor groups on the surface of a non-immunostimulating polysaccharide cross-linked network and immobilizing paramagnetic metal centers (Fe(III), Mn(II), etc.) as surface coordination groups. The microstructure of the surface donor group-metal-water (i) amplifies the magnetization fluctuations (electron-nuclear pair interactions) of water near the surface to induce T1 shortening, (ii) increases the outer sphere r1 through repeated diffusion-contact-detachment cycles, and (iii) strengthens the temporal correlation with nuclear spin through the slow rotation (increased τR) of the immobilized metal center. At the same time, PRR / BCR non-operation is achieved through high-substitution cross-linking that maintains the surface charge at a weak negative charge and lowers the spatial accessibility of epitopes per repeat, thereby inhibiting immune receptor occupancy, internalization, and cytokine / complement activation during the imaging window. This “Non-perturbative Contrast Generation” reflects inherent changes in pathophysiology (vascular permeability, interstitial retention, lymphatic drainage) without disturbance, thereby enhancing the stability of the input distribution in same-patient time series and multi-center comparisons.

[0104] (5) Handling of magnetic field strength (B0), temperature, and medium dependence

[0105] r1 / r2 depends on B0, temperature, and the viscosity and ionic strength of the medium. Generally, r1 increases with τR·τM optimization in medium magnetic fields (e.g., 1.5–3.0 T), but may decrease in high magnetic fields due to electron spin relaxation characteristics. The present invention does not aim for “absolute quantification” but performs semiquantitative, trend-based interpretation using a formulation-normalized enhancement index (F-NEM) utilizing formulation metadata (dosage, injection-scan interval, lot-specific r1 characteristics, magnetic field strength). ΔR1 is selectively used as an auxiliary index only when a T1 map obtained with the same parameters is available, and is adopted only under operating range, masking, and QC rules.

[0106] (6) Suppression of Aggregation and Protein Binding

[0107] Contrast agent aggregation and aggregation increase microscopic susceptibility gradients, thereby elevating r2* and undermining T1 contrast. The present invention reduces non-specific interparticle interactions and minimizes binding to serum proteins through a size of 2–8 nm, weak negative charge, and high-substitution crosslinking, thereby maintaining monodisperse behavior in the circulation and interstitium environment. This ensures temporal stability and spatial homogeneity of T1 contrast, thereby increasing the reliability of time-signal indicators such as F-NEM, PSE, WIR, TTP, and AUC, as well as morphological indicators such as ES, NF, and LV.

[0108] In summary, the contrast platform of the present invention stabilizes T1-positive contrast through a multilayer strategy comprising (i) enhancing S / V and exosphere r1 with a hydration diameter of 2–8 nm, (ii) inhibiting protein binding and aggregation with a surface of -20–0 mV, (iii) maximizing r1 by increasing τR with high-substitution crosslinking, (iv) promoting intra-exosphere exchange with surface donor group-metal center coordination, and (v) excluding nonlinear and T2* effects with QC-based operating range and masking. This closed loop of physics-chemistry-interpretation practically controls contrast agent concentration-signal nonlinearity and equipment and environment dependence, and ultimately enables semi-quantitative readings and time-series trend analysis centered on the formulation-normalized enhancement index (F-NEM).

[0109] The above mechanism is combined with the previously presented lymph-selective pharmacokinetics (avoidance of venous contamination, interstitial-to-lymph-priority drainage) and PRR / BCR non-operational design to provide stable T1-positive contrast within the imaging window—i.e., low-variance inputs. Therefore, the quantitative imaging framework of the present invention can sensitively and reproducibly capture clinically significant “changes” and “differences” while avoiding the limitations of absolute-value-centered approaches.

[0110]

[0111] 4. Non-immunostimulating polysaccharide cross-linked colloidal particle composition (Composition & Surface Coordination)

[0112] The composition of the present invention is a polysaccharide cross-linked colloidal particle designed to be stably dispersed in aqueous media, such as lymph nodes, without gelation or aggregation, and to satisfy pattern recognition receptor (PRR) and / or B cell receptor (BCR) non-operativity. Key design parameters are (i) a hydration diameter of 2–8 nm (allowable 2–10 nm), (ii) a weak negative ζ-potential of -20–0 mV, (iii) a compact, highly crosslinked three-dimensional spherical network formed by epoxide-based selective modification, and (iv) optionally metal ions (Fe₂) through surface donor groups (carboxylate / amine / phosphate / hydroxyl / catechol). 2+ / Fe 3+ , Gd 3+ , Mn 2+ It is a coordination bond of ) or iron oxide nanoparticles, and this combination induces pharmacokinetics characterized by lymphoselectivity, venous contamination avoidance, and renal excretion. The non-perturbative design enhances the time-series stability of the input signal by inhibiting immune receptor occupancy, internalization, and cytokine / complement activation during the imaging window.

[0113] The particles according to the present invention have a compact spherical three-dimensional network in which intramolecular and intermolecular crosslinking occurs through epoxide selective modification of monosaccharide -OH functional groups, so the rotational degrees of freedom are reduced, making it difficult for the three-dimensional arrangement of the surface to form high-affinity complementarity with binding sites of PRRs (e.g., TLRs, C-type lectin receptors), BCRs, or antibodies. Furthermore, a weak negative charge (ζ-potential -20 to 0 mV) achieved by post-modification of surface amine groups with -COOH inhibits electrostatic non-specific adsorption, opsonization, and complement activation. As a result, PRR / BCR-mediated signaling (e.g., NF-κB pathway activation, cytokine release) is significantly inhibited, so the immune system is not switched on, and ADA formation is not observed even in a repeated administration environment.

[0114] When the hydration diameter is limited to 2 to 10 nm (preferably 2 to 8 nm), it does not penetrate the normal capillary wall, and when injected into the interstitial space, at least 10%, preferably 20%, more preferably 30%, and even more preferably 50% of the injected amount remains in the interstitial space for more than 1 hour at the injection site, after which it is preferentially discharged into terminal lymphatic vessels without venous contamination. This induces expansion of the interstitial space or anatomical distension, and can function as a filler or space expander (Examples 4 to 7). The mechanism of extended residence time is interpreted as being due to the fact that the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention have a larger molecular weight and size compared to monomolecular drugs, and their diffusion within tissue structures such as the extracellular matrix (ECM) is restricted. In fact, it was confirmed that when a T1-MRI contrast agent based on the particles of the present invention is injected into the joint cavity, intradermis, or subcutaneously, it is excreted only into the lymphatic vessels and imaging is maintained for 1 to 2 hours after injection (see Examples 4-6). Furthermore, the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention serve as a localized drug delivery system, compensating for adverse pharmacokinetic / pharmacodynamic (PK / PD) characteristics that may arise from the rapid systemic diffusion of a therapeutic agent, and contributing to the reduction of side effects and the improvement of therapeutic efficacy. In addition, the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention can be modified into ligands, agonists, or antagonists capable of binding to specific cell surface receptors through amine groups or -COOH functional groups on the surface; in this case, they can selectively remain at the target site for at least 1 day, preferably at least 4 days, and sustain local action (Examples 4-7). In particular, when an antibody or small molecule ligand is bound, it can specifically bind to a target cell expression receptor and then be eliminated from the body via the lymphatic system, thus enabling simultaneous targeting and regulation of biodistribution.Therefore, the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention serve as a technology platform offering multifunctional advantages such as extended residence time at the injection site, expansion of anatomical structures, selective lymphatic drainage, cell targeting function, inhibition of systemic diffusion, and improved local diagnostic and therapeutic efficiency, making them highly suitable as diagnostic and therapeutic compositions or drug delivery systems.

[0115] The fraction that enters systemic circulation is filtered by the renal glomeruli and excreted in the urine; during this process, the uptake rate of the hepatic and splenic macrophage systems is low, which reduces the risk of long-term retention and immune side effects. In addition, a high cross-linking rate (at least 60% of the total number of monosaccharides, preferably at least 70%, more preferably at least 90% or 95%) confers resistance to endogenous glycoside hydrolases, thereby suppressing exposure to potential sugar-based immunogenic sites and minimizing the impact of structural heterogeneity on imaging and drug delivery performance.

[0116] According to the examples, the dextran crosslinker-based T1 contrast agent of the present invention showed (1) imaging findings of migration to central lymphatic vessels without long-term retention or capture in compartments with high immune cell density, such as lymph nodes (Fig. 18, indirect evidence of reduced interaction with immune cell surface receptors), (2) no significant activation of NF-κB reporters in a macrophage model (Fig. 4), and (3) suppression of cytokine release without reduced cell viability even at high concentration exposure (Fig. 6). (4) In animal repeated-dose toxicity tests (e.g., rats up to 1,500 mg / kg, repeated administration) and initial clinical safety assessments, no significant changes related to systemic inflammatory markers, hypersensitivity reactions, or hematological abnormalities were observed, and it was confirmed that the maximum allowable dose was not reached and no serious drug-related adverse events occurred (Tables 4, 7, and 8). These results support the fact that the platform of the present invention maintains a low risk of immune-mediated side effects even with repeated administration.

[0117] (1) Overview of Composition and Synthesis

[0118] After selectively modifying the -OH repeating units of a substrate polysaccharide (e.g., dextran, pullulan, inulin, hyaluronan derivatives, etc.) with an epoxide-based first crosslinking agent, a spherical three-dimensional network is formed by (a) direct bonding between the modified -OHs and / or (b) intramolecular and intermolecular crosslinking using a polyvalent amine second crosslinking agent. The crosslinking substitution rate relative to the total number of monosaccharides is set to at least 60% (preferably 70% or more, more preferably 90% or more, even more preferably 95% or more) to reduce the spatial accessibility of the repeating sugar epitopes and confer enzyme resistance. Subsequently, the surface-terminal amines are post-modified with an organic acid anhydride (e.g., succinic anhydride) or a carbonyl activator to adjust the -COOH ratio, thereby setting the final ζ-potential to -20 to 0 mV. The hydration diameter is tuned to a range of 2 to 10 nm (preferably 2 to 8 nm).

[0119] (2) Surface coordination chemistry (Fe / Mn / Gd and iron oxide)

[0120] On the surface of the cross-linked network, carboxylates (-COO) - ), amine (-NH2), phosphate (-PO4 2- ), hydroxyl (-OH), and catechol (-(OH)2-Ar) donor groups are distributed. Metal ions are coordinately bonded to these donor groups to form T1 (Fe 2+ / Fe 3+ , Gd 3+ , Mn 2+ It is possible to realize contrasts with paramagnetic iron oxide nanoparticles) or T2* (superparamagnetic iron oxide). Typically, metal salts (e.g., FeCl3, MnCl2) are titrated at pH 6.8–7.4 and 4–25 °C, and relaxivity and r2 / r1 are optimized by adjusting the surface metal / carbohydrate repeating unit ratio (e.g., 1:20–1:100). The introduction of catechol (e.g., dopamine-carbonyl bonds) increases the surface coordination site density, thereby affecting endurance sphere water content (q) and exchange rate (k). ex) is enhanced, and r1 enhancement and r2 / r1 suppression are simultaneously achieved. When superparamagnetic iron oxide is applied (core-shell introduction or surface anchoring), it can be operated as a T2 model, and the basic embodiment of the present invention is at 37 °C, corresponding B0, r2 / r1 ≤ 5 and metal reference r1 ≥ 1.0 mM so that T1 positive contrast predominates. -1 ·s -1 It aims for.

[0121] (3) Correlation between physical properties and functions and principles of operation

[0122] The nanometer (2–8 nm) size increases the surface-to-volume ratio, thereby increasing coordination site exposure per unit mass and inhibiting microvascular endothelial penetration, which minimizes venous capillary leakage. A weak negative charge of -20–0 mV reduces non-specific protein binding and opsonization, lowering hepatic and splenic macrophage capture and causing preferential tracking of the interstitial → lymphatic capillary → peripheral lymphatic → lymph node pathway. High-substitution crosslinking inhibits the geometric exposure of PRR / BCR binding motifs, supporting immuno-non-operation, while simultaneously increasing resistance to glycoside hydrolases to maintain structural stability in the imaging window. Consequently, the T1 signal is primarily proportional to the target physiology of vascular permeability, interstitial retention, and lymphatic drainage without contrast agent-origin disturbance, serving as the basis for calculating the formulation-normalized enhancement metric (F-NEM) normalized by formulation metadata.

[0123] (4) Manufacturing method (example)

[0124] The polysaccharide cross-linked colloidal particles of the present invention may be provided by a manufacturing method comprising the following steps as a non-limiting example:

[0125] Step 1: preparing an aqueous solution of linear polysaccharides, branched polysaccharides, or cyclic polysaccharides;

[0126] A second step of modifying the -OH functional groups of monosaccharides, which are the building blocks of polysaccharides, with the first crosslinking agent by adding a first crosslinking agent having an epoxide functional group that reacts with the hydroxyl group (-OH) of the polysaccharide with an alkaline aqueous solution;

[0127] A third step of dropwise adding a second crosslinking agent having two or more amine groups (-NH2) to produce polysaccharide crosslinked colloid particles having terminal amine groups derived from the second crosslinking agent on the surface, wherein spatially adjacent functional groups modified by the first crosslinking agent are intramolecularly and / or intermolecularly crosslinked through the second crosslinking agent;

[0128] A fourth step of administering an organic acid anhydride to polysaccharide crosslinked colloid particles having terminal amine groups derived from a second crosslinking agent on their surface to modify some or all of the terminal amine groups into carboxylic acid groups and / or carboxylate groups; and

[0129] Optionally, iron ions (Fe) in the water-dispersible polysaccharide cross-linked colloidal particles prepared in the previous step 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ Step 5: Preparing a composite in which the surface of iron oxide nanoparticles is modified into polysaccharide cross-linked colloidal particles by administering a precursor (e.g., iron chloride) or an aqueous solution of iron oxide nanoparticles that provides ).

[0130] In order to prevent an unintended increase in hydration size due to swelling of linear polysaccharide, branched polysaccharide, or cyclic polysaccharide molecules, and to reduce or minimize immunogenicity to immune cells relative to the polysaccharide being crosslinked without being hydrolyzed by enzymes in the body so as to be excreted without exposure to potential sugar-based immunogenic sites, the present invention crosslinks linear polysaccharides, branched polysaccharides, or cyclic polysaccharides intramolecularly and / or intermolecularly in an aqueous solution.

[0131] The non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention may be formed by intermolecularly cross-linking 1 to 3 linear polysaccharides or branched polysaccharides, or 1 to 30 cyclic polysaccharides, in an aqueous solvent.

[0132] The linear polysaccharides, branched polysaccharides, or cyclic polysaccharides to be crosslinked may be homopolysaccharides or heteropolysaccharides. Non-limiting examples of linear polysaccharides, branched polysaccharides, or cyclic polysaccharides include dextran, cyclodextrin, maltodextrin, and inulin, as illustrated in FIG. 8. In this specification, dextran also includes various derivatives thereof. Non-limiting examples of dextran derivatives include carboxymethyl dextran (CM dextran), dextran sulfate, and diethylaminoethyl dextran (DEAE-dextran).

[0133] The non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention are water-soluble colloidal amorphous nanoparticles in which linear polysaccharides (e.g., inulin) dispersed in an aqueous solvent, as well as complex branched polysaccharides (e.g., dextran) or cyclic polysaccharides (e.g., cyclodextrin) dispersed in an aqueous solvent, are cross-linked intramolecularly and / or intermolecularly to form a more complex three-dimensional network; or, as described above, at least 60%, at least 70%, at least 90%, or at least 95% of the total number of monosaccharides, which are the building blocks of polysaccharides dispersed in an aqueous solvent, are modified by a cross-linking agent, or at least one -OH functional group in at least one of two consecutive monosaccharides within the polysaccharide chain is largely modified by a cross-linking agent, thereby (i) not being hydrolyzed by enzymes in the body and (ii) having a significantly weakened binding affinity with surface receptors (PRR, BCR, etc.) of immune cells in the body, so that the immune response switch It is not activated, and as a result, immunogenicity to immune cells can be minimized so that B cell activation and antibody production are not induced.

[0134] That is, the non-immunostimulating polysaccharide cross-linked colloidal particles of the present invention are modified by a cross-linking agent in such a way that more than 50% of the monosaccharides, which are the building blocks of polysaccharides, are modified. This ensures that even if there is a certain degree of interaction with body tissues or body fluids, the binding affinity with surface receptors (PRR, BCR, etc.) of immune cells in the body is significantly weakened, thereby ensuring that unwanted immune responses are not induced.

[0135] Preferably, by controlling the crosslinking density more precisely according to the molecular weight of the polysaccharide, it can be not only not hydrolyzed by enzymes in the body, but also not hydrolyzed at the acidic pH of the stomach or lysosomes within cells.

[0136] For example, the present invention can produce polysaccharide crosslinked colloid particles in which 60% or more, 70% or more, 90% or more, or 95% or more of the total number of monosaccharides that are building blocks of polysaccharides are modified by the crosslinking agent at the -OH functional groups of monosaccharides that are building blocks of polysaccharides, by adding and reacting (i) a first crosslinking agent having an epoxide group that reacts with the hydroxyl group (-OH) of the polysaccharide and a functional group that chemically bonds with the hydroxyl group (-OH) and / or amine group (-NH2) and (ii) a second crosslinking agent having two or more amine groups (-NH2) to an aqueous solution of a linear polysaccharide, branched polysaccharide, or cyclic polysaccharide through the second and third steps, thereby modifying and crosslinking the monosaccharides that are building blocks of the linear polysaccharide, branched polysaccharide, or cyclic polysaccharide with the crosslinking agent.

[0137] In the second step, the first crosslinking agent having an epoxide group that reacts with a hydroxyl group (-OH) and a functional group that chemically bonds with a hydroxyl group (-OH) and / or an amine group (-NH2) is not limited in type as long as it modifies the -OH functional group site of the monosaccharide so that (a) between the functional group modified by the first crosslinking agent and the spatially adjacent -OH functional group, and (b) two spatially adjacent functional groups modified by the first crosslinking agent can react with the second crosslinking agent, and preferably may be a halo alkyl oxirane, for example, epichlorohydrin.

[0138] In the third step, the second crosslinking agent having two or more amine groups (-NH2) can be replaced with any crosslinking agent capable of covalently bonding to the functional group derived from the first crosslinking agent that modifies the -OH functional group site of the monosaccharide, and this also falls within the scope of the present invention.

[0139] In one embodiment of the present invention, when epichlorohydrin is used as a first crosslinking agent to modify the -OH functional groups of a monosaccharide, ethylenediamine or diethylenetriamine (DETA) may be used as a second crosslinking agent to participate in crosslinking between spatially adjacent modified functional groups (Figs. 1a and 1b).

[0140] When linear polysaccharides, branched polysaccharides, or cyclic polysaccharides react with epichlorohydrin as a first crosslinking agent and ethylenediamine or DETA as a second crosslinking agent, a series of chemical modifications occur that significantly change the structure and properties of the polysaccharides.

[0141] According to the present invention, when polysaccharides are modified (combined modifications) by using epichlorohydrin as a first crosslinking agent and ethylenediamine or DETA as a second crosslinking agent, the crosslinking increases and stability is improved.

[0142] In particular, terminal amine groups derived from the second crosslinking agent exposed on the surface of polysaccharide crosslinked colloid particles can impart new properties such as enhanced chelating ability, increased reactivity, and additional functionalization potential. Therefore, polysaccharide crosslinked colloid particles having terminal amine groups derived from the second crosslinking agent on their surface, generated in the third step, can enhance their ability to interact with other molecules, such as chelation or drug binding, through the amine groups (-NH2).

[0143] Polysaccharide crosslinked colloidal particles having terminal amine groups derived from the second crosslinking agent formed in the third step exposed on the surface can be modified through the fourth step to expose -COOH functional groups on the surface, thereby controlling the surface charge to -20 mV to 0 mV and enabling the particle to have a biocompatible pH. Additionally, the number of amine groups derived from the crosslinking agent exposed on the surface of the polysaccharide crosslinked colloidal particles of the present invention can be controlled through a reaction to introduce -COOH-containing functional groups.

[0144] For example, when amine groups derived from the crosslinking agent exposed on the surface of polysaccharide crosslinked colloid particles are modified into -COOH-containing functional groups, the surface charge can be freely controlled within the range of -20 mV to 0 mV by adjusting the type and / or degree of modification of the -COOH-containing functional groups. In the fourth step, for example, by adjusting the amount of succinyl anhydride (SA), the surface charge can be freely controlled within the range of -20 mV to 0 mV. As the amount of added SA increases, the surface charge takes on a negative value.

[0145] In the polysaccharide crosslinked colloid particles of the present invention, the interaction with water and other molecules is altered when the amine group derived from the crosslinking agent exposed on the surface is replaced with a carboxyl group.

[0146] For example, in the fourth step, the amount of organic acid anhydride administered can be controlled to control the number of amine groups derived from the crosslinking agent exposed on the surface of the polysaccharide crosslinked particles of the present invention. Accordingly, by controlling the reaction ratio with the drug, the number of functional molecules (e.g., drugs, targeting molecules) bound to the polysaccharide crosslinked colloid particles can also be controlled.

[0147] The polysaccharide crosslinked colloidal particles of the present invention can be bonded to functional nanoparticles, molecules, or metal ions by covalent or coordinate bonding through amine groups and / or -COOH functional groups derived from the crosslinking agent exposed on their surface.

[0148] The polysaccharide crosslinked colloidal particles of the present invention can be coordinately bonded to a core composed of iron oxide-based nanoparticles through crosslinking agent-derived amine groups and / or -COOH functional groups exposed on their surface. The polysaccharide crosslinked colloidal particles of the present invention enable strong coordination bonding between the crosslinking agent-derived functional groups (carboxyl groups or amine groups) and the core surface composed of iron oxide-based nanoparticles, thereby imparting colloidal stability.

[0149] In step 5, iron ions (Fe 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ Non-limiting examples of hydrophilic functional groups that coordinate with iron oxide nanoparticles include hydroxyl, carboxylic acid, carboxylate, amine, etc.

[0150] According to the present invention, iron ions (Fe 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ Polysaccharide cross-linked colloidal particles coordinated with ) can be designed and synthesized to function as a T1 MRI contrast agent that shows a bright signal in MRI images through steps 1 to 5, so that the location of the polysaccharide cross-linked colloidal particles can be tracked through MRI images after injection into the body.

[0151] The polysaccharide cross-linked colloid particles according to the present invention can be designed and synthesized to maintain a contrast effect for a relatively long period, thereby extending the scan time during MRI imaging and improving the spatial resolution of MRI, thus enabling imaging at a higher resolution. Accordingly, the polysaccharide cross-linked colloid particles according to the present invention can image lymphatic vessels via MRI without venous contamination, which are clinically very important in vivo but difficult to observe with conventional contrast agents.

[0152] In short, the polysaccharide crosslinked colloidal particles of the present invention have functional groups derived from the crosslinking agent exposed on their surface and iron ions (Fe 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ If ) is coordinately coupled, it can also function as a T1 or T2 MRI contrast agent.

[0153] In addition, metal oxide nanoparticles, such as iron oxide, designed to act as T1 or T2 MRI contrast agents can also perform their role as T1 or T2 MRI contrast agents by surface-modifying them into polysaccharide cross-linked colloidal particles in the fifth step, thereby enabling them to move between internal structures without aggregation upon injection into the body.

[0154] Accordingly, since the polysaccharide cross-linked colloid particles according to the present invention can perform the role of an MRI contrast agent, the location of the polysaccharide cross-linked colloid particle-based drug can be tracked via MRI imaging after injection into the body to determine whether it is selectively excreted into lymphatic vessels without venous contamination, whether it is phagocytosed by macrophages after injection into the body, whether it is metabolically degraded, whether it is circulated in the blood, whether it is delivered to the cellular parenchyma through capillaries, whether it accumulates in tissues, whether it is excreted in urine through the kidneys, whether it is excreted in feces, whether it is absorbed into the vascular circulatory system after injection into the body, whether it leaks through the blood vessel walls, whether it can be collected and reused through urine / feces, and whether it is translocated into cells.

[0155] Meanwhile, according to the present invention, polysaccharide crosslinked colloid particles formed by intramolecular and / or intermolecular crosslinking of linear polysaccharides, branched polysaccharides, or cyclic polysaccharides with a crosslinking agent may be subjected to chemical modifications such as adding, removing, or modifying functional groups (-COOH, -NH2, -OH, etc.) so that they are exposed on the surface. Such modifications may be performed to impart new chemical properties or to enhance existing properties.

[0156] In the present invention, the crosslinking agent-derived functional group exposed on the surface of the polysaccharide crosslinked colloidal particle may be the terminal functional group of the crosslinking agent itself or a modified / substituted functional group thereof. For example, the crosslinking agent-derived functional group exposed on the surface of the polysaccharide crosslinked colloidal particle may be one in which at least some of the functional groups of the crosslinking agent exposed on the surface have been modified / substituted.

[0157] In the present invention, the hydrophilic functional group may be derived from a functional group of a linear polysaccharide, branched polysaccharide, or cyclic polysaccharide that did not participate in the crosslinking reaction, a functional group of a crosslinking agent that did not participate in the crosslinking reaction, and / or a functional group obtained by further modifying a one-terminal of a crosslinking agent exposed after the crosslinking reaction.

[0158] Non-limiting examples of functional groups derived from the crosslinking agent or hydrophilic functional groups exposed on the surface of polysaccharide crosslinked colloid particles include amine groups, carboxyl groups, hydroxyl groups, and / or thiol groups.

[0159] The -OH functional groups of monosaccharides, which are the building blocks of polysaccharides, and / or functional groups derived from crosslinking agents can be modified through various covalent linkages as shown in Table 1 below.

[0160]

[0161] Reactive functional groups such as amines, thiols, carboxyls, and hydroxyls facilitate not only surface modification but also chemical binding with biopharmaceuticals or various types of small molecule drugs, such as ligands that specifically bind to receptors on specific cells, antibodies or fragments thereof, antigenic peptides, and nucleic acids (DNA, RNA, or fragments thereof). Additionally, in some cases, they may be degraded by the acidic atmosphere (pH ≤ 7) surrounding the tumor or by hydrolytic enzymes to release the active form of the drug. For example, non-limiting examples of acid-sensitive bonds that degrade in the acidic atmosphere (pH ≤ 7) surrounding the tumor include carbonate or ester bonds.

[0162] If the functional groups exposed on the surface of the polysaccharide cross-linked colloid particles of the present invention carry a positive charge such as amines, cytotoxicity may occur as with other cationic polymers, but this can be resolved by substituting some or all of the amine groups with carboxyl groups, methyl groups, ethyl groups, etc.

[0163] Non-limiting examples of hydrophilic functional groups that coordinate with metal ions (e.g., iron ions) or iron in iron oxide include amines, thiols, carboxyls (carboxylates and carboxylic acids), hydroxyls, etc.

[0164] Accordingly, the polysaccharide cross-linked colloidal particles of the present invention can be used to produce various polysaccharide cross-linked colloidal particle derivatives with unique properties suitable for specific applications through chemical modifications of functional groups such as -COOH, -NH2, and -OH exposed on their surface, and these also fall within the scope of the present invention.

[0165] (5) Selective functionalization

[0166] The basic formulation is based on the principle of non-targeting (minimizing binding to cell surface receptors), but tissue selectivity can be imparted if necessary by conjugating antibodies, peptides, or small molecule ligands to surface -COOH / amine / catechol donors at low density (e.g., occupying 0.1-1% of donor sites) using EDC / NHS or click chemistry. Even during conjugation, the final ζ-potential and hydration diameter are maintained within specifications to preserve lymph selectivity and venous contamination avoidance characteristics. The degree of -COOH substitution of surface amines is utilized as a means to fine-tune the number of bindings of drugs or targeting molecules, as well as their hydrophilicity and charge.

[0167] (6) Operating effect

[0168] The above property optimization (2-8 nm, ζ -20~0 mV, high-substitution crosslinking, controlled surface coordination) simultaneously achieves (i) minimization of immune disturbance based on PRR / BCR non-operation, (ii) avoidance of venous contamination due to the interstitial→lymphatic priority pathway, and (iii) reduction of tissue residue and metal accumulation due to renal excretion dominance. Accordingly, the variance of input signals is reduced in same-patient time series and inter-institutional comparisons, and reproducible calculation of time-signal indices (PSE, WIR, TTP, AUC), conformation indices (ES, NF, LV), and synthesis indices (TDI) is made possible with F-NEM normalized by formulation metadata as the axis.

[0169] (7) Scope of application

[0170] The present composition can be used for both intravenous whole-body T1 imaging and intradermal / subcutaneous micro-injection MR lymphangiography, and provides standardized input for early diagnosis and monitoring of treatment response in lymphatic diseases, inflammatory and fibrotic diseases, neoplastic lesions, and various other lesions. If necessary, it can be converted to a T2 model (iron oxide) for negative contrast application, but the basic implementation of the present invention focuses on the stable securing of T1 positive contrast.

[0171]

[0172] 5. Semi-quantitative T1 Workflow & Server System

[0173] In this specification, the 'formulation-normalized enhancement metric (F-NEM)' refers to a semi-quantitative metric that standardizes the enhancement signal by reflecting formulation metadata, and is used as an axis for calculating time-signal metrics (PSE, WIR, TTP, AUC), form metrics (ES, NF, LV), and synthesis metrics (TDI).

[0174] ΔR1 is an auxiliary indicator rather than a core basis for judgment in the present invention, and can be selectively referenced within a predefined operating range, masking, and quality control (QC).

[0175] The non-immunostimulating composition of the present invention satisfies PRR / BCR non-operability to minimize physiological disturbance during the imaging window and supports input stability for trend-based readings centered on F-NEM.

[0176] The “AI learning target and labeling scheme” can be designed as follows to be combined with the upper pipeline of the T1 semi-quantitative imaging workflow of the present invention and the server system (administration-acquisition-F-NEM calculation-time-signal / morphological indicator calculation-structured report-audit log). The prerequisites are semi-quantitative input centered on the formulation-normalized enhancement metric (F-NEM), limitation to frames and voxels that pass quality control (QC), and the use of ΔR1 as an optional auxiliary indicator only under operating range, masking, and QC.

[0177] First, the operating principle of input, preprocessing, and data standardization. After contrast agent administration and image acquisition, the server applies QC rules (automatic exclusion of frames in which venous contamination and extravasation are detected in time-resolved images, pre-defined SNR hurdle check, verification of flip angle and B1 correction suitability, and / or T2* influence interval masking), and then normalizes the signal with formulation metadata (administration amount, infusion-scan interval, r1 characteristics by lot and magnetic field strength) to generate an F-NEM map. The training input consists of (i) an F-NEM time series map (resampled to the same anatomical coordinate system), (ii) volumes or vectors of time-signal derived indicators (PSE, WIR, TTP, AUC), (iii) shape indicators (ES, NF, LV) and masks, and (iv) optionally a ΔR1 map, and all samples are combined with DICOM metadata normalization and HL7 FHIR-compatible clinical context (diagnosis, time of measurement, treatment event).

[0178] Second, the learning task is defined. The AI ​​module of the present invention consists of the following representative tasks: (1) Classification: Nodal filling delay, reflux presence / grade, bypass asymmetry, lesion activity (active / inactive), fibrosis co-occurrence, etc. are calculated as binary / multi-class probabilities. (2) Segmentation: Superficial and deep lymphatic networks, lymph nodes, dermal reflux regions, lesion boundaries, and necrotic cores are derived as 2D / 3D masks (based on the calculation of ES·NF·LV). (3) Temporal prediction: The PSE / WIR / TTP / AUC trajectories and TDI change rates at future k-times are estimated from the F-NEM time series to preemptively present Progression / Stable / Response flags. (4) Multi-target regression: Quantitative scores (e.g., F-NEM peak, TAR (if used in the relevant scenario), ES score) are predicted as continuous values ​​and directly input into the threshold-based judgment.

[0179] Third, there is the labeling system and domain. Based on the F-NEM standardized frame, the labels are stratified as follows: ① Functional labels: Lymph node filling time delay (minutes or relative frames), reflux frequency (ratio of reflux frames to total frames), bypass path asymmetry (left-right or lesion-control bypass network length / signal ratio), wash-out index. ② Morphologic labels: ES (boundary gradient median / quantile-based score), NF (ratio of under-enhancement area), LV (absolute volume or rate of change relative to reference). ③ Composite labels: Tissue Damage Index (TDI; weighted combination of F-NEM, ES, NF, and LV), progression flag (e.g., TDI ≥ 20% increase relative to reference). All labels are managed by a codebook containing definitions, measurement rules, and thresholds, and are orthogonally mapped to DICOM-SR templates and HL7 FHIR Observation resources.

[0180] Fourth is the annotation protocol and consensus procedure. For granularity, the gold standard is finalized through adjudication following dual reader-independent annotation; in cases of significant boundary discrepancies, probabilistic boundaries (soft labels) are maintained instead of minority or majority voting to reflect boundary uncertainty during training. For feature labels, the server performs initial automatic calculation (rule-based / model-assisted), and readers approve or modify them using a QA checklist. Label accuracy and reproducibility are periodically evaluated using Dice, ICC, and Cohen's κ; if bias is detected, the codebook is revised and backward-compatible maps are applied to past labels.

[0181] Fifth, there is the feature representation and model structure. The input consists of (a) a 3D F-NEM map (fixed resolution), (b) a time-series F-NEM stack (channel-axis time encoding), and (c) a multimodal pack of derived metric volumes / vectors. A 3D U-Net family is used for segmentation, a CNN / Vision Transformer (ViT) backbone + metadata branch (formulation / QC flag encoding) for classification and multi-target regression, and a spatiotemporal Transformer or TCN for time-series forecasting. All models are designed to naturally ignore unavailable regions by simultaneously feeding QC masks to the input channels.

[0182] Sixth is the loss function, correction, and uncertainty handling. For multi-objective learning, segmentation Dice / BCE, classification Focal / Calibrated CE, and regression Huber / MAE are optimized as a weighted sum, and label quality weights (loss-weighted downgrade for samples with low annotation agreement) are applied. Prediction probability correction combines temperature scaling and bin-wise isotonic, and prediction uncertainty (aleatoric / epistemic) is calculated using N-Samble or MC-Dropout to support "defer to human" logic.

[0183] Seventh is the data partitioning and validation scenario. Based on non-leak partitioning at the patient level, considering differences in distribution by institution, equipment, and formulation lot, the following scenarios are implemented: (i) time-series separation within the same institution, (ii) leave-one-site-out validation, and (iii) leave-one-vendor-out validation. All validation is limited to the QC-pass and F-NEM standardization framework, and the premise of input stability is specified in the metadata.

[0184] Eighth, output, reporting, and interoperability. The model output is standardized into (1) a lesion and lymph structure mask, (2) a function score (filling delay, reflux frequency, bypass asymmetry, etc.), (3) a shape score (ES, NF, LV), (4) a synthetic indicator (TDI and rate of change), (5) a progress / stability / response flag and confidence (e.g., 95% confidence interval), and is exported to DICOM-SR and HL7 FHIR. The server simultaneously records the model version, hyperparameters, random number seed, processing time, and QC pass history in the audit log to ensure traceability and enforceability.

[0185] Ninth is continuous learning and response to distribution shifts. New data collected during operation is automatically anonymized and loaded into a data lake, and changes in label quality and distribution (e.g., drift in F-NEM distribution, increase in QC-fail rate) are monitored. Semi-periodic retraining is performed when predefined thresholds are exceeded, while version differences and performance changes are publicly recorded on the Model Card, and clinical distribution follows a shadow mode → A / B stepwise rollout procedure.

[0186] Tenth, safeguards and usage limits are specified. The AI ​​is mandated to learn and infer only from data that has passed F-NEM standardization and QC, and inputs that do not meet these prerequisites are reported as "unanalyzable (QC-fail)." ΔR1 may be optionally input as an auxiliary channel, but model design and index calculation are configured to be completed even in the absence of ΔR1. As a SaMD that assists in interpretation, this module is confirmed by medical staff for the final clinical diagnosis.

[0187] According to the above system, the AI ​​pipeline of the present invention is consistently connected from label definition to learning, inference, and structured reporting based on non-immunostimulating contrast agents, F-NEM-centered semi-quantitative input, and strict QC, and is designed to provide time-series change rates, left-right comparisons, and baseline changes with reliability across lymphatic, inflammatory, and neoplastic indications.

[0188]

[0189] 6. ΔR1 as a Secondary Metric and Acquisition & Quality Control

[0190] Regarding the “premise of input stability for AI evaluation metrics and the use of ΔR1 as an auxiliary metric,” procedures and criteria for quantitatively verifying the clinical validity and operational reproducibility of metrics generated in the T1 semi-quantitative imaging pipeline of the present invention are described. In the present invention, AI performance evaluation must be standardized using the formulation-normalized enhancement metric (F-NEM) and is limited to frames and voxels that have passed pre-defined quality control (QC). ΔR1 (=1 / T1 post -1 / T1 pre ) is an auxiliary indicator used selectively only when the operating range, masking, and QC conditions are satisfied, and the evaluation, reporting, and inference pipeline of the present invention is designed to be completed centered on F-NEM even without ΔR1.

[0191] First, there are the prerequisites of input stability and QC gates. All samples input for evaluation must pass the following: (i) normalization with formulation metadata (dosage, infusion-scan interval, r1 characteristics by lot and magnetic field strength) to generate an F-NEM map; (ii) automatic exclusion of venous contamination and injection leakage; (iii) SNR thresholds (e.g., SNR ≥ 15 based on the lesion / background ratio); (iv) verification of flip angle and B1(+) correction adequacy; and (v) masking of metal-induced T2* effects and high-intensity saturated regions. Data that fails to pass these QC gates is classified as “QC-fail” and is not included in any evaluation metrics.

[0192] Second, there are the principles regarding the use of ΔR1 as an auxiliary indicator and its acquisition and quality control. ΔR1 is calculated only when baseline T1 and post-contrast T1 are acquired in the same anatomical coordinate system using “identical acquisition parameters (grid, FOV, slice thickness, TR / TE, bandwidth) and the same correction scheme,” and it may be utilized as an auxiliary indicator. When using ΔR1, the following are mandatory: (i) limiting it to within the linear approximation operating range (e.g., relative enhancement rate greater than 0% to less than or equal to 300%), (ii) automatic masking of voxels with rapid signal changes (changes of ≥4 times compared to the previous time point), and (iii) T2* masking of areas where high signal is maintained only in spin echoes relative to gradient echoes or where abnormal loss occurs with increasing echo time. This principle is intended to ensure that ΔR1 is utilized as an auxiliary indicator so that it is not affected by the nonlinearity of contrast agent concentration-signal and the saturation zone.

[0193] After contrast administration (post-contrast), repeated T1 maps are acquired at predetermined time points (e.g., 5, 10, 20, 40, 60 minutes) while maintaining the same imaging geometry and correction scheme. A QC preliminary check is performed on the first frame or with a low-dose test injection to automatically exclude frames with injection site leakage (diffuse high signal in the skin / subcutaneous area), venous contamination (early linear high signal along anatomical venous pathways), and registration failures due to changes in body position. Additionally, pixels below the signal-to-noise ratio (SNR) threshold, areas of fat-water separation failure, and areas adjacent to metal artifacts are masked and excluded from the calculation.

[0194] Calculate R1=1 / T1 in pixel units from the T1 map at each time point, and at the baseline, R1 from the prior T1 maps aligned to the same coordinate system pre Calculate (to reduce base fluctuations, it is recommended to record 1-2 additional pre-free passes and use the average value). In the later stages, R1 at each stage post After calculating (t), transient noise is mitigated by applying the median / average of adjacent 2-3 frames if necessary. Per pixel ΔR1(t)=R1 post (t)-R1 pre After calculating, anatomical masks such as lymph nodes, lymphatic vessels, and tumors are defined as segmentations for organ or lesion unit analysis, and the median, percentile, and spatial gradient (boundary sharpness) of the ΔR1 distribution within the corresponding mask are calculated as summary indicators.

[0195] ΔR1 operation must always simultaneously satisfy “linear approximate operating range, T2* masking, SNR hurdle, and exclusion of intravenous contamination / infusion leakage,” and if any of these are violated, the corresponding frame or voxel is excluded from ΔR1 calculation and all subsequent indicators. The present invention does not presuppose absolute quantification, and ΔR1 is used only as a semi-quantitative signal to support F-NEM (Formulation-Normalized Enhancement Index). However, if an approximate relationship between ΔR1 and the effective concentration C (ΔR1 ≈ r1 · C) holds in the low-concentration linear range, and the r1 characteristics under formulation, magnetic field strength, and temperature conditions are determined in advance during the manufacturing stage, limited interpretations such as “relative concentration estimation” or “time-series concentration change rate estimation” may be performed using the corresponding r1 reference value. Even in this case, the results are reported based on F-NEM, and it is specified that the ΔR1-based concentration estimate is auxiliary information under the premise of operating range, masking, and QC.

[0196] Third, there are the discrimination indicators and calculation rules. The primary evaluation of the classification task is performed using ROC-AUC, and PR-AUC and F1-score are included when the positive rate is low or discriminative power near the threshold is important. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) are indicated at each of the predefined thresholds (e.g., clinical operation threshold, Youden optimal point, cost-weighted threshold), and 95% confidence intervals are calculated using bootstrapping (patient-level resampling). The DeLong test is used for comparing ROC-AUC between models, and the McNimard test is used for comparing paired accuracy.

[0197] Fourth, there is calibration and reliability evaluation. The consistency of the predicted probability is reported using the Brier score, Expected Calibration Error (ECE), reliability diagram, and calibration-in-the-large (calibration slope). If necessary, temperature scaling or isotonic regression is applied, and indicators before and after calibration are presented together. Prediction uncertainty is estimated using MC-Dropout or ensemble to output confidence intervals and confidence scores, and a “Defer-to-Human” flag is assigned to the quantile samples with the highest uncertainty.

[0198] Fifth is the evaluation of clinical utility. Decision Curve Analysis (DCA) is used to calculate the net benefit within the clinical threshold range and quantify the benefit relative to the standard of care. Additionally, the confusion matrix and class-specific F1-scores are reported separately in the progression / stable / response trichotomy based on the rate of change in the Tissue Damage Index (TDI).

[0199] Sixth are efficiency and operational metrics. Inference latency, throughput, and memory / power usage are measured and presented according to standard input resolution and batch size. In server-based batches, model version, hyperparameters, random seed, processing time, and QC pass rate are stored in audit logs to ensure reproducibility and enforceability.

[0200] Seventh is the design of data partitioning and external validation. Based on patient-level non-export partitioning, external validation is performed in parallel using leave-one-site-out and leave-one-vendor-out methods. All validation is limited to “F-NEM standardized + QC passed” data, and this premise is specified in all performance tables and diagram captions. Distribution shift is monitored using the statistical distance of the F-NEM distribution between the training and validation sets (e.g., Wasserstein distance) and changes in the QC-fail rate; if a threshold is exceeded, retraining or threshold reset is performed.

[0201] Eighth, there is the interoperability of labels and output. The labels used for evaluation (e.g., nodal filling delay, reflux presence / grade, bypass asymmetry, active / inactive, fibrosis co-occurrence; ES / NF / LV; TDI) are exported to DICOM-SR / HL7 FHIR along with codebook definitions, and the result report structurally records the input premises (F-NEM standardization, QC pass, ΔR1 usage), accuracy / correction / efficiency indicators, reliability, and hold rate.

[0202] Ninth is the pre-specification of thresholds and clinical transferability. Operational thresholds are pre-specificated during the development phase, and inter-institutional transferability of thresholds is ensured by using F-NEM-based indicators normalized with identical formulation metadata in multi-center trials. A “weak localization” strategy is permitted, allowing for the adjustment of only institution-specific calibration slices when necessary.

[0203] In summary, the present invention takes F-NEM standardization and strict QC as irreversible prerequisites for input stability, and calculates and reports distinction indicators such as ROC / PR-AUC, F1, and PPV / NPV, correction indicators such as Brier / ECE, and efficiency and uncertainty indicators within a consistent framework. ΔR1 is selectively used as an auxiliary tool only under operating range, masking, and QC, and the performance verification and clinical reporting of the present invention are completed within a semi-quantitative system centered on F-NEM. This achieves evaluation-operational alignment with minimized distribution differences between datasets, and enables highly reliable decision support even during actual clinical implementation.

[0204]

[0205] 7. Formulation Metadata-Based Normalization: F-NEM Transformation Procedure (F-NEM Normalization)

[0206] To define the F-NEM (formulation-normalized enhancement metric) conversion procedure and its mechanism of action, which standardizes the contrast signal into a semi-quantitative metric that allows for comparison over time, left-right, and between organs using formulation metadata, rather than using the contrast signal for absolute concentration estimation. The core principle is that, since the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention are designed not to disturb the physiological system during imaging (lymph-selective distribution, avoidance of venous contamination, renal excretion dominance), the observed enhancement signal is mapped to a common axis excluding the influence of formulation (administration amount, infusion-scan interval, lot-specific r1, etc.) and acquisition deviation (SNR, B1, motion, etc.) and used as a semi-quantitative metric.

[0207] First, the input and premise are defined. The input data consists of time-series T1-weighted images or T1 maps aligned to the same anatomical coordinate system, masks of regions of interest (lymphatic vessels, lymph nodes, and perilesional interstitial tissue), and formulation metadata. The formulation metadata includes the actual dose, infusion-scan interval (τ), lot-specific r1 pre-contour values ​​under magnetic field intensity (B0), temperature, and medium conditions, the surface coordination point density or coordination ratio per metal of the formulation (if necessary), and the acquisition and calibration history (flip angle, whether B1 correction was applied, fat suppression quality). All signals are processed only for frames and voxels that have passed quality control (QC) rules—automatic exclusion of frames with venous contamination or infusion leakage, satisfaction of SNR hurdles, verification of B1 correction suitability, and masking of suspected T2* regions.

[0208] Second, the calculation of the original signal and the selection of the standardizable interval are performed. The average relative enhancement rate or region average signal at each time point is calculated within the region of interest, and only frames and voxels falling within the linear approximation operating range (e.g., relative enhancement rate greater than 0% and less than or equal to 300%) are selected to avoid non-linear and saturation effects of contrast agent concentration-signal. Voxels with rapid signal changes (changes of ≥4 times compared to the previous time point) are automatically masked to eliminate disturbances. If a T1 map is available, ΔR1 can be calculated as an optional auxiliary indicator under the premise of this operating range, masking, and QC; however, the focus of standardization and reporting is placed on F-NEM.

[0209] Third, the operating principle and conversion formula for formulation scaling. Since the observed enhancement is scaled by the formulation and time, it is converted into a virtual standard response at a unit dose, unit time, and reference magnetic field strength. To this end, (i) dose normalization (first-order scaling of the observed enhancement with respect to the dose), (ii) time normalization (correcting kinetic attenuation and amplification according to the infusion-scan interval τ with an empirical or predefined kernel), (iii) relaxivity normalization (converting the lot-specific r1 to a reference r1,ref), and (iv) magnetic field strength / temperature correction (scaling with a function of r1(B0,T) if necessary) are applied in stages. In practical implementation, a scaler of the form F-NEM = Eobs × (r1,ref / r1,lot) × (Dref / Dobs) × Kτ × KB0 is applied, where Eobs is defined as the region average enhancement (or relative enhancement rate) of the QC-passed frames, r1,lot as the lot-specific value determined during the manufacturing stage, D as the actual dose per metal, and Kτ·KB0 as a fixed correction factor for each protocol. In protocols where serum binding rate or surface coordination point density has a significant effect on the indicator, the corresponding term may be included as an additional scaler. This scaling is based on the premise that the purpose is not to convert the signal into “absolute concentration,” but to map it to a comparative quantification axis based on the same formulation.

[0210] Fourth, the composition of summary statistics and reported values. For points of interest or windows in the transformed F-NEM time series, peak F-NEM, hour-averaged F-NEM, initial slope of ascent (WIR; relative to F-NEM), time to reach maximum (TTP), and under-curve integral (AUC) are calculated. Shape indicators are calculated from the F-NEM map, including edge sharpness (ES), low-inflow area ratio (corresponding to NF), and volume (LV). If necessary, the critical area ratio (TAR) is used as an auxiliary value for co-estimation. All values ​​are expressed as unitless standardized figures (e.g., arbitrary units, au), and measurement uncertainty is co-stated as a 95% confidence interval using bootstrap or inter-frame variation.

[0211] Fifth is data integrity and interoperability. For each scan, the formulation metadata used (dosage, τ, r1, lot, B0, etc.), the applied scaler (Kτ, KB0), QC pass rates and reasons for exclusion (intravenous contamination, infusion leakage, insufficient SNR, B1 non-conformity, T2* masking), and audit logs including model, version, hyperparameters, random seed, and processing time are stored together. The final output is generated as a structured numerical report containing F-NEM and derived indicator groups (PSE, WIR, TTP, AUC, ES, NF, LV, optional TAR, synthetic indicator TDI), and exported in DICOM-SR and HL7-FHIR formats. This procedure is designed to maintain consistency in input distribution even if the institution, equipment, or lot changes, ensuring reproducibility in multi-institutional comparisons, time-series tracking, and left-right comparisons.

[0212] Sixth, the principles and limitations of interpretation are specified. F-NEM is a semi-quantitative indicator that reliably reflects the rate of change over time, left-right ratio, and inter-institutional differences under the assumptions of the same formulation, same operating range, and same QC. Conversely, frames and voxels suspected of high concentration saturation or metal-induced T2* effects, sections where venous contamination or infusion leakage is detected, and data that do not meet SNR hurdles should be excluded from interpretation. T1 map-based ΔR1 may be included as an auxiliary indicator only if the operating range and QC conditions are satisfied; however, F-NEM is the primary reading axis and does not claim absolute quantification.

[0213] In short, the F-NEM transformation procedure is based on the premise of “non-operative contrast generation.” It standardizes observation augmentations using scalers for formulation, time, and magnetic field strength to convert them into a unitless comparative quantitative axis, and reports time-signal, morphology, and synthetic indicators (TDI) centered on this axis in a consistent format. As a result, it enables quantitative interpretation with low variability in time series within the same patient, left-right comparisons, and cross-comparisons between institutions and equipment, while simultaneously ensuring input stability for clinical decision-making and AI learning and evaluation.

[0214]

[0215] 8. Mechanism for Ensuring Multi-institutional Reproducibility (F-NEM-centric Inter-site Reproducibility)

[0216] The present invention ensures reproducibility by mapping the input distribution to a common axis using formulation metadata-based normalization (F-NEM) and forced QC and masking rules, rather than attempting to absorb differences in scanners, institutions, and protocols through "harmonization." The operating principle is as follows: First, it presupposes input stability. All frames and voxels are processed only if they pass automatic exclusion of venous contamination / infusion leakage, SNR hurdles, B1 suitability, and T2* suspected segment masking. Second, it involves formulation normalization. Observation augmentation is scaled by dosage, infusion-scan interval, lot-specific r1, and magnetic field strength (B0) and converted into a unitless F-NEM, which is a common metric for semi-quantitative rather than absolute quantification. Third, it involves server-based unified processing. Output logic, parameters, and model versions are applied uniformly on a central server, and processing times, hyperparameters, and random seed numbers are stored in audit logs to enable reproducibility tracking. Fourth, it involves site onboarding and drift monitoring. New institutions align F-NEM distribution and QC pass rates to the standard range using a small number of pilot cases, and during operation, automatically alerts for distribution shifts through monthly control charts of indicator groups (PSE, WIR, TTP, AUC, ES, NF, LV) and F-NEM summary statistics (median / IQR) by institution, equipment, and sequence. Fifth is cross-institutional validation. Inter-institutional F-NEM differences within the same patient and stage population are monitored using Bland-Altman / CCC, and if the allowable deviation is exceeded, QC rules, omitted formulation metadata records, and protocol deviations are traced back. Sixth is reporting and interoperability. All results are distributed as DICOM-SR / HL7-FHIR structured figures, enabling immediate inter-institutional comparison. This closed-loop of F-NEM-centric standardization, mandatory QC, centralized processing, and drift monitoring can consistently guarantee low-variance comparative quantitative indicators in multi-institution, multi-equipment environments without ΔR1.

[0217]

[0218] 9. Standardized MRL-Volumetry-AI Pipeline

[0219] The present invention provides an integrated pipeline that combines non-agonistic, lymphoselective T1 contrast agents, formulation metadata-based normalization (F-NEM, formulation-normalized enhancement metric), and mandatory quality control (QC) rules into a single processing line to produce consistent numerical reports for MR lymphangiography (MRL), three-dimensional volumetric analysis, and AI-based interpretation. The basic mechanism of action is summarized as a cyclical procedure of “input stabilization → formulation-normalization → automatic metric calculation → structured reporting / learning.”

[0220] The first step is input stabilization. For T1-weighted time series or T1 maps acquired by intradermal / subcutaneous small-volume multi-point infusion protocols, automatic exclusion of venous contamination and infusion leakage, SNR hurdles, flip angle and B1 conformity checks, and T2* suspicious interval masking are applied. Only frames and voxels that pass QC become the sole input for subsequent outputs, and the reasons for exclusion are recorded in the audit log.

[0221] The second step is formulation-normalization. Time-signal change amounts (e.g., region average enhancement rate) are scaled by predefined formulation metadata, such as dosage, infusion-scan interval, lot-specific r1 characteristics, and magnetic field strength, and converted into dimensionless F-NEM. ΔR1 can be optionally calculated on the premise of identical coordinate and parameter acquisition and passing QC, but the primary indicator for this pipeline is F-NEM.

[0222] The third step involves the simultaneous calculation of MRL-specific indicators and 3D volumetric indicators. Segmentation of the lymphatic network (superficial / deep, lymph nodes / collecting ducts) is generated using an automated and manual combination method, and PSE, WIR, TTP, and AUC are calculated from the F-NEM time series. Morphological indicators such as boundary sharpness (ES), necrotic fraction (NF), and lesion volume (LV) are extracted, and lymph-specific functional indicators (nodal filling delay, reflux presence / grade, bypass asymmetry, active / inactive, and fibrosis co-occurrence) are quantified according to standard definitions. If necessary, the critical area ratio (TAR) is used as an auxiliary quantitative value for staging.

[0223] The fourth step is the registration and standardization of the 3D volumetric analysis. After rigid or non-rigid registration of the time series, the data is resampled into a unified coordinate system to ensure the orthogonality (temporal separation) of the temporal indices (PSE / WIR / TTP / AUC) and spatial indices (ES / NF / LV). Frames with registration quality below a threshold are automatically excluded, and a re-imaging recommendation flag is assigned if the exclusion rate exceeds the threshold.

[0224] The fifth step is the calculation of synthetic indicators and the judgment logic. A Tissue Damage Index (TDI) is calculated by weighting ES, NF, and LV with F-NEM as the axis, and a threshold for the rate of change of TDI relative to baseline (e.g., ≥20% increase) is marked as a progression flag. In lymphedema, inflammation, and tumor scenarios, response, stability, and progression are determined based on consistent criteria by referencing the declining / increasing patterns of F-NEM-based time-signal indicators together with the normalization / deterioration of morphological indicators.

[0225] The sixth step is structured reporting and interoperability. Output metrics (F-NEM, PSE, WIR, TTP, AUC, ES, NF, LV, TDI, and optional TAR), formulation metadata, QC checklists, and reasons for exclusion are exported as DICOM-SR and HL7 FHIR compliant figures. The same format ensures inter-agency comparisons, long-term tracking, and traceability for regulatory submissions.

[0226] The seventh step is AI integration. Using only QC-pass and F-NEM normalized data as input, segmentation / classification / time-series prediction models are trained and inferred. Labels are standardized with functional indicators such as nodal filling delay, reflux presence / grade, bypass asymmetry, active / inactive lesion activity, and fibrosis co-occurrence, as well as TDI / progression flags, and the output is immediately distributed to DICOM-SR / HL7 FHIR. Evaluation metrics (ROC-AUC, PR-AUC, F1, PPV / NPV, correction, uncertainty) are reported only under the premise of “input stability.”

[0227] Finally, there is server-based execution. The central server centrally manages output logic, model versions, parameters, random seeding, and processing times, and automatically alerts for distribution shifts in multi-institutional environments through site onboarding and drift monitoring (F-NEM distribution control diagrams by institution, equipment, and sequence). This closed-loop structure of “non-toxic contrast agent + F-NEM normalization + mandatory QC + standard reports + AI integration” minimizes variability across the entire MRL-Volumetric MRI-AI spectrum and continuously provides stable comparative quantitative indicators without ΔR1.

[0228]

[0229] 10. AI Training, Labeling & Evaluation Framework

[0230] The AI ​​frame of the present invention is designed with a closed-loop structure of “input stabilization-standard normalization-standard labeling-learning / inference-validation / auditing.” The core premise is that only frames and voxels that have passed F-NEM (formulation-normalized enhancement metric) calculated by formulation metadata-based normalization and forced quality control (QC) are used as model inputs, which ensures the reproducibility of learning and evaluation by minimizing distribution shifts between institutions and equipment.

[0231] Data collection and preprocessing begin with the aggregation of DICOM metadata and QC logs. Time series within the same patient are aligned to a unified coordinate system using rigid / non-rigid registration, while venous contamination, infusion leakage, suspected T2* intervals, and regions with insufficient SNR are automatically masked and excluded from training. The time-signal variation of each frame and voxel is scaled by formulation metadata (dosage, infusion-scan interval, lot-specific r1 characteristics, magnetic field strength, etc.) and converted into F-NEM time series; if necessary, ΔR1 is optionally added only as an auxiliary feature. Input tensors are structured in spatiotemporal blocks (e.g., xyzt), and superficial / deep lymph networks are combined with lymph node and collecting duct masks to create region-of-interest (ROI)-centered patch datasets.

[0232] The labeling system defines functional and morphological labels separately. Functional labels consist of standard indicators reflecting lymphatic system dynamics, including, but notably, (i) nodal filling delay (standardized time from infusion to filling), (ii) reflux frequency (ratio of frames in which dermal reflux patterns are observed), and (iii) bypass asymmetry (bypass network length / signal ratio between left and right or between lesion and control). Morphological labels consist of boundary sharpness (ES), necrotic fraction (NF), and lesion volume (LV), and are confirmed through two-reviewer consensus or an adjudication procedure following semi-automatic segmentation. Each label is standardized with the HL7 FHIR / DICOM-SR compliant codebook and value ranges (units, scales, and thresholds) to maintain semantic consistency when used across sites. Clinical evaluation labels use the rate of change of the tissue damage index (TDI) from baseline and a progression flag based on whether a predefined threshold (e.g., TDI ≥ 20% increase) is exceeded.

[0233] The training tasks consist of three axes. First, the segmentation / detection task aims for 3D segmentation of lymphatic vessels, lymph nodes, and collecting ducts, and is trained using a U-Net / Transformer hybrid by channel combining F-NEM maps and T1-weighted images (optional). Second, the classification / regression task aims for the automatic calculation of functional and morphological indicators, and predicts PSE, WIR, TTP, AUC, ES, NF, and LV using ROI-based time-series encoders (Temporal CNN / Transformer). Third, the time-series forecasting / risk modeling task learns trend changes at the patient level to predict future TDI change rates and progression flags. In all tasks, input is limited to F-NEM normalized and QC-passed data, and ΔR1 and raw signals are treated only as auxiliary features to avoid inducing domain drift.

[0234] Data partitioning adheres to the patient-level non-overlap principle and performs stratification to ensure that institution, equipment, and sequence distributions are balanced across the training, validation, and test sets. To prevent data leakage, rules are applied to ensure that images of the same patient at different time points or organs are not separated into different sets. Class imbalance is corrected using label distribution-based sample weighting, focal loss, hard example mining, and minority class augmentation (spatial transformation, time point dropping / warping, etc.).

[0235] Evaluation metrics are calculated separately according to their application. Classification and risk assessment tasks are based on ROC-AUC and PR-AUC, and F1 (harmonic mean) and PPV / NPV are reported alongside them to compensate for the impact of class imbalance. Regression tasks verify the reliability of prediction probabilities by presenting correction metrics (Brier score, calibration curve ECE / MCE) along with MAE / RMSE. Segmentation tasks utilize Dice / Jaccard and boundary-based metrics (HD95, ASSD) in parallel. All evaluations are valid only under the premise of “F-NEM normalization and QC-passed frames / voxels,” and results violating this premise (e.g., including vein-contaminated frames) are distinguished by separate notation. Uncertainty quantification uses Monte Carlo dropout or ensemble variance, and human-in-the-loop routing (expert review of high-uncertainty cases) is implemented through confidence-error correlation analysis.

[0236] Explainability and error analysis are included as essential steps to enhance the clinical reliability of the model. Temporal-spatial attention maps and occlusion sensitivity are used to visualize the anatomical regions and time zones referenced by the model, and contributions by indicator (e.g., TDI contribution of F-NEM vs. ES vs. NF) are calculated as Shapley values ​​and summarized in the structured report. Errors are classified and aggregated by type (undetected venous contamination, residual infusion leakage, registration failure, SNR degradation) and reflected in the continuous improvement of data collection and QC rules.

[0237] Model operation and drift management are performed on a server-based system. The model version, hyperparameters, random seed, training data summary statistics, and performance baseline at the time of deployment are fixed in the audit log, and an alert is triggered if the F-NEM distribution and metric distributions (PSE / WIR / TTP / AUC, ES / NF / LV, TDI) of the real-world data fall outside the threshold range. The site onboarding process is approved only when performance is within the tolerance range following a spot check with a small validation set, and distribution differences are corrected through domain adaptation (e.g., feature-wise affine calibration) if necessary. Retraining is performed only on new cohorts that satisfy QC and label quality requirements, and the non-inferiority of performance and calibration before and after the update is statistically verified.

[0238] Interoperability and reporting are unified under DICOM-SR / HL7 FHIR. Inputs (formulation metadata, QC checklists), output metrics (F-NEM, PSE, WIR, TTP, AUC, ES, NF, LV, TDI, progress flags), explainability summaries (attention map thumbnails, metric contributions), uncertainty statistics, and audit log keys are exported as standard fields. Integration with EHR, PACS, and eCRF ensures traceability and regulatory compliance throughout the entire clinical and research lifecycle. Personal information is processed in a pseudonymized and encrypted state, and patient-level keys adhere to the principle of local storage.

[0239] In short, this framework is based on the stable input of non-active contrast agents (F-NEM normalization and QC passing) and standardizes the entire process of learning, inference, verification, and auditing through standard labels, indicators, and reports. As a result, reproducible performance, calibration, and interpretable output are secured even in multi-institutional and multi-equipment environments, providing a reliable evidence system for both clinical decision-making and regulatory submission.

[0240]

[0241] 11. Clinical Use Cases I—Lymphatic Disorders

[0242] The contrast agent of the present invention is a non-immunostimulating polysaccharide cross-linked colloidal particle having a hydration diameter of 2 to 8 nm, a weak negative charge of -20 to 0 mV, and high substitution cross-linking (≥60% relative to monosaccharides, preferably ≥90%), and is designed to have a dominant T1 positive contrast by coordinating metal ions to surface carboxylate / amine / phosphate / catechol donor groups. This property yields two key operational results. First, due to PRR / BCR non-operation, immune receptor occupancy, internalization, and cytokine / complement activation are suppressed during the imaging window, thereby minimizing exogenous disturbances to physiological states (vascular permeability, interstitial retention, lymphatic drainage). Second, the size of 2–8 nm and weak negative charge inhibit venous capillary leakage and preferentially follow the physiological drainage pathway from the interstitial → lymphatic capillary → peripheral lymphatic vessel → lymph node, thereby forming lymphoselective pharmacokinetic characteristics that avoid venous contamination in MR lymphangiography. These non-perturbative and lymphoselective signatures reduce the temporal variability of the input images, enabling the calculation of a stable “formulation-normalized enhancement metric (F-NEM)” in within-patient time series and inter-institutional comparisons.

[0243] Based on the above operating principle, the present invention defines QC, operating range, and masking rules as follows to systematically suppress physical limitations of the contrast signal (density-signal nonlinearity, high-density saturation, metal-induced T2* effects) and variations in the imaging system (SNR degradation, B1 inhomogeneity, motion). Only frames and voxels that pass through these rules are used to calculate F-NEM, time-signal indices (PSE, WIR, TTP, AUC), shape indices (ES, NF, LV), and composite indices (TDI), and become the sole inputs for structured reports and AI learning and evaluation.

[0244] First is the establishment of a linear approximation operating range. The T1 contrast signal exhibits nonlinear characteristics, characterized by low signal in low-concentration regions, an increase in signal at moderate concentrations, and a subsequent decrease in signal upon saturation in high-concentration regions. To compensate for this, the signal at each time point is limited to a reasonable range where the relative enhancement rate (PSE) relative to the baseline exceeds 0%—e.g., 300% or less. Additionally, if a sudden surge or drop in signal (e.g., a change of more than four times compared to the previous time point) is detected within the same voxel, that voxel is masked. If metal-induced T2* effects are estimated (e.g., high signal is maintained only in the spin echo relative to the gradient echo, or abnormal signal loss occurs with increasing echo time), the corresponding region is automatically detected and excluded from analysis. This operating range limitation eliminates nonlinear regions caused by excessive local accumulation of metal complex concentrations or injection point leakage, ensuring that the F-NEM maintains a dose-time dependent monotonic relationship.

[0245] Second are the Signal-to-Noise Ratio (SNR) and instrument conformance hurdles. Frames subject to analysis must satisfy a predefined SNR threshold (e.g., SNR ≥ 15 based on lesion / background criteria); frames that do not meet this threshold are marked as QC-fail and excluded from reports and AI input. The application and conformity of flip angle and B1 heterogeneity correction are automatically checked using sequence metadata and built-in validation frames, and the corresponding time point is excluded if conformity is not met. Frames are similarly excluded if fat suppression failure, phase wrapping, or significant off-resonance distortion is detected, and the reasons for exclusion are recorded in the audit log.

[0246] Third, there is the automatic exclusion of venous contamination and infusion leakage. The present invention tracks the temporal filling of superficial lymphatic vessels → peripheral lymphatic vessels → lymph nodes under small-volume multi-point intradermal / subcutaneous injections and standardized infusion-scan intervals using T1 positive contrast. Venous contamination is defined as a pattern in which a linear early high signal appears along the anatomical venous pathway and is accompanied by a rapid wash-out synchronized with the deep venous system; the corresponding pathway and frame are excluded after automatic labeling. Infusion leakage is excluded based on the presence of wide spreading around the infusion point and the absence of temporal progression. As shown in Fig. 22b, these exclusion rules are designed to preserve, without distortion, the dermal backflow, lymphatic dilation, and diffusion enhancement areas observed as the lymphedema stage increases in the lymphedema limb, while retaining only lymphatic vessel / lymph node-centric filling in normal limbs.

[0247] Fourth, there is registration, deformation correction, and standard coordinate system resampling. Time-series images of the same anatomical region undergo rigid registration followed by non-rigid registration if necessary to correct motion, and are resampled into a unified anatomical coordinate system. Frames in which motion indices exceed a threshold are excluded, and if the exclusion rate exceeds a pre-defined standard, a re-imaging recommendation flag is applied to the entire scan. This registration step ensures that the temporal parameters of PSE, WIR, TTP, and AUC, and the spatial parameters of ES, NF, and LV are not affected by coordinate system variations.

[0248] Fifth, the formulation-normalized enhancement index (F-NEM) is calculated. F-NEM is a standardized value calculated by normalizing the time-signal change amount (e.g., PSE or region average enhancement rate) with formulation metadata (dosage, infusion-scan interval, metal ion r1 characteristics by lot and magnetic field strength, surface coordination point density, etc.). Normalization includes a procedure for scaling to unit dose, unit time, and reference magnetic field strength conditions, and lot-specific r1 characteristics are replaced with standard values ​​obtained during the manufacturing stage. If a T1 map is obtained, ΔR1 can be selectively calculated as an auxiliary indicator within the operating range, masking, and QC hurdles; however, the report and AI pipeline of the present invention are designed to operate based on F-NEM even without ΔR1. The F-NEM calculated in this way serves as a semi-quantitative input that minimizes the non-linearity and saturation effects of absolute concentration, and stably reflects time-series change rates, left-right comparisons, and changes relative to the baseline.

[0249] Sixth, the calculation and operational basis of lymphatic system-specific function indicators. For frames in which superficial / deep lymphatic networks are continuously visualized, lymph node filling delay (relative time from infusion to filling), dermal reflux frequency (ratio of frames in which reflux patterns occur), and bypass path asymmetry (bypass network length / signal ratio between left and right or lesion and control) are quantified based on F-NEM. The threshold area ratio (TAR) defined in Example 8 is calculated as the ratio of the enhancement area to the area of ​​interest and is used as an auxiliary quantitative value reflecting the severity of lymphedema; when the QC rule of the present invention is applied, the TAR tends to increase monotonically (linearly increase) according to the lymphedema stage (Stage 0→4). In the animal image of Fig. 22b, normal limbs show only restricted filling centered on lymphatic vessels / lymph nodes (Stage 0), whereas in lymphedematous limbs, dermal reflux and diffusion enhancement areas become more distinct as the Stage increases, which visually confirms that high qualitative and quantitative agreement is maintained between INV-MRL and NIRF-ICGL.

[0250] Seventh, synthetic indicators and clinical evaluation. The present invention defines a tissue damage index (TDI) by weighted combination of morphological indicators (ES, NF, LV) with F-NEM as the axis, and flags a threshold increase in TDI relative to a reference (e.g., ≥20%) as a progression signal. This synthetic procedure does not rely on the absolute value of contrast agent concentration, but adopts the time-series rate of change and the consistency of structural changes as evaluation criteria to enhance the reproducibility of clinical decision-making.

[0251] Eighth, there is reporting, auditing, and enforceability. For each scan, the applied QC checklist (automatic removal of venous contamination / infusion leak, passing SNR hurdles, flip angle / B1 correction suitability, T2* masking, alignment quality), the formulation metadata used (administration site / dosage / infusion-scan interval, lot information, reference r1 characteristics), output metrics (F-NEM, PSE, WIR, TTP, AUC, ES, NF, LV, optional TAR, TDI), and reasons for exclusion / masking are all stored in the audit log. In a server-based processing environment, the model version, hyperparameters, random seed, and processing time are recorded together to ensure compliance with regulatory and quality audits and enforceability of rights. Frames and voxels that fail QC are stored solely for research support purposes and are not used in clinical reports or for AI training and evaluation.

[0252] The above procedure starts from contrast agent properties (non-active and lymphoselective) and progressively eliminates error sources such as signal physics (non-linearity, saturation, and T2*), imaging systems (SNR, B1, and motion), and anatomical separation (venous contamination and infusion leakage), thereby increasing the consistency of the input distribution and enabling the reproducibility of F-NEM-centered comparative quantification and time-series trend readings throughout the entire animal-clinical cycle. As a result, as in Example 8, performance is reliably reproduced in which INV-MRL clearly distinguishes dermal reflux, lymphatic dilation, and diffusion enhancement regions from normal, and TAR and staging classification significantly match NIRF-ICGL. The QC, operating range, and masking rules of the present invention are key components that guarantee such reproducibility at legal and technical levels.

[0253]

[0254] 12. Clinical Use Cases II—Oncology, Inflammation, Fibrosis

[0255] The present invention provides input stability and clinical operability that are fundamentally different from GBCAs, agonist ligand nanoformulations, and general dextran / USPIO formulations through (i) physicochemical design of PRR / BCR non-agonistivity, (ii) lymphoselective pharmacokinetics and avoidance of venous contamination, (iii) a quantification frame of formulation-normalized comparative quantification (F-NEM-centered) that does not depend on the absolute value of ΔR1, and (iv) server-based QC and audit logging. Thus, it simultaneously supports clinical utility and industrial applicability across the entire spectrum of lymphatic, tumor, inflammation, and fibrosis requiring repeated administration.

[0256] The following describes the differences from the prior art and the mechanism of action regarding the non-immunostimulating polysaccharide cross-linked nanoparticle platform and the MRI T1 quantitative imaging method and system using formulation-normalized metadata of the present invention.

[0257] First, it is distinguished from gadolinium chelate (GBCA)-based contrast agents. As small molecule chelates, GBCAs exhibit extensive vascular leakage and rapid systemic distribution immediately after administration, making it difficult to separate lymphatic vessels from veins and difficult to avoid "venous contamination" in the evaluation of lymphedema. Furthermore, the T1 signal is strongly affected by concentration-signal nonlinearity and high-concentration saturation, resulting in low reproducibility for time-series comparative quantification. The contrast agent of the present invention has a structure in which metal ions are surface-coordinated to non-immunostimulating polysaccharide cross-linked colloidal particles having a hydration diameter of 2–8 nm, a weak negative charge of -20–0 mV, and high-substitution cross-linking (≥60% relative to monosaccharides, preferably ≥90%). This structure inhibits venous capillary leakage and preferentially follows the physiological drainage pathway of interstitial → lymphatic capillaries → peripheral lymphatic vessels → lymph nodes. As a result, lymph-specific filling is dominant, and venous contamination is structurally avoided. In addition, the present invention performs semi-quantitative and time-series readings using F-NEM, which standardizes the amount of time-signal change using formulation metadata without relying on absolute ΔR1, thereby ensuring reproducibility that is insensitive to variations between equipment and institutions.

[0258] Second, this distinguishes it from nano-formulations conjugated with agonistic target ligands. Nano-contrast agents conjugated with target ligands (e.g., TLR-agonistic peptides, immune cell surface protein ligands) induce physiological disturbance events such as target occupancy, internalization, cytokine release, and complement activation during the imaging window, thereby altering target density and the microenvironment. In this process, the time-signal curve is distorted by agent origin variables, and the inconsistency in data distribution across organs makes calibration for AI learning and validation difficult. The present invention achieves PRR / BCR non-agonism through epoxide-based selective modification, high-substitution crosslinking, and surface-COOH post-modification, thereby significantly suppressing immune receptor occupancy, internalization, and NF-κB activation during the imaging window. In other words, the contrast signal is designed to be primarily proportional to "changes in target physiology" rather than "agent agonism," thereby providing a non-perturbative input.

[0259] Third, this distinguishes it from general dextran and dextran derivative-based formulations. Prior dextran-based carriers are prone to PRR and BCR cross-linking due to the exposure of repeating sugar structures and the introduction of positively charged surface groups, and increased immunogenicity and retention / capture have been reported upon repeated administration. The present invention structurally reduces the spatial accessibility of repeating sugar epitopes through high-substitution cross-linking of at least 60% (preferably ≥90%) relative to the total number of monosaccharides, and significantly reduces PRR / BCR binding affinity by implementing a weak negative charge through post-modification of surface basic amines with -COOH. As a result, cytokine release and complement activation remain within acceptable limits, imaging findings of migration to central lymphatic vessels without capture or retention are reproduced even in immune cell-dense environments such as lymph nodes, and the tolerance for repeated administration is high.

[0260] Fourth, this distinguishes it from existing superparamagnetic iron oxide (SPIO / USPIO)-based contrast agents. Many SPIO formulations induce negative contrast due to T2* dephasizing dominance, making it difficult to quantify anatomical boundaries and functional signals; furthermore, nonlinearity and artifacts are exacerbated at high local accumulations. This invention provides stable T1 positive contrast even at low doses by optimizing surface coordination chemistry (carboxylate / amine / hydroxyl / phosphate donor density) to promote intra-extrasphere water exchange while maintaining a low r2 / r1 ratio. Simultaneously, the T2* influence region is automatically masked in the QC rule, allowing only the linear approximation region of the signal to be used for analysis.

[0261] Fifth, the quantification framework is distinctive. Prior technologies attempt absolute quantification by relying on scanner harmonization or phantom calibration, but they are vulnerable to dependence on equipment, magnetic field strength, and the environment. Based on the premise of overcoming these limitations of absolute quantification, the present invention calculates and reports PSE, WIR, TTP, AUC, ES, NF, LV, and the synthetic indicator TDI, centering on F-NEM, which standardizes time-signal change using formulation metadata (dosage, infusion-scan interval, and r1 characteristics by lot and magnetic field strength). ΔR1 is selectively used only as an auxiliary indicator within the operating range, masking, and QC hurdles when necessary. The adoption of such “formulation-normalized comparative quantification” mitigates distribution inconsistencies between datasets and fundamentally improves multi-institutional reproducibility.

[0262] Sixth, there is a difference in substantive efficacy in lymphatic system application. GBCAs or non-selective formulations face limitations in deriving functional indicators due to the difficulty in separating lymphatic vessels from veins and the short residence time in lymph nodes. Based on a QC system that includes the exclusion of venous contamination, infusion leakage, and T2* masking, the present invention reliably achieves continuous visualization of superficial / deep lymphatic networks, F-NEM-based quantification of delayed lymph node filling, frequency of dermal reflux, and bypass pathway asymmetry, as well as consistency between TAR and staging classification. This difference results in a reduction in re-imaging rates and improved multi-center reading consistency in clinical workflows.

[0263] Seventh, the differentiation lies in safety and the possibility of repeated administration. Actuating ligand formulations or some polymeric carriers carry a high risk of inducing ADA, systemic inflammation, and hypersensitivity reactions upon repeated administration. Due to the PRR / BCR non-actuating design and pharmacokinetics with renal excretion dominance, the safety of the present invention has been confirmed in preclinical and early clinical trials without significant abnormal changes in complete blood counts and inflammatory markers. This means that time-series F-NEM can be stably accumulated and compared during long-term follow-up of the same patients.

[0264] Eighth, there is a difference in server-based processing and execution capabilities. In the prior art system, processing and correction procedures are performed sporadically, resulting in low reproducibility and auditability. The present invention is designed to perform F-NEM calculation, PSE / WIR / TTP / AUC·ES / NF / LV calculation, TDI calculation, structured report (DICOM-SR / HL7-FHIR) generation, and audit log recording of model, version, hyperparameter, random number seed, and processing time collectively within the server under the premise of input stability (limited to QC-passed frames), thereby simultaneously ensuring compliance with regulatory and quality audits and enforceability of rights.

[0265] Ninth, this is the point where the effectiveness of the invention is proven by unexpected results. The triple design of high-displacement crosslinking, a small hydration diameter, and a weak negative charge goes beyond merely lowering the immune response to consistently generate pharmacokinetic characteristics such as lymph selectivity, avoidance of venous contamination, and renal excretion dominance. Consequently, F-NEM-based comparative quantification derived from this enables stable operation in multi-institutional and multi-equipment environments. This causal link of “immunoactivation → lymph selectivity → input stability → quantification reproducibility” is evaluated as a technical effect not predicted in the prior art.

[0266] Meanwhile, the operational points for inflammatory and fibrotic diseases and neoplastic lesions are as follows.

[0267] (1) Application to inflammatory and fibrotic diseases.

[0268] Inflammatory response is a pathological condition in which capillary dilation, increased permeability, interstitial edema, immune cell infiltration, matrix remodeling, and fibrosis overlap along the time axis. Since the contrast agent of the present invention does not disrupt the immune signaling network during the imaging window, the imaging signal is primarily correlated with changes in target physiology (permeability, interstitial retention, lymphatic drainage). Accordingly, time-signal and morphological indicators centered on F-NEM are utilized as follows. WIR (Initial Inflow Gradient) and TTP (Time to Reach Maximum) sensitively reflect increased capillary dilation and permeability, while AUC and lavage pattern quantify interstitial retention and drainage impairment. F-NEM normalized PSE provides the intensity of lesion contrast as a standardized value. The morphological indicator ES (Boundary Sharpness) indicates the irregularity and diffusion of the inflammatory boundary, while NF (Necrotic Fraction) and LV (Lesion Volume) quantitatively present necrotic and fibrotic changes. The synthetic indicator TDI is expressed as the rate of change relative to baseline by weighting the combination of F-NEM, ES, NF, and LV, and in treatment monitoring, a decrease in WIR and AUC, normalization of ES, and a reduction in NF act as early response signals. In multi-center comparisons, the portability of thresholds is maintained thanks to F-NEM based on identical formulation metadata correction.

[0269] (2) Application to neoplastic lesions.

[0270] The tumor microenvironment is a heterogeneous space where leaky neovascularization, elevated interstitial pressure, reduced lymphatic drainage, central necrosis, and fibrotic stroma coexist. The contrast agent of the present invention evades target receptor activation and provides stable T1-positive contrast through its size, charge, and water exchange characteristics, thereby non-activatingly capturing functional changes in the vascular-interstitial-lymphatic axis. F-NEM normalized WIR and TTP quantify perfusion-permeability heterogeneity, while AUC and the washout index quantify interstitial retention and poor lymphatic drainage. ES quantifies the sharpness / roughness of marginal infiltration to aid in the determination of local invasiveness, while NF indicates the degree of necrosis due to hypoperfusion / hypoxia, and LV provides tumor volume. The TDI, a weighted combination of these indicators, is utilized for early prediction of response and detection of residual / recurrence in neoadjuvant, targeted, and immunotherapy. In staging, the reliability of the T and N categories is enhanced by interpreting the primary lesion's F-NEM-PSE, WIR, and ES together with delayed fullness, heterogeneity, and residual metrics of lymph nodes.In indications where lymph node metastasis is a key variable, such as breast cancer, the lymphoselectivity of this contrast agent clearly reveals delayed filling of the axillary lymph node chain, reflux, and the appearance of bypass pathways, enabling standardized staging assistance using proximal-distal F-NEM gradient and bypass asymmetry indicators.

[0271]

[0272] 13. Design of Synthetic Indicators—Tissue Damage Index (TDI)

[0273] The Tissue Damage Index (TDI), a synthetic indicator of the present invention, is a semi-quantitative score that expresses the “activity (dynamics)” and “structural damage (morphological change)” of a single lesion as a single numerical value. The TDI does not presuppose ΔR1 and integrates the F-NEM (formulation-normalized enhancement metric) and the morphological indicator only for frames / voxels that have passed the QC and operating range rules.

[0274] Based on the terms, abbreviations, and QC and operating range rules throughout this specification, the definition of a synthetic index centered on F-NEM that can be calculated without ΔR1, weighted combination formula, threshold flag calculation, reporting rules, and clinical interpretation guide are described sequentially.

[0275] First, there are the components and the mechanism of action. The Tissue Damage Index (TDI) is calculated by weighted combination of the formulation-normalized enhancement metric (F-NEM) and three formulation indicators: edge sharpness (ES), necrotic fraction (NF), and lesion volume (LV). F-NEM is a standardized contrast signal intensity that reflects increased microvascular permeability and the magnitude of interstitial retention; ES quantifies increased discontinuity and roughness at the lesion margin to indicate invasiveness; NF indicates the progression of necrosis due to hypoperfusion and hypoxia; and LV indicates volume changes resulting from tissue destruction and fibrosis. Since input variability is suppressed by the non-immunostimulating contrast agent of the present invention and QC rules, the combination of the four indicators has a combined effect of “minimum disturbance and maximum reflection of target physiology.”

[0276] Second is the procedure for calculating ΔR1 independence. For each scan, using only QC-passed frames and voxels, it is calculated in the following order: 1) Obtain the lesion-unit F-NEM value by taking the median or upper quantile (e.g., the 75th quantile) of the lesion region of interest from the F-NEM-based enhancement map. 2) For ES, extract the median of the marginal gradient in the same coordinate system; however, since lower values ​​indicate greater damage, apply standardization to reverse the direction. 3) NF is defined as the ratio of the area of ​​the low-enhancement region within the lesion and is expressed in the range of 0 to 1. 4) LV is defined as the absolute volume of the lesion segmentation result, or as the percentage change relative to baseline when the purpose is time-series comparison. These raw values ​​are linearly normalized to a score of 0 to 100 against a pre-established reference distribution (reference groups for normal tissue, mild lesions, and severe lesions) or an institution-wide reference range. The direction of normalization is standardized to “damage increases with larger values,” and ES is mapped to the inversion score, while F-NEM, NF, and LV are mapped to the forward score.

[0277] Third, there are the weighted combination formula and base weights. The lesion-unit TDI is calculated by weighting the normalized partial scores S_F(NEM), S_ES, S_NF, and S_LV. The recommended base weights are F-NEM 0.35, ES 0.25, NF 0.20, and LV 0.20. This design aims to reflect functional changes (F-NEM) somewhat more heavily while considering morphological changes (ES, NF, LV) in a balanced manner. Weights may be adjusted based on clinical validation, such as when necrosis dominates the prognosis in specific indications, and the adjustment values ​​and reasons are recorded in the report metadata. The total patient-unit TDI is calculated by volume-weighting the TDI of all target lesions or by adopting the value of the dominant lesion as a representative value in accordance with clinical guidelines.

[0278] Fourth, there is baseline referencing and the calculation of the rate of change. In the tracking of treatment response and disease, the rate of change (percentage change) is calculated by comparing the lesion-unit or patient-unit TDI at the baseline and the follow-up point. The rate of change is defined as (TDI_follow-up-TDI_baseline) / max(TDI_baseline, ε)×100, where ε is set as a small constant to prevent the denominator from converging to zero. The rate of change relative to the baseline offsets absolute differences between individuals, thereby enhancing the consistency of assessment even in multi-institutional and multi-equipment environments.

[0279] Fifth, there are the threshold flags and judgment logic. A signal of high risk of progression is assigned when the rate of change in TDI relative to baseline exceeds a pre-defined threshold. The recommended thresholds are as follows: Progression Flag: TDI ≥ 20% increase or the absolute TDI value remains or increases at 70 points or higher. Partial Response: TDI ≥ 20% decrease with the absolute value stabilizing downward at 40 points or lower. Stable: Cases where neither of the above two conditions applies. The thresholds can be adjusted through prospective verification by indication, and the adjusted values ​​are disclosed together in the report. For patients with multiple lesions, patient-level progression is determined when the dominant lesion meets the progression flag, or when the average TDI of a lesion group accounting for 30% or more of the total volume rises above the threshold.

[0280] Sixth is uncertainty and robustness treatment. The measurement uncertainty of each partial indicator is estimated using a 95% confidence interval via bootstrapping (voxel resampling), and the confidence interval for the TDI is calculated using a propagation rule that considers weighted covariance. If a frame located at the QC boundary is included, a conservative rule is applied to lower the TDI or assign a "deferred judgment" status. To mitigate outliers, each partial score is aggregated using the upper and lower 2.5th quantile trimming (trimmed mean), and infusion leakage and venous contamination masking regions are excluded from the calculation.

[0281] Seventh is the rules for handling missing values. If a specific indicator is technically uncalculable (e.g., insufficient ES reliability due to microscopic lesions), the TDI is calculated using only the remaining indicators, but the weights are renormalized so that the sum of the remaining indicators equals 1. The reason for missing values ​​and the correction method are specified in the report metadata. When the same indicator is consecutively missing at two consecutive time points for the same lesion, that indicator is excluded from the rate of change determination.

[0282] Eighth, there are reporting rules and standard formats. Raw values, normalized scores, weights, calculated TDIs, and confidence intervals for F-NEM, ES, NF, and LV are displayed at the lesion level, while representative TDIs, rates of change, and diagnoses (progression / stability / response) at the patient level are exported as structured reports (DICOM-SR, HL7 FHIR). The report includes metadata for the formulation used (dosage, infusion-scan interval, lot information), QC checklist pass records, reasons for masking and exclusion, and audit logs for model / version / hyperparameters and processing times. To aid clinician interpretation, F-NEM heatmaps of the lesion map, necrosis and margin overlays, and time-series spaghetti plots can be provided.

[0283] Ninth is the clinical interpretation guide. TDI does not rely on absolute concentration or the absolute value of ΔR1 itself; instead, it is a semi-quantitative indicator that integrates functional and morphological changes, standardized by F-NEM, along a single axis. In inflammatory and fibrotic diseases, a decrease in TDI suggests activity calming and structural recovery, while in neoplastic lesions, a decrease in TDI implies a combined effect of vascular normalization, reduced necrosis, and volume reduction. Conversely, an increase in TDI signifies the deterioration of one or more of the following: increased permeability or retention, or marginal indistinctness, enlarged necrosis, or increased volume. It is advisable to evaluate the left-right comparison within the same patient, the rate of change from baseline, and the consistency of direction before and after treatment together. If a high-risk flag is detected, the clinician decides whether to modify treatment, perform additional tests, or re-imaging.

[0284] Tenth, weight optimization and verification by indication. In lymphedema, since marginal and volume changes are important for functional assessment, a profile is recommended to increase the weights of ES and LV, while in tumors, a profile is recommended to increase the weight of NF. Weight tuning is performed based on correlations with prognostic indicators (e.g., recurrence-free survival) and improvements in C-index in an independent validation cohort, and the final values ​​are fixed and applied to prospective operations.

[0285] According to the above design, TDI is calculated using only F-NEM and morphological indicators even in environments where it is difficult to secure ΔR1, and maintains the reproducibility of the change rate determination even in multi-institution and multi-equipment situations as long as QC, operating range, and masking rules are observed. Consequently, TDI functions as a key synthetic biomarker for clinical decision-making (progression / stability / response) in the contrast agent-quantification framework of the present invention.

[0286] The calculation of TDI can be implemented by the following general formula.

[0287] TDI = w1·F-NEM_norm + w2·ES_norm + w3·NF_norm + w4·PVC_norm

[0288] Here, Percent Volume Change is the percent change of segmentation-derived lesion volume relative to baseline and quantifies macroscopic expansion or regression of the lesion.

[0289] Here, ES_norm, NF_norm, and PVC_norm are values ​​of each indicator standardized to the 0-1 range, and weights w1…w4 are set based on clinical suitability, the balance of sensitivity / specificity by disease, and field validation data (e.g., Σw i =1). Alternatively, logistic regression, a normalized composite score, or a pre-trained ML regressor / classifier (e.g., LASSO, XGBoost) can be used, but in all cases, input metrics and weights (or model parameters) are stored as metadata for auditability.

[0290] The present invention provides non-immunostimulating polysaccharide cross-linked colloidal particles defined by a high-substitution cross-linking network based on epoxide selective modification, a weak negative charge of -20 to 0 mV, and a hydration diameter of 2–8 nm, thereby achieving pattern recognition receptor / B-cell receptor (PRR / BCR) non-operationality and minimizing disturbance of the immune-complement axis during imaging windows. As a result, the contrast signal depends primarily on target physiology (vascular permeability, interstitial retention, lymphatic drainage) rather than contrast agent origin variables, resulting in high time-series stability. Lymph-selective pharmacokinetics (avoidance of venous contamination, interstitial-to-lymphatic priority tracking) and renal-preferential excretion simultaneously achieve safety for repeated administration and reduced long-term retention. Fe / Mn / Gd or iron oxide nanoparticles coordinated to surface donor groups provide bright, uniform T1 contrast (or selective T2 contrast) even at low doses through r1 enhancement and low r2 / r1. Based on formulation metadata-based formulation-normalized enhancement indices (F-NEM), standardized calculations of PSE, WIR, TTP, AUC, ES, NF, LV, and the synthetic indices TDI can be performed, enabling comparisons within the patient, between left and right, and across institutions using semi-quantitative indices instead of absolute concentrations. Furthermore, the implementation of a server / system including QC rules such as automatic exclusion of venous contamination and infusion leakage, SNR hurdles, and B1 / T2* masking, as well as audit logs, enhances multi-institutional reproducibility and regulatory compliance. This platform is expandable within the same framework to include not only non-targeted contrast agent modes but also targeted ligand-bound and metal non-coordinating drug carrier modes, thereby connecting diagnosis, treatment, and monitoring across lymphedema, inflammation / fibrosis, and neoplastic lesions into a single standard pathway. Consequently, this invention simultaneously improves three axes—safety, reliability of quantitative interpretation, and clinical applicability (reporting and AI integration)—significantly reducing the gap between clinical trials and real-world usage environments.

[0291] Hereinafter, cross-linked dextran structures (nanoparticles) are referred to interchangeably as CDex, DNP, and C-DNP.

[0292] Figure 1 illustrates the structural conceptual diagrams of INV-001 and NEMO-103 strains, their manufacturing method, and their physicochemical properties.

[0293] Figure 2 is a comparison of the in vivo acute edema response of NEMO-103 using Crosslinked dextran (Cdex) versus Dextran T10.

[0294] Figure 3 illustrates the binding chemistry process of Cdex and Toll-like receptor (TLR) agonist (a), the Cdex@TLR chemical structure using each binding chemistry method (b), and the ratio of TLR agonist bound per Cdex using each binding chemistry method (c).

[0295] Figure 4 is an immunofluorescence staining image of the NF-κB p65 signal after 2 hours of treatment in Examples 2-5.

[0296] Figure 5 is an immunofluorescence staining image of the NF-κB p65 signal after treating cells with a TLR agonist for 4 and 6 hours, respectively.

[0297] Figure 6 shows Cdex, TLR A1 agonist, Cdex@TLR A1 and Cdex@TLR B1 After treating cells for 4 hours, cell viability according to concentration was confirmed using cell counting kit-8 (CCK-8).

[0298] Figure 7 shows the control group, Doxorubicin, and TLR A1 agonist, Cdex@TLR A1 , Cdex@TLR A1 X2 (double dose) and Cdex@TLR B1 This is the experimental design and results of the in vivo efficacy evaluation observed after administration to a xenograft model mouse.

[0299] Figure 8 is a structural formula of examples of branched polysaccharides or cyclic polysaccharides that are crosslinked to the polysaccharide crosslinked colloid particles in Example 3.

[0300] Figure 9 shows T1 MRI images (a) and signal-to-noise ratio (SNR, b) taken after intravenously administering to mice a substance synthesized by introducing iron into the crosslinks of Dextran T-5, Maltodextrin, α-cyclodextrin, β-cyclodextrin, and Inulin of Examples 3-11.

[0301] Figure 10 shows T1 MRI images confirming renal excretion after intravenous administration to mice of a substance synthesized by introducing iron into the crosslinks of Dextran T-5, Maltodextrin, α-cyclodextrin, β-cyclodextrin, and Inulin of Examples 3-12.

[0302] Figure 11 illustrates a comparison of MRA images taken 30 minutes (a) and 60 minutes (b) after administering NEMO-103 to humans in the Phase 1 / 2a clinical trial of NEMO-103, and the CNR (c), Distension (d), and Overall quality (e) of the images.

[0303] Figure 12 illustrates an MRA image taken about 30 minutes after administering NEMO-103 (a) and gadolinium contrast agent (GBCA, b) to the human body, and a comparison of the image in terms of contrast-to-noise ratio (CNR, c), Distension (d), and Overall quality (e).

[0304] Figure 13 shows MRA images taken 60 minutes after administering NEMO-103 (a,c) and GBCA (b,e) to the human body, and a comparison with GBCA in terms of CNR (c), Distension (d), and Overall quality (e) of the images.

[0305] Figure 14 shows the results of the evaluation of joint expansion, joint boundary clarity, and contrast before and after administering NEMO-103 to humans in the Phase 2b clinical trial of NEMO-103.

[0306] Figure 15 compares the effects of Gd-DOTA and INV-001 on popliteal lymph nodes (LN) and lymphangiography in Example 5. (a) 3D-TOF scan images before (0 min), 16 min, and 96 min after intradermal injection of Gd-DOTA and INV-001. (b) Analysis of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of popliteal LN and lymph vessels (LV) after administration of Gd-DOTA and INV-001.

[0307] Figure 16 shows the results of a dosage optimization study conducted in Example 5 to improve interstitial space and minimize lymph node enlargement. Visualization of the popliteal LN and LV was performed 16 minutes after injecting various concentrations and doses into the hind leg.

[0308] Figure 17 shows lymphatic images obtained via magnetic resonance lymphangiography (MRL) after injecting INV-001 and Gd-based contrast agents in various animal models (rats, miniature pigs, and beagle dogs). Each panel compares the degree of lymphatic visualization and the presence or absence of venous contamination after contrast agent injection. INV-001 demonstrates the characteristic of maintaining the lymphatic contrast effect compared to Gd contrast agents without causing venous contamination, and shows that its efficacy is maintained even in lymph node dissection disease models.

[0309] Figure 18 is an image showing contrast enhancement of INV-001 injected into the lower extremities of a Beagle dog to the central lymphatic system.

[0310] Figure 19 shows magnetic resonance lymphangiography (MRL) images acquired after injecting INV-001 into a human.

[0311] Figure 20 shows representative images of INV-MRL and NIRF-ICGL in the dorsal (d) and ventral (v) positions. Popliteal lymph nodes (LN) were observed in both modalities of the control limb (red arrow). The orange arrow indicates an area showing a diffusion pattern in NIRF-ICGL, where the contrast agent is widely spread across the skin surface, making it difficult to observe the basal lymphatic vessels (LV). However, in INV-MRL, hidden LVs in the same position can be observed.

[0312] Figure 21 shows the results of an edema test performed by administering Cdex compared to non-crosslinked standard dextran T-10. When standard dextran and Cdex were administered intravenously to rats at the same dosage (125, 250 mg / kg), edema was observed in the face and feet 30 minutes after administration only in the case of dextran.

[0313] Figure 22a shows the determination of the disease using the pattern change of the fluorescence signal in NIRF-ICGL.

[0314] Figure 22b compares images of INV-MRL and NIRF-ICGL according to each stage of normal and lymphedematous limbs.

[0315] Figure 22c shows representative images of INV-MRL and NIRF-ICGL in the dorsal (d) and ventral (v) positions. In the control limb, popliteal lymph nodes (LN) were observed in both modalities (red arrows). The orange arrows indicate areas showing a diffusion pattern in NIRF-ICGL, where the contrast agent is widely spread across the skin surface, making it difficult to observe the basal lymphatic vessels (LV). However, in INV-MRL, hidden LVs in the same position can be observed.

[0316] Figure 23 is a schematic diagram showing the qualitative stage difference between INV-MRL and NIRF-ICGL for each animal. Overall, the stage difference between the two imaging methods was not significant. Since the time difference between the two imaging measurements was minimal, it is presumed that this difference is due to the difference in imaging techniques rather than the actual change in the animals' lymphedema status.

[0317] Figure 24 shows the difference in TAR values ​​of the extremities by lymphedema stage between the control group and the control group in (A) INV-MRL and (B) NIRF-ICGL. As the lymphedema stage progressed from normal (control group) to more severe stages, the average TAR value increased in all imaging methods. In the case of abdominal NIRF-ICGL, the TAR value was 0 because no LV was observed in the control group extremities.

[0318] Figure 25 shows (A) the TAR correlation between INV-MRL and NIRF-ICGL in dorsal and ventral positions. INV-MRL and dorsal NIRF-ICGL were able to visualize the existing LV even in the control limb. Since lymphedema progresses by gradually spreading from the existing LV, they showed a higher correlation than abdominal NIRF-ICGL. (B) Correlation between TAR values ​​and qualitative stage results in each imaging aspect. For the reasons mentioned earlier, INV-MRL and dorsal NIRF-ICGL showed a higher correlation. This indicates that for an accurate diagnosis of lymphedema, the ROI should be placed in an area where existing lymph nodes can be observed even in a normal state. * represents P < 0.05, ** represents P < 0.01, *** represents P < 0.005, and **** represents P < 0.0001.

[0319] Figure 26 shows the Pearson correlation coefficients between the quantitative results of INV-MRL and NIRF-ICGL and the qualitative stage of lymphedema. The closer the value is to 1, the stronger the correlation, and generally, a correlation of 0.5 or higher was observed.

[0320] Figure 27 is an MRI image taken 24 hours after administration, showing that INV-001 was excreted from the injection site, lymphatic vessels, and lymph nodes.

[0321] Figure 28 shows T1 and T2 time mapping images taken 24 hours after administration, showing that INV-001 was excreted from the liver (L), medulla (M), and cortex (C).

[0322] Figure 29 shows the significant change in ankle diameter of the control group and lymphedematous limbs before and after surgery and radiation therapy. The insert shows a photograph of an ankle swollen due to lymphedema. **** means P<0.0001.

[0323] Figure 30 is an anatomical image showing (A) the popliteal lymph node of the hind leg and (B) the inguinal lymph node, and (C) a photograph of the result of forming a lymphedema animal model.

[0324] Figure 31 shows the absorption and emission spectra of ICG for NIRF-ICG imaging and the bandpass filter used.

[0325] Figure 32 shows (A) NIRF-ICGL of a control and lymphedematous limb and (B) animal posture and ROI location for obtaining images.

[0326] Figure 33 shows a schematic diagram and representative image of the qualitative staging assessment of lymphedema using NIRF-ICGL.

[0327] Figure 34 shows a schematic diagram and representative image of the qualitative staging assessment of lymphedema using INV-MRL.

[0328] Figure 35 shows the results of qualitative staging analysis of lymphedema in individual animals for INV-MRL and NIRF-ICGL.

[0329] Figure 36 shows the number of animals at each qualitative stage of INV-MRL and NIRF-ICGL.

[0330] Figure 37 shows the definition of the critical area ratio (TAR) in the quantitative evaluation of lymphedema in the extremities.

[0331] FIG. 38 shows a conceptual diagram of a lymphatic disease diagnosis system according to one embodiment of the present invention.

[0332] FIG. 39 shows a flowchart of a method for processing lymphatic vessel images according to one embodiment of the present invention.

[0333] FIG. 40 illustrates the detailed steps of step S1920 of FIG. 19.

[0334] FIG. 41 illustrates the detailed steps of step S2010 of FIG. 20.

[0335] FIG. 42 illustrates the detailed steps of step S2020 of FIG. 20.

[0336] FIG. 43 illustrates the detailed steps of step S1920 of FIG. 19.

[0337] FIG. 44 illustrates the detailed steps of step S1910 of FIG. 19.

[0338] FIG. 45 shows a detailed configuration of an electronic device according to one embodiment of the present invention.

[0339] FIG. 46 shows the detailed configuration of an electronic device and a server device according to one embodiment of the present invention.

[0340] The present invention will be explained in more detail below through examples. However, the following examples are intended only to clearly illustrate the technical features of the present invention and do not limit the scope of protection of the present invention.

[0341] Hereinafter, cross-linked dextran structures (nanoparticles) are referred to interchangeably as CDex, DNP, and C-DNP.

[0342] Example 1. Polysaccharide cross-linked colloidal particles with controlled size and surface charge

[0343] 1-1. Synthesis of C-DNP-3 cross-linked with 3 kDa molecular weight dextran

[0344] 1 g of dextran (molecular weight 3 kDa) was dissolved in 4.2 mL of distilled water, and 8.3 mL of NaOH solution was added. 3.3 mL of epichlorohydrin was added and stirred. After adding 7 mL of diethylenetriamine (DETA), the mixture was stirred for an additional 24 hours to synthesize C-DNP-3 cross-linked with 3 kDa dextran. After purification by ultrafiltration, the hydrodynamic size measured by DLS was 3 nm.

[0345] 1-2. Synthesis of C-DNP-5 Cross-linked with 5 kDa Molecular Weight Dextran

[0346] C-DNP-5 cross-linked with 5 kDa dextran was synthesized in the same manner as in Example 1-1, except that dextran (molecular weight 5 kDa) was used instead of dextran (molecular weight 3 kDa). After purification by ultrafiltration, the hydrodynamic size measured by dynamic light scattering (DLS) was 4 nm.

[0347] 1-3. Synthesis of C-DNP-10 Cross-linked with 10 kDa Molecular Weight Dextran

[0348] C-DNP-10 cross-linked with 10 kDa dextran was synthesized in the same manner as in Example 1-1, except that dextran (molecular weight 10 kDa) was used instead of dextran (molecular weight 3 kDa). After purification by ultrafiltration, the hydrodynamic size measured by DLS was 5 nm.

[0349] 1-4. Synthesis of C-CMDNP-10 cross-linked with 10 kDa molecular weight carboxymethyl dextran (CM-dextran)

[0350] C-CMDNP-10 cross-linked with 10 kDa CM dextran was synthesized in the same manner as in Example 1-1, except that CM dextran (molecular weight 10 kDa) was used instead of dextran (molecular weight 3 kDa). After purification by ultrafiltration, the hydrodynamic size measured by DLS was 5 nm.

[0351] 1-5. Substitution of functional groups exposed on the surface of C-DNP or C-CMDNP

[0352] 1-5-1. Amine group:

[0353] The functional group exposed on the surface of the C-DNP or cross-linked CM dextran nanoparticle (C-CMDNP) synthesized in Examples 1-1 to 1-4 is an amine group.

[0354] 1-5-2. Carboxyl group:

[0355] 30 mg of succinyl anhydride (SA) was added to 10 mL of C-DNP or C-CMDNP synthesized in Examples 1-1 to 1-4, and the mixture was stirred for 12 hours. The final C-DNP or C-CMDNP produced was purified by ultrafiltration.

[0356] 1-5-3. Thiol group:

[0357] N-succinimidyl S-acetylthioacetate was added to 10 mL of C-DNP or C-CMDNP synthesized in Examples 1-1 to 1-4, and the mixture was stirred for 12 hours. The final C-DNP or C-CMDNP produced was purified by ultrafiltration.

[0358] 1-5-4. Hydroxide group:

[0359] Nitrous acid was added to 10 mL of C-DNP or C-CMDNP synthesized in Examples 1-1 to 1-4, and the mixture was stirred for 12 hours. The final C-DNP or C-CMDNP produced was purified by ultrafiltration.

[0360] 1-6. Surface Charge Control of C-DNP / C-CMDNP

[0361] When 30 mg of SA is added according to Example 1-5-2 to 10 mL of C-DNP synthesized in Examples 1-1 to 1-3, the surface charge is -3 mV. When 50 mg of SA is added, the surface charge is -20 mV. If no SA is added, the surface charge is +5 mV. In experiments using C-DNP-3, C-DNP-5, and C-DNP-10, the surface charge according to the amount of added SA showed the same results under the same functional group substitution conditions. When 30 mg of SA is added according to Example 1-5-2 to 10 mL of C-CMDNP-10 with a surface charge of +5 mV synthesized in Example 1-4, the surface charge is -4 mV. As such, since the surface charge becomes negative as the amount of added SA increases, the surface charge can be controlled.

[0362] 1-7. Introduction of fluorescent molecules into C-DNP and control of the number of introduced molecules

[0363] 1 mg of 5 / 6-carboxyfluorescein succinimidyl ester (NHS-fluorescein) was dissolved in 1 mL of dimethyl sulfoxide (DMSO). 5, 10, and 25 μL of this solution were added to 1 mg of C-DNP-5 (surface charge: +5 mV) exposed to the amine group of Example 1-5-1. After 24 hours, the fluorescein-introduced C-DNP-5 was purified by ultrafiltration. The number of fluorescein bound to C-DNP-5 reacted with 5 mmol NHS-fluorescein / DMSO was 4, the number of fluorescein bound to C-DNP-5 reacted with 10 mmol NHS-fluorescein / DMSO was 9, and the number of fluorescein bound to C-DNP-5 reacted with 25 mmol NHS-fluorescein / DMSO was 15. As described above, it is possible to control the amount of the substance intended to bind to C-DNP by adjusting the reaction ratio.

[0364]

[0365] Example 2. Polysaccharide cross-linked colloidal particle-drug conjugate (Cdex@TLR as an anticancer agent)

[0366] The mechanism of cancer cell death by TLR agonists is described in detail in the literature (Remaute et al., Euur. J. Pharm. Biopharm, 2022, 172, 16.).

[0367] TLR agonists are known to activate various immune cells, such as dendritic cells (DCs), macrophages, and T cells. This activation can enhance the body's immune response against cancer cells.

[0368] 2-1. Preparation of Nanostructures (Cdex) Using Dextran T-10

[0369] 180 μmol of dextran T-10 (average molecular weight 10,000 Da) was dissolved in 9 mL of distilled water, and 75 mmol of epichlorohydrin and 75 mmol of NaOH were added. Then, 380 mmol of ethylenediamine was added and stirred at room temperature (RT) for 24 hours. 25 mg of succinic anhydride was added to this material at room temperature, and after a succinylation reaction for 24 hours, the substance was purified by dialysis using a 10 kDa molecular weight cutoff (MWCO) filter. The hydrodynamic size measured by DLS is 5 nm.

[0370] As a result of performing an edema test using NEMO-103 with Cdex, a crosslinked dextran of the crosslinking target dextran T-10, when dextran and Cdex were administered intravenously to rats at the same dosage (125, 250 mg / kg), edema was observed in the face and feet after administration only in the case of dextran (Fig. 2).

[0371] 2-2. Introduction of drugs capable of binding to Cdex (TLR 7 / 8 Agonists)

[0372] Currently, various types of TLR 7 agonists are commercially available, and materials containing amine groups capable of binding to the amine groups derived from the crosslinking agent exposed on the surface of Cdex in Example 2-1 were selected. Among these, two types of TLR 7 agonists (TLR 7) in which the introduced amine groups did not interfere with the chemical sites required for TLR 7 binding were selected. A1 , TLR B1 Selected an agonist.

[0373] As exemplified in Fig. 3, Cdex and a TLR agonist were bonded using the EDC (1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide), NHS (N-Hydroxysuccinimide)-PEG (polyethyleneglycol)-NHS, and BCN (Bicyclo[6.1.0]nonyne)-NHS / N3(azide)-NHS chemical bonding methods (Cdex@TLR, Fig. 3 a, b). For chemical bonding using N3-PEG-NHS, 30 mg of N3-PEG-NHS dissolved in 2 mL of DMSO was added to the Cdex material of Example 2-1, and the mixture was stirred at room temperature for 2 hours. 10 mg of BCN-NHS and TLR A1 10 mg of agonist was mixed and dissolved in 2 mL of DMSO, added, and stirred for 2 hours, after which it was purified by dialysis using a 10 kDa molecular weight cutoff (MWCO) filter. For chemical bonding using NHS-PEG-NHS, TLR was applied to the Cdex material. A1 10 mg of agonist and 20 mg of NHS-PEG-NHS were added and stirred for 2 hours, after which the mixture was purified by dialysis using a 10 kDa molecular weight cutoff (MWCO) filter.

[0374] As a result, TLR bound to Cdex A1 The number of agonist bonds was 3 for the NHS-BCN / N3-NHS chemical bond and 0.45 for the NHS-PEG-NHS chemical bond, confirming that the NHS-BCN / N3-NHS chemical bond method was the most efficient in terms of chemical bonding efficiency (Fig. 3c).

[0375] 2-3. Quantification of drugs bound to Cdex

[0376] The substance from Example 2-2 was diluted to 0.5 μg / mL, placed in a cuvette, and its absorption was confirmed using a UV-Vis Spectrophotometer at wavelengths ranging from 200 to 500 nm. While there was no absorption at 350 nm in the case of unbound Cdex, absorption was confirmed at 350 nm in the case of bound Cdex. Through this process, the amount of drug bound to Cdex was quantified to determine the TLR per mole of Cdex. A1 The number of agonist bindings was verified. More than 3 TLRs per Cdex. A1 It was confirmed that the agonist binds.

[0377] 2-4. Hydration Diameter and Colloidal Stability of Cdex@TLR

[0378] TLR in Example 2-2 A1 The agonist-bound Cdex was found to have a hydrodynamic size of 6 nm and was confirmed to be stably dispersed without clumping even in a 1 M NaCl solution, which is a much harsher condition than typical physiological conditions.

[0379] 2-5. Evaluation of in vitro efficacy of Cdex@TLR (Confirmation of NF-κB p65 signal)

[0380] Cdex@TLR A1 or Cdex@TLR B1 The efficacy of the polysaccharide cross-linked colloid particles of the present invention as a drug delivery system can be evaluated by examining whether the TLR signaling pathway (NF-κB signaling) is induced in cells expressing TLRs (Toll-like receptors) upon treatment.

[0381] To determine whether the NF-κB signaling pathway was induced through TLR signal activation, the presence or absence of TLR signaling activation following treatment with Cdex@TLR of Example 2-2 was confirmed using the RAW 264.7 macrophage cell line, a representative immune cell.

[0382] Specifically, the sample Cdex@TLR prepared in Example 2-2 A1 and Cdex@TLR B1 Two types, and control group Cdex, TLR 7 / 8 agonist (TLR A1 , TLR B1 After treating mouse RAW 264.7 macrophage cell line cells with the agonist, NFkB p65 was observed via immunofluorescence staining according to the treatment concentrations and times presented in Table 2.

[0383]

[0384] When observed via NF-κB p65 immunofluorescence staining after 2 hours of treatment, as shown in Figure 4, fluorescence was observed throughout the cytoplasm in the control group (cell only, Cdex), and the TLR 7 / 8 agonist (TLR A1 , TLR B1 agonist) and Cdex@TLR A1 , Cdex@TLR B1 It was confirmed that NF-κB p65 fluorescence was observed more strongly in the nucleus compared to the cytoplasm. It is known that when the TLR signaling pathway is activated, the NF-κB p65 fluorescent signal appears more strongly in the nucleus than in the cytoplasm. Cdex@TLR A1 and Cdex@TLR B1 TLR A1 , TLR B1Since the fluorescence intensity induced by NF-κB p65 is expressed relatively more strongly than that of the agonist, Cdex@TLR A1 and Cdex@TLR B1 It was confirmed to be effective in TLR signal activation. When comparing fluorescence intensity according to treatment concentration, it was observed that the NF-κB p65 fluorescence signal intensity at 6 μM was similar to that when treated at 9 μM.

[0385] Among various TLR 7 / 8 agonists, the optimal agonist was selected by confirming the extent of NF-κB p65 expression observed upon treatment to cells. Gardiquimod, Resiquimod, TLR A1 , TLR B1 6 μM of the agonist was administered to mouse RAW 264.7 macrophage cells for 4 hours. After 4 hours, NF-κB p65 signaling was evaluated by Western blot after separating the cytoplasmic and nuclear portions of each cell lysate. Lamin (nuclear) and β-actin (cytoplasmic) were used as housekeeping proteins for Western blot normalization. As a result of comparing the four agonists, TLR A1 In the case of the agonist, the highest NF-κB p65 expression was observed in the nucleus. Compared to treatment with an equal volume of DMSO, Resiquimod and TLR B1 When treated with an agonist, relatively low NF-κB p65 expression is observed.

[0386] Similar experiments with Gardiquimod, Resiquimod, and TLR A1 , TLR B1After treating mouse RAW 264.7 macrophage cells with 6 μM of the agonist for 4 and 6 hours, NF-κB p65 expression was observed by immunofluorescence staining (Fig. 5). TLR after 4 hours of treatment A1 In the case of the agonist, it was confirmed to have the greatest TLR signaling activation effect by showing the highest NF-κB 65 expression in the nucleus.

[0387] In short, FIG. 4 suggests that the polysaccharide cross-linked colloidal particles of the present invention, as drug carriers, provide an orientation that increases the affinity (avidity) of low-molecular-weight drugs, such as TLR 7 / 8 agonists, uniformly distributed on their surface without aggregation to cell surface receptors (TLRs).

[0388] 2-6. Evaluation of cell viability of TLR-expressing immune cells upon Cdex@TLR treatment

[0389] Verified ADC platform Cdex@TLR A1 and Cdex@TLR B1 Two types, and Cdex, TLR 7 / 8 agonist (TLR A1 , TLR B1 Cell viability was tested according to the treatment concentration and time of the agonist. Cells were evaluated using the TLR-expressing mouse RAW 264.7 macrophage cell line.

[0390] Specifically, the ADC platform was divided into an untreated group and low, medium, and high concentrations, and cell changes according to concentration and time were confirmed using the CCK-8 assay, a cell viability assay.

[0391] Looking at the CCK-8 assay results after 4 hours of treatment, it was confirmed that Cdex itself did not show any change in cell viability depending on the treatment concentration. TLR A1It was confirmed that cell viability decreased to approximately 89% starting from 30 μM for the agonist. Cdex@TLR compared to untreated A1 and Cdex@TLR B1 In this case, cell viability was enhanced by stimulating the NF-κB activation pathway (Fig. 6). From this, high concentrations of TLR A1 While the agonist was aggregated, Cdex@TLR A1 and Cdex@TLR B1 It can be seen that it activates immune cells expressing TLR without aggregation even at high concentrations.

[0392] 2-7. Confirmation of drug efficacy in cancer xenograft mice_In vivo efficacy

[0393] The efficacy of the drug was confirmed by monitoring changes in tumor size following drug infusion. Cdex@TLR A1 and Cdex@TLR B1 Cancer cell death caused by administration and the resulting immune response was confirmed in cancer xenograft mice.

[0394] Balb / c nude mice (6 weeks old) were used as experimental animals, and an in vivo efficacy was evaluated by creating a BT-474 breast cancer xenograft mouse model. 1 x 10⁶ were implanted subcutaneously in the dorsal area of ​​the mice. 6 Approximately 100 μL of BT-474 breast cancer cells and Matrigel were mixed in a 1:1 volume ratio and transplanted.

[0395] The size of the tumor is approximately 50 mm 3 When reached, Cdex@TLR A1 and Cdex@TLR B1The efficacy of the drug was confirmed by directly administering it to the tumor and measuring changes in tumor size for two weeks. The test substance was administered on day 0, day 4, day 7, day 12, and day 15 (Fig. 7). To ensure effective cancer cell death, Doxorubicin was administered to all groups except the control group on Day 1 and Day 9, and a Doxorubicin only group was established separately for comparison (Fig. 7, Table 3).

[0396]

[0397] Compared to the control group (phosphate buffered saline, PBS infusion), Doxorubicin, TLR A1 Cdex@TLR compared with the treated group A1 , Cdex@TLR A1 x2 and Cdex@TLR B1 In the treated group, a decrease in relative tumor volume over time was observed.

[0398] Cdex@TLR compared to the control group (PBS infusion) A1 , Cdex@TLR A1 x2 and Cdex@TLR B1 The anticancer therapeutic effect was confirmed by observing a decrease in relative tumor volume over time in the treated group. Doxorubicin only and TLR A1 In the agonist treatment group, the growth rate of the tumor was observed to slow down, but no reduction in tumor size was observed.

[0399] In short, the polysaccharide cross-linked colloid particles of the present invention, as drug carriers in tumor sites within the body, suggest that TLR agonists uniformly distributed on their surface provide an orientation that increases the avidity to cell surface receptors (TLRs) of immune cells without aggregation, thereby activating various immune cells such as dendritic cells (DCs), macrophages, and T cells, which can enhance the body's immune response to cancer cells.

[0400]

[0401] Example 3. Polysaccharide cross-linked colloidal particle-based T1 MRI contrast agent of the present invention

[0402] In this example, a study was conducted to form polysaccharide cross-linked colloid particles by cross-linking various polysaccharides including Dextran (Fig. 8) and to confirm whether they can be used as T1 MRI contrast agents.

[0403] 3-1. Formation of nanostructures using dextran T-5

[0404]

[0405] 180 μmol of dextran T-5 (Fig. 8 (ai), average molecular weight 5,000 Da) was dissolved in 9 mL of distilled water, and 75 mmol of epichlorohydrin and 75 mmol of NaOH were added. Then, 380 mmol of diethylenetriamine was added and stirred at room temperature (RT) for 24 hours, followed by purification using a 5 kD molecular weight cutoff (MWCO) filter.

[0406] 3-2. Synthesis of Nanostructures Using Maltodextrin

[0407] All experiments were conducted in the same manner as in Example 3-1, except that maltodextrin (Fig. 8 (a-ii), average molecular weight 990 Da) was used instead of dextran T-5.

[0408] 3-3. Synthesis of Nanostructures Using Alpha-Cyclodextrin

[0409] All experiments were conducted in the same manner as in Example 3-1, except that alpha-cyclodextrin (Fig. 8 (bi), average molecular weight 970 Da) was used instead of dextran T-5.

[0410] 3-4. Synthesis of Nanostructures Using Beta-Cyclodextrin

[0411] All experiments were conducted identically except that beta-cyclodextrin (Fig. 8(b-ii), average molecular weight 1100 Da) was used instead of dextran T-5 in Example 3-1.

[0412] 3-5. Synthesis of Nanostructures Using Inulin

[0413] All experiments were conducted in the same manner as in Example 3-1, except that inulin (Fig. 8 (c), average molecular weight 2500 Da) was used instead of dextran T-5.

[0414] To confirm the crosslinking of the materials of Examples 3-1 to 3-5, they were prepared at the same concentration, placed in a cuvette, and their absorption was checked using a UV-Vis Spectrophotometer at wavelengths of 200 to 400 nm. While general polymers do not absorb light in the visible light region, when nanoparticles are formed after crosslinking, the transmitted light is scattered, and consequently, absorption is observed to occur at short wavelengths. Although no absorption was observed in the observed range for the polymers before crosslinking, absorption was observed in the short wavelength region (200 to 250 nm) for the materials with formed nanostructures, confirming the formation of nanostructures by crosslinking.

[0415] The nanostructures synthesized according to Examples 3-1 to 3-5 have amine groups formed on their surfaces by diethylenetriamine. To quantify the number of amine groups on the surface per unit nanostructure, an o-phthalaldehyde assay was performed. The results showed that Dextran T-5 had 12.7 groups, Maltodextrin had 4.5 groups, Alpha-cyclodextrin had 8.1 groups, Beta-cyclodextrin had 9.0 groups, and Inulin had 8.0 groups. Therefore, it was confirmed that amine groups are introduced to the surface of all nanostructures when using the present synthesis method.

[0416] 3-6: Introduction of Carboxyl Groups in Nanostructures

[0417] 25 mg of succinic anhydride was added to the materials of Examples 3-1 to 3-5 at room temperature, and after a succinylation reaction for 24 hours, the materials were purified using a 5 kD molecular weight cutoff (MWCO) filter.

[0418] The number of residual amine groups in the materials of Examples 3-6 was determined through o-phthalaldehyde assay. The number of residual amine groups per unit nanostructure was found to be 0.7 for Dextran T-5, 0.2 for Maltodextrin, 0.2 for Alpha-Cyclodextrin, 0.2 for Beta-Cyclodextrin, and 0.4 for Inulin. A residual amine value of 1 or less indicates that all amine groups were substituted with carboxyl groups. Therefore, in the corresponding reaction, all amine groups in the nanostructures using various polymers were substituted with carboxyl groups.

[0419] 3-7. Introduction of Iron into Nanostructures

[0420] 45 μL of iron chloride hexahydrate solution was added to the materials of Examples 3-6. After adjusting the pH to 8 using 2.5 M NaOH, the nanostructures were synthesized by reacting at room temperature for 1 hour and then purifying with a 5 kD molecular weight cutoff (MWCO) filter.

[0421] The iron and polymer content of the materials in Examples 3-7 were analyzed. Iron bound to the surface was analyzed by inductively coupled plasma-optical emission spectroscopy (ICP-OES), and the content of each polymer was analyzed by the phenol-sulfuric acid method. As a result, the amount of iron bound to 1 mg of polymer was analyzed to be 0.03 mg for dextran T-5, 0.026 mg for maltodextrin, 0.033 mg for alpha-cyclodextrin, 0.026 mg for beta-cyclodextrin, and 0.029 mg for inulin, respectively. Through this reaction, it was confirmed that similar amounts of iron are introduced during the iron introduction reaction using various polymers.

[0422] 3-8. Measurement of Hydration Diameter and Surface Charge of Iron-Introduced Nanostructures

[0423] As a result of analyzing the hydrodynamic diameter of the materials in Examples 3-7 using DLS, it was confirmed that dextran T-5 had a hydrodynamic diameter of 3.6 nm, maltodextrin had 6.8 nm, alpha-cyclodextrin had 2.9 nm, beta-cyclodextrin had 2.8 nm, and inulin had 3.8 nm, indicating that they had similar hydrodynamic diameters. Regarding surface charge, dextran T-5 was found to have -3.01 mV, maltodextrin -6.62 mV, alpha-cyclodextrin -9.06 mV, beta-cyclodextrin -7.32 mV, and inulin -2.83 mV. Therefore, it was confirmed that the hydrodynamic diameter and surface charge of the iron-introduced nanostructures formed using this synthesis method were similar.

[0424] 3-9. Comparison of Viscosity Measurements of Nanostructures

[0425] The substances of Examples 3-7 were prepared at a concentration of 5 mg / mL, and their viscosity was measured at 25°C. The viscosities of the cross-linked polymers were similar, with dextran T-5 at 1.09 mPa.s, maltodextrin at 1.43 mPa.s, alpha-cyclodextrin at 1.26 mPa.s, beta-cyclodextrin at 1.3 mPa.s, and inulin at 1.14 mPa.s.

[0426] 3-10. Comparison of T1 MRI Contrast Effect Measurements of Nanostructures

[0427] To analyze the T1 MRI performance of the substances in Examples 3-7, the spin-spin relaxivity coefficient (r2) and the spin-lattice relaxivity coefficient (r1) were measured, respectively, and their ratio (r2 / r1 ratio) was calculated. The r2 / r1 ratio serves as a metric for determining whether a contrast agent is suitable as a T1 MRI or T2 MRI contrast agent. As a result of analyzing the substances in Examples 3-7 at 3.0 Tesla MRI, in the case of dextran, r1 was 1.90, r22.27, r2 / r1 ratio 1.19, maltodextrin is r14.60, r25.19, r2 / r1 ratio 1.13, alpha-cyclodextrin is r15.26, r25.78, r2 / r1 ratio 1.10, beta-cyclodextrin is r15.62, r26.24, r2 / r1 ratio 1.11, inulin is r14.33, r24.73, The r2 / r1 ratio was confirmed to be 1.09. Therefore, it was confirmed that all synthesized substances exhibit a T1 MRI contrast effect because they have an r2 / r1 ratio close to 1.

[0428] 3-11. Evaluation of Animal Contrast Efficacy of Iron-Introduced Nanostructures

[0429] After intravenously administering the substances of Examples 3-7 to mice, T1-weighted images were taken on a 9.4 Tesla MRI. Male Balb / c mice aged 5 weeks or older were anesthetized, and the substances of Examples 3-7 were administered into the tail veins. Subsequently, T1 MRI images were taken at time intervals, and the SNR was analyzed. As shown in Figure 9, it can be confirmed that all iron-introduced nanostructures of Examples 3-7 showed a bright signal in the blood vessel (jugular vein) about 3 to 5 minutes after injection.

[0430] 3-12. Elongation of Iron-Introduced Nanostructures

[0431] After intravenously administering the substances of Examples 3-7 to animals, T1-weighted imaging was performed on a 9.4 Tesla MRI for up to 1 hour to confirm the renal excretion of the administered substances. As shown in Fig. 10, T1 signals began to be observed in the bladder for most of the iron-introduced nanostructures within 30 minutes. This indicates that all administered substances passed through the kidneys and were excreted into the bladder.

[0432]

[0433] Example 4. MR arthrography using dextran crosslinker-based T1 MRI contrast agent NEMO-103

[0434] 4-1. Preparation of Dextran Crosslinker-Based T1 MRI Contrast Agent NEMO-103

[0435] After dissolving dextran T-10 (20 mM) in an aqueous solution, sodium hydroxide and epichlorohydrin (ECH) were added sequentially, followed by the addition of ethylenediamine (EDA) to induce crosslinking of dextran molecules (10,000 g / mol). At this time, EDA molecules that did not participate in crosslinking provided amine terminal groups to the core surface, and these amine groups reacted with succinic anhydride to be substituted with carboxyl groups.

[0436] The dextran crosslinker prepared in this way was purified using a 10 kDa molecular weight cutoff (MWCO) filter, and the purified sample was reacted with an aqueous solution of iron(II) chloride and iron(III) chloride to deposit iron ions onto the carboxyl groups on the core surface. It was purified again using a 10 kDa MWCO filter to finally obtain the water-soluble dextran crosslinker-based T1 contrast agent NEMO-103 injection.

[0437] NEMO-103 is an amorphous colloidal nanoparticle with iron (Fe) bonded to a dextran crosslinked core. The average hydrodynamic diameter of the prepared particles was 4.0±0.1 nm, and the surface charge was measured to be -11.63±2.7 mV (based on DLS analysis). Approximately 60% or more of the monosaccharide -OH functional groups were modified by the crosslinking agent, and this high degree of crosslinking prevented particle aggregation and provided structural stability through the formation of a high-density three-dimensional network.

[0438] This contrast agent remained stable without aggregation for more than 11 days under normal physiological conditions as well as at various pH (5, 7, 9) and salt concentrations (250, 500, 1,000 mM NaCl), which satisfies a stability criterion that is much longer than the 24-48 hour excretion period confirmed in non-clinical tests.

[0439] As a result of magnetic property analysis, the magnetization value in a 3 T magnetic field based on MPMS (Magnetic Property Measurement System) was measured to be 6.53 emu / gFe, and in MRI phantom experiments, the T1 and T2 relaxation coefficients according to iron concentration were measured, and the r2 / r1 ratio was 1.3, which indicates ideal relaxivity characteristics as a T1 contrast agent.

[0440] 4-2. Evaluation of Non-clinical Contrast Performance and Immunosafety of Dextran Crosslinker-based T1 Contrast Agent NEMO-103

[0441] The contrast agent NEMO-103 (identical name: INV-002) is a T1 MRI contrast agent based on water-soluble polysaccharide crosslinked colloid particles having a compact spherical three-dimensional network structure formed by selectively modifying the monosaccharide -OH functional groups of a dextran polymer through an epoxide-based first crosslinking agent, and then reacting a second crosslinking agent (e.g., containing polyhydric amine groups) intramolecularly or intermolecularly.

[0442] This structure is designed to maximize immune evasion properties by inhibiting non-specific binding to recognition receptors (BCR, PRR, etc.) on the surface of immune cells, in addition to physicochemical stability. In this embodiment, non-clinical data are presented, including safety evaluations related to the potential to induce an immune response, as well as its contrast imaging performance.

[0443] 4-2-1. Non-clinical Contrast Performance Evaluation (CNR-based Efficacy Analysis)

[0444] The image enhancement effect of NEMO-103 was evaluated in the following regions based on the contrast-to-noise ratio (CNR):

[0445] Evaluation sites: meniscus, anterior cruciate ligament, cartilage, bone

[0446] Average CNR improvement rate: Over 120% improvement compared to existing Gd-based contrast agents

[0447] Contrast duration: Increased by at least 4 times

[0448] These results suggest that NEMO-103 can obtain high-contrast images relative to tissue, providing a foundation for improving image resolution and the accuracy of joint diagnosis.

[0449] 4-2-2. Immunosafety-based toxicity assessment (GLP toxicology testing and immunological interpretation)

[0450] NEMO-103 met preclinical safety requirements according to U.S. FDA guidelines, and the following characteristics indirectly suggest that it is an immuno-non-immunostimulating particle:

[0451] HEK-293 Cytotoxicity Test: IC5*?* not determined, no cytotoxicity

[0452] hERG assay: No cardiac repolarization disorder

[0453] Platelet / Plasma Mix Test: No Coagulation Reaction

[0454] Plasma protein binding rate: Low levels measured, low likelihood of tissue accumulation

[0455] In particular, the cross-linked structure of NEMO-103 was designed based on the selective modification of monosaccharide -OH functional groups and precisely controlled cross-linking conditions, which indirectly suggests that it is a non-immunostimulating particle at the molecular level as follows:

[0456] Avoidance of binding with PRRs and BCRs: Reduced binding affinity to receptors that recognize polysaccharide structural specificity

[0457] Clustering inhibition: Does not act as a multivalent antigen, so immune signaling is not induced.

[0458] No induction of immune cell activation: No B-cell activation or antibody production even with repeated administration

[0459] Stability under pathological conditions: No hypersensitivity even in inflammatory lesions

[0460] 4-2-3. Summary of Key Toxicity Test Figures (Including Exposure Multiples Relative to Clinically Planned Dose)

[0461]

[0462] In addition, it received negative results in the reverse mutation test, chromosomal aberration test, and micronucleus test, and was determined to have no concerns regarding carcinogenicity, genotoxicity, or reproductive toxicity.

[0463] Example 4-2 indirectly demonstrated that NEMO-103, a T1 contrast agent based on non-immunostimulating polysaccharide cross-linked colloid particles, is a next-generation MRI contrast agent that secures immunological safety while simultaneously improving contrast performance compared to conventional Gd contrast agents. In particular, the fact that it does not induce an immune response even with repeated administration and that no toxic reactions are observed even at high doses suggests that the platform technology of the present invention can be widely applied in various fields of bioimaging and precision diagnosis.

[0464] 4-3. Comparison of Image Quality in Shoulder MR Arthrography

[0465] To evaluate the clinical efficacy of NEMO-103, image quality was compared after administering NEMO-103 into the shoulder joint cavity in 32 patients (a total of 80 MRA images) of the Phase 1 / 2a study.

[0466] Comparison 1: Comparison of image quality between Phase I (30 min) and Phase II (60 min) in the same patient showed no significant difference in CNR, capsular distension, and overall image quality (Fig. 11).

[0467] Comparison 2: In the comparison of NEMO-103 and GBCA-based images at Phase I, the CNR was similar, but the distension of the inferior joint capsule and axillary pouch was significantly higher in NEMO-103, and the overall image quality evaluation also showed statistically superior results (p < 0.05, Fig. 12).

[0468] Comparison 3: The difference was more pronounced in the comparison at the Phase II time point (approx. 54-55 minutes), and NEMO-103-based imaging was significantly superior to GBCA in CNR, inferior capsular distension, and axillary sac distension (p < 0.01, Fig. 13).

[0469] 4-4. Visual Turing Test (VTT)

[0470] Radiologists were evaluated on the ability to distinguish images taken at two different time points (Phase I, II) after NEMO-103 injection. The average discrimination accuracy among a total of 8 readers was 46.8% (146 / 312), which was not statistically significant compared to random guessing (p = 0.423).

[0471] 4-5. Safety Evaluation in Phase 1 / 2a Clinical Trials

[0472] The safety of NEMO-103 was evaluated in a clinical trial (Stage 1: n=9, Stage 2: n=23) involving 32 patients suspected of having a rotator cuff lesion due to shoulder pain.

[0473] 4-5-1. Adverse Reactions:

[0474] Two adverse events (ADRs) occurred, but both recovered, and there were no serious adverse events (SAEs) or drug-related dropouts. No dose-limiting toxicity (DLT) was observed.

[0475] 4-5-2. Residue and Tissue Safety:

[0476] Tissue residue of NEMO-103 in the liver and spleen was confirmed to be negative (100% disappearance) 24 hours after administration, and this was also confirmed radiologically. The signal difference between the liver and spleen MRIs before and after administration was positive (liver: 13.2 ± 4.0, spleen: 5.5 ± 0.9).

[0477] 4-5-3. Video Quality and Durability:

[0478] Similar quality was achieved in images acquired at 30 and 60 minutes after administration, and excellent performance was also demonstrated in the depiction of radiological structures (joint capsule, rotator cuff, cartilage, etc.). In particular, NEMO-103, with its long retention time, maintained an excellent CNR even after 2 hours and was completely eliminated from the joint cavity after 24 hours, excluding the possibility of retention in the body.

[0479] 4-6. Evaluation of the Correlation Between Joint Distension and MRA Image Quality in a Phase 2b Clinical Trial

[0480] This example is based on the results of a Phase 2b clinical trial conducted to evaluate the intra-articular distribution and contrast effect of NEMO-103, a dextran cross-linked T1 MRI contrast agent. In this trial, joint distension, articular sharpness, and image contrast were set as the primary evaluation items for image quality, and statistical significance (p-value < 0.0001) was confirmed in both the primary and secondary analysis groups (see Fig. 14).

[0481] As described below, this embodiment demonstrated the efficacy and safety of NEMO-103 as a T1 contrast agent for realizing high-resolution intra-articular MRA imaging in clinical practice, and in particular showed the possibility of AI-based quantitative image analysis based on the correlation between joint distension and image quality.

[0482] 1. Intra-articular Contrast Agent Distribution and Diagnostic Value

[0483] Joint distensibility is an imaging metric that quantifies the distribution uniformity and penetration range of T1 contrast agents. It enhances the visualization of intra-articular anatomical structures (cartilage, synovium, ligaments, etc.) and increases diagnostic sensitivity and specificity. In this study, improvements in joint distensibility were closely associated with enhanced image sharpness and contrast, indirectly reflecting the in vivo uniformity of contrast agent distribution and extended residence time.

[0484] 2. Clinical evidence regarding immunoinactivation properties

[0485] NEMO-103 was designed as a non-immunostimulating polysaccharide cross-linked colloidal particle-based contrast agent, and its characteristics were supported in this study through the following immunosafety indicators:

[0486] Absence of acute immune response: No immune response-based abnormalities, such as synovitis, edema, or pain, were observed after intra-articular administration.

[0487] Maintenance of normal joint distension: The stable increase in distension without excessive distension or internal joint cavity reactions indirectly suggests the absence of immune hypersensitivity.

[0488] Uniformity of contrast agent distribution and image stability: Along with the physicochemical properties of the contrast agent, stable image quality can be ensured even in an inflammatory environment within the body.

[0489] 3. Possibilities for AI-based video analysis and standardization

[0490] Quantitative imaging indicators such as joint distension, sharpness, and contrast can be utilized as standard datasets for the future development of AI-based image interpretation and automated diagnostic tools. In particular, the quantitative indicators obtained in this study enhance the reliability of contrast agent evaluation as a standard for repeatable image quality.

[0491]

[0492] Example 5. Magnetic resonance lymphangiography using INV-001

[0493] 5-1. Preparation and Characterization of Contrast Agents

[0494] Gd-DOTA (Control Group):

[0495] DOTAREM® (Guerbet, France) was diluted with physiological saline and used (final concentration: 15 mM).

[0496] INV-001 (Test Group):

[0497] Dextran T-5 (20 mM) with an average molecular weight of 5,000 Da was reacted with sodium hydroxide and epichlorohydrin, and then diethylenetriamine was added to prepare cross-linked dextran. Subsequently, the amine terminals were modified to carboxyl groups using succinic anhydride, and iron was coordinated to the surface by reacting with iron(III) chloride. The product was purified and concentrated using a 3 kDa MWCO filter.

[0498] Structural characteristics:

[0499] TEM analysis revealed that INV-001 is a spherical particle with an average diameter of 2.6 ± 0.3 nm (n=100), and negative staining was performed with uranyl acetate. According to dynamic light scattering analysis, the hydrodynamic size was 3.6 ± 0.2 nm, and the surface charge was measured to be -2.95 ± 0.20 mV. It was confirmed that more than 60% of the total number of monosaccharides constituting INV-001 were modified by a crosslinking agent, and an average of 5 iron atoms were present per particle.

[0500] Physical and chemical properties:

[0501] Magnetization: 14.25 emu / gFe (3 Tesla magnetic field)

[0502] Relaxivity: r₁ = 3.95 ± 0.35 mM -1 s -1 , r₂ = 4.73 ± 0.24 mM -1 s -1 , r2 / r1 = 1.20 ± 0.04

[0503] Others: pH = 8.23, Viscosity = 3.23 ± 0.03 cP, Density = 1.031 ± 0.001 g / cm³, Boiling point = 101°C, Freezing point = 0°C

[0504] Stability Assessment:

[0505] INV-001 maintained a stable colloidal state without aggregation under pH 5, 7, 9 and 250–1,000 mM NaCl conditions, and there was no change in particle size during the period from administration to excretion in the body.

[0506] 5-2. Hydrodynamic Magnitude and Magnetic Resonance Relaxation

[0507] The hydrodynamic size of INV-001 was measured to be 3.6 nm, and the T₁ and T₂ relaxation rates were measured at various concentrations (0.125, 0.25, 0.5, 1.0 mM) under a 9.4 T magnetic field condition. As a result, r₁ = 2.61 mM -1 s -1 , r₂ = 4.33 mM -1 s -1 , r2 / r1 = 1.66.

[0508] This means that since INV-001 exceeds the maximum size (2 nm) for intravenous invasion, it is suitable as a contrast agent that specifically remains in the lymphatic system without venous contamination.

[0509] 5-3. Video Performance Comparison: INV-001 vs. Gd-DOTA (See Fig. 15)

[0510] This example compares the pharmacokinetic characteristics and imaging performance of an iron-based contrast agent INV-001 and a gadolinium-based contrast agent (Gd-DOTA) within the lymphatic system by performing magnetic resonance lymphangiography (MRL) in Sprague-Dawley (SD) rats. In particular, INV-001 is a novel contrast agent designed to prevent venous contamination, and in this study, the optimal dosage and image quality for the visualization of lymph nodes and lymphatic vessels were evaluated.

[0511] After administering 1.125 μmol (75 μL, 15 mM) of INV-001 and Gd-DOTA, respectively, the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of lymph nodes and lymphatic vessels were compared.

[0512]

[0513] While Gd-DOTA showed a signal peak at 16 minutes after administration, INV-001 maintained a uniform high signal for a longer duration (16-32 minutes) and showed excellent contrast enhancement, particularly in lymphatic vessels (CNRLV).

[0514] In summary, INV-001 was confirmed to be promising as an iron-based MRL contrast agent capable of providing a stable contrast effect in lymph nodes and lymphatic vessels for a long time, with less venous contamination compared to Gd-based contrast agents.

[0515] 5-4. Qualitative Comparison of Venous Contamination, Cutaneous Reflux, and Lymph Node Congestion

[0516] In this example, after subcutaneously injecting INV-001 or Gd-DOTA into the lower extremities of Sprague-Dawley rats, 3D TOF-based magnetic resonance lymphangiography (MRL) was performed to qualitatively evaluate the presence of venous contamination, dermal backflow, and lymph node congestion. The evaluation was performed according to the qualitative evaluation criteria in Table 6.

[0517]

[0518] In the Gd-DOTA administration group, venous contamination was observed in both the 1.125 μmol and 2.25 μmol administration groups, with the most severe contamination observed in the 2.25 μmol group. In contrast, INV-001 visualized lymph nodes and lymphatic vessels well without venous contamination under all administration conditions. Additionally, tissue examination using methylene blue two days after MRL imaging confirmed that the lymphatic structures observed in the INV-001-based images were consistent with anatomical entities.

[0519] 5-5. Qualitative Image Quality Evaluation According to Various Injection Conditions

[0520] To comprehensively evaluate venous contamination and image quality, INV-001 and Gd-DOTA were injected subcutaneously under various conditions, MRL images of the popliteal lymph nodes and lymphatic vessels were acquired, and qualitative scoring was performed according to Table 6.

[0521] As a result of varying the concentration under fixed injection volume (75 μL) conditions, INV-001 obtained excellent scores of 4.2, 5.0, and 5.0 in the 0.75 μmol (10 mM), 1.125 μmol (15 mM), and 1.5 μmol (20 mM) administration groups, respectively. On the other hand, the Gd-DOTA 2.25 μmol (30 mM) administration group showed a low score of 3.8.

[0522] As a result of varying the injection amount under fixed concentration (15 mM) conditions, INV-001 showed scores of 3.6, 4.4, and 4.6 in the 0.3 μmol (20 μL), 0.45 μmol (30 μL), and 0.75 μmol (50 μL) administration groups, respectively. Gd-DOTA 1.125 μmol (75 μL) scored only 3.2.

[0523] As shown in Figure 16, INV-001-based MRL imaging showed no venous contamination or skin reflux compared to Gd-DOTA, and the image contrast of lymph nodes and lymphatic vessels was also excellent. In particular, the highest level of image quality was confirmed under 0.45 μmol and 0.75 μmol conditions.

[0524] Example 5-5 experimentally demonstrated that the iron-based contrast agent INV-001 possesses superior biocompatibility and lymphatic-specific imaging characteristics compared to Gd-DOTA. In particular, INV-001 has a hydrodynamic diameter of approximately 3.6 nm, which exceeds the venous penetration limit (approx. 2 nm), thereby preventing venous contamination at the source. It also exhibited a long duration of visualization in lymph nodes and excellent contrast effects. As a result of the biosafety evaluation, no residue of INV-001 was observed in tissues within 24 and 48 hours after administration, and no changes in liver or kidney signals were observed, confirming rapid elimination from the body.

[0525] Therefore, INV-001 has high potential for use as a T1 contrast agent for magnetic resonance lymphangiography (MRL) and is a promising candidate capable of providing stable image quality over a long period without venous contamination in the diagnosis and monitoring of lymphatic system diseases.

[0526] 5-6. Evaluation of Lymphangiographic Efficacy of INV-001 in Various Animal Models

[0527] In this embodiment, the lymphatic-specific imaging ability of the INV-001 contrast agent was evaluated in various animal models (Sprague-Dawley rats, miniature pigs, and Beagle dogs) and compared with a gadolinium-based contrast agent (Gd-DOTA). Each experiment was evaluated based on the degree of lymphatic visualization and the presence of venous contamination, and the results of the image analysis are shown in Fig. 17.

[0528] 1) Sprague-Dawley rat (MRL video)

[0529] 3D TOF-based MRL imaging was performed after subcutaneous injection of INV-001 or Gd-DOTA into the hind legs of SD rats. In the group injected with INV-001, peripheral lymphatic vessels with a diameter of 300 μm or less were clearly visualized, and no venous contamination was observed. In contrast, venous contamination was clearly confirmed when Gd-DOTA was administered under the same conditions.

[0530] 2) Mini pig model

[0531] MRL was performed after injecting INV-001 or Gd-DOTA under conditions corresponding to the low dose of the clinical trial (15 mM, 0.6 mL × 3 sites). When INV-001 was administered, abdominal lymph nodes and leg lymphatic vessels were clearly visualized, and no venous contamination was observed. On the other hand, when Gd-DOTA was administered, lymphatic branching appeared faintly, and venous structures were visualized, clearly showing signs of contamination.

[0532] 3) Beagle dog model

[0533] Visualization of the lymphatic vessels after injecting INV-001 into the dorsum of the foot revealed them clearly. Notably, lymphatic vessel structures were observed only in the INV-001 administered group compared to the untreated control group. This suggests that the contrast-enhancing effect of INV-001 was effectively delivered to both skin lymphatic vessels and intramuscular lymphatic vessels.

[0534] 4) Verification of efficacy in disease models

[0535] In a lymph node dissection disease model using miniature pigs, the lymphangiographic effects of the left lymph node dissection group and the right normal group were compared after the injection of INV-001. Imaging results showed that no lymph nodes were identified in the resected left side, whereas lymphatic branching and lymph nodes were clearly visualized in the right normal group. This demonstrates that INV-001 can sensitively reflect changes in lymphatic structure even in diseased states.

[0536] 5-7. Observation of contrast enhancement to the central lymphatic system after peripheral administration of INV-001 in Beagles

[0537] This embodiment evaluated whether the iron-based T1 MRI contrast agent INV-001 reached the central lymphatic system as well as the peripheral lymphatic system using magnetic resonance lymphangiography (MRL) after subcutaneously injecting the lower extremity of a beagle dog.

[0538] As shown in Fig. 18, INV-001 exhibited extensive contrast enhancement starting from the injection site at the extremity, ascending along the popliteal lymph nodes and peripheral lymphatic vessels, and reaching the central lymphatic region of the abdomen. In particular, diffusion to the central lymphatic system was observed without signal loss in the lymph nodes, suggesting that the contrast agent can continuously move along the lymphatic flow without being captured by immune cells in the lymph nodes.

[0539] These results indirectly demonstrate that the contrast agent component of INV-001 significantly inhibits the interaction with immune cells in the body, particularly immune cells within lymph nodes such as B cells and dendritic cells. This is attributed to the structural characteristics of the non-immunostimulating polysaccharide cross-linked colloid particles of the present invention. Specifically, (a-1) the -OH functional group of a monosaccharide is selectively modified with a first cross-linking agent of the epoxide series, (a-2) a three-dimensional network is formed between the modified -OH functional groups through a second cross-linking agent (e.g., polyvalent amine) or a direct reaction, and (a-3) the surface charge is controlled to a range of -20 mV to 0 mV by further modifying the amine functional group derived from the cross-linking agent into a carboxylic acid group, thereby reducing the binding affinity with the PRR or BCR of immune cells.

[0540] As a result, INV-001 (i) minimizes receptor-ligand interactions with multiple immune cells within the lymph node, thereby preventing signal transduction via receptor clustering, and (ii) does not induce B cell activation or antibody production. (iii) while maintaining resistance to hydrolysis by endogenous enzymes, (iv) stable elimination is possible without a hypersensitive immune response even under inflammatory conditions.

[0541] These technical characteristics are demonstrated through the imaging experiment of Fig. 18 that INV-001 can penetrate the lymph node structure where immune cells are densely concentrated and reach the central lymphatic system along the flow of lymph fluid, which strongly supports the applicability of the contrast agent of the present invention as a non-immunostimulating T1 lymphoid contrast agent platform.

[0542] 5-8. GLP Non-clinical Toxicity Evaluation and Validation of Immunoevasion Design of Non-Immunostimulating Polysaccharide Cross-linked Colloidal Contrast Agent INV-001

[0543] This example summarizes the results of a non-clinical toxicity test based on GLP standards, conducted to confirm that INV-001, the iron-based T1 contrast agent of the present invention, exhibits excellent biosafety under high-dose and repeated administration conditions without inducing an immune response in the body.

[0544] Conventional polysaccharide-based drugs or nanoparticles can interact with immune cells (such as B cell receptors and pattern recognition receptors) to induce unexpected immune responses (e.g., antibody production, inflammation, hypersensitivity reactions, allergies). In particular, cross-linked polysaccharide-based contrast agents may cause problems with accumulation in the body and long-term retention due to their polymeric structure, raising concerns about immunotoxicity and reduced efficacy of the contrast agent.

[0545] INV-001 of the present invention is designed as a non-immunostimulating polysaccharide cross-linked colloidal particle platform to enable B cell activation and evasion of antibody production, inhibition of receptor clustering, and stable release without retention in organs.

[0546] As shown in the table below, no toxic reactions were observed in the cardiovascular, central nervous, respiratory, immune, and major organ systems of INV-001 even at doses tens of times higher than the planned clinical dose, and it was confirmed that it is completely excreted from the body without accumulation.

[0547]

[0548] In addition, negative results were observed in the reverse mutation test, chromosomal aberration test, and micronucleus test, ruling out the possibility of genotoxicity.

[0549] INV-001 of the present invention applies a molecular design strategy that blocks non-specific binding to immune cell receptors to overcome the immunological and pharmacokinetic limitations of existing gadolinium-based contrast agents. Immuno-evasion and reduced residue of this contrast agent were demonstrated through imaging results (see Fig. 18 and Examples 5-7) showing that the agent moves stably to the central lymphatic system without being phagocytosed or retained by immune cells even after passing through lymph nodes.

[0550] Furthermore, GLP test results have proven that it is a non-immunostimulating contrast agent that is safely excreted without antibody production or tissue reaction even with repeated administration, and significantly reduces the possibility of immune-based side effects (allergies, inflammation, etc.) associated with polysaccharide-based drugs.

[0551] Through this embodiment, it was confirmed that INV-001 is a highly safe contrast agent that is excreted without accumulating in vivo, and that it possesses the characteristic of suppressing immune activation upon repeated administration based on structural properties in which its binding affinity to immune cell receptors is weakened. Therefore, INV-001 can be utilized as a lymph-specific T1 contrast agent with a low risk of inducing an immune response and ensured safety even under conditions of high dose and repeated administration.

[0552]

[0553] Example 6. Evaluation of lymphatic-specific imaging efficacy and immunoinactivation-based safety of INV-001 in Phase 1 clinical trial

[0554] This embodiment relates to the results of a Phase 1 clinical trial to evaluate the image quality and safety, in particular the responsiveness to the immune system, when INV-001 is applied to humans.

[0555] After administering INV-001 intradermally or subcutaneously to the distal end of the foot, high-resolution magnetic resonance lymphangiography (MRL) was performed. As a result, INV-001, originating from the injection site, traveled along the peripheral lymphatic vessels, and clear and sharp images of the lymphatic vessels were observed (see Fig. 19). The images reflected the distribution of the contrast agent within the lymphatic system and enabled lymphatic-specific, high-resolution visualization over time following injection.

[0556] In parallel, the safety of INV-001 was evaluated in a Phase 1 clinical trial involving three dose groups (0.7 mL / site, 1.0 mL / site, and 1.3 mL / site) administered to a total of nine healthy adult volunteers (see Table 8).

[0557]

[0558] As a result, the following key indicators related to immune safety were identified:

[0559] Maximum Tolerated Dose (MTD) Not Reached: The maximum tolerable dose (MTD) attributable to toxicity was not observed in any dose group, suggesting that immune system-related adverse reactions, such as systemic inflammatory responses, febrile responses, and cytokine storms, did not significantly occur even at high doses.

[0560] No Adverse Drug Reactions (ADRs) or Serious Adverse Events (SAEs): No drug-related adverse reactions (ADRs) or serious adverse events (SAEs) were observed in any of the study subjects. This means that immune-based adverse effects, such as allergies, rashes, and vasculitis, were not clinically identified.

[0561] Successful Lymphatic System-Specific Imaging: Despite lymph nodes and lymphatic vessels being tissues densely populated with immune cells, INV-001 stably visualized the lymph nodes, and no image degradation or local inflammatory response due to interaction with immune cells was observed. This indirectly demonstrates that INV-001 was not recognized as an immunogenic substance by T cells, B cells, macrophages, etc., in the lymph nodes.

[0562] In particular, INV-001 is a water-soluble polysaccharide cross-linked colloidal particle with a compact spherical three-dimensional network structure formed by selectively modifying monosaccharide -OH functional groups and precisely controlling cross-linking agent conditions. It has a significantly low binding affinity to immune cell receptors such as B cell receptors (BCR) and PRRs, making it highly unlikely to induce an immune response. These structural characteristics are directly linked to the non-immunostimulant properties confirmed in Phase 1 clinical trials.

[0563] The results above suggest that INV-001 demonstrates both safety and efficacy as a T1 MRI contrast agent capable of precisely visualizing the lymphatic system in humans without immune system-related adverse reactions, supporting its potential for future clinical application in the diagnosis and monitoring of lymphatic diseases.

[0564]

[0565] Example 7. INV-MRL-based contrast imaging technique for lymphedema staging: Application of non-immunostimulating polysaccharide cross-linked colloidal particles

[0566] This embodiment describes the results of effectively visualizing lymphatic lesions in an animal model of lymphedema and applying them to stage assessment using Magnetic Resonance Lymphangiography (hereinafter INV-MRL) based on chemically cross-linked non-immunostimulating polysaccharide cross-linked colloidal particle-based contrast agent INV-001.

[0567] The contrast agent used in this embodiment is a colloidal particle having a highly precise three-dimensional network structure formed by selectively modifying the -OH functional groups of a monosaccharide using an epoxide-based first crosslinking agent, and then inducing a second crosslinking agent (including polyvalent amines) or a direct crosslinking reaction between adjacent -OH functional groups. This particle is designed to avoid interaction with immune cells in vivo (e.g., BCR, PRR, etc.) by substituting some or all of the amine groups derived from the crosslinking agent with -COOH functional groups to control the surface charge to within the range of -20 mV to 0 mV. This structural design prevents the clustering of immune cell receptors and imparts non-immunostimulating characteristics that do not induce immune activation even upon repeated administration.

[0568] Figure 20 is a representative image showing the results of applying INV-MRL and NIRF-ICGL, respectively, in an animal model with induced lymphedema. INV-MRL is advantageous for capturing structural abnormalities that are difficult to identify with NIRF-ICGL, as it clearly visualizes deep lymphatic vessels even in areas where the contrast agent diffuses along the skin surface. In particular, as indicated by the orange arrow in Figure 20, INV-MRL was able to image even the hidden collateral lymphatic vessels (collateral LV) located beneath the diffused contrast agent.

[0569] INV-MRL was also able to qualitatively identify structural abnormalities such as lymphatic leakage, formation of accessory pathways, and lymphatic dilation according to the stage of lymphedema classified from 0 to 4, and these abnormalities showed a high correlation with the splash, stardust, and diffusion patterns defined in NIRF-ICGL. The Thresholded Area Ratio (TAR) values ​​derived from the same ROI showed a quantitative correlation that increased with stage progression, and the TAR values ​​measured by INV-MRL demonstrated higher sensitivity than those of NIRF-ICGL (R 2 =0.5919 vs. 0.5309, dorsal NIRF).

[0570] In particular, INV-MRL offers the following technical advantages:

[0571] Based on non-immunostimulating polysaccharide cross-linked colloidal particles, retention in immune organs such as lymph nodes is limited.

[0572] Unlike GBCA-based MRL, selectively visualizes only lymphatic flow without venous contamination.

[0573] Low dependence on ROI, and reproducible deep lymphoid structures under the same conditions.

[0574] The 3D reconstruction function enables quantitative analysis of the degree of lymphatic diffusion according to disease progression.

[0575] Through this technical verification, INV-MRL is evaluated as a superior imaging platform capable of visualizing deep lymphatic structures compared to the existing optical imaging-based NIRF-ICGL, and has demonstrated the applicability of non-immunostimulating polysaccharide cross-linked colloidal particles in the field of lymphatic system diagnosis.

[0576] This embodiment demonstrates at a preclinical level that the design logic and structural characteristics of non-immunostimulating polysaccharide cross-linked particles, which are key technical components of the present invention, can provide a new paradigm for lymphatic imaging technology, and provides a technical foundation capable of establishing new standards for high-precision lymphatic imaging and staging assessment in future clinical applications.

[0577]

[0578] Example 8. Evaluation of the efficacy of INV-001 using a lymphedema animal model

[0579] The efficacy of INV-001 was evaluated through preclinical studies using a lymphedema animal model. Images obtained from INV-001-enhanced magnetic resonance lymphangiography (INV-MRL) were compared with images obtained from another lymph-targeted imaging technique, near-infrared indocyanine green lymphangiography (NIRF-ICGL). Additionally, a quantitative evaluation method based on the two imaging techniques was developed to compare correlations and assess potential utility in a clinical setting.

[0580] Although all animals underwent the same surgical procedure and radiation exposure, the degree of lymphedema may vary among animals. Therefore, a sufficient number of animals were used, and the severity was classified into four levels based on the degree of postoperative lymphedema. Considering potential interactions between different contrast agents, Evans Blue, commonly used to visualize the lymphatic system during surgery, was not used for model construction. NIRF-ICGL was performed three days after the MRL experiment to account for the time required for the excretion of INV-001 (less than 48 hours). In each lymphangiography experiment, the contrast agent was injected into both the control group and the limbs affected by lymphedema, and measurements were taken on both sides.

[0581] 8-1. Formation of hindlimb lymphedema animal models

[0582] To create a hindlimb lymphedema model, a total of 40 male Sprague-Dawley rats weighing 280–320 g (8–9 weeks of age) were used. Ankle edema was evaluated for one week after the procedure and irradiation, and 29 of the 40 animals suitable for lymphedema development were used in this experiment (Fig. 29). Twenty-six lymphedema animal models were used for INV-MRL and NIRF-ICGL, and three of them were used to verify the excretion of INV-001. Lymphedema was considered to have occurred if the ankle diameter of the affected limb was at least 1 mm larger than that of the unaffected limb. Statistical analysis of ankle cross-sectional area also revealed a significant difference between the limbs affected by lymphedema and the unaffected limbs (control group) (P<0.0001).

[0583] All animal procedures were performed in a specially designated area, and surgical and radiological techniques adhered to established procedures for constructing an animal limb lymphedema model. Prior to surgery, the animals were initially sedated with 4% isoflurane gas and then anesthetized with a dose of 50 mg / kg of a combination of tiletamine / zolazepam (Zoletil 50; Virbac, France) and Xylazine (Rompun; Bayer Korea, Seoul, South Korea) in a 5:1 volume ratio. During anesthesia, the hair on the hind legs was shaved using electric clippers. The surgical site was then disinfected with 75% ethyl alcohol to prepare the stage for the microsurgical lymph node removal procedure. A circumferential incision was made through the skin along the left groin to the subcutaneous layer to locate the inguinal and popliteal lymph nodes (LN), which are the dominant lymph nodes of the hind leg. The lymph nodes, along with the surrounding adipose tissue, were removed using an electrocautery device (Bovie Medical Corporation, Item No. 18010-00, FL) without damaging other tissues. The cut edges of the skin were cauterized using the device and fixed to the basal fascia with 4-0 nylon sutures spaced 2 mm apart. Immediately after surgery, ketoprofen (1 mg / kg; SCD Ketoprofen Inj., Samchundang Pharmaceutical, Seoul, South Korea) was administered via intramuscular injection (Fig. 30). Two days after surgery, an X-ray irradiator (X-Rad 320, Precision X-Ray Inc., CT) was used to irradiate the incision site with a concentrated radiation dose of 20 Gy (10 segments at a rate of 1 Gy / min), and the rest of the body was protected with an 8 mm lead shield.

[0584] 8-2. INV-MRL

[0585] After confirming model formation, unenhanced MRL and INV-MRL images were obtained from the animal model. During the imaging scan, the animal model was anesthetized with 2-2.5% isoflurane (Terrell™ Isoflurane, Piramal Critical Care, India) mixed with 70% nitrous oxide and 30% oxygen. Body temperature was maintained using a heated bed, and stable respiration was continuously monitored. Unenhanced MRL images were obtained prior to the injection of INV-001. Next, a 2 μl dose of 15 mM INV-001 was injected into the intradermal / subcutaneous space between the 3rd and 4th toes of the control (normal) and lymphedematous (affected) limbs using a 26 Gauge BD Ultra-Fine TM Lymph flow was visualized using INV-MRL with a 0.5 mL II insulin syringe (BD Medical-Diabeters Care, California). After 30 minutes of injection, INV-MRL was performed every 16 minutes up to 96 minutes in a coronary 3D Time-of-Flight (TOF) sequence with saturation bands. 3D TOF MRI scans were performed using a 9.4T Bruker preclinical MRI scanner (PharmaScan 70 / 16, Bruker BioSpin GmbH, MA) equipped with a 72 mm penetration volume coil and a surface receiver coil designed for small animals. INV-MRL images were acquired in the axial, sagittal, and coronal planes, and the MRI parameters were as follows: TR / TE=10 / 2.54 ms, flip angle=40°, matrix size=256×256×192, FOV=70×70×30 mm 2 MRI image analysis was performed using ImageJ software (ImageJ 1.53cv, http: / rsbweb.nih.gov / ij / ; NIH, Bethesda, MD).

[0586] 8-3. Qualitative evaluation of lymphedematous limbs in NIRF-ICGL using INV-MRL with cutaneous reflux staging and established descriptions

[0587] NIRF-ICGL images were obtained using a custom imaging system for animal experiments and compared with INV-MRL results. The imaging system included a 2-inch bandpass filter (FF01-832 / 27-50-D; Semrock, West Henrietta, NY) transmitting a wavelength band of 823–837 nm and a 4.2-watt power LED with a 730 nm peak (LST1-01G01-FRD1-00; Opulent Americas, Raleigh, NC) (Fig. 31). Prior to imaging, hair on the animals' hind legs, including the region of interest (ROI), was removed using electric scissors and depilatory cream to prevent light scattering while sedated with 4% isoflurane gas. Subsequently, 2 μL of ICG fluorescent dye solution was injected intradermally and subcutaneously into the paws in the same manner as for INV-MRL. Similar to the procedure for INV-MRL, images were taken 30 minutes after ICG injection. Due to the penetration depth limitation of NIRF-ICGL, images were acquired separately at the dorsal (d) and ventral (v) positions, unlike INV-MRL (Fig. 32).

[0588] In clinical practice, cutaneous reflux staging using NIRF-ICGL assessed the severity of lymphedema based on patterns observed in static images. Under normal conditions, interstitial fluid from the extracellular matrix is ​​collected and transported to the lymphatic system. Therefore, ICG agents mixed in the interstitial fluid travel primarily along lymphatic vessels (LVs) and appear as linear patterns in NIRF-ICGL images. When lymphedema occurs due to lymphatic occlusion, lymphatic drainage in the collecting lymphatic vessels is impaired. To compensate for this, the lymphatic system attempts to find bypass routes through secondary pathways, which appear as a splash pattern. As lymphatic occlusion worsens, leakage occurs not only in the collecting LVs but also in secondary lymphatic vessels, appearing as a stardust pattern characterized by scattered dots. As occlusion deteriorates, the interstitial fluid becomes immobile and accumulates in the matrix, appearing as a diffusion pattern. Ultimately, there is no movement at all, and the contrast agent is not observed beyond the injection site; in the most severe cases, this is referred to as a blackout pattern. In this embodiment, the description of cutaneous reflux staging used in clinical practice was adopted as defined after being slightly modified to suit the animal model (Table 9, Fig. 33).

[0589] [Correction pursuant to Rule 91 Dec. 29, 2025]

[0590] Next, an explanation for INV-MRL for qualitative assessment was established using maximum intensity projection (MIP). The staging criteria explanation for the qualitative assessment of lymphedema severity was established based on lymphatic vessel visibility, the development of collateral pathways, lymphatic vessel dilation, and the diffusion of lymphatic fluid into surrounding tissues (Table 10, Fig. 34). The criteria based on changes in LV and drainage patterns share the same pathophysiological basis as previous studies using cutaneous reflux. Of the 26 animals used for qualitative assessment and staging, 11 (animals 1, 2, 3, 10, 14, 15, 16, 17, 25, 27, and 28) were excluded from the evaluation because, as expected, they failed to form a lymphedema model or died during the experiment for unknown reasons (Figs. 35, 36). Three independent researchers evaluated the classification based on the explanation in a blind manner.

[0591] [Correction pursuant to Rule 91 Dec. 29, 2025]

[0592] 8-4. Quantitative evaluation for lymphedema limbs

[0593] In both imaging methods applying qualitative evaluation criteria, it was observed that as lymphedema worsened, the contrast-enhancing area (white area on a black background) transitioned from a linear pattern to a diffuse pattern due to lymphatic vessel dilation, lymphatic leakage, and pooling. The threshold area ratio (TAR) was defined by setting a threshold using the area higher than the median based on the histogram value of total brightness. TAR was calculated by dividing the threshold area by the entire region of interest (ROI) (Equation 1).

[0594] [Equation 1]

[0595]

[0596] TAR quantitatively represents the diffusion of lymph fluid within the interstitial matrix within a given ROI and can serve as a potential parameter for quantitatively evaluating the progression of lymphedema. An anatomically consistent ROI was established from the animal's ankle to the hip joint. Using ImageJ software, a threshold region for exclusively selecting the LV was determined in the control limb and similarly applied to the lymphedematous limb. The range selected in the threshold function was set to level 0 to ensure that INV-MRL and dorsal NIRF-ICGL selected only the LV. The same threshold range was then applied to the evaluation of all other images (Fig. 37). Since the LV could not be observed in the normal state (level 0), the value of abdominal NIRF-ICGL was 0.

[0597] In fact, when TAR was applied to representative images of each skin reflux pattern in previous NIRF-ICGL results, it was observed that TAR increased with increasing severity of lymphedema (Fig. 22a). Based on this, TAR was measured for each step of the NIRF-ICGL and INV-MRL images obtained in this embodiment. The TAR values ​​were calculated using software according to the definition, and animal information was blinded during the analysis.

[0598] 8-5. Statistical Analysis and Correlation Analysis Between Variables

[0599] Statistical analysis was performed to confirm the significance of ankle diameters in both hind legs of each animal, and statistical differences before and after surgery were investigated between the limbs affected by lymphedema and the control limbs after model formation. To investigate the correlation between the quantitative parameters of INV-MRL and the dorsal and ventral NIRF-ICGL, as well as the correlation between these parameters and qualitative staging results, a significance test of the Pearson correlation coefficient using a two-tailed t-test was performed. In these tests, statistically significant linear relationships were confirmed by calculating the R-squared and P-values ​​derived from Pearson's r. The significant P-values ​​for each correlation were P<0.05(*), P<0.01(**), P<0.005(***), and P<0.0001(****), respectively. This analysis was performed using GraphPad Prism 10 (version 10.1.2; GraphPad Software, Inc, Boston, Massachusetts) and Microsoft Excel 2019 (version 2404; Microsoft Corporation, Redmond, Washington).

[0600] 8-6. Release of INV-001 from the injection site, lymphatic vessels, and lymph nodes

[0601] Excretion was performed on 5 rats (right: normal, left: lymphedema) 24 hours after administration. In the MRI images taken 24 hours later, the T1 MRI signals observed at the time of administration were not observed in the injection site (i.e., foot), lymphatic vessels, and lymph nodes, so it was determined that all of the administered INV-001 was excreted (Fig. 27).

[0602] 8-7. Excretion of INV-001 in the Liver and Kidney

[0603] To evaluate the excretion of INV-001 in the liver and kidneys, changes in T1 and T2 times were analyzed. T1 and T2 time maps obtained from the liver and kidneys (medulla and cortex) were used for analysis and compared with values ​​before administration (hour 0). T1 and T2 times did not show significant changes compared to hour 0 even after 24 hours. Therefore, it was determined that INV-001 was effectively excreted from the body within 24 hours (Fig. 28).

[0604] [result]

[0605] 1. Proof of Concept of INV-MRL

[0606] As can be seen from the results of INV-MRL, INV-001 selectively visualized only the lymphatic vessels (LV) and drainage from the injection site to the brachial lymph nodes without venous contamination.

[0607] It also showed changes in lymphatic drainage occurring in lymphedematous limbs. This demonstrates that INV-MRL can be used to obtain lymph-selective imaging similar to NIRF-ICGL in a clinical setting.

[0608] 2. INV-MRL and NIRF-ICGL imaging results of lymphedematous limbs

[0609] INV-MRL and NIRF-ICGL effectively visualized the LV and popliteal lymph nodes (red arrows, Figs. 22b and 22c) collected from the control limb. Both imaging modalities exhibited nearly identical morphologies, showing high concordance particularly in the dorsal position images of NIRF-ICGL, which displayed the collected LV. These results highlighted the utility of this technique in providing complementary visual information regarding lymphatic structure and function. This indicates that the pathophysiological characteristics of lymphedema—spreading from the site of existing lymphatic vessels to surrounding areas due to lymphatic leakage and occlusion—were observed similarly in both imaging modalities across all stages. INV-MRL showed abnormal lymphatic drainage in the lymphedematous limb, which was similar to NIRF-ICGL. In particular, INV-MRL was able to visualize the deep collateral lymphatic vessels (LV), which were difficult to detect with NIRF-ICGL because the contrast agent spreads across the skin surface due to lymphatic leakage. In the area indicated by the orange arrow in Fig. 22c, INV-MRL was able to visualize the collateral LV, whereas it was difficult to observe with NIRF-ICG. This result demonstrates the advantage of INV-MRL in providing comprehensive depth and detail when imaging deeper structures of the lymphatic system, which are important for the complete evaluation of lymphedema.

[0610] 3. Qualitative assessment of lymphedema stage

[0611] Stage 0 represents a normal state according to the description of NIRF-ICGL, and starting from Stage 1, abnormal patterns begin to appear in the order of splash, stardust, and diffuse. These patterns gradually spread from the site where the lymph node was resected to the surrounding area, and subsequent patterns indicate further progression of lymphedema (Fig. 33). In the case of INV-MRL, Stage 0 represents a normal state, similar to NIRF-ICGL. At Stage 0, the previously collected LV from the injection site to the lymph node was visualized. Starting from Stage 1, collateral pathways begin to gradually appear around the existing LV, similar to the splash pattern observed in NIRF-ICGL. As the stages progress, the area where collateral pathways are visible gradually expands. Stage 4 is characterized by the observation of lymphatic fluid diffusion and pooling in some areas (Fig. 34).

[0612] When images obtained from the two modalities were classified according to their respective evaluation criteria, a total of 21 lymphedematous limbs were distributed across all stages except stage 5, with at least 4 animals observed in each stage. The number of animals was generally similar across stages, with the highest number observed in stage 3. The imaging sequence involved first imaging with INV-MRL, followed by imaging with NIRF-ICGL after INV-001 was removed. However, since the measurements were taken within 3 days of each other, it is unlikely that significant physiological changes occurred during this period. While the stage assessment results between the two imaging techniques were generally consistent, slight differences in qualitative stage assessment were observed even within the same animal (Fig. 23).

[0613] 4. Correlation between Quantitative Assessment and Qualitative Staging of Lymphedema Based on TAR

[0614] Based on the definition of Tough Area (TAR), the TAR of all images collected from INV-MRL and NIRF-ICGL was calculated. Figure 24 shows the average TAR for each qualitative stage of lymphedema. In both INV-MRL and NIRF-ICGL, the TAR increased proportionally with higher lymphedema stages classified by qualitative assessment. Differences in each TAR value could be distinguished according to the qualitative staging criteria.

[0615] In addition, the correlation between TAR values ​​obtained from INV-MRL and NIRF-ICGL images at dorsal and ventral positions was investigated. The quantitative results of INV-MRL showed a relatively higher correlation with the dorsal NIRF-ICGL results than with the ventral results (P-value: **0.0038 < *0.1547). The Pearson correlation coefficient squared (R2) was 0.2614 and 0.1787, respectively (Fig. 25A). Next, to evaluate the relationship between qualitative classification and quantitative parameters for lymphedema assessment, the correlation between TAR values ​​and staging results for each animal was investigated. TAR values ​​measured by INV-MRL and dorsal NIRF-ICGL increased according to the lymphedema staging classification of each animal. This indicates a very high correlation between the qualitative and quantitative assessments (P-value <0.0001). The TAR values ​​for ventral NIRF-ICGL also showed a high correlation, although not as high as the results of the previous two images. The R2 values ​​were 0.5919 for INV-MRL, 0.5309 for dorsal NIRF-ICGL, and 0.3622 for ventral NIRF-ICGL (Fig. 25B).

[0616] [discussion]

[0617] Previous studies confirmed that INV-001, an iron-based contrast agent, possesses lymphatic system specificity, remaining in the lymphatic system longer than conventional GBCA and not causing venous contamination. Due to these characteristics, it is advantageous for obtaining in vivo imaging of the lymphatic system. However, to determine whether these characteristics could be useful for diagnosing lymphatic diseases such as lymphedema in clinical settings using animal models of diseases that share the pathophysiological characteristics of the human lymphatic system, Example 8 aimed to evaluate the potential of INV-MRL using a hindlimb lymphedema animal model. INV-MRL not only clearly visualized the lymphatic vessels (LV) and lymph nodes of the control limb but was also able to effectively image lymphatic dilation (LV dilation), collateral pathway expansion, and lymphatic leakage around the LV caused by impaired lymphatic circulation in the lymphedematous limb.

[0618] These abnormal lymphatic flow patterns were similar to those observed in NIRF-ICGL, an imaging method already widely used for lymphedema assessment in clinical settings. Therefore, NIRF-ICGL images were also obtained from the same animals for comparative qualitative evaluation. The stage of the lymphedema animal model was classified using NIRF-ICGL images and clinically established staging criteria based on cutaneous reflux patterns, and staging criteria for INV-MRL were established by comparing them with INV-MRL images. When comparing the same animals and the same ROI, the staging criteria for INV-MRL were found to be consistent with the staging criteria for NIRF-ICGL. Considering that NIRF-ICGL is known to have high accuracy in evaluating lymphedema in clinical settings, it is expected that INV-MRL will also demonstrate higher accuracy compared to NIRF-ICGL when applied clinically.

[0619] In addition, a quantitative evaluation method using TAR, a value representing the degree of contrast agent diffusion within tissues, was employed to assess the severity of lymphedema in these two imaging techniques. As lymphedema progresses and leaked lymph spreads to surrounding tissues, the lymphangiography image was changed from a linear form (visualizing only the LV) to a diffusion zone form. Consequently, TAR values ​​tended to increase according to the severity of the lymphedema stage classified by the extent of disease progression. A comparative analysis of TAR values ​​and qualitative stages for each animal revealed that the values ​​increased as the qualitative stage progressed. This suggests that TAR can be used as a quantitative parameter corresponding to the qualitative assessment for lymphedema diagnosis.

[0620] Figure 26 shows a correlation heat map visualizing the relationship between TAR values ​​obtained from INV-MRL and dorsal / ventral NIRF-ICGL in individual animals and the corresponding lymphedema stages. The TAR values ​​of INV-MRL showed a stronger correlation with dorsal NIRF-ICGL than with ventral NIRF-ICGL in each animal. This is presumed to be because both INV-MRL and dorsal NIRF-ICGL can visualize previously collected lymphatics even in a normal state. While TAR values ​​from all imaging methods generally showed a good correlation with stage classification, INV-MRL and dorsal NIRF-ICGL showed particularly good correlations. Since lymphedema often develops along existing collecting lymphatics and gradually spreads to wider areas due to lymph leakage, obtaining images containing ROIs where these lymphatics are clearly visible is crucial for detecting abnormalities in lymphatic flow. This suggests that imaging methods capable of visualizing normal lymphatics allow for a clearer assessment of lymphedema progression.

[0621] INV-MRL and NIRF-ICGL enabled the qualitative staging and quantitative assessment of lymphedema by visualizing lymphatic anatomy and function. However, INV-MRL had fewer restrictions on ROI selection compared to NIRF-ICGL. While INV-MRL could visualize deeper lymphatic vessels and drainage through its 3D reconstruction capabilities, NIRF-ICGL was specialized in imaging superficial lymphatic vessels and drainage due to limitations based on penetration depth. Due to these limitations, NIRF-ICGL requires careful selection of ROIs where the existing LV is visualized, even in a normal state, to clearly visualize the onset and progression of lymphedema. In contrast, INV-MRL can visualize the existing LV within a consistent ROI, allowing for earlier and clearer prediction of lymphedema onset.

[0622] In Example 8, using a disease animal model, we evaluated how well INV-MRL could visualize the onset and progression of lymphedema compared to NIRF-ICGL, a conventional lymphangiography method. Based on preclinical results, it was confirmed that INV-001 has significant potential as a lymphatic system-specific contrast agent. Using INV-MRL, we proposed staging criteria for the qualitative assessment and quantitative parameters of lymphedema. This example utilized a hindlimb lymphedema animal model commonly used in lymphedema research, which closely mimics the pathophysiological characteristics of lymphedema onset and progression in humans. Therefore, the results of the preclinical study can be applied clinically. Furthermore, INV-MRL effectively visualized the anatomical structure and flow of lymph, as well as disease-induced changes, without leg vein contamination, unlike conventional MRLs using GBCA. Thus, INV-MRL is expected to become an essential imaging method for evaluating lymphatic circulation disorders, including lymphedema.

[0623]

[0624] Example 9. Lymphatic system diagnosis system

[0625] Hereinafter, examples of utilizing the embodiments described above will be explained in detail with reference to FIGS. 38 to 46.

[0626] FIG. 38 shows a conceptual diagram of a lymphatic disease diagnosis system according to one embodiment of the present invention.

[0627] Referring to FIG. 38, a lymphatic disease diagnostic system (1000) may include an electronic device (100) and a medical device (200). The lymphatic disease diagnostic system (1000) may acquire an image of a patient's (10) lymphatic vessels using a composition for lymphatic disease imaging diagnosis. The lymphatic disease diagnostic system (1000) may provide diagnostic information regarding lymphatic disease by analyzing the acquired lymphatic vessel image.

[0628] The patient (10) may be a subject requiring diagnosis of a lymphatic disease. A lymphatic disease imaging diagnostic composition may be injected into the patient (10) before taking a lymphatic vessel image.

[0629] The medical device (200) may be a device that generates medical images by photographing the inside of the patient's (10) body. The medical device (200) may scan the patient (10), into whom a composition for imaging diagnosis of lymphatic system diseases has been injected, using a magnetic resonance imaging method. The medical device (200) may transmit raw data or lymphatic vessel images obtained through scanning to an electronic device (100).

[0630] In one embodiment of the present invention, the medical device (200) may be a magnetic resonance imaging (MRI) device.

[0631] The electronic device (100) can obtain lymphatic vessel information by processing data obtained from the medical device (200). The electronic device (100) may include at least one processor including a processing circuit and at least one memory storing at least one instruction. The electronic device (100) can perform the operation of processing a lymphatic vessel image by having at least one processor execute at least one instruction stored in the memory.

[0632] The electronic device (100) can acquire an image of a lymphatic vessel through which a composition for imaging diagnosis of lymphatic system diseases passes. Based on the acquired lymphatic vessel image, the electronic device (100) can acquire lymphatic vessel information corresponding to the lymphatic vessel. The lymphatic vessel information may enable a quantitative evaluation of lymphatic system diseases, such as lymphedema, by including parameters regarding lymphatic flow, structural abnormalities, and specific abnormalities of the lymphatic vessel.

[0633] Lymphatic vessel images processed by the electronic device (100) can be obtained using a composition for imaging diagnosis of specific lymphatic system diseases. The composition for imaging diagnosis of lymphatic system diseases may contain water-soluble polysaccharide cross-linked colloid particles and a contrast agent.

[0634] In one embodiment of the present invention, the electronic device (100) may be a computing device such as a server, a personal computer (PC), a workstation, a tablet computer, or a smartphone. The electronic device (100) may be connected to the medical device (200) via a wired or wireless communication network. The electronic device (100) may also be implemented in an integrated form as part of the medical device (200).

[0635] FIG. 39 shows a flowchart of a method for processing lymphatic vessel images according to one embodiment of the present invention.

[0636] Referring to FIGS. 38 and 39, a method for processing lymphatic vessel images can be performed by an electronic device (100).

[0637] In step S1910, the electronic device (100) can obtain a lymphatic vessel image of a lymphatic vessel through which a composition for imaging diagnosis of lymphatic diseases passes. In one embodiment of the present invention, the composition for imaging diagnosis of lymphatic diseases may contain a contrast agent comprising water-soluble polysaccharide cross-linked colloidal particles and metal ions or iron oxide particles.

[0638] In step S1920, the electronic device (100) can obtain lymphatic vessel information corresponding to the lymphatic vessel based on a lymphatic vessel image. In one embodiment of the present invention, the lymphatic vessel information may include parameters for at least one of lymphatic flow, structural abnormalities, and specific abnormalities of the lymphatic vessel. Parameters for lymphatic flow may include at least one of the speed of movement of a composition within the lymphatic vessel, the direction of movement, and the volume of movement per unit time. Parameters for structural abnormalities may include at least one of the diameter of the lymphatic vessel, tortuosity, the number of branches, and the number of collateral vessels.

[0639] In one embodiment of the present invention, lymphatic information may indicate the severity of lymphedema. The severity of lymphedema may include a zero stage corresponding to a normal state, and first to n stages (n ≥ 1) corresponding to an abnormal state.

[0640] In one embodiment of the present invention, the method for processing lymphatic vessel images may further include the step of generating a report containing diagnostic information regarding lymphatic system diseases based on lymphatic vessel information.

[0641] FIG. 40 illustrates the detailed steps of step S1920 of FIG. 39.

[0642] Referring to FIG. 40 together with FIG. 38 and 39, the step (S1920) of obtaining lymphatic vessel information corresponding to a lymphatic vessel based on a lymphatic vessel image can be performed by an electronic device (100).

[0643] In step S2010, the electronic device (100) can obtain a state value corresponding to the state of the lymphatic vessel by using at least one artificial intelligence model that takes a lymphatic vessel image as input. The electronic device (100) can extract visual features such as morphological features of the lymphatic vessel, a contrast agent distribution pattern, and signal intensity from the lymphatic vessel image. The artificial intelligence model can analyze the extracted features. Based on the analysis results, the artificial intelligence model can calculate a quantitative state value representing the current state of the lymphatic vessel.

[0644] In one embodiment of the present invention, at least one artificial intelligence model may be trained to update weights using training lymphatic vessel images and state values ​​labeled for each of the training lymphatic vessel images as input. For example, the artificial intelligence model may be trained using a plurality of lymphatic vessel image data previously read by an expert and diagnostic results for the images (e.g., normal, lymphedema stage).

[0645] In one embodiment of the present invention, an artificial intelligence model may be composed of a plurality of neural network layers. Each of the plurality of neural network layers has a plurality of weight values ​​and performs neural network operations through operations between the operation result of a previous layer and the plurality of weights. The plurality of weights possessed by the plurality of neural network layers may be optimized by the learning results of the deep neural network model. For example, the plurality of weights may be updated so that the loss value or cost value obtained from the deep neural network model during the learning process is reduced or minimized. For example, the deep neural network model may include, but is not limited to, a CNN (Convolutional Neural Network), DNN (Deep Neural Network), RNN (Recurrent Neural Network), RBM (Restricted Boltzmann Machine), DBN (Deep Belief Network), BRDNN (Bidirectional Recurrent Deep Neural Network), or Deep Q-Networks.

[0646] In one embodiment of the present invention, the state value may be a value identifying at least one of the presence or absence of lymphedema within the lymphatic vessel, quantitative parameters of lymphedema, and the severity of lymphedema. For example, quantitative parameters of lymphedema may include the diameter of the lymphatic vessel, the contrast agent diffusion area, the depth of skin reflux, the curvature of the lymphatic vessel, etc.

[0647] In step S2020, the electronic device (100) can identify the state of the lymphatic vessel as one of a plurality of predefined states based on the state value. The electronic device (100) can compare the state value obtained in step S2010 with a predefined classification criterion. The electronic device (100) can finally determine the state of the lymphatic vessel based on the comparison result. The plurality of predefined states can be set as a normal state (Stage 0), lymphedema Stage 1, lymphedema Stage 2, lymphedema Stage 3, lymphedema Stage 4, etc., according to the lymphedema staging classification criteria.

[0648] In one embodiment of the present invention, the state value may include a probability value for each of a plurality of previously defined states. The step of identifying the state of a lymphatic vessel may include identifying the state corresponding to the highest probability value among the probability values ​​as the state of the lymphatic vessel. For example, an artificial intelligence model may output a state value with a probability of 'normal' of 5%, a probability of 'first stage' of 85%, and a probability of 'second stage' of 10% for an input lymphatic vessel image. The electronic device (100) may identify the 'first stage' corresponding to the highest probability value of 85% as the final state of the lymphatic vessel.

[0649] Referring to Tables 9 and 10 described above, a method for processing lymphatic vessel images according to one embodiment of the present invention may include criteria for quantitatively evaluating the severity of lymphedema. An electronic device (100) may identify the condition of a lymphatic vessel as one of a plurality of predefined conditions based on the lymphatic vessel image. Tables 9 and 10 may exemplify a plurality of conditions that define the stage of lymphedema. The severity of lymphedema may include a stage 0 corresponding to a normal state and stages 1 through 5 corresponding to an abnormal state.

[0650] Referring again to Table 9, it defines the stages of lymphedema based on the pattern of lymphatic reflux into the skin. Stage 0 may be a normal state where there is no dermal reflux and only a linear pattern is observed. Stage 1 may be a state where a splash pattern appears around the popliteal and inguinal lymph nodes. Stage 2 may be a state where the splash pattern extends to the entire limb and a stardust pattern appears around the distal end. Stage 3 may be a state where the stardust pattern extends proximally. Stage 4 may be a state where a diffuse pattern is present along with a stardust pattern. Stage 5 may be the most severe state where no lymphatic drainage is observed.

[0651] Referring again to Table 10, Table 10 can define the stages of lymphedema based on structural changes in the lymphatic vessels. Stage 0 may be a normal state where there is no dermal reflux and collecting lymphatic vessels (LVs) are identified. Stage 1 may be a state where a collateral pathway is observed near the existing lymphatic vessels. Stage 2 may be a state where the collateral pathway is dilated. Stage 3 may be a state where the collateral pathway is dilated and dilatation of the lymphatic vessels (LVs) occurs together. Stage 4 may be a state where a diffusing area appears around the lymphatic vessels. Stage 5 may be a state where no lymphatic drainage is observed.

[0652] FIG. 41 illustrates the detailed steps of step S2010 of FIG. 40.

[0653] With reference to FIG. 41, along with FIG. 38 to 40, the detailed steps of step S2010 can be performed by an electronic device (100).

[0654] In step S2110, the electronic device (100) may obtain a first state value corresponding to a skin reflux pattern of lymph fluid flowing in the lymphatic vessel by using a first artificial intelligence model that takes a lymphatic vessel image as input. For example, the first artificial intelligence model may identify a linear pattern indicating normal lymph flow, a splash pattern where the lymphatic system bypasses a secondary path, a stardust pattern where leakage occurs in the lymphatic v...

Claims

1. A non-immunostimulating polysaccharide cross-linked colloidal particle composition that is dispersible in an aqueous medium and satisfies pattern recognition receptor (PRR) and / or B cell receptor (BCR) non-operational properties, wherein (a) In an aqueous solvent, the hydration diameter is 2 to 10 nm (preferably 2 to 8 nm), and (b) The ζ-potential is -20 mV to 0 mV, and (c) having a compact spherical three-dimensional network structure that is intramolecularly and / or intermolecularly crosslinked by selectively modifying the -OH functional groups of the monosaccharide repeating units of the crosslinking target polysaccharide with an epoxide-based first crosslinking agent, (d) At least 60% of the total number of the monosaccharides (preferably 70% or more, more preferably 90% or more, even more preferably 95% or more) are modified by a crosslinking agent, and (e) post-modifying some or all of the crosslinking agent-derived amine functional groups exposed on the surface with -COOH to achieve the ζ-potential of (b), and (f) Optionally, iron ions (Fe) coordinately bonded to amine functional groups or -COOH functional groups derived from the crosslinking agent exposed on the surface of polysaccharide crosslinked colloidal particles 2+ / 3+ ), gadolinium ion (Gd 3+ ) or manganese ions (Mn 2+ A particle composition characterized by including ) or iron oxide nanoparticles. Here, PRR / BCR non-operativity is defined in human PBMC or monocyte / macrophage-based in vitro tests as not showing an increase in TNF-α, IL-6, IL-1β secretion relative to the vehicle exceeding a pre-defined threshold, and / or signal amplification in TLR reporters being below the threshold, or C3a / C5a complement activation being within the acceptable range.

2. A particle composition according to claim 1, designed such that, under conditions without the addition of a targeting ligand, the binding affinity with a PRR or BCR is lower than a predetermined Kd upper limit (e.g., above the μM range) so that significant endocytosis does not occur.

3. A particle composition according to claim 1 in which the anti-drug antibody (ADA) is negative even after repeated administration, or clinical signs of hypersensitivity, allergy, or edema are below the pre-defined score criteria.

4. In paragraph 1, r2 / r1 ≤ 5 and metal reference r1 ≥ 1.0 Mm at 37 ℃ so that T1 positive contrast is dominant through metal coordination. -1 ·s -1 Particle composition designed to satisfy 5. A particle composition according to claim 1, designed to inhibit venous capillary leakage and to preferentially follow the physiological drainage pathway of interstitial → lymphatic capillary → peripheral lymphatic vessel → lymph node.

6. A particle composition designed to avoid venous contamination in MR lymphangiography, in any one of claims 1 to 5.

7. An MRI T1 semi-quantitative imaging method using a particle composition of any one of claims 1 to 6, wherein (a) Administer the above particle composition intradermally, subcutaneously, or via other peripheral routes, and (b) From T1-weighted images or T1 map-based data at different time points in the same anatomical region, a formulation-normalized enhancement metric (F-NEM) is calculated using formulation metadata (dosage, infusion-scan interval (τ), formulation-specific r1 reference value, etc.), and (c) Calculate one or more of the time-signal indices (PSE, WIR, TTP, AUC) and / or shape indices (ES, NF, LV) with respect to the above F-NEM as an axis, and (d) Method of providing calculated indicators as standardized numerical reports.

8. A method according to claim 7, which automatically excludes frames in which (i) vein contamination is detected and (ii) extravasation is detected in the time-resolved image, and applies quality control (QC) rules including whether a predefined SNR hurdle is satisfied, flip angle / B1 correction suitability, and / or T2* influence interval masking.

9. A method according to claim 7, for calculating a Tissue Damage Index (TDI) defined as a weighted combination of formulation-normalized enhancement index (F-NEM), edge sharpness (ES), necrotic fraction (NF), and / or lesion volume (LV), and for marking a threshold increase in TDI relative to reference (e.g., ≥20%) as a progression flag.

10. A method according to claim 7 for continuously visualizing superficial and / or deep lymphatic networks in MR lymphangiography (MRL) and quantifying nodal filling delay, reflux presence / grade, bypass asymmetry, lesion activity (active / inactive), and / or fibrosis co-occurrence based on the formulation-normalized enhancement index (F-NEM).

11. Receive input data (video time series, formulation metadata, QC logs), A semi-quantitative processing system configured to (i) calculate the formulation-normalized enhancement index (F-NEM), (ii) calculate the percentage signal enhancement (PSE), wash-in rate (WIR), time-to-peak (TTP), area under the curve (AUC), edge sharpness (ES), necrotic fraction (NF), and / or lesion volume (LV), (iii) calculate the tissue damage index (TDI) (weighted combination), (iv) determine QC pass / mask status, (v) generate a DICOM-SR / HL7-FHIR structured report, and (vi) record and store audit logs including model, version, hyperparameter, random seed, processing time, etc.

12. A system according to paragraph 11, wherein the output report follows a standardized numerical format for comparison between equipment and institutions and includes labels (nodal filling delay, reflux presence / grade, bypass asymmetry, active / inactive, and / or fibrosis co-occurrence, etc.) and outcome indicators (F-NEM, PSE, WIR, TTP, AUC, ES, NF, LV, TDI, progression flag).

13. A system in which control logic is implemented to perform performance evaluation and reporting limited to formulation-normalized enhancement index (F-NEM) normalization and QC-passed data as a prerequisite for input stability in Clause 11.

14. A system according to claim 11 that includes a QC module that automatically removes frames in which vein contamination and extravasation are detected in time-resolved images, and enforces pre-defined SNR hurdle checks, flip angle / B1 correction suitability checks, and / or T2* influence interval masking in the form of a checklist.

15. A computer-readable recording medium storing an instruction that directs a method step of any one of claims 7 through 10 to be performed when executed by a processor.

16. Use of the particle composition of any one of claims 1 to 6 for MR lymphangiography to visualize superficial and / or deep lymphatic networks while avoiding venous contamination.

17. An imaging kit comprising a particle composition of any one of claims 1 to 6, and further comprising instructions for use (administration site, dosage, infusion-scan interval), a formulation metadata record sheet, and a standard report form for calculating the formulation-normalized enhancement index (F-NEM) and / or tissue damage index (TDI).

18. An artificial intelligence (AI) learning method according to any one of claims 7 to 10, wherein the method performs superficial and / or deep lymph network segmentation, nodal filling delay, reflux presence / grade, bypass asymmetry, active / inactive lesion activity, and / or fibrosis co-occurrence labeling on a learning dataset, and trains a classification, segmentation, and time-series prediction model using a formulation-normalized enhancement index (F-NEM) and a group of derived indices as input.

19. An artificial intelligence (AI) learning method according to claim 18, which predicts the rate of change of TDI during the inference stage and automatically marks a progress flag when a threshold is exceeded and exports it to DICOM-SR / HL7-FHIR.

20. Use of a particle composition according to any one of claims 1 to 6 for tracking the rate of change of vascular permeability, interstitial retention, and lymphatic drainage impairment (PSE / WIR / TTP / AUC) and evaluating morphological changes (ES / NF / LV) in neoplastic lesions.

21. A particle composition of any one of claims 1 to 6 for use in evaluating activity (acute / inactive), extent (focal / multifocal), and / or tissue damage based on baseline-referenced change and / or side-to-side comparison (ipsilateral vs. contralateral) in inflammatory and fibrotic diseases.

22. Use of a particle composition of any one of claims 1 to 6 for the continuous visualization of superficial / deep networks and the formulation-normalized enhancement index (F-NEM)-based quantification of functional indicators in lymphatic circulation disorders.

Images