System and method for reduced dosage of gadolinium-based contrast agents for MRI examinations
A physics-informed framework for MRI enhances diagnostic quality by synthesizing full-dose-equivalent images from reduced-dose acquisitions, addressing dose reduction challenges and maintaining image quality through tissue relaxation rate modeling and neural network estimation.
Patent Information
- Application Number
- PCT/US2026/027650
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-09
- Filing Date
- 2026-05-11
- Publication Date
- 2026-11-12
AI Technical Summary
Existing methods for reducing gadolinium-based contrast agent (GBCA) doses in MRI examinations suffer from loss of diagnostic quality, lack of generalizability, and unreliable signal-to-noise ratio, particularly at ultra-low doses, without a rigorous statistical framework for determining the minimum effective dose.
A physics-informed computational framework that exploits the dependence of tissue relaxation rates on gadolinium concentration to synthesize full-dose-equivalent or super-contrast T1 and T2 maps and contrast-weighted images from pre-contrast and substantially reduced-dose acquisitions, using multiparametric quantitative MRI and a Joint Parameter Estimation Network to estimate baseline tissue properties and relaxivity maps.
Enables a 5-fold reduction in administered GBCA dose while maintaining diagnostic visibility, providing quantitatively accurate images and flexible contrast synthesis, with improved generalizability and interpretability by incorporating MR physics.
Smart Images

Figure US2026027650_12112026_PF_FP_ABST
Abstract
Description
Atty. Dkt. No. 125141.05024. MGH2025-020SYSTEM AND METHOD FOR REDUCED DOSAGE OF GADOLINIUM-BASED CONTRAST AGENTS FOR MRI EXAMINATIONS CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is based on, claims priority to, and incorporates herein by reference in its entirety, US Provisional Application Serial No.63 / 803,462, filed May 9, 2025.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH
[0002] N / ABACKGROUND
[0003] The present disclosure relates to systems and methods for medical imaging. More particularly, the disclosure relates to systems and methods for magnetic resonance imaging [MRI] using gadolinium-based contrast agents (GBCAs).
[0004] Modern clinical medicine relies on a variety of medical imaging modalities and techniques. One important imaging modality is magnetic resonance imaging or " MR” imaging. When a substance, such as human tissue, is subjected to a uniform magnetic field (polarizing field Bo), the individual magnetic moments of the nuclei in the tissue attempt to align with this polarizing field, but precess about it in random order at their characteristic Larmor frequency. If the substance, or tissue, is subjected to a magnetic field (excitation field Bi) that is in the x-y plane and that is near the Larmor frequency, the net aligned longitudinal moment, Mz, may be rotated, or "tipped”, into the x-y plane to produce a net transverse magnetic moment, Mxy. A signal is emitted by the excited nuclei or "spins”, after the excitation signal Bi is terminated, and this signal may be received and processed to form an image.
[0005] When utilizing these " MR” signals to produce images, magnetic field gradients (Gx, Gy, Gz) are employed. Typically, the region to be imaged is scanned by a sequence of measurement cycles in which these gradients vary according to the particular localization method being used. The resulting set of received MR signals are digitized and processed to reconstruct the image using one of many well-known reconstruction techniques.
[0006] Contrast-enhanced MRI is a central tool in modern diagnostic medicine, with gadolinium-based contrast agents (GBCAs) administered in over 30 million procedures -1- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020worldwide each year and more than 300 million total doses since their introduction in 1987. GBCAs are valuable for the detection, grading, and therapeutic monitoring of a variety of clinical applications, from brain tumors and multiple sclerosis, to the characterization of myocardial infarction and cardiomyopathy, staging of breast and liver malignancies, and evaluation of vascular, infectious, and inflammatory disorders. A common clinical dose is approximately 0.1 mmol / kg of a GBCA administered prior to scanning.
[0007] After administration, the contrast agent distributes throughout tissues and shortens the longitudinal (Tl) and transverse (T2) relaxation times of nearby water protons, producing high signal intensity in conventionally weighted images. Despite their clinical indispensability, GBCAs have undergone fundamental safety reassessment following the discovery of gadolinium deposition in the brain, bone, and skin even in patients with intact renal function. Current clinical guidelines, therefore, advocate use of the minimum effective dose to mitigate the unknown risks of cumulative systemic exposure. Reducing the administered dose, however, diminishes contrast in conventional images and lowers the visibility of pathology.
[0008] Some prior approaches to mitigate diagnostic quality loss at reduced doses have leveraged deep learning to synthesize "full-dose” image appearances from low-dose or non-contrast acquisitions, and more recently to perform "contrast boosting" of standarddose images to mimic double-dose protocols. While such methods can improve aesthetic metrics such as contrast-to-noise ratio (CNR), they remain fundamentally qualitative image- to-image translation tools that operate primarily on conventional weighted images. Because they operate as fully supervised black-box models without explicit MR physics, they rely heavily on training data distributions, limiting generalizability and interpretability. Furthermore, as dose reduction progresses toward ultra-low regimes, signal-to-noise ratio [SNR reaches a floor where estimation becomes unreliable, and the field has lacked a rigorous statistical framework for determining the minimum contrast concentration required for reliable parametric mapping.
[0009] There is therefore a need for a contrast-enhanced MRI methodology that is grounded in MR physics and pharmacokinetics, that can identify a principled lower bound on dose, and that can synthesize quantitatively accurate full-dose-equivalent or supercontrast images from low-dose acquisitions while preserving generalizability and -2- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020interpretabilitySUMMARY
[0010] The present disclosure overcomes the aforementioned drawbacks by providing systems and methods for physics-informed computational frameworks that can exploit the dependence of tissue relaxation rates on gadolinium concentration to synthesize full-dose-equivalent or "super-contrast” T1 and T2 maps, and contrast-weighted images derived therefrom, from a pre-contrast acquisition and a substantially reduced-dose GBCA acquisition.
[0011] In accordance with one aspect of the disclosure, a method is provided for generating a full-dose-equivalent contrast-enhanced magnetic resonance image of a subject. The method includes receiving pre-contrast MR data acquired from the subject using a magnetic resonance imaging (MRF) scanner configured to perform a pulse sequence selected for a contrast-enhanced imaging study and, following administration of a dose of a contrast agent that is less than a standard clinical dose for the contrast-enhanced imaging study, receiving low-dose MR data acquired with the MRI scanner and the pulse sequence selected for the contrast-enhanced imaging study. The method also includes computing, using a processor, pre-contrast and low-dose (Tl) or transverse (T2) relaxation maps from the precontrast MR data and the low-dose MR data, estimating, using the processor, a relaxivity using a relaxivity model relating relaxation rate to gadolinium concentration, and generating, using the estimated relaxivity, a full-dose-equivalent T1 map, T2 map, or contrast-weighted image of the subject.
[0012] In accordance with another aspect of the disclosure, a magnetic resonance imaging [MRI] system is provided that includes a magnet system configured to generate a polarizing magnetic field about at least a portion of a subject, a plurality of gradient coils configured to apply magnetic field gradients to the polarizing magnetic field, and a radio frequency RF] system configured to apply RF excitation fields to the subject and acquire MR image data. The system also includes a computer system programmed to control the plurality of gradient coils and the RF system to perform a pulse sequence selected for a contrast-enhanced imaging study to acquire pre-contrast MR data from the subject. Following administration of a dose of a contrast agent that is less than a standard clinical -3- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020dose for the contrast-enhanced imaging study, the processor is configured to control the plurality of gradient coils and the RF system to perform the pulse sequence selected for the contrast-enhanced imaging study to acquire low-dose MR data from the subject. The processor is also configured to determine a pre-contrast and low-dose (Tl) or transverse (T2) relaxation map from the pre-contrast MR data and the low-dose MR data and generate a full-dose-equivalent T1 map, T2 map, or contrast-weighted image of the subject using a relaxivity model relating relaxation rate to gadolinium concentration.
[0013] In accordance with yet another aspect of the disclosure, a non-transitory computer-readable storage medium is provided that has instructions stored thereon. When the instructions are executed by a processor, the processor is caused to acquire pre-contrast magnetic resonance (MR) data of the subject and acquire low-dose MR data of the subject after administering a dose of a contrast agent to the subject that is less than a prescribed contrast dose associated with a contrast-enhanced imaging protocol used to acquire the low- dose MR data. The processor is also caused to provide the pre-contrast MR data and the low- dose MR data to a trained neural network and generate, using the trained neural network, a full-dose-equivalent image or quantitative map of the subject based on the pre-contrast MR data and the low-dose MR data reflecting the prescribed contrast dose.BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Fig. 1 is a block diagram of an exemplary magnetic resonance imaging (MRI) system configured in accordance with the present disclosure.
[0015] Fig. 2 is schematic illustration providing an overview of one non-limiting example of a framework in accordance with the present disclosure, in which three QALAS sequences are acquired (pre-contrast, 20% dose, and full dose for validation), T1 and T2 maps are computed via dictionary matching, and a physical model exploiting the linear relationship between Rl, R2 and gadolinium dose, regularized by a denoising prior, predicts full-dose maps that are validated against the full-dose acquisition.
[0016] Fig. 3A is a schematic illustration of a physics-informed framework in accordance with the present disclosure.
[0017] Fig. 3B is a flow chart setting forth some non-limiting steps for a process in accordance with the present disclosure for use in a clinical setting..4.QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020
[0018] Fig.4A is a set of imaging showing a phantom containing six vials with varying gadolinium concentrations (4, 2, 1, 0.5, 0.25, and 0 mmol) scanned with a 3D-QALAS sequence (1 mm isotropic resolution, TR = 4.5 s, turbo factor = 123, scan time = 7 min 20 s, FOV - 238 x 240 x 192 mm).
[0019] Fig.4B is a set of graphs crated from the data acquired from the phantoms of Fig. 3A and showing T1 and T2 verified against multi-TE and multi-IR acquisitions, confirming the linear relationship between R and gadolinium concentration.
[0020] Fig. 5A is a schematic of a simulation pipeline in accordance with the present disclsoure.
[0021] Fig. 5B is a set of predicted T1 maps across dose levels from 10% to 50%.
[0022] Fig. 5C is a set of predicted T2 maps across dose levels from 10% to 50%.
[0023] Fig. 6A is a set of predicted high-dose T1 maps in a healthy volunteer in axial, coronal, and sagittal planes, with columns for 0% Gd, 20% Gd, the proposed method, and 100% Gd reference, including offline reconstruction with 3D SENSE, ESPIRiT sensitivity maps, coil compression, TV-wavelet regularization, and MPPCA denoising.
[0024] Fig. 6B is a set of predicted high-dose T2 maps in a healthy volunteer in axial, coronal, and sagittal planes, with columns for 0% Gd, 20% Gd, the proposed method, and 100% Gd reference, including offline reconstruction with 3D SENSE, ESPIRiT sensitivity maps, coil compression, TV-wavelet regularization, and MPPCA denoising.
[0025] Fig. 7A is graphs showing absolute uncertainty,
[0026] Fig. 7B is a graph showing relative uncertainty
[0027] Fig. 7C is a graph showing absolute uncertainty
[0028] Fig. 7D is a graph showing relative uncertainty o(r2e^) / r2.
[0029] Fig. 8 is a set of synthetic high-dose Tl-weighted images (TI = 1000 ms, TR = 2000 ms) in a healthy volunteer in axial, coronal, and sagittal planes for 0% Gd, 20% Gd, the proposed predicted images, and 100% Gd reference.DETAILED DESCRIPTION
[0030] The present disclosure provides a physics-informed framework, and corresponding systems and methods, that exploit the linear dependence of tissue relaxation-5- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020rates on gadolinium concentration to synthesize full-dose-equivalent or "super-contrast” T1 and / or T2 maps, and contrast-weighted images derived therefrom, from a pre-contrast acquisition and a substantially reduced-dose GBCA acquisition. In one non-limiting example, reduced-dose may be approximately 20% of standard GBCA for a given clinical acquisition.
[0031] In one aspect, multiparametric quantitative MRI acquisitions (e.g., 3D-QALAS) may be used to obtain MR data at a pre-contrast (0%) state and at a low-dose state. T1 and T2 maps may be computed for each acquisition, and a linear relaxivity model may be used together with a denoising prior, formulated as a maximum a posteriori (MAP) estimation solved using, for example, alternating direction method of multipliers (ADMM) with a total variation (TV) denoiser, to predict per-voxel full-dose T1 and T2 maps. In one non-limiting example, the T1 and / or T2 may may be generated via dictionary matching. In another, nonlimiting example, the model may take a form such as R, =+ rt[Gd], i = 1,2.
[0032] In another aspect, the present disclosure can include a unified end-to-end forward modellinking microvascular pharmacokinetics (Tofts model), molecular relaxivity, and Bloch-equation-based 3D-QALAS signal encoding, mapping biophysical parameters 0 = {T ° / o, T^0, PD, IE, r1(r2, Cp, Ktrans, ve) to a five-point signal vector [Sx,..., S5]7’. From this model, a Cramer-Rao Bound (CRB) analysis or similar analysis can be computed to derive a fundamental statistical precision limit and to identify a physical information plateau at approximately 20% of the standard clinical dose (corresponding to approximately 3.33 mM plasma Gd for a 10 mmol injection in a 3 L plasma volume), establishing a principled lower bound for dose reduction.
[0033] In a further aspect, a physics-informed deep-learning architecture — a Joint Parameter Estimation Network — accepts dual-acquisition 3D-QALAS signals (pre-contrast and low-dose) and simultaneously estimates baseline tissue properties (Tl, T2, PD), systemspecific inversion efficiency (IE), and effective longitudinal and transverse relaxivity maps (rx,r2); the estimated parameters are then propagated through the forward model to synthesize images of any arbitrary pulse sequence (e.g., MPRAGE, Tlw) at any desired contrast dose, including full-dose-equivalent and super-contrast (greater-than-standard- dose) images.
[0034] As will be explained, the systems and methods provided herein may belower-6- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020a multi-fold (e.g., 5-fold) reduction of administered GBCA dose while producing images that can match or exceed the diagnostic visibility of standard-dose protocols, thereby reducing patient drug exposure and environmental burden. The generated images can be used for clinical diagnosis, treatment planning, or therapeutic monitoring of conditions including brain tumors, multiple sclerosis, myocardial infarction, and malignancies. Further, the systems and methods provided herein may facilitate interpretability and generalizability arising from explicit incorporation of MR physics rather than purely data-driven image translation. Further still, the systems and methods provided herein may provide doseflexible synthesis of desired contrast-weighted sequences from a single acquisition protocol. Further yet, the systems and methods provided herein may provide a principled, CRB-based determination of a minimum reliable dose for a given clinical application or applications.
[0035] Referring now to Fig. 1, one example of an MRI system 100 is provided that may be configured, programmed, or otherwise utilized in accordance with the present disclosure. The MRI system 100 includes an operator workstation 102, which will typically include a display 104, one or more input devices 106 (such as a keyboard and mouse or the like), and a processor 108. The processor 108 may include a commercially available programmable machine running a commercially available operating system. The operator workstation 102 provides the operator interface that enables scan prescriptions to be entered into the MRI system 100. In general, the operator workstation 102 may be coupled to multiple servers, including a pulse sequence server 110; a data acquisition server 112; a data processing server 114; and a data store server 116. The operator workstation 102 and each server 110, 112, 114, and 116 are connected to communicate with each other. For example, the servers 110, 112, 114, and 116 may be connected via a communication system 140, which may include any suitable network connection, whether wired, wireless, or a combination of both. As an example, the communication system 140 may include both proprietary or dedicated networks, as well as open networks, such as the internet.
[0036] The pulse sequence server 110 functions in response to instructions downloaded from the operator workstation 102 to operate a gradient system 118 and a radiofrequency (RF) system 120. Gradient waveforms to perform the prescribed scan are produced and applied to the gradient system 118, which excites gradient coils in an assembly 122 to produce the magnetic field gradients Gx, Gy, Gzused for position encoding magnetic -7- QB\125141.05024\102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020resonance signals. The gradient coil assembly 122 forms part of a magnet assembly 124 that includes a polarizing magnet 126 and a whole-body RF coil 128.
[0037] RF waveforms are applied by the RF system 120 to the RF coil 128, or a separate local coil (not shown in Fig. 1), in order to perform the prescribed magnetic resonance pulse sequence. Responsive magnetic resonance signals detected by the RF coil 128, or a separate local coil, are received by the RF system 120, where they are amplified, demodulated, filtered, and digitized under direction of commands produced by the pulse sequence server 110. The RF system 120 includes an RF transmitter for producing a wide variety of RF pulses used in MRI pulse sequences. The RF transmitter is responsive to the scan prescription and direction from the pulse sequence server 110 to produce RF pulses of the desired frequency, phase, and pulse amplitude waveform. The generated RF pulses may be applied to the whole-body RF coil 128 or to one or more local coils or coil arrays.
[0038] The RF system 120 also includes one or more RF receiver channels. Each RF receiver channel includes an RF preamplifier that amplifies the magnetic resonance signal received by the coil 128 to which it is connected, and a detector that detects and digitizes the I and Q quadrature components of the received magnetic resonance signal. The magnitude of the received magnetic resonance signal may, therefore, be determined at any sampled point by the square root of the sum of the squares of the I and Q components:M = √I2+ Q2
[0001] and the phase of the received magnetic resonance signal may also be determined according to the following relationship:φ = tan-1(Q / I)
[0039] The pulse sequence server 110 also optionally receives patient data from a physiological acquisition controller 130. By way of example, the physiological acquisition controller 130 may receive signals from a number of different sensors connected to the patient, such as electrocardiograph (ECG) signals from electrodes, or respiratory signals from a respiratory bellows or other respiratory monitoring device. Such signals are typically-8- QB\125141.05024\102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020used by the pulse sequence server 110 to synchronize, or "gate,” the performance of the scan with the subject’s heartbeat or respiration.
[0040] The pulse sequence server 110 also connects to a scan room interface circuit 132 that receives signals from various sensors associated with the condition of the patient and the magnet system. It is also through the scan room interface circuit 132 that a patient positioning system 134 receives commands to move the patient to desired positions during the scan.
[0041] The digitized magnetic resonance signal samples produced by the RF system 120 are received by the data acquisition server 112. The data acquisition server 112 operates in response to instructions downloaded from the operator workstation 102 to receive the real-time magnetic resonance data and provide buffer storage, such that no data are lost by data overrun. In some scans, the data acquisition server 112 does little more than pass the acquired magnetic resonance data to the data processor server 114. However, in scans that require information derived from acquired magnetic resonance data to control the further performance of the scan, the data acquisition server 112 is programmed to produce such information and convey it to the pulse sequence server 110. For example, during prescans, magnetic resonance data are acquired and used to calibrate the pulse sequence performed by the pulse sequence server 110. As another example, navigator signals may be acquired and used to adjust the operating parameters of the RF system 120 or the gradient system 118, or to control the view order in which k-space is sampled.
[0042] The data processing server 114 receives magnetic resonance data from the data acquisition server 112 and processes it in accordance with instructions downloaded from the operator workstation 102. Such processing may, for example, include one or more of the following: reconstructing two-dimensional or three-dimensional images by performing a Fourier transformation of raw k-space data; performing other image reconstruction techniques, such as iterative or backprojection reconstruction techniques; applying filters to raw k-space data or to reconstructed images; generating functional magnetic resonance images; calculating motion or flow images; and so on.
[0043] Images reconstructed by the data processing server 114 are conveyed back to the operator workstation 102. Images may be output to operator display 112 or a display 136 that is located near the magnet assembly 124 for use by attending clinician. Batch mode -9- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020images or selected real time images are stored in a host database on disc storage 138. When such images have been reconstructed and transferred to storage, the data processing server 114 notifies the data store server 116 on the operator workstation 102. The operator workstation 102 may be used by an operator to archive the images, produce films, or send the images via a network to other facilities.
[0044] The MRI system 100 may also include one or more networked workstations 142. By way of example, a networked workstation 142 may include a display 144, one or more input devices 146 (such as a keyboard and mouse or the like), and a processor 148. The networked workstation 142 may be located within the same facility as the operator workstation 102, or in a different facility, such as a different healthcare institution or clinic. The networked workstation 142 may include a mobile device, including phones or tablets.
[0045] The networked workstation 142, whether within the same facility or in a different facility as the operator workstation 102, may gain remote access to the data processing server 114 or data store server 116 via the communication system 140. Accordingly, multiple networked workstations 142 may have access to the data processing server 114 and the data store server 116. In this manner, magnetic resonance data, reconstructed images, or other data may be exchanged between the data processing server 114 or the data store server 116 and the networked workstations 142, such that the data or images may be processed remotely by a networked workstation 142. This data may be exchanged in any suitable format, such as in accordance with the transmission control protocol (TCP), the internet protocol (IP), or other known or suitable protocols.
[0046] The systems described above may be used to create an apparatus or perform methods for synthesizing full-dose or super-contrast MR images and corresponding quantitative T1 / T2 maps from a pre-contrast acquisition combined with a substantially reduced-dose acquisition of a GBCA, by exploiting the dependence of tissue relaxation rates on Gd concentration and embedding this physics into either an iterative model-based reconstruction or a physics-informed neural network. An procedural overview of one process illustrating the systems and methods of the present disclosure is provided in Fig. 2. As illustrated at 200, QALAS sequences were acquired at 0% (Input 1), 20% of dose (Input 2)), and 100% of dose (Validation). At 202, T1 maps and T2 maps were computed for each sequence using dictionary matching, with the full-dose acquisition served as a validation -10- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020reference. To predict high-dose T1 and T2 maps from low-dose acquisitions, a maximum a posteriori (MAP) estimation problem was created, at 204. The 0% and 20% Gd T1 and T2 maps were used as inputs to the MAP Estimate model, and an optional denoising prior was imposed to regularize the solution. The predicted full-dose maps were then compared against the actual full-dose maps. Thus, the three QALAS acquisitions were obtained (precontrast, low-dose at approximately 20% of standard dose, and an optional full-dose acquisition reserved for validation), T1 and T2 maps are computed for each acquisition by dictionary matching, and a physical model leveraging the linear relationship between the relaxation rates Rl, R2 and the gadolinium concentration (regularized by a denoising prior) was used to predict full-dose-equivalent T1 and T2 maps that are validated against the fulldose acquisition.
[0047] Referring now to Fig.3A, based on this concept, a physics-informed framework 300 for MRI contrast enhancement can be provided. The framework 300 can include an end- to-end forward modeling 302. In one non-limiting example, the forward modeling 302 may combine Tofts pharmacokinetics, molecular relaxivity, and 3D-QALAS signal encoding. The end-to-end forward model coupling microvascular pharmacokinetics, molecular relaxivity, and 3D-QALAS signal encoding provides robust descriptions of contrast agent dynamics. That is, the pharmacokinetic model can be linked to tissue relaxation changes through linear relaxivity relationships, and propagated through the Bloch-based signal equations of the 3D- quantification using an interleaved Look-Locker acquisition sequence with a T2 preparation pulse (3D-QALAS).
[0048] The framework 300 may also include CRB analysis of effective relaxivity defining theoretical estimation uncertainty across clinical doses 304. This unified biophysical chain can be leveraged in two complementary ways. First, a statistical precision bound, Cramer-Rao bound (CRB), can be derived to quantify the statistical limits of parametric estimation. The analysis shows that estimation variance increases rapidly at low doses, exhibiting a clear inflection point below 20% of the standard dose, which can be taken as a practical lower bound for statistically reliable mapping. Second, these governing equations can be embedded into a neural network or deep learning architecture to empower the network to function as a physics-constrained estimator rather than a black-box synthesis model.-11- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020
[0049] The framework 300 may include a joint parameter estimation network 306. The network 306 may integrate dual-acquisition 3D-QALAS signals to estimate MR parameters. The joint parameter estimation network 306 can integrate dual-acquisition 3D- QALAS signals to estimate baseline tissue properties together with effective longitudinal and transverse relaxivities. Example MR parameters may include, for example, Tl, T2, PD, IE, 6 f f effr, and r2}. Also, the framework 300 may include a synthetic contrast synthesis system 308, which may deliver reconstructed images at a range of doses, such as series at 0%, 20%, and 100%-estimated, or other desired amounts. The resulting synthetic contrast synthesis are illustrated by a reconstructed MPRAGE series at 0% dose, 20% dose, 100%-estimated dose, and 100%-real dose.
[0050] Throughout this description, "low dose” may refer to an administered GBCA dose that is less or substantially less than the standard clinical dose of, for example, approximately 0.1 mmol / kg. In some non-limiting examples, "low dose” may be approximately 90%, 80%, 70%, 60%, 50%, 40%, 30% or 20% of the standard dose.
[0051] A process 310 for using the systems and methods described herein within a clinical setting is provided in Fig. 3B. The process 310 beings with the acquisition of data from a patient or imaging subject at process block 312. For example, data acquisition may be performed using an MRI scanner and a pulse sequence selected for a contrast-enhanced imaging study to acquire a pre-contrast MR data from the subject. Also, the MRI scanner and the pulse sequence selected for the contrast-enhanced imaging study may be used to acquire low-dose MR data from the subject following administration of a dose of a contrast agent that is less than a standard clinical dose for the contrast-enhanced imaging study.
[0052] The pulse sequence selected for a contrast-enhanced imaging study may be selected to acquire multiparametric quantitative MRI data, such as using a 3D-QALAS pulse sequence or an interleaved Look-Locker acquisition with a T2-preparation pulse. The acquisition, regardless of the specific pulse sequence used, may collect multiple distinct 3D image volumes (e.g., five: SI,..., S5) within a single scan. Some non-limiting acquisition parameters, as employed when imaging the phantom illustrated in Fig. 4A to generate the validation reflected in Fig. 4B, include 1 mm isotropic resolution, TR = 4.5 s, turbo factor = 123, scan time approximately 7 min 20 s, and FOV = 238 x 240 x 192 mm.-12- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020
[0053] As described, the data acquisition 312 may be performed while the patient is in multiple sates. For example, two or more states may be utilized. In one non-limiting example, a pre-contrast (0% Gd) acquisition may be acquired followed by and a low-dose acquisition following administration of the contrast agent. In one, non-limiting example, a low-dose acquisition may be performed at approximately 20% of the standard clinical Gd dose. However, other does below the standard clinical dose may be used.
[0054] As described, for validation, a full-dose (100%) acquisition was obtained, but this is not needed in clinical practice. An IRB-approved phantom and in vivo experiments described herein used Gd concentrations spanning 0, 0.25, 0.5, 1, 2, and 4 mmol, as further depicted in Figs. 4A and 4B, which shows a phantom containing six vials of varying gadolinium concentration scanned with the 3D-QALAS sequence and verified against multi- TE and multi-IR reference acquisitions.
[0055] Referring again to Fig. 3B, at process block 314, the data is processed to generate parameter maps. For example pre-contrast and low-dose (Tl) or transverse (T2) relaxation maps from the pre-contrast MR data and the low-dose MR data may be generated. That is, for each acquisition, T1 or T2 maps can be computed from the acquired pre-contrast and low-does MR data. In one non-limiting example, dictionary matching against a precomputed Bloch-equation-derived dictionary can be used to generate the quantitative maps. Dictionary matching results may be verified against multi-TE and multi-IR reference acquisitions. In another configuration, self-supervised learning for rapid multiparameter estimation can be performed, for example, using 3D-QALAS (e.g., SSL-QALAS).
[0056] At process block 316, the process continues by estimating relaxivity using a relaxivity model relating relaxation rate to gadolinium concentration. This can be achieved using, as a non-limiting example, a model relating relaxation rates with Gd concentration. The model may be a machine learning or other model or may be algorithmic, such as:R= 1,2.
[0057] where R ° / o(l / T°%) is the pre-contrast relaxation rate, rtis the molecular relaxivity (slope), and [Gd] is the gadolinium concentration in mM. This relationship has been confirmed in phantom experiments with vials of varying Gd concentration, as illustrated in graphs of Fig. 4B, which display the monotonic changes in T1 (T2) and the-13- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020corresponding linear changes in R1 (R2) across the dose range, thereby confirming the linear dependence of relaxation rate on gadolinium concentration that underlies the estimation framework.
[0058] Referring again to Fig. 3B, at process block 318, a faull-dose-equivalent T1 map, T2 map, or contrast-weighted image of the subject is then generated using the estimated relaxivity. This generation can be achieve using a machine learning model, or an algorithmic implementation. In one non-limiting example, full-dose Tl, T2 maps (.100%) are generating from the low-dose acquisitions by formulating a maximum a posteriori (MAP) estimation problem:
[0059] where the first term is the data-fidelity term enforcing the physical model on the 0% and 20% Gd measurements, (•) is a denoising operator, and X balances data fidelity against smoothness of the predicted high-dose maps. In one non-limiting example, the MAP problem can be solved using a Plug-and-Play ADMM scheme. Once n is estimated, T^oo%can be reconstructed via the relaxivity model at full dose.
[0060] Referring now to Figs.5A-5C, the accuracy of this model-based reconstruction across a range of dose fractions was demonstrated. Fig, 5A provides a schematic depiction of the simulation pipeline in which a segmented phantom was assigned baseline T and r values, post-contrast T values were generated via the physical model, and QALAS measurements were simulated through Bloch-equation modeling with added noise. The simulation pipeline of Fig.5A was used to generate the predicted Tl and T2 maps, of Figs. 5B and 5C, respectively, across simulated dose levels ranging from 10% to 50% of the standard dose.
[0061] Referring now to Figs.6A and 6B, optional offline k-space reconstruction may include 3D SENSE with ESPIRiT sensitivity maps, coil compression, TV-wavelet regularization, and MPPCA denoising, as employed in the in vivo reconstructions. Fig. 6A shows predicted high-dose Tl maps and Fig. 6B shows predicted high-dose T2 maps in a healthy volunteer in axial, coronal, and sagittal planes, with successive columns corresponding to 0% Gd, 20% Gd, the proposed method, and the 100% Gd reference. The bottom row highlights the change relative to no contrast in order to emphasize Gd-affected -14- QB\125141.05024\102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020regions.
[0062] As described above, a machine learning or neural network, or other artificial intelligence model, may be used to implement a physics-informed process. In one nonlimiting example, a joint parameter estimation network may be configured to receive as input dual-acquisition signals (pre-contrast and low-dose) and to simultaneously determine or estimate (i) baseline tissue properties (e.g., T^%, T and / or proton density PD); (ii) a system-specific inversion efficiency (IE); and (iii) effective longitudinal and transverse relaxivity mapsand r2e^). The network can be constrained by an end-to-end forward model f(0) mapping a biophysical parameter vector 0, as will be described.
[0063] That is, the biophysical parameter vector 0 = {T%, T^0, PD, IE, n, rz, CP, Ktrans, ve} can be mapped to the signal vector S = [SI,..., S5]T, for example, derived from solving the Bloch equations, such as under the sequence’s T2-preparation, inversion, and Look-Locker readout pulses. That is, tissue concentration Ctft} is described by the Tofts pharmacokinetic model:
[0064] where Ktransis the volume transfer constant and veis the extravascular extracellular space (EES) volume fraction. For purposes of signal formation, Ct(t] can be treated as quasi-static during a stable post-contrast interval. The effective relaxivity can be defined as — R%+ rtCt, i G 1,2,, characterizing system sensitivity per unit administered dose.
[0065] Because the network functions as a physics-constrained estimator rather than a black-box image-to-image translator, the estimated parameters can be propagated through the forward model to synthesize images of any arbitrary pulse sequence (e.g., MPRAGE, Tl- weighted) at any arbitrary contrast dose, including doses exceeding the administered low dose, the standard dose, or even greater "super-contrast” levels. This dose-flexible synthetic contrast synthesis is illustrated in Fig.2 via the reconstructed MPRAGE series 206 including a pre-contrast (0%) image, a low-dose (20%) image, a 100%-estimated synthetic full-dose enhancement generated by propagating the network-estimated parameters through the forward model, and the actual 100%-real full-dose ground truth for comparison.-15- QB\125141.05024\102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020
[0066] The systems and methods provided herein can include computing a Cramer- Rao Bound (CRB) on the variance of any unbiased estimator of the biophysical parameters, for example, derived from the forward model f(0) and the noise characteristics of the acquisition. CRB curves can be evaluated across a physiologically relevant range of plasma Gd concentrations and tissue classes (white matter, gray matter, blood) under varying BBB conditions (intact, moderate, and high leakage), assuming an exemplary standard clinical dose corresponding to 3.33 mM (10 mmol injected into ~3 L of plasma).
[0067] This analysis is illustrated in Figs. 7A-7D, in which Figs. 7A and 7B, respectively, depict the absolute uncertainty cr(ry^) and the relative uncertaintyof the effective longitudinal relaxivity, and Figs. 7C and 7D, respectively, depict the absolute uncertainty cr(rf^) and the relative uncertaintyof the effective transverse relaxivity, with separate curves for each combination of tissue class and BBB condition, and a black-dashed vertical line at 3.33 mM marking the 20%-of-standard-dose threshold at which the CRB transitions into a stable high-precision plateau. In this non-limiting example, across all conditions, the CRB exhibited a rapid initial decay transitioning into a stable plateau at approximately 20% of the standard dose, identifying a principled minimum reliable dose. In this way, a system can be provided that selects the administered low dose at or above the CRB-derived inflection point.
[0068] As described, after estimation tissue or MR parameters (such as T1 and T2 maps and PD or IE), synthetic contrast-weighted images are generated. This can be done by inserting the parameters into the analytical signal equation of a target pulse sequence at a target effective dose or using a trained model. In one example, synthetic Tl-weighted images can be generated using TI = 1000 ms and TR = 2000 ms, as depicted in Fig. 8, which shows synthetic high-dose Tl-weighted images in a healthy volunteer in axial, coronal, and sagittal planes, with successive columns corresponding to 0% Gd, 20% Gd, the proposed predicted images, and the 100% Gd reference, and a bottom row highlighting the change relative to no contrast to emphasize Gd-affected regions; the predicted images closely replicate the fulldose contrast appearance while utilizing only a low-dose input.
[0069] As described above, systems, methods, and computer-readable storage mediums are provided that can receive a multiparametric MR data acquired pre- and post--16- QB\125141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020administration of a low dose of GBCA, generate tissue maps from the MR data, and execute a model-based reconstruction or physics-informed neural network to estimate full-dose- equivalent maps and / or biophysical parameters and synthesize contrast- weighted images at one or more target effective doses. Although 3D-QALAS may be used, other multiparametric quantitative MRI sequences may be used. Tissue or MR parameters may include Tl, T2, and / or PD. Thus, maps may include Tl, T2, and myelin water maps. Parameter estimation may use dictionary matching, self-supervised learning (e.g., SSL- QALAS], or other physics-constrained estimators. The denoising prior may be a TV denoiser or any equivalent regularizer suitable for Plug-and-Play ADMM. The administered low dose may range from approximately 10% to approximately 90% of the standard dose, with approximately 20% being one selected dose based on the CRB plateau.
[0070] As used herein, the term "full-dose-equivalent” can refer to an image or map having contrast characteristics comparable to those that would be obtained from a standard clinical dose of contrast agent, regardless of the actual dose administered. The term "contrast-weighted image” can refer to any MR image in which signal intensity is influenced by the presence of a contrast agent, including Tl-weighted, T2-weighted, and other weighted image types.
[0071] As used herein, the term "full-dose-equivalent” can refer to an image or map having contrast characteristics comparable to those that would be obtained from a standard clinical dose of contrast agent, regardless of the actual dose administered. The term "contrast-weighted image” can refer to any MR image in which signal intensity is influenced by the presence of a contrast agent, including Tl-weighted, T2-weighted, and other weighted image types.
[0072] In some implementations, devices or systems disclosed herein can be utilized or installed using methods embodying aspects of the invention. Correspondingly, description herein of particular features or capabilities of a device or system is generally intended to include disclosure of a method of using such features for intended purposes and of implementing such capabilities. Similarly, express discussion of any method of using a particular device or system, unless otherwise indicated or limited, is intended to inherently include disclosure, as embodiments of the invention, of the utilized features and implemented capabilities of such device or system.-17- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020
[0073] As used herein, unless otherwise limited or defined, "or" indicates a nonexclusive list of components or operations that can be present in any variety of combinations, rather than an exclusive list of components that can be present only as alternatives to each other. For example, a list of " A, B, or C" indicates options of: A; B; C; A and B; A and C; B and C; and A, B, and C. Correspondingly, the term "or" as used herein is intended to indicate exclusive alternatives only when preceded by terms of exclusivity, such as "only one of," or "exactly one of." For example, a list of "only one of A, B, or C" indicates options of: A, but not B and C; B, but not A and C; and C, but not A and B. In contrast, a list preceded by "one or more" (and variations thereon) and including "or" to separate listed elements indicates options of one or more of any or all of the listed elements. For example, the phrases "one or more of A, B, or C" and "at least one of A, B, or C" indicate options of: one or more A; one or more B; one or more C; one or more A and one or more B; one or more B and one or more C; one or more A and one or more C; and one or more A, one or more B, and one or more C. Similarly, a list preceded by “a plurality of’ (and variations thereon) and including “or” to separate listed elements indicates options of one or more of each of multiple of the listed elements. For example, the phrases “a plurality of A, B, or C” and "two or more of A, B, or C” indicate options of: one or more A and one or more B; one or more B and one or more C; one or more A and one or more C; and one or more A, one or more B, and one or more C.
[0074] The present invention has been described in terms of one or more preferred embodiments, and it should be appreciated that many equivalents, alternatives, variations, and modifications, aside from those expressly stated, are possible and within the scope of the invention.-18- QBM25141.050241102311882 1
Claims
1. Atty. Dkt. No. 125141.05024. MGH2025-020CLAIMS1. A method for generating a full-dose-equivalent contrast-enhanced magnetic resonance image of a subject, the method comprising:(a) receiving pre-contrast MR data acquired from the subject using a magnetic resonance imaging (MRI) scanner configured to perform a pulse sequence selected for a contrast-enhanced imaging study;(b) following administration of a dose of a contrast agent that is less than a standard clinical dose for the contrast-enhanced imaging study, receiving low-dose MR data acquired with the MRI scanner and the pulse sequence selected for the contrast-enhanced imaging study;(c) computing, using a processor, pre-contrast and low-dose (Tl) or transverse (T2) relaxation maps from the pre-contrast MR data and the low-dose MR data;(d) estimating, using the processor, a relaxivity using a relaxivity model relating relaxation rate to gadolinium concentration; and(e) generating, using the estimated relaxivity, a full-dose-equivalent T1 map, T2 map, or contrast-weighted image of the subject.
2. The method of claim 1, wherein the pulse sequence selected for the contrast- enhanced imaging study comprises a 3D-QALAS pulse sequence.
3. The method of claim 1, wherein the dose of the contrast agent is approximately 20% of the standard clinical dose.
4. The method of claim 1, wherein computing the pre-contrast and low-dose relaxation maps comprises dictionary matching against a pre-computed Bloch-equation- derived dictionary.-19- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-0205. The method of claim 1, wherein the relaxivity model comprises one of a linear relationship between relaxation rate and gadolinium concentration and a trained machine learning model.
6. The method of claim 1, wherein generating the full-dose-equivalent T1 map, T2 map, or contrast-weighted image comprises formulating a maximum a posteriori (MAP) estimation problem.
7. The method of claim 6, wherein the MAP estimation problem is solved using an alternating direction method of multipliers (ADMM) with a denoising prior.
8. The method of claim 1, wherein estimating the relaxivity comprises using a joint parameter estimation network configured to receive dual-acquisition signals and simultaneously estimate baseline tissue properties and effective relaxivity maps.
9. The method of claim 8, wherein the joint parameter estimation network is constrained by an end-to-end forward model combining microvascular pharmacokinetics, molecular relaxivity, and MRI signal encoding.
10. The method of claim 1, further comprising selecting the dose of the contrast agent based on a analysis process identifying a minimum reliable dose.
11. The method of claim 1, further comprising generating a super-contrast image corresponding to a dose exceeding the standard clinical dose.
12. A magnetic resonance imaging (MRI) system comprising:a magnet system configured to generate a polarizing magnetic field about at least a portion of a subject;a plurality of gradient coils configured to apply magnetic field gradients to the polarizing magnetic field;-20- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-020a radio frequency (RF) system configured to apply RF excitation fields to the subject and acquire MR image data; anda computer system programmed to:control the plurality of gradient coils and the RF system to perform a pulse sequence selected for a contrast-enhanced imaging study to acquire pre-contrast MR data from the subject;following administration of a dose of a contrast agent that is less than a standard clinical dose for the contrast-enhanced imaging study, control the plurality of gradient coils and the RF system to perform the pulse sequence selected for the contrast-enhanced imaging study to acquire low-dose MR data from the subject; determine a pre-contrast and low-dose (Tl) or transverse (T2) relaxation map from the pre-contrast MR data and the low-dose MR data;generate a full-dose-equivalent T1 map, T2 map, or contrast-weighted image of the subject using a relaxivity model relating relaxation rate to gadolinium concentration.
13. The MRI system of claim 12, wherein the pulse sequence selected for the contrast-enhanced imaging study comprises a 3D-QALAS pulse sequence.
14. The MRI system of claim 12, wherein the dose of the contrast agent is approximately 20% of the standard clinical dose.
15. The MRI system of claim 12, wherein the computer system is programmed to determine the pre-contrast and low-dose relaxation map using dictionary matching against a pre-computed Bloch-equation-derived dictionary.
16. The MRI system of claim 12, wherein the relaxivity model comprises a linear relationship between relaxation rate and gadolinium concentration.
17. The MRI system of claim 12, wherein the computer system is programmed to generate the full-dose-equivalent T1 map, T2 map, or contrast-weighted image by formulating a maximum a posteriori (MAP) estimation problem.-21- QBM25141.050241102311882 1Atty. Dkt. No. 125141.05024. MGH2025-02018. The MRI system of claim 12, wherein the computer system is programmed to use a joint parameter estimation network configured to receive dual-acquisition signals and simultaneously estimate baseline tissue properties and effective relaxivity maps.
19. A non-transitory computer-readable storage medium having stored thereon instructions that, when executed by a processor, cause the processor to:acquire pre-contrast magnetic resonance [MR] data of the subject;acquire low-dose MR data of the subject after administering a dose of a contrast agent to the subject that is less than a prescribed contrast dose associated with a contrast- enhanced imaging protocol used to acquire the low-dose MR data;provide the pre-contrast MR data and the low-dose MR data to a trained neural network; andgenerate, using the trained neural network, a full-dose-equivalent image or quantitative map of the subject based on the pre-contrast MR data and the low-dose MR data reflecting the prescribed contrast dose.
20. The storage medium of claim 19, wherein the trained neural network comprises a physics-informed neural network constrained by an end-to-end forward model.
21. The storage medium of claim 20, wherein the end-to-end forward model combines microvascular pharmacokinetics, molecular relaxivity, and MRI signal encoding.
22. The storage medium of claim 19, wherein the trained neural network comprises a joint parameter estimation network configured to simultaneously estimate baseline tissue properties and effective relaxivity maps.
23. The storage medium of claim 19, wherein generating the full-dose-equivalent image or quantitative map comprises propagating parameters estimated by the trained neural network through a forward model.-22- QBM25141.050241102311882 1
