Multi-modal image-based method and system for dynamic monitoring of nanomaterial urinary repair effect

Abstract

This invention belongs to the field of electrical data processing, specifically relating to a method and system for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging. It aims to address the problems of difficulty in deep fusion of multimodal image data, inability to quantitatively map the microscopic repair process, and lack of dynamic prediction capabilities. The method includes: acquiring and preprocessing multimodal images of the target region; achieving cross-modal feature fusion and alignment through cascaded deep networks and attention mechanisms; mapping the fused features to quantitative parameters of microscopic repair indices using a pre-trained quantification model library; smoothing the parameter time series and predicting future trends based on a state-space model, and generating a visualization report. This application achieves accurate, quantitative, dynamic monitoring and forward-looking evaluation of the urinary repair effect of nanomaterials.

Description

Technical Field

[0001] This invention belongs to the field of electrical data processing, specifically relating to a method and system for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging. Background Technology

[0002] As the application of nanomaterials in the treatment of urinary system diseases deepens, the accurate assessment of their repair effects has become a crucial step in clinical translation. Nanomaterials, with their excellent biocompatibility, targeted delivery capabilities, and controlled release characteristics, have been widely explored for applications such as urethral injury repair, bladder function reconstruction, and renal tubular regeneration. However, traditional efficacy assessments primarily rely on postoperative tissue biopsies, urinary biochemical indicators, or single-modal imaging (such as ultrasound or CT), making it difficult to dynamically capture the entire process of nanomaterial distribution, degradation kinetics, and interactions with the tissue microenvironment. Such static or discrete assessment methods not only suffer from latency but may also lead to misjudgments of the repair process due to sampling bias, failing to meet the needs of personalized precision medicine for real-time, non-invasive, and high-dimensional information feedback.

[0003] Among these, multimodal medical imaging fusion analysis technology offers a new approach to solving the aforementioned problems. This technology integrates the advantages of multiple imaging modalities, including magnetic resonance imaging (MRI), positron emission tomography (PET), optical coherence tomography (OCT), and contrast-enhanced ultrasound, to simultaneously acquire composite information at the structural, functional, and molecular levels across different time scales and spatial resolutions. Its core objective is to construct a cross-modal, temporally continuous visualization framework to track the migration trajectory, residence time, local concentration changes, and induced tissue regeneration responses of nanomaterials within the urinary system.

[0004] While some studies have attempted to use multimodal imaging for tumor diagnosis and monitoring, current technologies face multiple bottlenecks in the specific scenario of urinary tract repair. First, different modalities of imaging exhibit inherent differences in spatiotemporal resolution, signal-to-noise ratio, and imaging depth, lacking adaptive registration and fusion mechanisms for dynamic deformations of the urinary system's anatomical structures (such as the bladder filling-emptying cycle). Second, the signal characteristics of nanomaterials are easily affected by urine flow, intestinal gas, and artifacts from surrounding tissues, resulting in weak cross-modal signal correlation and difficulty in establishing stable material-effect mapping models. Third, current systems mostly employ offline post-processing, failing to generate quantitative assessment indicators of repair efficacy in real-time during treatment, thus hindering timely adjustments to clinical interventions. These issues collectively result in existing monitoring methods being insufficient in terms of sensitivity, specificity, and timeliness to support closed-loop optimization of nanomaterial-based urinary tract repair. Therefore, there is an urgent need for intelligent monitoring methods and systems capable of deeply integrating multimodal images, dynamically analyzing material behavior, and outputting repair efficacy in real-time. Summary of the Invention

[0005] This invention provides a method and system for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging. It aims to address the technical problems in existing technologies, such as incomplete information from a single image modality, lack of real-time tracking of the distribution and metabolic processes of nanomaterials within the urinary system, inability to accurately quantify the dynamic response of tissue repair, resulting in assessment lag, high misjudgment rates, and a lack of individualized treatment guidance. This invention achieves high-precision, full-cycle, and quantifiable dynamic monitoring of the nanomaterial-mediated urinary repair process by constructing a multi-source medical image fusion analysis architecture and combining temporal modeling of the specific imaging enhancement mechanism of nanomaterials with the pathophysiological characteristics of urinary tissues.

[0006] A dynamic monitoring method for the urinary repair effect of nanomaterials based on multimodal imaging includes the following steps: S1: Acquire multimodal medical imaging data of subjects before and after nanomaterial intervention therapy; S2: Extract original feature maps related to the urinary system tissue structure, blood perfusion status, and spatial distribution of nanomaterials from the multimodal medical imaging data; S3: Perform spatiotemporal alignment processing on the original feature map to generate a multimodal fused image sequence with a synchronous spatiotemporal reference; S4: Based on the multimodal fused image sequence, construct a nanomaterial-tissue interaction dynamic model and output dynamic parameters; S5: Based on the aforementioned kinetic parameters, generate a dynamic evaluation index set for urinary tract repair effects; S6: Input the dynamic evaluation index set of urinary tract repair effect into the pre-trained repair process prediction network, and output the optimal image acquisition time point and nanomaterial replenishment recommendation threshold for the next monitoring cycle.

[0007] Preferably, in step S1, the multimodal medical imaging data includes magnetic resonance imaging data, ultrasound imaging data, and optical coherence tomography (OCT) data; wherein, the magnetic resonance imaging data is acquired using T1-weighted and T2-weighted dual-sequence acquisition, the ultrasound imaging data is acquired using color Doppler and superharmonic imaging modes, and the OCT data is acquired through an intracavitary scanning catheter.

[0008] Preferably, in step S2, extracting the original feature map includes: For magnetic resonance imaging data, a semantic segmentation algorithm based on the U-Net architecture was used to extract the three-dimensional contour masks of the renal pelvis, ureter, bladder and urethra, and the spatial distribution maps of T1 relaxation time and T2 relaxation time were calculated based on exponential fitting of multi-echo sequences. For ultrasound imaging data, radio frequency signal inversion processing is performed, including bandpass filtering, cross-correlation displacement estimation, strain tensor calculation and Hooke's law inversion, to generate tissue elastic modulus map and blood flow velocity vector field; For optical coherence tomography (OCT) data, interferometric signal demodulation is performed, including fast Fourier transform, window function weighting, and edge detection, to generate epithelial thickness, basement membrane continuity, and submucosal scattering coefficient characteristics.

[0009] Preferably, in step S3, the spatiotemporal alignment process includes: Spatial alignment: Using magnetic resonance imaging (MRI) data as the reference grid, voxel coordinates of ultrasound imaging and optical coherence tomography (OCT) data are mapped to this grid through an affine transformation matrix. The transformation parameters are optimized using the mutual information maximization criterion, with an iteration step size of 0.01 and a convergence threshold of 10. -5 ; Time alignment: A dynamic time warping algorithm is adopted, with a window constraint of ±6 hours. Euclidean distance is used as the distance metric, and cubic spline interpolation is used to ensure temporal continuity. Finally, a multimodal fused image sequence with a synchronous spatiotemporal reference is generated.

[0010] Preferably, in step S4, constructing the nanomaterial-tissue interaction dynamics model includes: Define the signal intensity function of nanomaterials as: , in The initial signal strength, The clearance rate constant, The background noise baseline; Define the epithelial regeneration function as follows: , in To maximize repair potential, This is the regeneration rate constant; The signal intensity temporal curve S(t) and epithelial thickness temporal curve R(t) of the target region in the multimodal fused image sequence were fitted using the Levenberg-Marquardt nonlinear least squares algorithm, and the solution was obtained. and The coupling relationship.

[0011] Preferably, step S4 further includes calculating the cross-correlation coefficient. As a measure of the interaction strength between nanomaterial retention and tissue regeneration, cov represents the covariance. and They are respectively and The standard deviation.

[0012] Preferably, the kinetic parameters output in step S4 also include the retention index. Elimination half-life and organizational penetration depth ,in It is determined by the maximum vertical distance at which the nanomaterial signal penetrates the epithelial layer in optical coherence tomography.

[0013] Preferably, in step S5, the dynamic evaluation index set for urinary tract repair effect includes: Epithelial integrity recovery rate: The ratio of the current continuous epithelial length to the baseline damage length is multiplied by 100%, where the continuous epithelial length is extracted by a morphological skeletonization algorithm; Basement membrane reconstruction index: defined as the integral mean of the proportion of pixels in an optical coherence tomography image whose gray-level gradient of the basement membrane layer is greater than a preset threshold over the monitoring period; Local inflammatory response intensity score: Based on the proportion of areas with a Young's modulus greater than 8 kPa in the ultrasound elastic modulus map, it is quantified from 0 to 4. Microvessel density change rate: Calculate the difference between the current microvessel pixel density and the baseline microvessel pixel density, divided by the baseline value. The microvessel pixel density is determined by the proportion of pixels with a velocity greater than 0.5 cm / s in the ultrasound blood flow velocity vector field.

[0014] Preferably, in step S6, the repair process prediction network is a gated recurrent unit network. Its input layer receives the time-series vector of the dynamic evaluation index set of urinary tract repair effect, the hidden layer contains two layers of gated recurrent units, using the tanh activation function, and the output layer is a fully connected layer that outputs the next image acquisition time point. Recommended threshold for nanomaterial supply .

[0015] The present invention also provides a system for realizing the above-mentioned method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging, comprising: A multimodal imaging acquisition unit is used to acquire multimodal medical imaging data of subjects before and after nanomaterial intervention therapy. The multimodal imaging acquisition unit includes a magnetic resonance scanner, a high-frequency ultrasound probe, and an intracavitary optical coherence tomography catheter. The magnetic resonance scanner is equipped with a dedicated urological imaging coil, the center frequency of the ultrasound probe takes into account both penetration depth and resolution, and the optical coherence tomography catheter has 360-degree rotation scanning capability. The feature extraction unit is used to extract the original feature map from the multimodal medical image data. The feature extraction unit has built-in magnetic resonance semantic segmentation submodule, ultrasound elastic inversion submodule and optical coherence tomography reconstruction submodule, which respectively perform semantic segmentation, radio frequency signal inversion and interference signal demodulation processing. The spatiotemporal alignment unit is used to perform spatiotemporal alignment processing on the original feature map to generate a multimodal fusion image sequence. The spatiotemporal alignment unit is equipped with a mutual information optimization engine and a dynamic time warping engine, which run independently to optimize spatial coordinate mapping and temporal node interpolation. An interactive dynamics modeling unit is used to construct a nanomaterial-tissue interaction dynamics model based on the multimodal fused image sequence and output dynamic parameters. The interactive dynamics modeling unit is equipped with a parameter fitting engine, integrates the Levenberg-Marquardt algorithm library, and supports parallel fitting of multiple target regions. The dynamic evaluation index generation unit is used to generate a set of dynamic evaluation indicators for urinary tract repair effect based on the kinetic parameters. The dynamic evaluation index generation unit has a built-in rule base that stores the calculation formulas and clinical threshold ranges of the evaluation indicators and supports real-time scoring. The repair process prediction unit is used to input the dynamic evaluation index set of urinary tract repair effect into the pre-trained repair process prediction network, and output the optimal image acquisition time point and nanomaterial replenishment recommendation threshold for the next monitoring cycle. The repair process prediction unit is connected to the hospital information system to realize the automatic distribution and execution of prediction results.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: First, by integrating multimodal data such as magnetic resonance imaging, ultrasound imaging, and optical coherence tomography, this method solves the problems of limited information and delayed evaluation in traditional single-image modal data. It enables dynamic capture of the distribution, metabolism, and tissue repair response of nanomaterials in the urinary system throughout the entire cycle, thereby significantly reducing the false judgment rate and improving the real-time and non-invasive nature of monitoring.

[0017] Secondly, cross-modal feature fusion and alignment are achieved through cascaded deep networks and attention mechanisms. Combined with a nanomaterial-tissue interaction dynamics model, the fused features are mapped to quantitative parameters (such as retention index and clearance half-life), enabling precise quantification of the microscopic repair process. This avoids inaccurate assessments caused by sampling bias or signal interference, providing highly sensitive and specific quantitative indicators for clinical use.

[0018] Finally, the prediction network based on the state-space model can prospectively assess future repair trends and automatically generate optimal image acquisition time points and nanomaterial replenishment recommendation thresholds, realizing closed-loop optimization from monitoring to intervention. This not only supports the dynamic adjustment of individualized treatment plans but also enhances the convenience and reliability of clinical decision-making through visual reports, ultimately promoting the advancement of nanomaterial urinary repair therapy towards precision medicine. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall technical architecture of the dynamic monitoring system for the urinary repair effect of nanomaterials based on multimodal imaging proposed in this invention. Figure 2 This is a schematic diagram of the overall technical solution of the dynamic monitoring method for the urinary repair effect of nanomaterials based on multimodal imaging proposed in this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This invention provides a method and system for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging. It aims to address the technical problems in existing technologies, such as incomplete information from a single image modality, lack of real-time tracking of the distribution and metabolic processes of nanomaterials within the urinary system, inability to accurately quantify the dynamic response of tissue repair, resulting in assessment lag, high misjudgment rates, and a lack of individualized treatment guidance. This method constructs a multi-source medical image fusion analysis architecture, combining the time-series modeling of the specific imaging enhancement mechanism of nanomaterials with the pathophysiological characteristics of urinary tissues, to achieve high-precision, full-cycle, and quantifiable dynamic monitoring of the nanomaterial-mediated urinary repair process.

[0022] The method includes the following steps: S1, Acquire multimodal medical imaging data of subjects before and after nanomaterial intervention therapy; S2, extract the original feature maps related to the urinary system tissue structure, blood perfusion status and spatial distribution of nanomaterials from the multimodal medical image data; S3, perform spatiotemporal alignment processing on the original feature map to unify the spatial coordinate system and temporal sampling nodes of each modality image, and generate a multimodal fusion image sequence with synchronous spatiotemporal reference; S4. Based on the multimodal fused image sequence, construct a dynamic model of nanomaterial-tissue interaction; S5. Based on the output parameters of the nanomaterial-tissue interaction dynamics model, generate a dynamic evaluation index set for urinary repair effect; S6, input the dynamic evaluation index set of urinary tract repair effect into the pre-trained repair process prediction network, and output the optimal image acquisition time point and nanomaterial replenishment recommendation threshold for the next monitoring cycle.

[0023] In step S1, multimodal medical imaging data of the subjects before and after nanomaterial intervention therapy are acquired. This multimodal medical imaging data includes at least magnetic resonance imaging (MRI) data, ultrasound imaging data, and optical coherence tomography (OCT) data. Specifically, after the nanomaterial is injected via the urethra or vein, MRI scans, high-frequency ultrasound scans, and intracavitary OCT scans are performed sequentially at preset time points: 0, 6 hours, 24 hours, 72 hours, and 7 days. MRI uses T1-weighted and T2-weighted dual sequences to simultaneously capture the longitudinal and transverse relaxation characteristics of tissues; ultrasound imaging uses color Doppler and superharmonic imaging modes, the former to characterize hemodynamic status and the latter to improve the signal-to-noise ratio of the scattered signals from the microbubble contrast agent or nanomaterial; OCT ensures high-resolution imaging of the epithelial microstructure. All imaging data are stored in a digital medical imaging communication standard format and include precise timestamps and spatial location metadata, providing a basis for subsequent spatiotemporal alignment.

[0024] In step S2, original feature maps related to the urinary system tissue structure, blood perfusion status, and spatial distribution of nanomaterials are extracted from the multimodal medical imaging data. For magnetic resonance imaging data, a semantic segmentation algorithm based on the U-Net architecture is used to extract three-dimensional contour masks for the renal pelvis, ureter, bladder, and urethra. The U-Net encoder consists of multiple convolutional blocks, and the decoder fuses the high-dimensional feature maps of each stage of the encoder with the upsampling path through skip connections, ultimately outputting a pixel-level tissue category probability map. Based on this, the spatial distribution maps of T1 relaxation time and T2 relaxation time are calculated. The calculation relies on exponential fitting of multi-echo or multi-reversal time series, with the fitting residual controlled within 5%. For ultrasound imaging data, radio frequency signal inversion processing is performed: first, the original radio frequency echo signal is bandpass filtered to remove low-frequency motion artifacts and high-frequency noise; then, the tissue displacement field is estimated using a cross-correlation algorithm; next, the strain tensor is calculated using the finite difference method; finally, the tissue elastic modulus map is inverted according to Hooke's law. Simultaneously, based on color Doppler frequency shift information, the blood flow velocity vector field is reconstructed, its direction and amplitude determined by the autocorrelation phase difference and the peak power spectrum. For optical coherence tomography (OCT) data, interferometric signal demodulation is performed: the original interferometric spectrum is converted into a frequency domain signal via fast Fourier transform; then, axial resolution is improved through window function weighting and zero-filling to generate the A-scan signal; multiple A-scans are stitched together along the rotation direction of the scanning catheter to form a B-scan cross-sectional image; based on this image, edge detection operators are used to identify the upper and lower boundaries of the epithelial layer and calculate the epithelial layer thickness; gray-level gradient analysis is used to determine the continuity of the basement membrane; and the scattering coefficient of the submucosal region is statistically calculated, its value being jointly estimated by local contrast and attenuation slope.

[0025] In step S3, the original feature maps undergo spatiotemporal alignment to unify the spatial coordinate systems and temporal sampling nodes of each modality image, generating a multimodal fused image sequence with a synchronized spatiotemporal reference. Spatial alignment uses the spatial resolution of magnetic resonance imaging as the reference grid, due to its most uniform voxel isotropy and maximum coverage. The voxel coordinates of ultrasound imaging and optical coherence tomography are mapped to this reference grid through an affine transformation matrix. The affine transformation parameters are optimized using the maximization of mutual information criterion, with the initial transformation matrix being an identity matrix, the iteration step size set to 0.01, and the convergence threshold set to 10. -5 The optimization process terminates when the mutual information increment over three consecutive iterations is less than a threshold. Time alignment employs a dynamic time warping algorithm with a window constraint of ±6 hours to prevent excessive distortion of the physiological timeline. Euclidean distance is used as the distance metric, and corresponding pixel values ​​in each feature map are matched point-by-point. Cubic spline interpolation ensures the continuity of the second derivative of the time series curve. After this processing, feature maps from all modalities are resampled to the same spatiotemporal grid, forming a multimodal fused image sequence, with each time point corresponding to a spatially aligned set of feature maps.

[0026] In step S4, a nanomaterial-tissue interaction dynamics model is constructed based on the multimodal fused image sequence. This model calculates the retention index, clearance half-life, and tissue penetration depth of the nanomaterial in the target region by coupling the signal attenuation curve of the nanomaterial with the regeneration rate function of the urinary epithelial tissue. Specifically, the nanomaterial signal intensity function is defined as:

[0027] in, As the initial signal intensity, in the spatiotemporally aligned, zero-time multimodal fusion image sequence, the initial enriched target region (ROI) of the nanomaterial is automatically identified through threshold segmentation (e.g., regions with signal intensity higher than the overall image average plus three standard deviations) or by combining a nanomaterial contrast agent-specific imaging channel. Subsequently, the average signal intensity of all pixels (or voxels) within this ROI is calculated; this value is the initial signal intensity. ; The clearance rate constant reflects the rate at which nanomaterials are metabolically excreted from the target region; As the background noise baseline, on the fused image at the same time (time zero), a normal tissue region with similar anatomical structure but confirmed to be free of nanomaterials is selected as the background ROI outside the target enriched region. The average signal intensity of all pixels within this background ROI is calculated; this value is the background noise baseline. Meanwhile, the epithelial regeneration function is defined as:

[0028] in, For the maximum repair potential, the maximum repair potential This is a preset physiological constant or determined through individualized methods. As a preferred embodiment, The average urinary epithelial thickness of healthy individuals was taken from a historical database. As another example, It can also be obtained by calibrating the epithelial thickness of the contralateral healthy site (such as the contralateral healthy ureter or bladder trigone) before treatment in the same subject; exp is used to quantify the exponential laws of "decay" and "growth" that change over time. The regeneration rate constant represents the biological activity of tissue repair. The signal intensity temporal curve of the target region in a multimodal fused image sequence was fitted using the least squares method. Time-series curve of epithelial thickness Solve and The coupling relationship. The fitting process calls the Levenberg-Marquardt nonlinear least squares algorithm, with initial parameters set to... , The maximum number of iterations is 100, and the iteration terminates when the sum of squared residuals is less than 0.001. Furthermore, a cross-correlation coefficient is introduced. As a measure of interaction strength:

[0029] in, Let represent the covariance, and and be the standard deviations of signal strength and regenerated thickness, respectively. The closer the value is to 1, the stronger the positive correlation between nanomaterial retention and tissue regeneration, indicating good biocompatibility and repair-promoting efficacy; if... A value close to 0 or negative suggests potential inflammatory interference or ineffective retention. Based on... , and Further calculate the retention index Elimination half-life and organizational penetration depth The latter is determined by the maximum vertical distance at which the nanomaterial signal penetrates the epithelial layer in optical coherence tomography.

[0030] In step S5, a dynamic evaluation index set for urinary tract repair effect is generated based on the output parameters of the nanomaterial-tissue interaction dynamics model. This evaluation index set includes epithelial integrity recovery rate, basement membrane reconstruction index, local inflammatory response intensity score, and microvascular density change rate. The epithelial integrity recovery rate is defined as the ratio of the current continuous epithelial length to the baseline damage length multiplied by 100%. The continuous epithelial length is extracted using a morphological skeletonization algorithm: first, the epithelial mask is refined, retaining a single-pixel-wide centerline; then, isolated branches shorter than 10 pixels are removed, leaving the remaining main trunk length as the continuous length. The basement membrane reconstruction index is defined as the integral mean of the basement membrane continuity function over the total monitoring time. The basement membrane continuity function is the proportion of pixels in the optical coherence tomography (OCT) image where the gray-level gradient of the basement membrane layer is greater than a preset threshold; a higher proportion indicates a more complete basement membrane structure. The intensity of local inflammatory response is scored based on the proportion of regions with a Young's modulus greater than 8 kPa in the ultrasound elastic modulus map, quantified from 0 to 4: less than 5% is grade 0, 5% to 15% is grade 1, 15% to 30% is grade 2, 30% to 50% is grade 3, and greater than 50% is grade 4. The rate of change in microvessel density is defined as the difference between the current microvessel pixel density and the baseline microvessel pixel density divided by the baseline value. Microvessel pixel density is determined by the proportion of pixels with a velocity greater than 0.5 cm / s in the ultrasound blood flow velocity vector field to the total area of ​​the region of interest.

[0031] By calculating the signal of nanomaterials With epithelial regeneration thickness Cross-correlation coefficient of time series curves Assess the degree of coupling between the two. If there is severe edema or nonspecific tissue thickening caused by inflammation (leading to...) (abnormally high), but it is related to the retention signal of nanomaterials. If there is no temporal correlation, then The value will decrease significantly (approaching 0 or becoming negative). This low The value will serve as an alert, indicating the currently fitted regeneration rate constant. The calculated repair indicators may be affected by non-specific factors and need to be carefully evaluated in conjunction with other indicators.

[0032] As described in step S5, this method specifically calculates a score of local inflammatory response intensity from ultrasound elastography. This index is related to the epithelial regeneration function. Parallel computing, independent output. If If the repair progress is good, but the local inflammatory response intensity score is also at a high level, it suggests that the repair process may be accompanied by a significant inflammatory response. Clinical interpretation requires comprehensive judgment to avoid misjudging simple edema or inflammatory hyperplasia as effective regeneration.

[0033] In step S6, the dynamic evaluation index set of urinary tract repair effect is input into a pre-trained repair process prediction network, which outputs the optimal image acquisition time point for the next monitoring cycle and the suggested threshold for nanomaterial replenishment. The repair process prediction network is a gated recurrent unit network. Its input layer receives a temporal vector of the dynamic evaluation index set of urinary tract repair effect, which contains four index values ​​from the three most recent time points, totaling 12 dimensions. The hidden layer contains two layers of gated recurrent unit cells, each with 64 cells, using the tanh activation function. The output layer is a fully connected layer, outputting two scalars: the next image acquisition time point. Recommended threshold for nanomaterial supply The network uses complete monitoring sequences of historical cases for end-to-end supervised learning during the training phase. Deployed on a cloud server, the prediction results are automatically triggered by the image acquisition appointment system and pharmacy management system after being reviewed by clinicians, thus achieving closed-loop intervention.

[0034] The system comprises a multimodal image acquisition unit, a feature extraction unit, a spatiotemporal alignment unit, an interactive dynamics modeling unit, a dynamic evaluation index generation unit, and a repair process prediction unit. The multimodal image acquisition unit includes a magnetic resonance imaging (MRI) scanner, a high-frequency ultrasound probe, and an intracavitary optical coherence tomography (OCT) catheter. The MRI scanner is equipped with a dedicated urological imaging coil to improve the signal-to-noise ratio and spatial resolution; the ultrasound probe's center frequency balances penetration depth and resolution; and the OCT catheter has a 360-degree rotation scanning capability, suitable for imaging within the urethra and bladder. The feature extraction unit incorporates three parallel processing submodules: an MRI semantic segmentation submodule, an ultrasound elastography submodule, and an OCT reconstruction submodule, each executing the feature extraction tasks described in S2. The spatiotemporal alignment unit performs spatial coordinate mapping and temporal node interpolation in S3, with its mutual information optimization engine and dynamic time warping engine operating independently to ensure processing efficiency. The interactive dynamics modeling unit is equipped with a parameter fitting engine, integrating the Levenberg-Marquardt algorithm library, supporting parallel fitting of multiple target regions. The dynamic evaluation indicator generation unit has a built-in rule base that stores the calculation formulas and clinical threshold ranges for each indicator, supporting real-time scoring. The repair process prediction unit interfaces with the hospital information system to achieve automatic distribution and execution of prediction results.

[0035] Throughout the system's operation, the data stream originates from the multimodal image acquisition unit, passes through the feature extraction unit to generate raw feature maps, is then fused into a sequence with a unified spatiotemporal reference by the spatiotemporal alignment unit, inputs it to the interaction dynamics modeling unit to calculate core parameters, subsequently outputs a quantitative score by the dynamic evaluation index generation unit, and finally generates clinical decision recommendations by the repair progress prediction unit. All units are interconnected via a high-speed internal bus, and data transmission employs an encrypted compression protocol to ensure security and real-time performance. The system supports concurrent processing of multiple patients, with each patient's data stream independently isolated to avoid cross-contamination. Anomaly handling mechanisms include image quality verification, retrying feature extraction failures, and model fitting divergence alerts to ensure robustness throughout the entire process.

[0036] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging, characterized in that, Includes the following steps: S1: Acquire multimodal medical imaging data of subjects before and after nanomaterial intervention therapy; S2: Extract original feature maps related to the urinary system tissue structure, blood perfusion status, and spatial distribution of nanomaterials from the multimodal medical imaging data; S3: Perform spatiotemporal alignment processing on the original feature map to generate a multimodal fused image sequence with a synchronous spatiotemporal reference; S4: Based on the multimodal fused image sequence, construct a nanomaterial-tissue interaction dynamic model and output dynamic parameters; S5: Based on the aforementioned kinetic parameters, generate a dynamic evaluation index set for urinary tract repair effects; S6: Input the dynamic evaluation index set of urinary tract repair effect into the pre-trained repair process prediction network, and output the optimal image acquisition time point and nanomaterial replenishment recommendation threshold for the next monitoring cycle.

2. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S1, the multimodal medical imaging data includes magnetic resonance imaging data, ultrasound imaging data, and optical coherence tomography (OCT) data; wherein, the magnetic resonance imaging data is acquired using T1-weighted and T2-weighted dual-sequence acquisition, the ultrasound imaging data is acquired using color Doppler and superharmonic imaging modes, and the OCT data is acquired through an intracavitary scanning catheter.

3. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S2, extracting the original feature map includes: For magnetic resonance imaging data, a semantic segmentation algorithm based on the U-Net architecture was used to extract the three-dimensional contour masks of the renal pelvis, ureter, bladder and urethra, and the spatial distribution maps of T1 relaxation time and T2 relaxation time were calculated based on exponential fitting of multi-echo sequences. For ultrasound imaging data, radio frequency signal inversion processing is performed, including bandpass filtering, cross-correlation displacement estimation, strain tensor calculation and Hooke's law inversion, to generate tissue elastic modulus map and blood flow velocity vector field; For optical coherence tomography (OCT) data, interferometric signal demodulation is performed, including fast Fourier transform, window function weighting, and edge detection, to generate epithelial thickness, basement membrane continuity, and submucosal scattering coefficient characteristics.

4. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S3, the spatiotemporal alignment process includes: Spatial alignment: Using magnetic resonance imaging (MRI) data as the reference grid, voxel coordinates of ultrasound imaging and optical coherence tomography (OCT) data are mapped to this grid through an affine transformation matrix. The transformation parameters are optimized using the mutual information maximization criterion, with an iteration step size of 0.01 and a convergence threshold of 10. -5 ; Time alignment: A dynamic time warping algorithm is adopted, with a window constraint of ±6 hours. Euclidean distance is used as the distance metric, and cubic spline interpolation is used to ensure temporal continuity. Finally, a multimodal fused image sequence with a synchronous spatiotemporal reference is generated.

5. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S4, constructing the nanomaterial-tissue interaction dynamics model includes: Define the signal intensity function of nanomaterials as: ， in The initial signal strength, The clearance rate constant, The background noise baseline; Define the epithelial regeneration function as follows: ， in To maximize repair potential, This is the regeneration rate constant; The signal intensity temporal curve S(t) and epithelial thickness temporal curve R(t) of the target region in the multimodal fused image sequence were fitted using the Levenberg-Marquardt nonlinear least squares algorithm, and the solution was obtained. and The coupling relationship.

6. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 5, characterized in that, Step S4 also includes calculating the cross-correlation coefficient. As a measure of the interaction strength between nanomaterial retention and tissue regeneration, cov represents the covariance. and They are respectively and The standard deviation.

7. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 5, characterized in that, The dynamic parameters output in step S4 also include the retention index. Elimination half-life and organizational penetration depth ,in It is determined by the maximum vertical distance at which the nanomaterial signal penetrates the epithelial layer in optical coherence tomography.

8. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S5, the dynamic evaluation index set for urinary tract repair effect is generated, including: Epithelial integrity recovery rate: The ratio of the current continuous epithelial length to the baseline damage length is multiplied by 100%, where the continuous epithelial length is extracted by a morphological skeletonization algorithm; Basement membrane reconstruction index: defined as the integral mean of the proportion of pixels in an optical coherence tomography image whose gray-level gradient of the basement membrane layer is greater than a preset threshold over the monitoring period; Local inflammatory response intensity score: Based on the proportion of areas with a Young's modulus greater than 8 kPa in the ultrasound elastic modulus map, it is quantified from 0 to 4. Microvessel density change rate: Calculate the difference between the current microvessel pixel density and the baseline microvessel pixel density, divided by the baseline value. The microvessel pixel density is determined by the proportion of pixels with a velocity greater than 0.5 cm / s in the ultrasound blood flow velocity vector field.

9. The method for dynamic monitoring of the urinary repair effect of nanomaterials based on multimodal imaging as described in claim 1, characterized in that, In step S6, the repair process prediction network is a gated recurrent unit network. Its input layer receives the time-series vector of the dynamic evaluation index set of urinary tract repair effect, the hidden layer contains two layers of gated recurrent units, using the tanh activation function, and the output layer is a fully connected layer, outputting the next image acquisition time point. Recommended threshold for nanomaterial supply .

10. A system for implementing the method for dynamic monitoring of the urinary tract repair effect based on multimodal imaging of nanomaterials as described in any one of claims 1 to 9, characterized in that, include: A multimodal imaging acquisition unit is used to acquire multimodal medical imaging data of subjects before and after nanomaterial intervention therapy. The multimodal imaging acquisition unit includes a magnetic resonance scanner, a high-frequency ultrasound probe, and an intracavitary optical coherence tomography catheter. The magnetic resonance scanner is equipped with a dedicated urological imaging coil, the center frequency of the ultrasound probe takes into account both penetration depth and resolution, and the optical coherence tomography catheter has 360-degree rotation scanning capability. The feature extraction unit is used to extract the original feature map from the multimodal medical image data. The feature extraction unit has built-in magnetic resonance semantic segmentation submodule, ultrasound elastic inversion submodule and optical coherence tomography reconstruction submodule, which respectively perform semantic segmentation, radio frequency signal inversion and interference signal demodulation processing. The spatiotemporal alignment unit is used to perform spatiotemporal alignment processing on the original feature map to generate a multimodal fusion image sequence. The spatiotemporal alignment unit is equipped with a mutual information optimization engine and a dynamic time warping engine, which run independently to optimize spatial coordinate mapping and temporal node interpolation. An interactive dynamics modeling unit is used to construct a nanomaterial-tissue interaction dynamics model based on the multimodal fused image sequence and output dynamic parameters. The interactive dynamics modeling unit is equipped with a parameter fitting engine, integrates the Levenberg-Marquardt algorithm library, and supports parallel fitting of multiple target regions. The dynamic evaluation index generation unit is used to generate a set of dynamic evaluation indicators for urinary tract repair effect based on the kinetic parameters. The dynamic evaluation index generation unit has a built-in rule base that stores the calculation formulas and clinical threshold ranges of the evaluation indicators and supports real-time scoring. The repair process prediction unit is used to input the dynamic evaluation index set of urinary tract repair effect into the pre-trained repair process prediction network, and output the optimal image acquisition time point and nanomaterial replenishment recommendation threshold for the next monitoring cycle. The repair process prediction unit is connected to the hospital information system to realize the automatic distribution and execution of prediction results.

Images