Dynamic monitoring method for residual focus of lymphoma

Abstract

The invention discloses a dynamic monitoring method for residual focus of lymphoma, and particularly relates to the technical field of focus monitoring. The method comprises the following steps: acquiring multi-modal medical detection original data of a lymphoma patient to generate a multi-modal medical detection data set; the method comprises the following steps: constructing a phase space distribution model of a residual lymphoma focus, extracting space phase structure feature data, identifying subclone structure features, and generating amplification trend feature data; performing dynamic stability analysis based on the amplification trend characteristic data and the space phase structure characteristic data to obtain subcloning competition dynamic stability data; heterogeneity diffusion risk assessment data is formed by analyzing the heterogeneity diffusion relationship between different subcloning structures; and finally, according to the heterogeneity diffusion risk assessment data, dynamically updating a focus progress monitoring index, and outputting a dynamic monitoring result of a patient focus evolution trend in real time. According to the invention, accurate monitoring and dynamic prediction of the evolution state of the residual focus of lymphoma are realized.

Description

Technical Field

[0001] This invention relates to the field of lesion monitoring technology, and more specifically, to a method for dynamic monitoring of residual lymphoma lesions. Background Technology

[0002] Lymphoma is a common type of malignant tumor, and dynamic monitoring of residual lesions is crucial during treatment. In clinical medical practice, multimodal imaging techniques are commonly used to follow up and evaluate patient lesions.

[0003] Existing methods for monitoring residual lymphoma lesions are mostly based on overall imaging parameters or macroscopic changes, which do not fully reflect the complex spatial heterogeneity and dynamic evolution process between different subclonal structures within the lesion, resulting in limited ability to identify the true progression trend of the lesion. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method for dynamic monitoring of residual lymphoma lesions to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for dynamic monitoring of residual lymphoma lesions includes the following steps: S1. Collect raw data of multimodal medical tests from lymphoma patients, and preprocess the raw data of multimodal medical tests to generate a multimodal medical test dataset. S2, based on a multimodal medical detection dataset, constructs a phase spatial distribution model of residual lymphoma lesions and extracts spatial phase structure feature data between different subclonal structures within residual lymphoma lesions; S3, based on spatial phase structure feature data, identifies the subclonal structure features inside residual lymphoma lesions and generates amplification trend feature data between different subclonal structures; S4. Based on amplification trend characteristic data and spatial phase structure characteristic data, dynamic stability analysis is performed on the competitive relationship between different subclonal structures to obtain subclonal competition dynamic stability data. S5, based on subclonal competition dynamic stability data, analyzes the heterogeneity diffusion relationship between different subclonal structures and generates heterogeneity diffusion risk assessment data; S6, based on heterogeneous spread risk assessment data, dynamically updates the progression monitoring indicators of residual lymphoma lesions and outputs dynamic monitoring results.

[0006] In a preferred embodiment, S1 specifically refers to: Magnetic resonance imaging data, positron emission tomography (PET) scan data, and computed tomography (CT) image data of lymphoma patients were collected as raw data for multimodal medical testing. The raw data of multimodal medical testing are processed by data format conversion, spatial registration, noise removal and data normalization to generate a multimodal medical testing dataset.

[0007] In a preferred embodiment, S2 specifically refers to: The multimodal medical testing dataset is rigidly and non-rigidly registered in three-dimensional space to obtain three-dimensional registered image data in a unified coordinate system. Based on three-dimensional registered image data in a unified coordinate system, an image segmentation algorithm was used to identify the residual lymphoma lesion area and extract the voxel gray value of the lesion area. Phase spectrum evaluation is performed on the voxel gray values ​​of the lesion area to generate phase value data for each voxel; A phase spatial distribution model of residual lymphoma lesions was constructed based on phase value data; Based on the phase spatial distribution model of residual lymphoma lesions, the phase difference matrix between voxel sets corresponding to different subclonal structures is calculated, and spatial phase structure feature data is output.

[0008] In a preferred embodiment, S3 specifically refers to: Density clustering was performed on the spatial phase structure feature data, and the candidate voxel set of subclone structure was divided according to the phase difference threshold between voxels and the three-dimensional Euclidean distance threshold. Voxel morphology centroid coordinates are calculated based on the candidate voxel set of subcloned structures to obtain voxel morphology centroid data. The spatial correspondence between candidate voxel sets of subclonal structures at different time points was determined by voxel morphological centroid data, forming voxel set association pairs of subclonal structures. Calculate the relative change rate of the number of endovoxels associated with each subclonal structure voxel set to obtain the subclonal structure amplification rate parameter. The amplification rate parameters of the subclonal structures are arranged in chronological order to form amplification trend characteristic data among different subclonal structures.

[0009] In a preferred embodiment, S4 specifically refers to: Spatial phase structure feature data and subclonal structure amplification trend feature data are synchronously paired to construct a time-series feature matrix; Singular spectral decomposition is performed on the time-series feature matrix to extract nonlinear dynamic eigenmodes, resulting in a competitive dynamic eigenmode sequence. Calculate the local Lyapunov exponent sequence and spectral entropy sequence of the competitive dynamic intrinsic mode sequence to form a set of competitive dynamic instability indices; Aggregate the set of competition dynamic instability indicators to generate competition dynamic stability curves; The time labels of the stability threshold crossover points are extracted from the competitive dynamic stability curve to form subclone competitive dynamic stability data.

[0010] In a preferred embodiment, S5 specifically refers to: The subclonal competition dynamic stability data and spatial phase structure feature data are matched according to time labels to construct a heterogeneity diffusion assessment matrix; The diffusion probability matrix is ​​obtained by using a random walk graph traversal algorithm based on the heterogeneous diffusion evaluation matrix to calculate the diffusion probability between each subclone structure. Perform spectral clustering on the diffusion probability matrix to extract a set of heterogeneous sub-clusters; Calculate the maximum diffusion intensity parameter, average diffusion intensity parameter, and diffusion intensity variance parameter of the heterogeneous sub-cluster set to generate a heterogeneous diffusion index set; The heterogeneity diffusion index set is transformed into heterogeneity diffusion risk assessment data based on a preset threshold determination strategy.

[0011] In a preferred embodiment, S6 specifically refers to: Transform heterogeneous diffusion risk assessment data into progress monitoring indicators to update the input vector; Load the progress monitoring indicator library and initialize the progress monitoring indicator set; Based on the progress monitoring indicator update input vector, hierarchical recursive decision rule operation is performed to update the progress monitoring indicator set, resulting in the updated progress monitoring indicator set. The updated set of progress monitoring indicators is merged with the corresponding time tags to generate dynamic monitoring result data.

[0012] The technical effects and advantages of the dynamic monitoring method for residual lymphoma lesions of the present invention are as follows: By preprocessing multimodal medical testing datasets, high-quality fusion of different types of imaging data is achieved. Based on the fused data, a phase spatial distribution model is constructed and spatial phase structure features are extracted, revealing microscopic phase differences within lesions and reflecting the spatial heterogeneity of subclonal structures. By identifying subclonal voxel sets and calculating amplification trends based on phase structure feature data, the proliferation rate of each subclonal population can be quantified, providing dynamic indicators for lesion evolution. By combining amplification trends and phase features to conduct competitive dynamic stability analysis, the stable and unstable transition moments of interactions between subclonal populations can be captured, revealing the competitive evolutionary laws within lesions. Based on competitive stability data analysis, heterogeneous diffusion relationships are analyzed and diffusion risks are assessed, enabling quantitative prediction of potential local progression and recurrence risks. Finally, based on the diffusion risk assessment results, progression monitoring indicators are dynamically updated and dynamic monitoring results are output, providing real-time and accurate lesion progression information for clinicians, improving the accuracy and timeliness of residual lesion monitoring and prognostic assessment. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a dynamic monitoring method for residual lymphoma lesions according to the present invention. Detailed Implementation

[0014] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention. Example

[0015] Figure 1 This invention provides a method for dynamic monitoring of residual lymphoma lesions, comprising the following steps: S1. Collect raw data of multimodal medical tests from lymphoma patients, and preprocess the raw data of multimodal medical tests to generate a multimodal medical test dataset. S2, based on a multimodal medical detection dataset, constructs a phase spatial distribution model of residual lymphoma lesions and extracts spatial phase structure feature data between different subclonal structures within residual lymphoma lesions; S3, based on spatial phase structure feature data, identifies the subclonal structure features inside residual lymphoma lesions and generates amplification trend feature data between different subclonal structures; S4. Based on amplification trend characteristic data and spatial phase structure characteristic data, dynamic stability analysis is performed on the competitive relationship between different subclonal structures to obtain subclonal competition dynamic stability data. S5, based on subclonal competition dynamic stability data, analyzes the heterogeneity diffusion relationship between different subclonal structures and generates heterogeneity diffusion risk assessment data; S6, based on heterogeneous spread risk assessment data, dynamically updates the progression monitoring indicators of residual lymphoma lesions and outputs dynamic monitoring results.

[0016] S1. Collect raw multimodal medical test data from lymphoma patients, and preprocess the raw multimodal medical test data to generate a multimodal medical test dataset, including: Magnetic resonance imaging data, positron emission tomography (PET) scan data, and computed tomography (CT) image data of lymphoma patients were collected as raw data for multimodal medical testing. Specifically, medical imaging examinations for lymphoma patients include magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT). MRI uses the interaction of magnetic fields and radiofrequency pulses to obtain magnetic resonance images of the tissues and organs within the lymphoma patient's body, including detailed structural information on different tissue types in the lymphoma lesion area, such as internal grayscale differences, signal intensity distribution, and structural contours. PET uses a specific radioactive contrast agent to obtain imaging data on the metabolic activity of the lymphoma lesion area, taking advantage of differences in the contrast agent's metabolic distribution. This metabolic activity imaging data can reveal metabolic differences between tumor tissue and normal tissue, such as the location, extent, and activity intensity of high-metabolic areas in the tumor tissue. CT utilizes the signal attenuation caused by X-rays penetrating the patient's body to obtain three-dimensional spatial anatomical images of the lymphoma lesion site, including the precise spatial location of the tumor tissue within the patient's body, the size of the lesion, and its spatial relationship with surrounding tissues.

[0017] The raw data of multimodal medical testing are subjected to data format conversion, spatial registration, noise removal and data normalization to generate a multimodal medical testing dataset. Specifically, data format conversion processing refers to converting the proprietary raw data formats output from magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT) devices into a unified data format that can be processed uniformly. For example, the proprietary image formats output by different devices can be uniformly converted into a common format for medical digital imaging and communication standards, ensuring that data from different modalities can be stored and processed uniformly. Spatial registration refers to the process of using registration algorithms to map data from different modalities into a unified three-dimensional coordinate space by calculating the correspondence of structural features between image data. This ensures that the spatial coordinates of magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT) data are consistent. During spatial registration, rigid spatial registration algorithms can be used for coarse registration, which involves translating and rotating different image data to initially establish spatial correspondences. After coarse registration, non-rigid spatial registration algorithms are used for fine registration. These algorithms calculate the grayscale similarity, feature point similarity, and structural contour similarity of each spatial location in the image data to achieve precise spatial matching, ultimately achieving accurate registration of different modalities in the spatial dimension. Noise removal refers to the use of image denoising algorithms to reduce random and systematic noise in image data, thereby improving image quality. For example, median filtering can be used to remove random noise by sorting the gray values ​​of each spatial point and its surrounding points in the image data and taking the median value of the sorted results as the denoised gray value. Another method is anisotropic diffusion filtering, which uses a diffusion equation to preserve the edges of lesion structures while smoothing the internal regions of the lesions, thus removing random noise from the image data. Data normalization refers to the unified mapping of the grayscale value range of denoised image data, adjusting the grayscale values ​​to a predetermined range so that magnetic resonance imaging data, positron emission tomography (PET) scan data, and computed tomography (CT) scan data have a unified standard in grayscale value distribution. For example, linear normalization methods can be used, which involve subtracting the minimum grayscale value from the grayscale value of each spatial location in the image data and then dividing by the difference between the maximum and minimum grayscale values, so that the grayscale distribution range of the image data is between zero and one. Nonlinear normalization methods can also be used, such as using grayscale value histogram matching, to map the grayscale distribution of different modal data to the same or similar statistical distributions, thereby ensuring that the grayscale values ​​between different modal medical test data can be directly compared.

[0018] S2, based on a multimodal medical detection dataset, constructs a phase spatial distribution model of residual lymphoma lesions, and extracts spatial phase structural feature data between different subclonal structures within the residual lymphoma lesions, including: The multimodal medical testing dataset is rigidly and non-rigidly registered in three-dimensional space to obtain three-dimensional registered image data in a unified coordinate system. Specifically, rigid registration refers to the process of aligning data from magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT) scans in spatial coordinates using spatial translation and rotation transformations. Spatial translation involves calculating the positional offsets of the image data in the horizontal, vertical, and vertical directions based on the differences in the anatomical structure center positions between different modalities. Each spatial position in the data is then moved and adjusted according to these offsets until all data are aligned. Spatial rotation involves rotating the image data in three dimensions based on anatomical markers or structural contours. The rotation aims to minimize the differences in structural contours between different modalities, ensuring precise alignment of the image data along all axes in the spatial coordinate system. Non-rigid registration refers to a process that considers individual differences and structural deformations in the patient's internal tissue structure, performing more precise spatial alignment on initially registered image data. Non-rigid registration achieves matching at each spatial location by calculating the similarity of local structural features or gray-level correlation of each voxel position in the image data. The non-rigid registration method employs elastic deformation registration, specifically: first, structural feature points are defined in the image data; by calculating the spatial positional differences of each structural feature point across different image data, the displacement field of the structural feature points is obtained; based on the displacement field of the structural feature points, interpolation methods are used to calculate the local displacement of each spatial location, achieving local elastic deformation; the image data after elastic deformation processing achieves spatial alignment of different modal data at the voxel level, forming three-dimensional registered image data in a unified coordinate system.

[0019] Based on three-dimensional registered image data in a unified coordinate system, an image segmentation algorithm was used to identify the residual lymphoma lesion area and extract the voxel gray value of the lesion area. Specifically, candidate locations of lesion areas are selected, and the locations that may be lesion areas are identified based on the differences in gray value distribution in the image data. The difference between the gray value of each voxel in the candidate lesion area and the gray value of the voxels in the surrounding normal tissue area is calculated, and the boundary range of the voxel set of the lesion area is determined by using a gray value threshold. Finally, the gray values ​​of all voxels within the determined boundary range are extracted to form the voxel gray value set of the lesion area. The voxel gray value set of the lesion area contains the structural feature information of the residual lymphoma lesion area, which can reflect the gray value feature differences of the internal tissue structure of the lesion area in different modal image data.

[0020] Phase spectrum evaluation is performed on the voxel gray values ​​of the lesion area to generate phase value data for each voxel; A phase spatial distribution model of residual lymphoma lesions was constructed based on phase value data; Specifically, phase spectrum assessment refers to processing the voxel gray values ​​of the lesion region using a phase-frequency domain transformation method. This involves performing a Fourier transform on the voxel gray values ​​at each spatial location within the lesion region. Specifically, this involves calculating the product of the voxel gray value with a sine function and a cosine function, integrating the product over the entire spatial region to obtain the spectral information corresponding to the gray value at each spatial location. Based on this spectral information, the phase information of the frequency domain signal is calculated. This phase information represents the phase shift characteristics of the voxel gray spatial distribution in the frequency domain, thus reflecting the spatial structural differences of the voxels in the frequency domain. After Fourier transform processing, the voxel phase value data corresponding to each spatial location within the lesion region is obtained. Based on three-dimensional spatial coordinates, each voxel at a spatial location has a corresponding phase value. By spatially reconstructing the spatial locations of all voxels and their corresponding phase values, a phase spatial distribution model is constructed. The model reflects the phase difference characteristics and spatial structural relationships between subclonal structural regions within the lesion. For example, voxel regions with similar phase values ​​within the lesion region may correspond to the same subclonal structural region, while spatial regions with significantly different phase values ​​may correspond to different subclonal structural regions. The spatial differences between different subclonal regions can be intuitively analyzed through the phase spatial distribution model.

[0021] Based on the phase spatial distribution model of residual lymphoma lesions, the phase difference matrix between voxel sets corresponding to different subclonal structures is calculated, and spatial phase structure feature data is output. Specifically, in the phase space distribution model, the set of all voxel positions corresponding to each subclonal structural region is determined; the difference in phase values ​​of voxel positions between different subclonal structural regions is calculated, and the absolute values ​​of the differences are summed and then divided by the number of voxel pairs involved in the calculation between the two subclonal structural regions to determine the average phase difference value between subclonal structural regions; finally, the average phase difference values ​​between different subclonal structural regions are summarized to form spatial phase structure feature data; the spatial phase structure feature data can reflect the differences in the phase space state between different subclonal structures within the residual lymphoma lesion area.

[0022] S3, based on spatial phase structure feature data, identifies subclonal structure features within residual lymphoma lesions, and generates amplification trend feature data among different subclonal structures, including: Density clustering was performed on the spatial phase structure feature data, and the candidate voxel set of subclone structure was divided according to the phase difference threshold between voxels and the three-dimensional Euclidean distance threshold. Specifically, density clustering refers to grouping voxels based on the distribution density and local density of their phase structure feature data in spatial locations; calculating the absolute value of the phase difference between any two spatial voxels within the lesion region and the three-dimensional Euclidean distance between the two spatial voxel positions; using the absolute value of the phase difference and the three-dimensional Euclidean distance between voxels as the criteria, and pre-setting the phase difference threshold and the three-dimensional Euclidean distance threshold; only when the absolute value of the phase difference between voxels is less than the phase difference threshold and the three-dimensional Euclidean distance is less than the three-dimensional Euclidean distance threshold, are the two spatial voxels determined to belong to the same subclonal structure candidate voxel set; through the above clustering determination, all voxels that meet the conditions are merged into one or more subclonal structure candidate voxel sets, thereby dividing multiple subclonal structure candidate voxel sets.

[0023] Voxel morphology centroid coordinates are calculated based on the candidate voxel set of subcloned structures to obtain voxel morphology centroid data. Specifically, the coordinates of the spatial position of each voxel within each candidate voxel set of subclonal structures are determined, including the coordinates of the voxel in the lateral, longitudinal, and height directions. The lateral coordinates of all voxels within each candidate voxel set are summed and divided by the total number of voxels in each set to obtain the lateral morphological centroid coordinates. Similarly, the coordinates in the longitudinal and height directions are summed and divided by the total number of voxels in each candidate voxel set to obtain the lateral and height morphological centroid coordinates. The lateral, longitudinal, and height morphological centroid coordinates are combined to obtain the voxel morphological centroid data of the candidate voxel set. The voxel morphological centroid data represents the central position of the candidate voxel set within the lesion region.

[0024] The spatial correspondence between candidate voxel sets of subclonal structures at different time points was determined by voxel morphological centroid data, forming voxel set association pairs of subclonal structures. Specifically, different time points refer to image data collected and processed from the same lymphoma patient within consecutive time intervals, such as different time points between the initial medical examination and subsequent medical examinations. When determining spatial correspondence, firstly, the voxel morphology centroid data of the candidate voxel sets of subclonal structures obtained at each time point is calculated. Then, the voxel morphology centroid data obtained at different time points are matched one-to-one: the three-dimensional Euclidean distance between the voxel morphology centroids of each candidate voxel set of subclonal structures at different time points is calculated in the spatial coordinate system. Next, a spatial position matching threshold is preset. If the three-dimensional Euclidean distance between the voxel morphology centroids of two candidate voxel sets of subclonal structures at two different time points is less than the preset spatial position matching threshold, then the two candidate voxel sets of subclonal structures are determined to be corresponding and combined to form a subclonal structure voxel set association pair. The subclonal structure voxel set association pair reflects the spatial positional relationship of the candidate voxel sets of subclonal structures at different time points.

[0025] Calculate the relative change rate of the number of endovoxels associated with each subclonal structure voxel set to obtain the subclonal structure amplification rate parameter. Specifically, the number of voxels in each subclonal structure voxel set at different time points within the subclonal structure voxel set association pair is counted. The number of voxels at the next time point in the subclonal structure voxel set association pair is subtracted from the number of voxels at the previous time point to obtain the voxel number difference. The voxel number difference is then divided by the number of voxels at the previous time point to obtain the relative change rate of the voxel number. The relative change rate reflects the increase or decrease of the number of voxels in the subclonal structure between different time points, that is, it reflects the trend of the subclonal structure amplification rate, and is called the subclonal structure amplification rate parameter. The amplification rate parameter is calculated separately within each subclonal structure voxel set association pair. The obtained amplification rate parameter can be used to describe the dynamic change trend of the subclonal structure inside the lesion.

[0026] Arrange the amplification rate parameters of subclonal structures in chronological order to form amplification trend characteristic data among different subclonal structures; Specifically, all subclonal structure amplification rate parameters are sorted according to their corresponding time points. A trend relationship is established with time points as the x-axis and subclonal structure amplification rate parameters as the y-axis, reflecting the dynamic changes in the amplification of each subclonal structure within the lesion region over continuous time intervals. For example, if the amplification rate of the subclonal structure voxel set continuously increases at multiple consecutive time points, it indicates that the number of subclonal structure voxels is gradually increasing; if the amplification rate of the subclonal structure voxel set continuously decreases, it indicates that the number of subclonal structure voxels tends to stabilize or decrease.

[0027] S4. Based on amplification trend characteristic data and spatial phase structure characteristic data, a dynamic stability analysis is performed on the competitive relationship between different subclonal structures to obtain subclonal competition dynamic stability data, including: Spatial phase structure feature data and subclonal structure amplification trend feature data are synchronously paired to construct a time-series feature matrix; Specifically, based on time tags corresponding to multiple medical tests for each lymphoma patient, spatial phase structure feature data and subclonal structure amplification trend feature data are extracted for each time tag. The spatial phase structure feature data is represented as a numerical matrix of spatial phase differences between multiple subclonal structures, with each value representing the degree of phase difference between two subclonal structures within a spatial region. The subclonal structure amplification trend feature data is represented as the relative change rate trend of the number of voxels in multiple subclonal structures. Through time tag matching, the spatial phase structure feature data and subclonal structure amplification trend feature data corresponding to the same time point are combined and paired. For example, the spatial phase difference values ​​and amplification trend feature data values ​​of two subclonal structures under a specific time tag are combined accordingly. After synchronous pairing, multiple consecutive time tag data combinations are used as a set of data. Each set of data contains spatial phase structure feature values ​​and subclonal structure amplification trend feature values. The data combinations of multiple consecutive time tags are arranged sequentially to form a time-series feature matrix with spatial phase structure feature values ​​and subclonal structure amplification trend feature values ​​as elements. The time-series feature matrix reflects the dynamic correlation between the spatial phase features and voxel amplification trends of subclonal structures in residual lymphoma lesions at multiple time points.

[0028] Singular spectral decomposition is performed on the time-series feature matrix to extract nonlinear dynamic eigenmodes, resulting in a competitive dynamic eigenmode sequence. Specifically, the temporal feature matrix is ​​reconstructed by constructing a covariance matrix, which transforms the variation information of elements in the temporal feature matrix under different time labels into matrix covariance features. The eigenvalues ​​and corresponding eigenvectors of the covariance matrix are calculated. The eigenvalues ​​are obtained by solving the matrix's eigenvalue equations, and the eigenvectors represent the matrix distribution direction corresponding to the eigenvalues. Based on the eigenvalues ​​and eigenvectors of the covariance matrix, the eigenvalues ​​are sorted from largest to smallest, and several representative eigenvalues ​​and their corresponding eigenvectors are selected to form the intrinsic modes after matrix decomposition. Each intrinsic mode is then reconstructed in reverse, mapping the intrinsic modes from the covariance matrix space back to the original temporal feature matrix space through the correspondence between eigenvectors and eigenvalues, forming multiple nonlinear dynamic intrinsic modes. These multiple intrinsic modes are then arranged sequentially according to the eigenvalue sorting relationship, forming an intrinsic mode sequence that reflects the competitive dynamic relationship between different subclonal structures. This competitive dynamic intrinsic mode sequence can reflect the competitive dynamic relationship between different subclonal structures within residual lymphoma lesions in terms of spatial location, phase structure, and amplification trend changes over continuous time intervals.

[0029] Calculate the local Lyapunov exponent sequence and spectral entropy sequence of the competitive dynamic intrinsic mode sequence to form a set of competitive dynamic instability indices; Specifically, phase space reconstruction is performed for each competing dynamic eigenmode. This involves: determining the distance relationship between adjacent states in the dynamic space; calculating the evolution trend of the neighborhood of each phase space state point in the reconstructed phase space; obtaining the local Lyapunov exponent for each state point by calculating the logarithmic average of the rate of change of distance between adjacent state points over time; and then arranging all local Lyapunov exponents sequentially according to the time series to obtain the local Lyapunov exponent sequence, which reflects the changing trend of the influence of local small perturbations on state evolution in the dynamic competition between subclones. The spectral entropy sequence is calculated as follows: for competing dynamic eigenmodes... The modes undergo Fast Fourier Transform (FFT) to obtain spectral data, which reflects the energy distribution of the intrinsic modes in the frequency domain. The energy proportions of different frequency components in the spectral data are then normalized probabilities. The logarithm of each frequency component's probability is multiplied by its own probability, and the sum of these products is taken, yielding the spectral entropy value for each intrinsic mode. The spectral entropies corresponding to multiple intrinsic modes are arranged sequentially according to time series to form a spectral entropy sequence. This sequence reflects the complexity of the competitive dynamic intrinsic mode sequence in the frequency domain. Combining the local Lyapunov exponent sequence with the spectral entropy sequence forms a set of competitive dynamic instability indices that reflect the degree of instability in the competitive relationship.

[0030] Aggregate the set of competition dynamic instability indicators to generate competition dynamic stability curves; Specifically, the local Lyapunov index sequence and spectral entropy sequence of the competitive dynamic instability index are comprehensively calculated. This involves performing nonlinear joint operations on the local Lyapunov index sequence and spectral entropy sequence, for example, unifying the two sequence spaces through exponential mapping; then, through nonlinear mapping, the data values ​​at corresponding positions of the two sequences are fused to obtain an aggregated numerical sequence that reflects the overall instability; the aggregated numerical sequence is then plotted with continuous time labels as the abscissa to show the competitive dynamic stability curve over time; the competitive dynamic stability curve reflects the comprehensive trend of the stability of the competitive relationship between subclonal structures within the residual lymphoma lesion over time.

[0031] Time labels at the crossover points of stability thresholds are extracted from the competitive dynamic stability curve to form subclone competitive dynamic stability data; Specifically, a stability threshold is preset along the vertical axis of the competitive dynamic stability curve; the intersection between the competitive dynamic stability curve and the preset stability threshold is determined, specifically: the intersection point on the competitive dynamic stability curve where the stability curve changes from below the preset stability threshold to above the preset stability threshold or vice versa is located; then the horizontal coordinate corresponding to the intersection point, i.e., the time label, is extracted to form stability threshold intersection point time label data; the extracted stability threshold intersection point time label data reflects the time position information of the subclonal structure's competitive relationship changing from a stable state to an unstable state or from an unstable state to a stable state.

[0032] S5, based on subclonal competition dynamic stability data, analyzes the heterogeneous diffusion relationships among different subclonal structures and generates heterogeneous diffusion risk assessment data, including: The subclonal competition dynamic stability data and spatial phase structure feature data are matched according to time labels to construct a heterogeneity diffusion assessment matrix; Specifically, subclonal competition dynamic stability data reflects the stable and unstable transition times and trends of the competitive relationship between different subclonal structures over time; spatial phase structural feature data reflects the spatial positional relationship and spatial phase change patterns among subclonal structures within the lesion region; the subclonal competition dynamic stability data and spatial phase structural feature data are matched with time labels as the horizontal axis, forming a pair of spatial phase structural feature data and corresponding subclonal competition dynamic stability data under each time label, and the paired data are arranged sequentially to form a heterogeneity diffusion assessment matrix; each row of the heterogeneity diffusion assessment matrix includes the spatial phase feature value and competition dynamic stability value of each subclonal structure under the time label; the heterogeneity diffusion assessment matrix can describe the spatial heterogeneity and competitive dynamic evolution state of each subclonal structure within the lesion over time.

[0033] The diffusion probability matrix is ​​obtained by using a random walk graph traversal algorithm based on the heterogeneous diffusion evaluation matrix to calculate the diffusion probability between each subclone structure. Specifically, the heterogeneity diffusion assessment matrix is ​​used as the input to the graph structure. Each subclonal structure corresponds to a graph node, and the connection weights between nodes in the graph are jointly determined by each set of spatial phase structural feature data and subclonal competition dynamic stability data. Starting from each subclonal structure, a random walk is performed according to the connection weights between nodes, traversing all nodes with a finite step size. The diffusion probability between each pair of subclonal structures is calculated by statistically analyzing the total number of traversals from one node to other nodes and the total path probability. The diffusion probability reflects the relative likelihood of a subclonal structure spreading to other subclonal structures under the influence of lesion spatial distribution and dynamic evolution. The diffusion probabilities between all pairs of subclonal structures are arranged in a matrix to form a diffusion probability matrix. Each element of the diffusion probability matrix reflects the dynamic spatial diffusion capability between specific subclonal structures; a higher diffusion probability indicates a greater diffusion risk.

[0034] Perform spectral clustering on the diffusion probability matrix to extract a set of heterogeneous sub-clusters; Specifically, by calculating the eigenvalues ​​and eigenvectors of the diffusion probability matrix, the nodes in the diffusion probability matrix are divided into several sub-clusters according to their diffusion probabilities. The subclonal structures within each sub-cluster are relatively close in the diffusion probability space, indicating that there are spatial and dynamic competitive connections between the subclonal structures. The Laplace matrix of the diffusion probability matrix is ​​constructed, the first few eigenvectors are calculated, the eigenvectors are projected onto a low-dimensional space, and a clustering algorithm is used to classify the eigenvectors, thereby obtaining a set of heterogeneous sub-clusters. The set of heterogeneous sub-clusters is used to reflect the aggregation and separation of subclonal structures within the lesion region during spatial diffusion and dynamic competitive evolution.

[0035] Calculate the maximum diffusion intensity parameter, average diffusion intensity parameter, and diffusion intensity variance parameter of the heterogeneous sub-cluster set to generate a heterogeneous diffusion index set; Specifically, within each heterogeneous sub-cluster set, the diffusion probabilities between all subclones are statistically analyzed, and the maximum value is taken as the maximum diffusion intensity parameter of the sub-cluster. The diffusion probabilities between all subclones within the same sub-cluster are summed and divided by the square of the number of subclones in the sub-cluster to obtain the average diffusion intensity parameter. The sum of the squares of the differences between the diffusion probabilities between each subclone within the sub-cluster and the average diffusion intensity parameter is calculated and then divided by the square of the number of subclones in the sub-cluster to obtain the variance parameter. The maximum diffusion intensity parameter, the average diffusion intensity parameter, and the diffusion intensity variance parameter reflect the maximum, overall level, and distribution dispersion of the diffusion capacity between subclones within the heterogeneous sub-cluster. The maximum diffusion intensity parameter, the average diffusion intensity parameter, and the diffusion intensity variance parameter of all heterogeneous sub-cluster sets are arranged sequentially according to the time label and the sub-cluster index number to form a heterogeneous diffusion index set.

[0036] The heterogeneity diffusion index set is transformed into heterogeneity diffusion risk assessment data based on a preset threshold judgment strategy. Specifically, threshold ranges are preset for the maximum diffusion intensity parameter, average diffusion intensity parameter, and diffusion intensity variance parameter. For each heterogeneous sub-cluster set, if any one of the maximum diffusion intensity parameter, average diffusion intensity parameter, or diffusion intensity variance parameter exceeds the corresponding preset threshold, the diffusion risk of the heterogeneous sub-cluster set is considered to be high-risk, and the sub-cluster set is recorded as high-risk for the current time label. If the maximum diffusion intensity parameter, average diffusion intensity parameter, or diffusion intensity variance parameter is all below the preset threshold, the diffusion risk of the heterogeneous sub-cluster set is considered to be low-risk, and it is recorded as low-risk; otherwise, it is recorded as medium-risk. Through the above determination method, the risk levels of all heterogeneous sub-cluster sets are sequentially marked under all time labels, ultimately forming heterogeneous diffusion risk assessment data.

[0037] S6, based on heterogeneous spread risk assessment data, dynamically updates the progression monitoring indicators for residual lymphoma lesions and outputs dynamic monitoring results, including: Transform heterogeneous diffusion risk assessment data into progress monitoring indicators to update the input vector; Specifically, the heterogeneity diffusion risk assessment data is represented by the diffusion risk levels of subclonal structures under multiple time labels, including high-risk, medium-risk, and low-risk levels. Each risk level is converted into numerical data according to a predefined mapping relationship. During the conversion, the highest value corresponds to the high-risk level, the lowest value corresponds to the low-risk level, and the value corresponding to the medium-risk level is in the middle. The converted risk levels are arranged according to the sequential combination of each heterogeneous sub-cluster under different time labels, and are used as elements of a vector to form the progress monitoring indicator update input vector. The progress monitoring indicator update input vector can reflect the trend of diffusion risk changes of subclonal structures over time.

[0038] Load the progress monitoring indicator library and initialize the progress monitoring indicator set; Specifically, the progression monitoring indicator library pre-stores various progression monitoring indicators needed for the dynamic evolution monitoring of lymphoma lesions, including multidimensional indicator types such as lesion spatial structure characteristics, subclonal structure competition dynamic characteristics, spread risk characteristics, and amplification trend characteristics. When the progression monitoring indicator set is initialized, each indicator is set with an initial value. The initial value is obtained through the analysis of the patient's first medical examination and multimodal medical test data, which is used to represent the baseline monitoring status of each subclonal structure in the lesion area in the initial state of the patient, forming a benchmark reference.

[0039] Based on the progress monitoring indicator update input vector, hierarchical recursive decision rule operation is performed to update the progress monitoring indicator set, resulting in the updated progress monitoring indicator set. Specifically, the hierarchical recursive decision rule operation updates the progress monitoring indicator set step by step in a hierarchical manner, including multiple recursive levels. Each recursive level corresponds to a specific type of progress monitoring indicator for updating. First, each value is extracted sequentially from the progress monitoring indicator update input vector and compared with the corresponding indicator in the current progress monitoring indicator set. Specifically: if the value in the input vector is higher than the corresponding indicator in the current indicator set, the corresponding progress monitoring indicator in the current indicator set is updated to the value in the input vector; if the value in the input vector is lower than the corresponding indicator value in the current indicator set, the corresponding indicator in the current progress monitoring indicator set remains unchanged; if the value in the input vector is equal to the current indicator value... For each indicator in the target set, the decision-making rules at the next level are used to recursively determine whether to update it. After each recursive decision, the overall indicator state of the updated progress monitoring indicator set is recalculated as the initial state for the next recursive decision-making operation. This step-by-step recursive update ensures that the changes between indicators at each level of the progress monitoring indicator set are consistent and logically coherent during the update process, avoiding potential inconsistencies and logical conflicts between indicators. After all levels of recursive decision-making rules are completed, an updated progress monitoring indicator set is formed. The updated progress monitoring indicator set reflects the latest changes in the subclonal structure of the patient's lesion area in terms of spatial location, competitive dynamics, spread risk, and amplification trend.

[0040] The updated set of progress monitoring indicators is merged with the corresponding time tags to generate dynamic monitoring result data. Specifically, the updated set of progression monitoring indicators, after being processed by hierarchical recursive decision rules, represents the latest progression status of each subclonal structure in the patient's lesion region. To accurately track and assess the dynamic changes in the patient's lesions, the updated set of progression monitoring indicators is merged with the corresponding medical test time labels. Specifically, the indicator set corresponding to each time label is paired with the time label to form multiple sets of time-indicator set correspondences, which are called dynamic monitoring result data. Each set of dynamic monitoring result data includes a medical test time label and all the latest progression monitoring indicators of subclonal structures in the patient's lymphoma lesion region under the time label. This realizes the historical record and real-time tracking and monitoring of the patient's lesion evolution status, which can be used by clinicians to judge the evolution trend of the patient's condition.

[0041] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0042] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0043] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0044] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0045] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0046] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0047] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] 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.

[0049] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for dynamic monitoring of residual lymphoma lesions, characterized in that, Includes the following steps: S1. Collect raw data of multimodal medical tests from lymphoma patients, and preprocess the raw data of multimodal medical tests to generate a multimodal medical test dataset. S2, based on a multimodal medical detection dataset, constructs a phase spatial distribution model of residual lymphoma lesions and extracts spatial phase structure feature data between different subclonal structures within residual lymphoma lesions; S3, based on spatial phase structure feature data, identifies the subclonal structure features inside residual lymphoma lesions and generates amplification trend feature data between different subclonal structures; S4. Based on amplification trend characteristic data and spatial phase structure characteristic data, dynamic stability analysis is performed on the competitive relationship between different subclonal structures to obtain subclonal competition dynamic stability data. S5, based on subclonal competition dynamic stability data, analyzes the heterogeneity diffusion relationship between different subclonal structures and generates heterogeneity diffusion risk assessment data; S6, based on heterogeneous spread risk assessment data, dynamically updates the progression monitoring indicators of residual lymphoma lesions and outputs dynamic monitoring results.

2. The method for dynamic monitoring of residual lymphoma lesions according to claim 1, characterized in that, S1, specifically: Magnetic resonance imaging data, positron emission tomography (PET) scan data, and computed tomography (CT) image data of lymphoma patients were collected as raw data for multimodal medical testing. The raw data of multimodal medical testing are processed by data format conversion, spatial registration, noise removal and data normalization to generate a multimodal medical testing dataset.

3. The method for dynamic monitoring of residual lymphoma lesions according to claim 2, characterized in that, S2, specifically: The multimodal medical testing dataset is rigidly and non-rigidly registered in three-dimensional space to obtain three-dimensional registered image data in a unified coordinate system. Based on three-dimensional registered image data in a unified coordinate system, an image segmentation algorithm was used to identify the residual lymphoma lesion area and extract the voxel gray value of the lesion area. Phase spectrum evaluation is performed on the voxel gray values ​​of the lesion area to generate phase value data for each voxel; A phase spatial distribution model of residual lymphoma lesions was constructed based on phase value data; Based on the phase spatial distribution model of residual lymphoma lesions, the phase difference matrix between voxel sets corresponding to different subclonal structures is calculated, and spatial phase structure feature data is output.

4. The method for dynamic monitoring of residual lymphoma lesions according to claim 3, characterized in that, S3, specifically: Density clustering was performed on the spatial phase structure feature data, and the candidate voxel set of subclone structure was divided according to the phase difference threshold between voxels and the three-dimensional Euclidean distance threshold. Voxel morphology centroid coordinates are calculated based on the candidate voxel set of subcloned structures to obtain voxel morphology centroid data. The spatial correspondence between candidate voxel sets of subclonal structures at different time points was determined by voxel morphological centroid data, forming voxel set association pairs of subclonal structures. Calculate the relative change rate of the number of endovoxels associated with each subclonal structure voxel set to obtain the subclonal structure amplification rate parameter. The amplification rate parameters of the subclonal structures are arranged in chronological order to form amplification trend characteristic data among different subclonal structures.

5. The method for dynamic monitoring of residual lymphoma lesions according to claim 4, characterized in that, S4, specifically: Spatial phase structure feature data and subclonal structure amplification trend feature data are synchronously paired to construct a time-series feature matrix; Singular spectral decomposition is performed on the time-series feature matrix to extract nonlinear dynamic eigenmodes, resulting in a competitive dynamic eigenmode sequence. Calculate the local Lyapunov exponent sequence and spectral entropy sequence of the competitive dynamic intrinsic mode sequence to form a set of competitive dynamic instability indices; Aggregate the set of competition dynamic instability indicators to generate competition dynamic stability curves; The time labels of the stability threshold crossover points are extracted from the competitive dynamic stability curve to form subclone competitive dynamic stability data.

6. The method for dynamic monitoring of residual lymphoma lesions according to claim 5, characterized in that, S5, specifically: The subclonal competition dynamic stability data and spatial phase structure feature data are matched according to time labels to construct a heterogeneity diffusion assessment matrix; The diffusion probability matrix is ​​obtained by using a random walk graph traversal algorithm based on the heterogeneous diffusion evaluation matrix to calculate the diffusion probability between each subclone structure. Perform spectral clustering on the diffusion probability matrix to extract a set of heterogeneous sub-clusters; Calculate the maximum diffusion intensity parameter, average diffusion intensity parameter, and diffusion intensity variance parameter of the heterogeneous sub-cluster set to generate a heterogeneous diffusion index set; The heterogeneity diffusion index set is transformed into heterogeneity diffusion risk assessment data based on a preset threshold determination strategy.

7. The method for dynamic monitoring of residual lymphoma lesions according to claim 6, characterized in that, S6, specifically: Transform heterogeneous diffusion risk assessment data into progress monitoring indicators to update the input vector; Load the progress monitoring indicator library and initialize the progress monitoring indicator set; Based on the progress monitoring indicator update input vector, hierarchical recursive decision rule operation is performed to update the progress monitoring indicator set, resulting in the updated progress monitoring indicator set. The updated set of progress monitoring indicators is merged with the corresponding time tags to generate dynamic monitoring result data.

Images