A method, device, equipment, storage medium and product for evaluating brain age based on deep learning intelligent analysis of brain myelin development

Abstract

The application discloses a method and device for evaluating brain age based on deep learning intelligent analysis of brain myelin development, equipment, storage medium and products, relates to the technical field of medical image processing, and the method comprises the following steps: acquiring a brain multi-modal magnetic resonance image; preprocessing the brain multi-modal magnetic resonance image to obtain a target brain image; measuring the correlation of the target brain image and the standardized features of each age template in a preset continuous age brain template library; and selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result based on the calculated correlation measurement results. When processing the brain multi-modal magnetic resonance image, the data inconsistency is solved by registration to the continuous age template space, thereby realizing accurate brain age prediction.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method, device, equipment, storage medium and product for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development. Background Technology

[0002] In pediatric clinical practice and research, accurately assessing the brain age of infants and young children is crucial for the early detection of developmental delays, neurological disorders, and for tracking the effectiveness of interventions. Currently, brain age prediction primarily relies on medical imaging analysis. Traditional methods often depend on physicians' experience to visually assess brain MRI images. This approach is highly subjective, has low repeatability, and is difficult to quantify and standardize. With the development of medical imaging technology, some statistical or machine learning models based on voxel morphology, brain region volume, and other features have emerged. However, these methods typically rely on single-modal imaging, lack sufficient utilization of multimodal information, and are poorly adapted to the rapidly changing brain development patterns of infants and young children.

[0003] In recent years, deep learning models have demonstrated good performance in brain age prediction. However, their "black box" nature leads to a lack of interpretability in the prediction results, making it difficult to gain the full trust of clinicians. Furthermore, the performance of these models heavily relies on large-scale, high-quality labeled data, resulting in limited generalization ability in data-scarce scenarios. Existing methods often use static templates or single-age-group templates for registration and comparison, leading to large registration errors and poor feature comparability, which in turn affects prediction accuracy and robustness. Summary of the Invention

[0004] The main purpose of this application is to provide a method, device, equipment, storage medium and product for assessing brain age based on deep learning intelligent analysis of myelin sheath development, aiming to solve the technical problem of poor accuracy in brain age prediction.

[0005] To achieve the above objectives, this application proposes a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development. The method includes:

[0006] Acquire multimodal magnetic resonance imaging of the brain; The multimodal magnetic resonance imaging of the brain is preprocessed to obtain the target brain image; The target brain image is correlated with the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0007] In one embodiment, the step of preprocessing the multimodal magnetic resonance imaging of the brain to obtain the target brain image includes: The brain multimodal magnetic resonance images were subjected to decranialization to obtain brain tissue images and corresponding brain tissue masks. Morphological operations are performed on the brain tissue mask to fill holes and ensure the connectivity and integrity of the mask, resulting in an optimized brain mask; Based on the optimized brain mask, precise brain region voxel data are extracted from the brain tissue image; The voxel data of the brain regions from different modalities are registered intermodally so that the image spaces of all modalities are aligned to the same reference modal space. The spatially aligned brain region voxel data are registered to the standard space corresponding to the preset continuous age brain template library through a hierarchical registration process that includes affine transformation and nonlinear deformation field, and the complete spatial transformation parameters are saved. The images registered to the standard space are resampled to generate the final target brain image used for matching.

[0008] In one embodiment, before the step of registering the spatially aligned brain region voxel data to the standard space corresponding to the preset continuous-age brain template library through a hierarchical registration process including affine transformation and nonlinear deformation field, and saving the complete spatial transformation parameters, the following steps are included: From the continuous age brain template library, obtain predefined brain region atlases related to white matter development; Using the saved spatial transformation parameters, the predefined brain region atlas is inversely mapped to the individual space of the target brain image to obtain individualized brain region labels; Based on the individualized brain region labels, brain region feature masks are generated for subsequent correlation measurement, wherein the brain region feature masks are preferably whole-brain white matter masks.

[0009] In one embodiment, the step of measuring the correlation between the target brain image and the standardized features of each age template in a preset continuous age brain template library, and selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result based on the calculated correlation measurement results includes: Within the brain region feature mask, the voxel intensity values ​​of the target brain image are extracted to obtain the first original feature vector; Within the brain region feature mask, voxel intensity values ​​of template images for each age are extracted to obtain the second original feature vector; The first original feature vector of the target brain image is subjected to standardization processing based on mean and standard deviation to obtain the first standardized feature vector; The second original feature vector of each age template is standardized based on the mean and standard deviation to obtain the second standardized feature vector; The linear correlation between the first standardized feature vector of the target brain image and the second standardized feature vector of each age template is calculated, and the strength of the calculated linear correlation is used as the correlation measurement result.

[0010] In one embodiment, the step of acquiring multimodal magnetic resonance imaging of the brain includes, prior to: A three-dimensional convolutional neural network deep learning model is constructed as a brain age prediction model, wherein the input configuration of the brain age prediction model is to receive standardized and spatially registered multi-channel three-dimensional brain images. Using an age-labeled brain image dataset, the brain age prediction model was trained end-to-end by calculating the prediction error through forward propagation and updating the parameters through back propagation until the model converged. Save the parameters of the trained brain age prediction model to obtain the trained brain age prediction model.

[0011] In one embodiment, after the step of selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result, the method further includes: The trained brain age prediction model is invoked to process the target brain image and output an auxiliary brain age prediction value. Calculate the optimal relevance measure corresponding to the brain age prediction result and determine it as the first confidence level; Calculate the model confidence level corresponding to the assisted brain age prediction value and determine it as the second confidence level; Based on the first confidence level and the second confidence level, the brain age prediction result and the auxiliary brain age prediction value are weighted and fused to generate the final brain age prediction value.

[0012] Furthermore, to achieve the above objectives, this application also proposes a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development, wherein the device comprises: The acquisition module is used to acquire multimodal magnetic resonance imaging of the brain; The preprocessing module is used to preprocess the multimodal magnetic resonance images of the brain to obtain the target brain image; The prediction module is used to measure the correlation between the target brain image and the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0013] Furthermore, to achieve the above objectives, this application also proposes a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described above.

[0015] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development as described above.

[0016] One or more technical solutions proposed in this application have at least the following technical effects: Compared to related technologies that rely on static templates, single modalities, or lack fine-grained processing of the continuity and structural heterogeneity of infant brain development, resulting in poor accuracy in brain age prediction, this application acquires multimodal magnetic resonance imaging (MRI) images of the brain; preprocesses the multimodal MRI images to obtain target brain images; and performs correlation measurement between the target brain images and the standardized features of each age template in a pre-set continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result. It is understood that this application employs a technical solution that integrates systematic preprocessing, continuous dynamic template matching, and multimodal feature standardization measurement. When processing multimodal magnetic resonance imaging (MRI) images of infants' brains, craniotomy, multimodal registration, and hierarchical standardized registration to a continuous age template space are employed. This transforms heterogeneous clinical imaging data into target brain images with high spatial and intensity comparability to a standard template library, fundamentally addressing the impact of data inconsistency on accuracy. Furthermore, by measuring the correlation between the target brain image and the standardized features of each template in a continuous age-based brain template library covering the entire developmental cycle, point-by-point, quantitative pattern comparison with the entire developmental process is achieved, overcoming the significant errors caused by using a single static template that cannot match rapid development. Finally, by selecting the template age with the highest correlation measurement result as the prediction result, the developmental reference point most similar to the target brain image pattern is essentially found, thus achieving accurate, objective, and anatomically significant brain age prediction. Based on systematic, continuous, and quantitative techniques, high-precision and robust quantitative assessment of infants' brain development status can be achieved, thereby determining the objective and interpretable physiological developmental age that best matches individual brain imaging characteristics, ultimately completing accurate brain age prediction and effectively solving the problem of insufficient prediction accuracy in existing technologies. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development according to this application. Figure 2This is a flowchart illustrating a second embodiment of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development according to this application. Figure 3 This is a flowchart illustrating Embodiment 3 of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development according to this application. Figure 4 This is a flowchart illustrating Embodiment 4 of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development according to this application; Figure 5 This is a flowchart illustrating Embodiment 5 of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development according to this application; Figure 6 This is a flowchart illustrating Embodiment Six of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development according to this application. Figure 7 This is a schematic diagram of the module structure of a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development, according to an embodiment of this application. Figure 8 This is a schematic diagram of the hardware operating environment involved in a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development, as described in an embodiment of this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] The main solution in this application embodiment is: Acquire multimodal magnetic resonance imaging of the brain; The multimodal magnetic resonance imaging of the brain is preprocessed to obtain the target brain image; The target brain image is correlated with the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0024] In this embodiment, this application uses a device for assessing brain age by analyzing myelin sheath development based on deep learning as the main execution subject. For ease of description, it will be referred to as "device" in detail below.

[0025] Existing technologies rely on static templates, single modalities, or lack fine processing of the continuity and structural heterogeneity of infant brain development, resulting in poor accuracy in predicting brain age.

[0026] This application provides a solution employing a technical approach that integrates systematic preprocessing, continuous dynamic template matching, and multimodal feature standardization measurement. When processing multimodal magnetic resonance imaging (MRI) images of infants' brains, craniotomy, multimodal registration, and hierarchical standardized registration to a continuous age template space transform heterogeneous clinical imaging data into target brain images with high spatial and intensity comparability to a standard template library, thus resolving the fundamental impact of data inconsistency on accuracy. Furthermore, by measuring the correlation between the target brain image and the standardized features of each template in a continuous age-based brain template library covering the entire developmental cycle, point-by-point, quantitative pattern comparison with the entire developmental process is achieved, overcoming the significant errors caused by the inability to match rapid development using a single static template. Finally, by selecting the template age with the highest correlation measurement result as the prediction result, essentially finding the developmental reference point most similar to the target brain image pattern, thereby achieving accurate, objective, and anatomically significant brain age prediction. Based on systematic, continuous, and quantitative technical means, it is possible to achieve high-precision and robust quantitative assessment of the brain development status of infants and young children, thereby determining the objective and interpretable physiological developmental age that best matches the individual's brain imaging characteristics, and finally completing the accurate prediction of brain age, effectively solving the problem of insufficient prediction accuracy of existing technologies.

[0027] Based on this, embodiments of this application provide a method for assessing brain age by intelligent analysis of myelin sheath development using deep learning, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development according to this application.

[0028] In this embodiment, the method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes steps S10-S30: Step S10: Acquire multimodal magnetic resonance imaging of the brain; It should be noted that multimodal magnetic resonance imaging of the brain refers to multiple sequences of image data reflecting different tissue characteristics obtained by scanning the brain of the same subject using magnetic resonance imaging equipment. These include at least T1-weighted and T2-weighted images, and may also include, but are not limited to, fluid attenuation inversion recovery sequences and diffusion-weighted imaging. T1-weighted images are primarily useful for observing anatomical structures, while T2-weighted images are more sensitive to changes in tissue water content.

[0029] Understandably, this step aims to collect foundational imaging data for brain age analysis. By acquiring multimodal images, the complementary information provided by different sequences can be comprehensively utilized, providing a comprehensive data foundation for subsequent fusion analysis. This overcomes the limitations of single-modal information and creates conditions for improving the robustness and accuracy of the prediction model.

[0030] Step S20: Preprocess the multimodal magnetic resonance imaging of the brain to obtain the target brain image; It should be noted that preprocessing refers to a series of standardized operations performed to ensure that the raw image data meets the requirements for subsequent quantitative analysis and comparison. Target brain images refer to image data that, after preprocessing, are located in a uniform standard space, have calibrated intensity features, and primarily contain information about brain tissue.

[0031] Understandably, this step is a crucial one in standardizing the original heterogeneous image data. By performing operations such as craniotomy, registration, spatial standardization, and intensity correction on the images, the non-biological variations caused by differences in different scanning equipment, parameters, and individual anatomy can be effectively eliminated. This makes the image data of all samples comparable in both spatial and intensity dimensions, laying a reliable foundation for subsequent accurate feature extraction and quantitative comparison, and is a prerequisite for ensuring the repeatability and accuracy of the entire method.

[0032] Step S30: The target brain image is correlated with the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0033] It should be noted that the continuous age brain template library is a collection of images representing standard brain anatomy at different age points (e.g., at monthly or daily intervals), covering the entire age range of the target prediction. Standardized features refer to image feature vectors extracted within a specific brain region mask after intensity normalization. Correlation measurement is a calculation process used to quantify the degree of linear correlation between two feature vectors.

[0034] Understandably, this step is the core of this method for achieving interpretable predictions. By performing age-point pattern matching and correlation calculations between individual target images and a continuous age template library, a reference template most similar to the individual's current brain developmental state can be found. Essentially, this method transforms the brain age prediction problem into finding the optimal matching point along a continuous developmental trajectory. The prediction results have clear physical meaning; that is, the predicted age is the age of the reference template with the most similar image pattern. This provides intuitive interpretability lacking in traditional black-box models, while directly utilizing continuous prior information from the entire developmental process, which helps improve prediction accuracy.

[0035] For example, refer to Figure 2 , Figure 2 This outlines the main workflow of the method. The "Preprocessing" section at the top includes "skull removal" and "template space registration." The "Age Prediction" section at the bottom focuses on "similarity calculation" and "prediction results (taking the average of T1 and T2)." This figure highlights the key stages from the original image to the predicted result.

[0036] For example, such as Figure 3 The detailed, systematic processing pipeline shown in this application describes a complete implementation flow of a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development: The entire process begins with format standardization and de-craniography of the acquired raw cranial MRI images to obtain clean brain tissue images. Subsequently, the process enters the multimodal spatial integration stage. First, the T2-weighted images are precisely registered to the T1-weighted image space of the same subject, achieving anatomical alignment between different image sequences. Then, using this aligned T1-weighted image as a reference, it is further mapped to a standard space defined by a pre-defined continuous age-based brain template library covering the target age range through a hierarchical registration strategy incorporating affine and nonlinear transformations. This process simultaneously generates and saves complete spatial transformation parameters from the individual image space to the standard template space.

[0037] While completing spatial standardization and obtaining the target brain image, the system uses the aforementioned saved transformation parameters to inversely map the predefined brain region atlases (e.g., atlases highlighting white matter regions) attached to the template library back to the individual's original image space, thereby obtaining brain region segmentation labels that precisely correspond to the individual's anatomical structure. Based on these individualized labels, specific brain region masks can be generated for subsequent quantitative analysis.

[0038] The core prediction process employs a strategy of independently analyzing and then fusing the T1 and T2 modalities. For each modality's target brain image, the system extracts voxel intensity features within the generated brain region mask (preferably a whole-brain white matter mask) and standardizes them based on the mean and standard deviation to eliminate global intensity differences. Next, the linear correlation between this standardized feature and the standardized features of each age template in the continuous age template library is calculated. By traversing the entire template library, similarity sequences between T1 and T2 images and each age template can be obtained, and the age corresponding to the template with the highest similarity is selected as the preliminary prediction result for that modality. Simultaneously, developmental curves can be plotted.

[0039] Finally, the system fuses the preliminary prediction results obtained from the T1 and T2 modalities (e.g., calculates the average value) and integrates the volume indicators of each region calculated from individualized brain region labels to generate a comprehensive report that includes the final predicted brain age, detailed quantitative data of brain structure, and visualized developmental curves. Figure 3 It clearly demonstrates the fully automated, end-to-end processing logic, from raw data input, through systematic preprocessing and standardization, parallel feature matching analysis, to result fusion and output.

[0040] This embodiment provides a method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development. It employs a technical solution integrating systematic preprocessing, continuous dynamic template matching, and standardized multimodal feature measurement. When processing multimodal magnetic resonance imaging (MRI) images of infants' brains, craniotomy, multimodal registration, and hierarchical standardized registration to a continuous age template space transform heterogeneous clinical imaging data into target brain images with high spatial and intensity comparability to a standard template library, thus resolving the fundamental impact of data inconsistency on accuracy. Furthermore, by measuring the correlation between the target brain image and the standardized features of each template in a continuous age brain template library covering the entire developmental cycle, point-by-point, quantitative pattern comparison with the entire developmental process is achieved, overcoming the significant errors caused by using a single static template that cannot match rapid development. Finally, by selecting the template age with the highest correlation measurement result as the prediction result, essentially finding the developmental reference point most similar to the target brain image pattern, thus achieving accurate, objective, and anatomically significant brain age prediction. Based on systematic, continuous, and quantitative technical means, it is possible to achieve high-precision and robust quantitative assessment of the brain development status of infants and young children, thereby determining the objective and interpretable physiological developmental age that best matches the individual's brain imaging characteristics, and finally completing the accurate prediction of brain age, effectively solving the problem of insufficient prediction accuracy of existing technologies.

[0041] In one feasible implementation, the step of preprocessing the multimodal magnetic resonance imaging of the brain to obtain the target brain image includes: The brain multimodal magnetic resonance images were subjected to decranialization to obtain brain tissue images and corresponding brain tissue masks. Morphological operations are performed on the brain tissue mask to fill holes and ensure the connectivity and integrity of the mask, resulting in an optimized brain mask; Based on the optimized brain mask, precise brain region voxel data are extracted from the brain tissue image; The voxel data of the brain regions from different modalities are registered intermodally so that the image spaces of all modalities are aligned to the same reference modal space. The spatially aligned brain region voxel data are registered to the standard space corresponding to the preset continuous age brain template library through a hierarchical registration process that includes affine transformation and nonlinear deformation field, and the complete spatial transformation parameters are saved. The images registered to the standard space are resampled to generate the final target brain image used for matching.

[0042] It should be noted that descaling refers to the image segmentation process of automatically identifying and removing non-brain tissue structures such as the scalp, skull, and dura mater from raw brain images. Methods include, but are not limited to, traditional threshold-based algorithms or semantic segmentation models based on deep learning. A brain tissue mask is a binary image where voxels labeled "1" represent brain tissue regions, and voxels labeled "0" represent background or non-brain tissue regions. Morphological operations are algorithms that process image pixel sets based on shape, including but not limited to dilation, erosion, opening, and closing operations, used to optimize the topological structure of the mask. Brain region voxel data refers to brain tissue image data composed of a three-dimensional pixel array, where each pixel has an intensity value. Intermodal registration refers to the process of geometrically aligning the anatomical structures of images from different modalities originating from the same subject through spatial transformation, typically including rigid and nonlinear transformations. Layered registration is a multi-stage registration strategy that first performs a global linear transformation and then performs local nonlinear deformation. Spatial transformation parameters refer to the coefficients or fields that describe the mathematical transformation required to map points in one image space to another, including affine transformation matrices and nonlinear deformation fields. Resampling refers to the process of calculating the intensity value for each new voxel location in an image based on a new spatial coordinate grid using interpolation algorithms (such as linear interpolation or cubic spline interpolation).

[0043] Understandably, this implementation first eliminates non-brain tissue interference and ensures the integrity of brain regions through craniotomy and masking optimization, providing clean and accurately defined input for subsequent analysis. Subsequently, through two registrations within the modality and the template space, internal inconsistencies between different scan sequences and large-scale anatomical differences between individuals and standard developmental templates are gradually eliminated. This series of operations works together to transform original clinical multimodal images from diverse sources with different geometric shapes and intensity distributions into a unified and comparable standard analytical space. This fundamentally solves the problems of inconsistent feature extraction and matching errors caused by data heterogeneity, providing a crucial data foundation for subsequent accurate and reliable quantitative feature comparison and similarity measurement.

[0044] In one feasible implementation, before the step of registering the spatially aligned brain region voxel data to the standard space corresponding to the preset continuous-age brain template library through a hierarchical registration process including affine transformation and nonlinear deformation field, and saving the complete spatial transformation parameters, the following steps are included: From the continuous age brain template library, obtain predefined brain region atlases related to white matter development; Using the saved spatial transformation parameters, the predefined brain region atlas is inversely mapped to the individual space of the target brain image to obtain individualized brain region labels; Based on the individualized brain region labels, brain region feature masks are generated for subsequent correlation measurement, wherein the brain region feature masks are preferably whole-brain white matter masks.

[0045] It should be noted that predefined brain region atlases refer to data files in a standard template space that have been manually or automatically divided and labeled with different functions or anatomical structures of brain regions. Their formats include, but are not limited to, NIfTI format labeled images. Atlases related to white matter development specifically refer to atlases that highlight key white matter sub-regions or the entire white matter. Inverse mapping refers to the process of projecting information located in the standard space (such as atlases) back to the individual's original space using a forward registration transformation from individual space to standard space, through mathematical inversion or direct application of inverse transformation. Personalized brain region labels refer to the results of identifying different brain regions in the individual's own brain imaging space after inverse mapping. Brain region feature masks are three-dimensional binary images extracted from personalized brain region labels according to specific needs (such as focusing on white matter) to define the regions for subsequent calculations.

[0046] Understandably, this implementation overcomes the potential errors and inconsistencies that may arise from directly segmenting brain regions on individual images by establishing a spatial correspondence between individual images and a standard template, and then projecting the authoritative and consistent brain region definitions from the standard template back into the individual space. Prioritizing the generation of a whole-brain white matter mask is based on the crucial biological prior knowledge that myelination in infant brain development primarily occurs in white matter. This process ensures that subsequent feature extraction and similarity measurement are performed within a stable region with clear anatomical significance and high correlation to developmental progress, thereby significantly improving the specificity of features and the physiological interpretability and accuracy of the final prediction results.

[0047] In one feasible implementation, the step of performing correlation measurement between the target brain image and the standardized features of each age template in a preset continuous age brain template library, and selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result based on the calculated correlation measurement results includes: Within the brain region feature mask, the voxel intensity values ​​of the target brain image are extracted to obtain the first original feature vector; Within the brain region feature mask, voxel intensity values ​​of template images for each age are extracted to obtain the second original feature vector; The first original feature vector of the target brain image is subjected to standardization processing based on mean and standard deviation to obtain the first standardized feature vector; The second original feature vector of each age template is standardized based on the mean and standard deviation to obtain the second standardized feature vector; The linear correlation between the first standardized feature vector of the target brain image and the second standardized feature vector of each age template is calculated, and the strength of the calculated linear correlation is used as the correlation measurement result.

[0048] It should be noted that voxel intensity values ​​refer to the signal intensity values ​​corresponding to each three-dimensional pixel in a magnetic resonance image, reflecting the physical or chemical properties of the corresponding biological tissue. The first and second original feature vectors are one-dimensional arrays formed by arranging the intensity values ​​of all voxels within a defined region in a specific spatial order. Standardization based on mean and standard deviation is a data preprocessing method that subtracts the mean of each value in the feature vector and then divides it by its standard deviation. Its functions include, but are not limited to, scaling the data to a similar range and eliminating global brightness differences between different samples. The first and second standardized feature vectors are new feature vectors obtained after the above standardization process, with a mean of 0 and a standard deviation of 1. Linear correlation is a metric used to statistically evaluate the degree and direction of the linear relationship between two variable sequences.

[0049] Understandably, this implementation method ensures that the compared image information is of the same biological significance by extracting features within a predefined anatomical mask. Statistical standardization of the original feature vectors effectively eliminates non-biological intensity biases caused by differences in scanning equipment and parameters, making the feature values ​​between different images directly comparable. Finally, by calculating the linear correlation between the standardized feature vectors, the similarity between the target individual's brain image and each reference template in the overall image pattern of key regions can be objectively and quantitatively assessed. This series of operations transforms the complex image matching problem into a computable mathematical comparison problem, and its output directly and objectively reflects the proximity of developmental states, providing accurate and reliable quantitative evidence for ultimately determining the best-matching predicted age.

[0050] For example, such as Figure 4 As shown, the construction process for the standardized, labeled, continuous-age brain template library and training dataset upon which the method of this application relies can be described as follows: The process begins with the initial processing of a large-scale collection of raw MRI images of infants' brains. First, the image data is anonymized and de-identified to protect privacy, and then categorized according to the scan sequence. Subsequently, the raw DICOM format images are converted to standard analysis formats such as NIfTI, during which key metadata such as imaging parameters, slice thickness, and voxel spacing are fully parsed and preserved.

[0051] Next, craniotomy is performed on each standardized brain image. Image processing algorithms automatically identify and remove non-brain structures such as the skull to generate an initial brain tissue mask. To further ensure mask quality, morphological operations are then used to refine the mask, such as filling in holes caused by sulci or noise, thereby obtaining a complete and connected definition of the brain tissue region.

[0052] The core of the process is to construct a representative average brain template using an iterative optimization strategy. In the first stage, all craniotomized brain images undergo initial spatial registration, aligning them to an initial reference space. The aligned images are then subjected to voxel-level intensity averaging to generate the "first average template." In the second stage, using the first average template as a new reference, a higher-order nonlinear registration algorithm is employed to perform further finer spatial alignment on all individual images. A second voxel-level intensity averaging is then performed, resulting in a more accurate "second average template" with clearer anatomical structures.

[0053] To enrich the templates with anatomical semantic information, brain region labels manually annotated by experts or defined in authoritative atlases need to be mapped to the templates. By calculating the spatial transformation from the label space to the final average template, and then transforming and resampling the labels accordingly, precise registration between the labels and the templates is achieved. During this process, the mapped labels undergo boundary correction and topological logic checks to ensure their anatomical rationality. Ultimately, this process produces a standard template dataset containing continuous age points, each age point including high-quality average brain images and their corresponding precise anatomical region labels, which is the "pre-defined continuous age brain template library" mentioned in this application. This standardized dataset provides a reliable benchmark for subsequent registration, feature extraction, and quantitative comparison of individual brain images.

[0054] In one feasible implementation, the step of acquiring multimodal magnetic resonance imaging of the brain includes, prior to: A three-dimensional convolutional neural network deep learning model is constructed as a brain age prediction model, wherein the input configuration of the brain age prediction model is to receive standardized and spatially registered multi-channel three-dimensional brain images. Using an age-labeled brain image dataset, the brain age prediction model was trained end-to-end by calculating the prediction error through forward propagation and updating the parameters through back propagation until the model converged. Save the parameters of the trained brain age prediction model to obtain the trained brain age prediction model.

[0055] It's important to note that a 3D convolutional neural network (CNN) deep learning model is a deep learning model specifically designed for processing 3D volumetric data (such as medical images). Its core operations include 3D convolution and 3D pooling, which automatically extract spatial hierarchical features from the input data. End-to-end training refers to a learning method where the entire transformation process from the original input data to the final output prediction result is performed by the same model, and all parameters are adjusted at once through optimization algorithms. Model convergence refers to the point during training where the model's performance metrics (such as prediction error) no longer significantly improve on the validation set or reach a preset stable state. Model parameters refer to all adjustable values ​​such as weights and biases in the deep learning process of a neural network, which determine the specific behavior of the model.

[0056] Understandably, this implementation, by pre-training a 3D convolutional neural network deep learning model, can automatically learn complex nonlinear feature representations highly correlated with brain age from standardized multimodal images. This path is independent of the aforementioned interpretable path based on template matching. In subsequent steps, the trained model can be invoked as a parallel prediction module, and its output prediction results will complement the template matching results. This design introduces a different, data-driven reasoning capability to the entire brain age prediction system, aiming to capture subtle developmental patterns that may not be covered by explicit template matching, thereby providing a more comprehensive information foundation for the final integrated decision-making and helping to further improve the performance and generalization ability of the overall prediction system.

[0057] Reference Figure 5 The complete implementation process of the deep learning model-based assisted brain age prediction path in this application can be described in detail below.

[0058] First, the acquired raw multimodal MRI images of the brain underwent a series of preprocessing steps aimed at improving data quality. These included image denoising to improve the signal-to-noise ratio, applying off-field correction algorithms to eliminate intensity inhomogeneities generated during scanning, and performing high-precision craniectomy to obtain clean brain tissue images. These preliminary processes provided clearer and more consistent input data for subsequent analysis.

[0059] Subsequently, these pre-processed multimodal images (such as T1-weighted and T2-weighted images) are input into a standardized processing pipeline. This pipeline spatially registers the images from different modalities and aligns them to a unified standardized space, ultimately generating spatially aligned multi-channel 3D brain imaging data. This standardized data will then serve as direct input to a deep learning model.

[0060] Next, a deep learning model specifically designed for processing 3D medical images is constructed, such as a 3D convolutional neural network deep learning model. The input layer of this model is configured to receive the aforementioned normalized and spatially registered multichannel 3D brain images. The model's architecture includes multiple convolutional layers, pooling layers, and fully connected layers, designed to automatically learn age-related, complex hierarchical features from the images.

[0061] The model was then trained under supervision using a large-scale brain imaging dataset with accurate age labels. The training process was end-to-end, calculating the error between the model's predicted age and the actual age label using forward propagation, and iteratively updating all parameters in the network using backpropagation to minimize the prediction error. This process continued until the model converged on an independent validation dataset, achieving stable predictive performance.

[0062] After training, all parameters of the optimized deep learning model are saved, forming a deployable brain age prediction model. In practical applications, a target brain image of any individual under test, processed using the same preprocessing and standardization procedures, is input into this model. The model can automatically extract its deep features and directly regress to output a high-precision auxiliary brain age prediction value. This approach provides a data-driven, automated prediction scheme capable of capturing nonlinear complex patterns.

[0063] In one feasible implementation, after the step of selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result, the method further includes: The trained brain age prediction model is invoked to process the target brain image and output an auxiliary brain age prediction value. Calculate the optimal relevance measure corresponding to the brain age prediction result and determine it as the first confidence level; Calculate the model confidence level corresponding to the assisted brain age prediction value and determine it as the second confidence level; Based on the first confidence level and the second confidence level, the brain age prediction result and the auxiliary brain age prediction value are weighted and fused to generate the final brain age prediction value.

[0064] It should be noted that "calling" refers to the process of starting and running a trained model for inference. The auxiliary brain age prediction value refers to the brain age prediction result independently output by the deep learning model path, without fusion. The optimal relevance measure refers to the highest relevance measure among all age templates calculated in the template matching path. The first confidence score is a measure of the inherent consistency of the template matching path itself, and its value directly reflects the reliability of the best match. Model confidence is an estimate of the uncertainty of the deep learning model's own output prediction value; its quantification methods include, but are not limited to, the entropy of the predicted probability distribution, the variance of the ensemble model output, or the variance of multiple forward propagations during exit inference. Weighted fusion refers to the operation of linearly combining two or more input values ​​according to preset rules or dynamically calculated coefficients to generate a single output value.

[0065] Understandably, this implementation describes a decision-making mechanism that integrates the advantages of two different prediction paths. Instead of simply averaging the results, this mechanism introduces a dynamic weighting strategy based on confidence levels. By separately evaluating the matching quality of the interpretable template matching path and the predictive determinism of the data-driven deep learning path, the system can adaptively allocate confidence weights between the prediction results of the two paths. When the prediction confidence of a particular path is higher, its influence in the final decision increases accordingly. This design enables the entire prediction system to be context-adaptive, flexibly balancing interpretability and complex pattern capture capabilities in cases with varying data inputs or characteristics. This promises to achieve more stable and accurate final prediction results than any single path in most cases, improving the overall robustness and practical performance of the system.

[0066] Reference Figure 6 After obtaining the standardized target brain image through preprocessing, the processing flow splits into two parallel paths: one path enters the "3D CNN" module in the diagram, which calls the trained brain age prediction model to process the target brain image and directly regress to output an auxiliary brain age prediction value; the other path corresponds to the "Data Analysis" module in the diagram, which executes the template matching process to obtain a brain age prediction result based on the most similar template and its optimal correlation metric (i.e., the first confidence level). Simultaneously, the system obtains the model confidence level (i.e., the second confidence level) corresponding to the auxiliary brain age prediction value from the 3D CNN module or its output. Subsequently, the system does not directly output the result of either single path, but instead enters a fusion decision node. Based on the first and second confidence levels, weights are dynamically calculated, and the prediction results of the two paths are weighted and fused to ultimately generate a more robust and accurate final brain age prediction value. This process clearly demonstrates how interpretable template matching is combined with data-driven deep learning prediction, and how the advantages of both are adaptively integrated through a confidence weighting mechanism.

[0067] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the method of assessing brain age based on deep learning intelligent analysis of myelin sheath development. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0068] This application also provides a device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development; please refer to [reference needed]. Figure 7 The device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: Acquisition module 10 is used to acquire multimodal magnetic resonance images of the brain; Preprocessing module 20 is used to preprocess the multimodal magnetic resonance images of the brain to obtain target brain images; The prediction module 30 is used to measure the correlation between the target brain image and the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0069] Or, the device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: The first processing module is used to perform decranialization processing on the multimodal magnetic resonance imaging of the brain to obtain brain tissue images and corresponding brain tissue masks. The first operation module is used to perform morphological operations on the brain tissue mask to fill holes and ensure the connectivity and integrity of the mask, thereby obtaining an optimized brain mask. The first extraction module is used to extract precise brain region voxel data from the brain tissue image based on the optimized brain mask. The first registration module is used to perform intermodal registration of the brain region voxel data of different modalities, so that the image space of all modalities is aligned to the same reference modal space. The second registration module is used to register the spatially aligned brain region voxel data to the standard space corresponding to the preset continuous age brain template library through a hierarchical registration process that includes affine transformation and nonlinear deformation field, and save the complete spatial transformation parameters. The first resampling module is used to resample the image registered to the standard space to generate the final target brain image for matching.

[0070] Or, the device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: The first acquisition module is used to acquire predefined brain region maps related to white matter development from the continuous age brain template library; The first mapping module is used to reverse map the predefined brain region atlas to the individual space of the target brain image using the saved spatial transformation parameters, so as to obtain individualized brain region labels. The first generation module is used to generate a brain region feature mask for subsequent correlation measurement based on the individualized brain region labels, wherein the brain region feature mask is preferably a whole-brain white matter mask.

[0071] Or, the device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: The second extraction module is used to extract the voxel intensity values ​​of the target brain image within the brain region feature mask to obtain the first original feature vector. The third extraction module is used to extract the voxel intensity values ​​of the template images of each age within the brain region feature mask to obtain the second original feature vector. The second processing module is used to perform standardization processing based on mean and standard deviation on the first original feature vector of the target brain image to obtain the first standardized feature vector. The third processing module is used to perform standardization processing based on mean and standard deviation on the second original feature vector of each age template to obtain the second standardized feature vector; The first calculation module is used to calculate the linear correlation between the first standardized feature vector of the target brain image and the second standardized feature vector of each age template, and use the strength of the calculated linear correlation as the correlation measurement result.

[0072] Or, the device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: The first construction module is used to construct a three-dimensional convolutional neural network deep learning model as a brain age prediction model, wherein the input configuration of the brain age prediction model is to receive standardized and spatially registered multi-channel three-dimensional brain images. The first training module is used to train the brain age prediction model end-to-end using a brain image dataset labeled with age, by calculating the prediction error through forward propagation and updating the parameters through back propagation, until the model converges. The first output module is used to save the parameters of the trained brain age prediction model, thereby obtaining the trained brain age prediction model.

[0073] Or, the device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development includes: The first calling module is used to call the trained brain age prediction model, process the target brain image, and output an auxiliary brain age prediction value. The second calculation module is used to calculate the optimal correlation metric corresponding to the brain age prediction result and determine it as the first confidence level. The third calculation module is used to calculate the model confidence level corresponding to the assisted brain age prediction value and determine it as the second confidence level; The first fusion module is used to perform weighted fusion of the brain age prediction result and the auxiliary brain age prediction value based on the first confidence level and the second confidence level to generate the final brain age prediction value.

[0074] This application provides a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development. It employs a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in the above embodiments, thus solving the technical problem of poor brain age prediction accuracy. Compared with the prior art, the beneficial effects of the device for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in this application are the same as those of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in the above embodiments. Furthermore, other technical features of the device for assessing brain age based on deep learning intelligent analysis of myelin sheath development are the same as those disclosed in the method of the above embodiments, and will not be repeated here.

[0075] This application provides a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in Embodiment 1 above.

[0076] The following is for reference. Figure 8 The diagram illustrates a structural schematic of a device suitable for implementing embodiments of this application, based on deep learning intelligent analysis of myelin sheath development to assess brain age. The device for assessing brain age based on deep learning intelligent analysis of myelin sheath development in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, tablets, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital televisions and desktop computers. Figure 8 The device shown is merely an example of a deep learning-based intelligent analysis device for assessing brain age by analyzing myelin sheath development. It should not impose any limitations on the functionality and scope of use of the embodiments in this application.

[0077] like Figure 8As shown, a device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0078] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0079] This application provides a device for assessing brain age based on deep learning intelligent analysis of myelin sheath development. It employs a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in the above embodiments, thus solving the technical problem of poor brain age prediction accuracy. Compared with the prior art, the beneficial effects of the device for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in this application are the same as those of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in the above embodiments. Furthermore, other technical features of this device for assessing brain age based on deep learning intelligent analysis of myelin sheath development are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0080] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0081] 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.

[0082] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development in the above embodiments.

[0083] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0084] The aforementioned computer-readable storage medium may be included in a device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development; or it may exist independently and not incorporated into a device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development.

[0085] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by a device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development, cause the device to: Acquire multimodal magnetic resonance imaging of the brain; The multimodal magnetic resonance imaging of the brain is preprocessed to obtain the target brain image; The target brain image is correlated with the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

[0086] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0087] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0088] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0089] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for assessing brain age based on deep learning intelligent analysis of myelin sheath development, thereby solving the technical problem of poor brain age prediction accuracy. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in the above embodiments, and will not be repeated here.

[0090] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described above for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development.

[0091] The computer program product provided in this application can solve the technical problem of poor accuracy in brain age prediction. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the method for assessing brain age based on deep learning intelligent analysis of myelin sheath development provided in the above embodiments, and will not be repeated here.

[0092] All acquisition of signals, information, or actions in this application are carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the relevant device owner.

[0093] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included in the scope of protection of this application.

Claims

1. A method for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development, characterized in that, The method includes: Acquire multimodal magnetic resonance imaging of the brain; The multimodal magnetic resonance imaging of the brain is preprocessed to obtain the target brain image; The target brain image is correlated with the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

2. The method as described in claim 1, characterized in that, The step of preprocessing the multimodal magnetic resonance imaging of the brain to obtain the target brain image includes: The brain multimodal magnetic resonance images were subjected to decranialization to obtain brain tissue images and corresponding brain tissue masks. Morphological operations are performed on the brain tissue mask to fill holes and ensure the connectivity and integrity of the mask, resulting in an optimized brain mask; Based on the optimized brain mask, precise brain region voxel data are extracted from the brain tissue image; The voxel data of the brain regions from different modalities are registered intermodally so that the image spaces of all modalities are aligned to the same reference modal space. The spatially aligned brain region voxel data are registered to the standard space corresponding to the preset continuous age brain template library through a hierarchical registration process that includes affine transformation and nonlinear deformation field, and the complete spatial transformation parameters are saved. The images registered to the standard space are resampled to generate the final target brain image used for matching.

3. The method as described in claim 2, characterized in that, Before the step of registering the spatially aligned brain region voxel data to the standard space corresponding to the preset continuous-age brain template library through a hierarchical registration process including affine transformation and nonlinear deformation field, and saving the complete spatial transformation parameters, the following steps are included: From the continuous age brain template library, obtain predefined brain region atlases related to white matter development; Using the saved spatial transformation parameters, the predefined brain region atlas is inversely mapped to the individual space of the target brain image to obtain individualized brain region labels; Based on the individualized brain region labels, brain region feature masks are generated for subsequent correlation measurement, wherein the brain region feature masks are preferably whole-brain white matter masks.

4. The method as described in claim 3, characterized in that, The step of performing correlation measurement between the target brain image and the standardized features of each age template in the preset continuous age brain template library, and selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result based on the calculated correlation measurement results includes: Within the brain region feature mask, the voxel intensity values ​​of the target brain image are extracted to obtain the first original feature vector; Within the brain region feature mask, voxel intensity values ​​of template images for each age are extracted to obtain the second original feature vector; The first original feature vector of the target brain image is subjected to standardization processing based on mean and standard deviation to obtain the first standardized feature vector; The second original feature vector of each age template is standardized based on the mean and standard deviation to obtain the second standardized feature vector; The linear correlation between the first standardized feature vector of the target brain image and the second standardized feature vector of each age template is calculated, and the strength of the calculated linear correlation is used as the correlation measurement result.

5. The method as described in claim 1, characterized in that, Prior to the step of acquiring multimodal magnetic resonance imaging of the brain, the following steps are included: A three-dimensional convolutional neural network deep learning model is constructed as a brain age prediction model, wherein the input configuration of the brain age prediction model is to receive standardized and spatially registered multi-channel three-dimensional brain images. Using an age-labeled brain image dataset, the brain age prediction model was trained end-to-end by calculating the prediction error through forward propagation and updating the parameters through back propagation until the model converged. Save the parameters of the trained brain age prediction model to obtain the trained brain age prediction model.

6. The method according to any one of claims 1-5, characterized in that, Following the step of selecting the age corresponding to the age template with the highest correlation measurement result as the brain age prediction result, the following is also included: The trained brain age prediction model is invoked to process the target brain image and output an auxiliary brain age prediction value. Calculate the optimal relevance measure corresponding to the brain age prediction result and determine it as the first confidence level; Calculate the model confidence level corresponding to the assisted brain age prediction value and determine it as the second confidence level; Based on the first confidence level and the second confidence level, the brain age prediction result and the auxiliary brain age prediction value are weighted and fused to generate the final brain age prediction value.

7. A device for assessing brain age by intelligently analyzing myelin sheath development based on deep learning, characterized in that, The device includes: The acquisition module is used to acquire multimodal magnetic resonance imaging of the brain; The preprocessing module is used to preprocess the multimodal magnetic resonance images of the brain to obtain the target brain image; The prediction module is used to measure the correlation between the target brain image and the standardized features of each age template in the preset continuous age brain template library. Based on the calculated correlation measurement results, the age corresponding to the age template with the highest correlation measurement result is selected as the brain age prediction result.

8. A device for assessing brain age based on deep learning-based intelligent analysis of myelin sheath development, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of a method for assessing brain age based on deep learning intelligent analysis of myelin sheath development as described in any one of claims 1 to 6.

Images