A method, system and device for assessing cognitive outcome in infants with white matter injury
By assessing infantile white matter lesions using structural disconnection scoring and decision tree models, this approach addresses the shortcomings of existing technologies that neglect spatial heterogeneity of lesions and disruption of fiber pathways, enabling accurate prediction and early assessment of infant cognitive development.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- THE FIRST AFFILIATED HOSPITAL OF MEDICAL COLLEGE OF XIAN JIAOTONG UNIV
- Filing Date
- 2025-11-27
- Publication Date
- 2026-05-22
Smart Images
Figure CN121313141B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical technology, specifically to a method, system, and device for assessing the cognitive prognosis of infantile white matter injury. Background Technology
[0002] With the advancement of perinatal medicine, the spectrum of white matter lesions in premature infants has undergone significant changes. The incidence of severe cystic lesions (periventricular leukomalacia) has decreased dramatically (<5%), while non-cystic focal white matter lesions (PWML) have become the most common form of lesion visible on magnetic resonance imaging (MRI). PWML typically presents as punctate, clustered, or linear T1 hyperintensities in the periventricular white matter region.
[0003] Current prognostic assessments of PWML primarily rely on routine MRI to evaluate the lesion itself, emphasizing the correlation between lesion burden, location, and motor impairments (such as cerebral palsy). However, recent imaging evidence suggests a key phenotypic shift in PWML, characterized by progressively posteriorly shifting lesion distribution and reduced lesion severity. This shift has led to a greater clinical focus on more subtle neurodevelopmental disorders, particularly in the cognitive domain, which require closer attention. These changes also challenge the accuracy of historical predictive models currently used for high-risk infants. Some scholars have proposed a prognostic assessment method based on conventional MRI that combines volume and location of PWML (Posterior Comminuted Lung Lesion). This method first marks PWML lesions using T1-weighted imaging (T1WI) and calculates the total lesion volume. Second, for PWML location assessment, medical image viewing software (3DSlicer) is used to identify the posterior commissure and left and right eyeballs on the T1WI image, reconstructing the T1WI image into a posterior commissure-binocular plane corrected image (i.e., three landmarks aligned on one axial plane, and two eyeball landmarks aligned on the coronal plane). After image reconstruction, a specific axial image perpendicular to the T1WI-eye plane is used for PWML assessment. Above the plane where the bilateral lateral ventricles change from a straight to a curved shape, the length of both lateral ventricles is measured, and the midline position is determined. Based on this midline position, the location of the PWML is divided into anterior or posterior. This scheme suggests that anterior PWML distribution can predict preschool motor and cognitive prognosis. However, the existing technical solutions mentioned above have some limitations when faced with the new characteristics of "smaller PWML lesions and a more posterior distribution": 1) The total volume of the lesion takes into account the severity of the injury, but ignores the characteristics and heterogeneity of its distribution; 2) In terms of lesion location assessment, although image plane reconstruction is performed, it only roughly divides it into the anterior and posterior parts of the lateral ventricle midline. Given the current posterior distribution of PWML lesions, the applicability of this assessment scheme is greatly reduced; 3) This scheme only assesses the PWML lesion itself and ignores the impact of the injury on the connectivity of potential white matter structures.
[0004] In summary, the methods for prognostic assessment of PWML based on conventional MRI have certain limitations: (1) Traditional qualitative assessment: relies on radiologists to perform rough localization and classification of the anterior, middle and posterior parts by visual inspection, which is highly subjective and cannot quantify the severity of the damage. (2) Simple volume analysis: although it can calculate the total volume of the lesion, it ignores the spatial heterogeneity of the lesion and its functional impact on the destruction of key nerve fiber pathways. (3) Probabilistic damage atlas: applicable to lesions with a large damage range, but for PWML lesions that are scattered and punctate, it is difficult to generate a reliable statistical atlas due to the low overlap of the lesions, resulting in insufficient predictive efficacy. Summary of the Invention
[0005] To address the shortcomings in predictive efficacy of existing technologies that neglect the spatial heterogeneity of lesions and their functional impact on the disruption of key neural fiber pathways, this invention proposes a method, system, and device for assessing the cognitive prognosis of infantile white matter injury. The method utilizes structural disconnection scores for the prognostic assessment of PWML (Physical Mild Brain Lesion) as an important indicator for evaluating cognitive prognostic decision tree models, thereby solving the problems existing in the prior art.
[0006] A method for assessing the cognitive prognosis of infantile white matter lesions includes the following steps:
[0007] Acquire T1-weighted imaging and diffusion tensor images of the brain of the infant to be evaluated;
[0008] Focal white matter lesion regions are delineated on T1-weighted images to generate binarized individual lesion masks; diffusion tensor images are preprocessed to generate individual partial anisotropy maps.
[0009] By registering the standard fiber tract map to the individual partial anisotropy map, individualized fiber tract labels are obtained. The individual lesion mask is overlaid with the individual fiber tract label to obtain the lesion volume located on each fiber tract. The individual lesion mask is used as a seed point to track the connections between the individual lesion mask and the whole brain fiber tracts pre-constructed based on a healthy infant population. The intersection map of the tracking results is taken as the individual disconnection map. The white matter structure disconnection score is calculated based on the individual disconnection map and the preset disconnection core area.
[0010] The lesion volume involving fiber bundles and the white matter structure disconnection score are input into a pre-trained decision tree model, and the risk level of cognitive development delay in infants is output according to the preset judgment rules.
[0011] Furthermore, the process of registering a standard fiber bundle map to an individual partial anisotropy map to obtain an individualized fiber bundle label, and then performing an overlay operation between the individual lesion mask and the individualized fiber bundle label to obtain the lesion volume located on each fiber bundle, specifically includes the following steps:
[0012] A set of diffusion tensor imaging data of the brains of infants with normal cognitive development was obtained and defined as a standard control group; based on the diffusion tensor image data of this standard control group, a local partial anisotropy template was constructed.
[0013] Register the standard white matter fiber bundle pattern to the local partial anisotropic template;
[0014] The individual partial anisotropy map of the infant to be evaluated is registered with the T1-weighted image, and the local partial anisotropy template is registered with the registered result to obtain individualized fiber bundle labels.
[0015] The volume of the lesion on each fiber bundle is calculated by overlaying the individual lesion mask with the individualized fiber bundle label.
[0016] Furthermore, the step of generating an individual disconnection map by tracking individual lesion masks against a pre-constructed whole-brain fiber tract connectivity based on a group of healthy infants specifically includes the following steps:
[0017] Whole-brain probabilistic fiber tracking based on diffusion tensor data from a standard control group;
[0018] The individual lesion mask of the infant to be evaluated is registered to the fiber tracing space of each control case as a seed point for tracing, and a binarized individual disconnection map is generated.
[0019] Register all individual disconnection maps to the local partial anisotropic template and take the intersection to generate the final individual disconnection map.
[0020] Furthermore, the lesion volume involving the fiber bundles includes the lesion volumes of the left mandibular fasciculus, the bilateral superior longitudinal fasciculi, and the left arcuate fasciculus.
[0021] Furthermore, the calculation of the white matter structure disconnection score based on the individual disconnection map and the preset disconnection core region specifically includes the following steps:
[0022] Based on the individual disconnection maps of all infants to be evaluated, a population-level disconnection probability distribution map is generated, and a probability threshold is set, with regions greater than 0.5 being designated as disconnection core regions;
[0023] The ratio of the number of voxels in an individual disconnection map to the total number of voxels in the disconnection core region is calculated to obtain the white matter structure disconnection score.
[0024] Furthermore, the preset determination rules include:
[0025] If the disconnection score is ≥0.58 and the lesion volume involving the left mandibular occipital fasciculus is ≥0.44mm³, it is considered to be at high risk of cognitive developmental delay.
[0026] If the disconnection score is ≥0.58, the lesion volume involving the left mandibular fasciculus is <0.44mm³, and the lesion volume involving the right superior longitudinal fasciculus is ≥11.93mm³, then it is considered a high risk of cognitive developmental delay.
[0027] If the above conditions are not met, the risk of cognitive developmental delay is determined to be low.
[0028] Furthermore, the decision tree model is internally validated using Bootstrap resampling and evaluated using the area under the curve, calibration curve, and clinical decision curve.
[0029] This invention also includes a cognitive prognostic assessment system for infant white matter injury, comprising:
[0030] The acquisition module is used to acquire T1-weighted imaging and diffusion tensor images of the brain of the infant to be evaluated;
[0031] The preprocessing module is used to delineate focal white matter lesion regions on T1-weighted images and generate binarized individual lesion masks; and to preprocess diffusion tensor images to generate individual partial anisotropy maps.
[0032] The calculation module is used to obtain individualized fiber bundle labels by registering the standard fiber bundle map to the individual part anisotropy map, and to obtain the lesion volume on each fiber bundle by overlaying the individual lesion mask with the individual fiber bundle labels. The module also tracks the connections between the individual lesion mask as seed points and the whole brain fiber bundles pre-constructed based on a healthy infant population, takes the intersection map of the tracking results as the individual disconnection map, and calculates the white matter structure disconnection score based on the individual disconnection map and the preset disconnection core area.
[0033] The assessment module is used to input the lesion volume involving fiber bundles and the white matter structure disconnection score into a pre-trained decision tree model, and output the risk level of cognitive development delay in infants according to the preset judgment rules.
[0034] The present invention also includes a computer device for assessing the cognitive prognosis of infant white matter injury, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for assessing the cognitive prognosis of infant white matter injury.
[0035] The present invention also includes a readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform the steps of the described method for assessing the cognitive prognosis of infant white matter injury.
[0036] This invention provides a method for assessing the cognitive prognosis of infantile white matter injury, which has the following beneficial effects:
[0037] This invention reduces reliance on visual qualitative assessment by extracting quantitative features of disconnection scores and the volume of lesions involving specific fiber tracts. The structural disconnection score used is the first time it has been applied to the prognostic assessment of PWML (Polyspinal Microvascular Lung Disease). This feature can quantify the potential damage of lesions to the core white matter connectivity network of the brain, revealing the pathophysiological mechanism of cognitive impairment from the essential level of functional connectivity. By quantifying the volume of specific fiber tracts involved, it can accurately capture scattered, punctate lesions distributed in different brain regions and precisely measure the severity of these lesions, thus improving predictive efficacy. This method is not affected by the overall posterior shift in lesion distribution and maintains high assessment sensitivity and applicability for PWML lesions with new features. Moreover, it can be extended to conventional MRI for analysis without the need to collect diffusion tensor image data, providing a new solution for early prognostic assessment and has potential for widespread application. Attached Figure Description
[0038] Figure 1 This is a schematic diagram of T1WI displaying PWML lesions (white arrows) and the masked lesions after annotation in an embodiment of the present invention;
[0039] Figure 2 This is a schematic diagram comparing the volume of fiber bundles involved in lesions in different PWML cognitive developmental outcome groups in this embodiment of the invention;
[0040] Figure 3 This is a diagram showing the probability distribution of disconnection, the core area of disconnection, and a comparison of disconnection scores among different PWML cognitive developmental outcome groups in this embodiment of the invention.
[0041] Figure 4 This is a diagram of the decision tree model architecture for determining the cognitive developmental outcome of PWML in an embodiment of the present invention;
[0042] Figure 5 This is a graph showing the evaluation index of the decision tree model in this embodiment of the invention. Figure 5 In the figure, (A) represents the receiver operating characteristic curve of the decision tree model; Figure 5 (B) in the figure represents the calibration curve of the decision tree model; Figure 5 (C) in the figure represents the decision curve of the decision tree model;
[0043] Figure 6 This is a schematic diagram of the process for assisting in the cognitive prognosis assessment of infant white matter injury in an embodiment of the present invention. Detailed Implementation
[0044] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0045] This invention proposes a method for auxiliary assessment of cognitive prognosis in infants with white matter lesions based on brain magnetic resonance imaging, such as... Figure 6 As shown, the method specifically includes the following steps:
[0046] S1. Data acquisition and preprocessing.
[0047] Obtain brain MRI data of preterm infants with corrected gestational age ≤45 weeks and PWML, including at least T1-weighted imaging (T1WI), T2-weighted imaging, and diffusion-weighted imaging (DTI) sequences.
[0048] T1WI image processing: Experienced radiologists manually delineate PWML lesions on T1WI images using ITK-SNAP (version 3.6.0) software to obtain a binarized individual lesion mask. Figure 1 ).
[0049] DTI image processing: The DTI data was preprocessed using FSL software. The preprocessing steps included head motion correction, eddy current correction, and skull removal. Finally, a partial anisotropy (FA) map was calculated and generated for subsequent analysis.
[0050] S2, Feature Calculation: Parallel calculation of the volume of lesions involving fiber bundles and the score of white matter structural disconnection.
[0051] 1) Quantification of the volume of lesions involving fiber bundles:
[0052] ①Based on the DTI data of the control group infants, a local FA template was constructed.
[0053] ② Register the Johns Hopkins University neonatal white matter atlas to the above-mentioned local FA template.
[0054] ③ For each PWML patient, their individual FA map is registered with the T1WI image, and then the fiber bundles in the local FA template space are transformed to the individual's T1WI space, ultimately obtaining 20 individualized major fiber bundle labels.
[0055] ④ Perform overlap relationship calculations on the individual lesion mask obtained in step 1 and the individualized fiber bundle label to calculate the lesion volume located on each fiber bundle.
[0056] The results obtained in this section showed that the lesion volume involving the left mandibular fasciculus, bilateral superior longitudinal fasciculations, and left arcuate fasciculus in the PWML cognitive developmental delay group was significantly higher than that in the cognitive developmental normal group. Figure 2As shown, the lesion volumes involving fiber tracts in different PWML cognitive developmental outcome groups include the left anterior thalamic radiation, right anterior thalamic radiation, left corticospinal tract, right corticospinal tract, left cingulate gyrus cingulate gyrus, right cingulate gyrus cingulate gyrus, left cingulate gyrus hippocampus, right cingulate gyrus hippocampus, splenium of corpus callosum, genu of corpus callosum, left mandibular fasciculus, right mandibular fasciculus, left inferior longitudinal fasciculus, right inferior longitudinal fasciculus, left superior longitudinal fasciculus, right superior longitudinal fasciculus, left uncus fasciculus, right uncus fasciculus, left arcuate fasciculus, and right arcuate fasciculus.
[0057] 2) Structural disconnection score calculation aims to quantify the potential white matter connectivity disruption caused by lesions:
[0058] ① Based on the DTI data of the control group infants, whole-brain probabilistic fiber tracking was performed for each control group using BedpostX and ProbtrackX (FSL toolkit).
[0059] ② The lesion mask of each PWML infant was registered to the FA space of each control case using the image processing software ANTS.
[0060] ③ Using the registered lesion mask as seed points, fiber tract tracing is performed based on the whole-brain probabilistic fiber tracing of each control case to assess the possible white matter fiber interruption caused by the lesion and generate a binary disconnection map.
[0061] ④ The fiber tract tracking results generated for each PWML infant on all controls were registered to the above local FA template using ANTS, and the intersection was taken to finally generate the individual's disconnection map.
[0062] ⑤ Calculate the probability distribution of the disconnection map for all PWML infants (probability range 0-1), and set a threshold of 0.5, setting regions >0.5 as disconnection core regions.
[0063] ⑥ Divide the number of voxels on the disconnection map of each PWML infant by the total number of voxels in the core disconnection region. The resulting value (range 0-1) is the disconnection score for that infant. A higher score indicates more severe damage to the core brain connectivity. Figure 3 As shown in (a).
[0064] The results in this section show that the disconnection score in the PWML cognitive developmental delay group was significantly higher than that in the cognitive developmental normal group, such as... Figure 3 As shown in (b).
[0065] S3. Decision tree model construction: Construct a decision tree model based on the differential results obtained in steps 1 and 2 above (analyzed with Python 3.9). Node splitting is guided by Gini impurities. The prediction result of the decision tree model is the risk level (high risk / low risk) of cognitive developmental delay in the PWML infant at 6 months of age.
[0066] To enhance model stability, 1000 bootstrap resampling operations were used for internal validation.
[0067] The model was evaluated using the area under the curve (AUC), calibration curve, and clinical decision curve.
[0068] The constructed decision tree model ultimately incorporated three indicators: disconnection score (importance 0.78), lesion volume involving the left mandibular fasciculus (importance 0.12), and lesion volume involving the right superior longitudinal fasciculus (importance 0.10). The model's final mean AUC was 0.78 (95% CI 0.71–0.85), sensitivity was 71.5% (95% CI 59.6%–81.8%), and specificity was 73.7% (95% CI 66.8%–80.4%). Figure 5 As shown in (A); Figure 5 As shown in (B), the calibration curve shows that the model's predicted probabilities trend in line with the ideal curve throughout the range. However, in the range of higher predicted probabilities, the curve deviates and the confidence interval widens. This may be due to statistical fluctuations caused by the relatively small number of actual high-risk samples in the dataset; however, as Figure 5 The decision curve analysis shown in (C) indicates that even with the aforementioned biases, this model still demonstrates significant clinical net benefit over a wide range of threshold probabilities (10%-50%).
[0069] like Figure 4 As shown, the rules of the decision tree model are as follows:
[0070] If the disconnection score is ≥0.58 and the lesion involving the left mandibular fasciculus has a volume ≥0.44 mm³, it is considered a high risk of cognitive developmental delay.
[0071] If the disconnection score is ≥0.58, the lesion volume involving the left mandibular fasciculus is <0.44mm³, and the lesion volume involving the right superior longitudinal fasciculus is ≥11.93mm³, then the individual is considered to be at high risk of cognitive developmental delay.
[0072] If none of the above conditions are met, the individual is considered to have a low risk of cognitive developmental delay.
[0073] The present invention has the following advantages:
[0074] 1) Improved objectivity and reproducibility of PWML prognostic assessment: This invention reduces reliance on doctors' visual qualitative assessment by extracting quantitative features of disconnection scores and lesion volume involving specific fiber bundles, providing more objective and quantitative indicators for clinical judgment.
[0075] 2) This invention provides a new technical approach for the early assessment of developmental outcomes in PWML: The structural disconnection score used in this invention is the first time it has been used for prognostic assessment of PWML and is an important indicator for evaluating cognitive prognostic decision tree models. Its advantages lie not only in providing a new quantitative indicator beyond overt PWML lesions, but also in its applicability to conventional MRI analysis without the need for DTI data collection, offering a new solution for early prognostic assessment and possessing potential for widespread application.
[0076] 3) Internal validation demonstrates promising application potential: Internal validation using the Bootstrap method shows that the decision tree model of this invention exhibits stable discriminative power (AUC: 0.78) in distinguishing cognitive outcomes in 6-month-old infants with PWML, achieving a relatively balanced result between sensitivity and specificity. Decision curve analysis (DCA) suggests that the model may have net clinical benefits within a certain risk threshold range.
[0077] The implementation of this invention relies on an optimized, specific processing flow and empirically validated key decision threshold parameters (disconnection score and specific thresholds for lesion involvement of specific fiber bundles). These validated parameter combinations are crucial for ensuring the effectiveness of the method. Its application value includes: 1) Optimizing medical resource allocation: It is expected to help clinicians identify children with PWML requiring focused attention earlier, thereby enabling targeted follow-up and providing a reference for early intervention; 2) Forming core technology: The method flow and model of this invention can be integrated into dedicated software or analytical services, forming the core technological foundation of related products or services.
[0078] Based on the same inventive concept, this invention also proposes a cognitive prognostic assessment system for infant white matter lesions, comprising:
[0079] The acquisition module is used to acquire T1-weighted imaging and diffusion tensor images of the brain of the infant to be evaluated.
[0080] The preprocessing module is used to delineate the focal white matter lesion region on the T1-weighted image and generate a binarized individual lesion mask; and to preprocess the diffusion tensor image to generate an anisotropy map of the individual part.
[0081] The calculation module is used to obtain individualized fiber bundle labels by registering the standard fiber bundle map to the individual partial anisotropy map, and to obtain the lesion volume on each fiber bundle by overlaying the individual lesion mask with the individual fiber bundle label. The module also tracks the connections between the individual lesion mask as seed points and the whole brain fiber bundles pre-constructed based on a healthy infant population, takes the intersection map of the tracking results as the individual disconnection map, and calculates the white matter structural disconnection score based on the individual disconnection map and the preset disconnection core area.
[0082] The assessment module is used to input the lesion volume involving fiber bundles and the white matter structure disconnection score into a pre-trained decision tree model, and output the risk level of cognitive development delay in infants according to the preset judgment rules.
[0083] The present invention also proposes a computer device for cognitive prognostic assessment of infant white matter injury, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the cognitive prognostic assessment method for infant white matter injury.
[0084] The present invention also proposes a readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform steps of a method for assessing the cognitive prognosis of infant white matter injury.
[0085] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for assessing the cognitive prognosis of infantile white matter injury, characterized in that, Includes the following steps: Acquire T1-weighted imaging and diffusion tensor images of the brain of the infant to be evaluated; Focal white matter lesion regions are delineated on T1-weighted images to generate binarized individual lesion masks; Preprocess the diffusion tensor image to generate anisotropy maps of individual parts; By registering the standard fiber tract map to the individual partial anisotropy map, individualized fiber tract labels are obtained. The individual lesion mask is overlaid with the individual fiber tract label to obtain the lesion volume located on each fiber tract. The individual lesion mask is used as a seed point to track the connections between the individual lesion mask and the whole brain fiber tracts pre-constructed based on a healthy infant population. The intersection map of the tracking results is taken as the individual disconnection map. The white matter structure disconnection score is calculated based on the individual disconnection map and the preset disconnection core area. The lesion volume involving fiber bundles and the white matter structure disconnection score are input into a pre-trained decision tree model, and the risk level of cognitive development delay in infants is output according to the preset judgment rules.
2. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 1, characterized in that, The process involves registering a standard fiber bundle map to an individual partial anisotropy map to obtain an individualized fiber bundle label. Then, an overlay operation is performed between the individual lesion mask and the individualized fiber bundle label to obtain the lesion volume located on each fiber bundle. This process specifically includes the following steps: A set of diffusion tensor imaging data of the brains of infants with normal cognitive development was obtained and defined as a standard control group; based on the diffusion tensor image data of this standard control group, a local partial anisotropy template was constructed. Register the standard white matter fiber bundle pattern to the local partial anisotropic template; The individual partial anisotropy map of the infant to be evaluated is registered with the T1-weighted image, and the local partial anisotropy template is registered with the registered result to obtain individualized fiber bundle labels. The volume of the lesion on each fiber bundle is calculated by overlaying the individual lesion mask with the individualized fiber bundle label.
3. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 2, characterized in that, The method involves tracking individual lesion masks as seed points and connecting them to whole-brain fiber bundles pre-constructed based on a group of healthy infants, then taking the intersection map of the tracking results as the individual disconnection map. Specifically, this includes the following steps: Whole-brain probabilistic fiber tracking based on diffusion tensor data from a standard control group; The individual lesion mask of the infant to be evaluated is registered to the fiber tracing space of each control case as a seed point for tracing, and a binarized individual disconnection map is generated. Register all individual disconnection maps to the local partial anisotropic template and take the intersection to generate the final individual disconnection map.
4. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 1, characterized in that, The lesion volume involving the fiber bundles includes the lesion volume of the left mandibular fasciculus, the bilateral superior longitudinal fasciculus, and the left arcuate fasciculus.
5. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 1, characterized in that, The calculation of white matter structure disconnection score based on individual disconnection maps and preset disconnection core regions specifically includes the following steps: Based on the individual disconnection maps of all infants to be evaluated, a population-level disconnection probability distribution map is generated, and a probability threshold is set, with regions greater than 0.5 being designated as disconnection core regions; The ratio of the number of voxels in an individual disconnection map to the total number of voxels in the disconnection core region is calculated to obtain the white matter structure disconnection score.
6. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 1, characterized in that, The preset determination rules include: If the disconnection score is ≥0.58 and the lesion volume involving the left mandibular occipital fasciculus is ≥0.44mm³, it is considered to be at high risk of cognitive developmental delay. If the disconnection score is ≥0.58, the lesion volume involving the left mandibular fasciculus is <0.44mm³, and the lesion volume involving the right superior longitudinal fasciculus is ≥11.93mm³, then it is considered a high risk of cognitive developmental delay. If the above conditions are not met, the risk of cognitive developmental delay is determined to be low.
7. The method for assessing the cognitive prognosis of infantile white matter injury according to claim 1, characterized in that, The decision tree model was internally validated using Bootstrap resampling and evaluated using the area under the curve, calibration curve, and clinical decision curve.
8. A cognitive prognostic assessment system for infant white matter injury, characterized in that, include: The acquisition module is used to acquire T1-weighted imaging and diffusion tensor images of the brain of the infant to be evaluated; The preprocessing module is used to delineate focal white matter lesion areas on T1-weighted images and generate binarized individual lesion masks. Preprocess the diffusion tensor image to generate anisotropy maps of individual parts; The calculation module is used to obtain individualized fiber bundle labels by registering the standard fiber bundle map to the individual partial anisotropy map, and to obtain the lesion volume on each fiber bundle by overlaying the individual lesion mask with the individual fiber bundle labels. The module also tracks the connections between the individual lesion mask as seed points and the whole brain fiber bundles pre-constructed based on a healthy infant population, takes the intersection map of the tracking results as the individual disconnection map, and calculates the white matter structure disconnection score based on the individual disconnection map and the preset disconnection core area. The assessment module is used to input the lesion volume involving fiber bundles and the white matter structure disconnection score into a pre-trained decision tree model, and output the risk level of cognitive development delay in infants according to the preset judgment rules.
9. A computer device for assessing the cognitive prognosis of infantile white matter injury, characterized in that, include: A memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the infant white matter injury cognitive prognostic assessment method according to any one of claims 1-7.
10. A readable storage medium, characterized in that, The readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform the steps of the infant white matter injury cognitive prognosis assessment method according to any one of claims 1-7.