Autism spectrum disorder identification method and system based on image generation, equipment, medium

Abstract

The application provides an autism spectrum disorder identification method and system based on image generation, equipment and a medium, can synthesize an FA image from a conventional MRI image without increasing the scanning time or changing the imaging protocol, extract a quantitative index reflecting the white matter development difference, and realize the accurate distinction between autism spectrum disorder and overall developmental delay. The method avoids the high time cost of DTI scanning, is particularly suitable for children who are not easy to cooperate, and the generated image has good anatomical consistency and biological interpretability, which is helpful for doctors to understand and accept. Compared with the existing multi-modal or EEG method, the application has the advantages of simple structure, stable identification process and wide application range, can be quickly deployed on a conventional imaging device, and significantly improves the efficiency and accessibility of autism early screening and auxiliary diagnosis.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method, system, device, and medium for identifying autism spectrum disorder based on image generation. Background Technology

[0002] Autism spectrum disorder (ASD) is a group of neurodevelopmental disorders that begin in childhood, often manifesting as delayed language development, social impairment, repetitive and stereotyped behaviors, and sensory abnormalities. Its clinical manifestations overlap to some extent with global developmental delay (GDD), making early identification and differentiation crucial. Because their intervention strategies differ, early identification helps in developing personalized intervention plans and improving long-term prognosis.

[0003] Currently, ASD identification still relies primarily on clinical behavioral scales, which suffer from high subjectivity, dependence on experience, and limited accuracy. There is an urgent need for objective biomarkers, such as imaging studies, as auxiliary diagnostic tools. Research indicates that fractional anisotropy (FA), as an important indicator of white matter integrity, can reveal white matter developmental abnormalities in patients with autism spectrum disorder and has potential value in differentiating ASD from GDD. However, FA images are derived from diffusion tensor imaging (DTI) sequences, which have long acquisition times, require high levels of cooperation from pediatric patients, and have low clinical adoption rates, limiting the practical application of FA in primary care hospitals.

[0004] Existing research, such as Chinese patent application number CN202411947168.3 entitled "A Method for Recognizing Autism Brain Maps through Multimodal Brain Map Information Fusion," discloses a method for autism recognition based on multimodal brain map information fusion. This method utilizes structural MRI, functional MRI, and phenotypic information to construct brain region maps and completes the recognition task through a graph neural network. While this method improves recognition accuracy, it has high requirements for imaging conditions and data complexity, making it unsuitable for widespread adoption at the grassroots level.

[0005] For example, Chinese patent application number CN202311763463.9, entitled "A Method and Device for Autism Risk Assessment Based on Small Sample Data Augmentation," discloses a method for small sample data augmentation based on electroencephalogram (EEG) data, utilizing functional connectivity maps for autism risk assessment. This method relies on high-quality EEG for signal acquisition and is limited by factors such as child cooperation and environmental interference, resulting in insufficient clinical operability.

[0006] In summary, while existing technologies have made some progress in recognition accuracy, they generally suffer from the following shortcomings: First, they rely on multimodal images or functional signals, resulting in high imaging costs and complex acquisition processes, making them difficult to promote and use in routine clinical settings; second, some methods require a high degree of cooperation from the children, and are especially unsuitable for young children or children with behavioral abnormalities; third, there is currently no publicly available technology that can directly generate FA images based on conventional MRI images and use them for structural abnormality recognition of ASD and GDD.

[0007] Therefore, there is an urgent need for a new method that relies on conventional MRI sequences, has generative and recognition capabilities, and combines accuracy with clinical applicability, in order to make up for the shortcomings of existing technologies and meet the practical needs of early auxiliary diagnosis of neurodevelopmental disorders in children. Summary of the Invention

[0008] To achieve the above-mentioned objectives and other advantages of the present invention, a first objective of the present invention is to provide an image-based autism spectrum disorder identification method, comprising the following steps:

[0009] The original T1WI, T2WI, T2-FLAIR and anisotropic fractional map DICOM data were preprocessed and divided into training and test sets.

[0010] The anisotropic fractional image generation model is trained using the training set data.

[0011] Construct anisotropic fraction templates for children and perform template standardization, then extract the anisotropic fraction skeleton and anisotropic fraction values;

[0012] The autism spectrum disorder identification model is trained using the anisotropic scores, and the model with the best performance is selected and its parameters are saved.

[0013] The test set is input into the trained anisotropic fraction image generation model to obtain anisotropic fraction generated images;

[0014] Extract the anisotropic fraction features from the generated anisotropic fraction image, input the anisotropic fraction features into the trained autism spectrum disorder recognition model, calculate the evaluation index, and output the autism spectrum disorder recognition result based on the evaluation index.

[0015] Furthermore, the preprocessing steps for the original T1WI, T2WI, T2-FLAIR, and anisotropic fractional map DICOM data include:

[0016] Convert the original T1WI, T2WI, T2-FLAIR and anisotropic fractional plot DICOM data into NIfTI format;

[0017] Register the anisotropic fractional map, T2WI, and T2-FLAIR images to the T1WI image and unify their sizes;

[0018] Remove the background and keep the foreground brain portion;

[0019] z-score normalization was performed on the foreground brain region;

[0020] Cross-sectional slices of T1WI, T2WI, T2-FLAIR, and anisotropic fractional images were extracted and preprocessed.

[0021] Furthermore, the step of training the anisotropic fractional image generation model using the training set data includes:

[0022] The training set data is input into the PT-GAN model for training. The generator learns the mapping relationship for generating anisotropic fractional images from regular sequences, and the discriminator is used to collaboratively guide and improve the realism and structural accuracy of the generated images.

[0023] Furthermore, the step of constructing a child anisotropic fraction template and performing template standardization includes:

[0024] Target anisotropic fraction images are selected from the subject images without head motion artifacts. All anisotropic fraction images are aligned to the target image after linear and nonlinear registration and merged to generate a four-dimensional image to construct an anisotropic fraction template for children.

[0025] The anisotropic fraction templates for children were registered to the anisotropic fraction templates and brain atlases provided by JHU, and the FSL toolkit was used to further register all anisotropic fraction images to the preschool template space to achieve image standardization.

[0026] Furthermore, the steps of extracting the anisotropic fraction skeleton and anisotropic fraction values ​​include:

[0027] The anisotropic fraction images after registration are averaged to generate an average anisotropic fraction image and an anisotropic fraction skeleton is extracted. The images of each subject are projected onto the skeleton to form standardized 4D skeleton data.

[0028] The true and generated anisotropy scores for each subject were extracted using a standardized JHU brain atlas.

[0029] Furthermore, the step of training the autism spectrum disorder identification model using the anisotropic scores, selecting the best-performing model, and saving the parameters includes:

[0030] The obtained anisotropic scores are input into the AutoGluon classification framework for training.

[0031] Five-fold cross-validation is used to optimize the model hyperparameters on the training set, and the best-performing model is selected and its parameters are saved.

[0032] Furthermore, the step of calculating the evaluation index includes:

[0033] Calculate the accuracy, sensitivity, precision, and area under the receiver operating characteristic curve.

[0034] The second objective of this invention is to provide an image-based autism spectrum disorder (ASD) recognition system, employing the aforementioned method, including a data preprocessing module, an anisotropic fractional image generation model training module, an anisotropic fractional feature extraction module, an ASD recognition model training module, an anisotropic fractional image generation model testing module, and an ASD recognition module; wherein,

[0035] The data preprocessing module is used to preprocess the original T1WI, T2WI, T2-FLAIR and anisotropic fractional graph DICOM data, and divide them into training set and test set.

[0036] The anisotropic fractional image generation model training module is used to train the anisotropic fractional image generation model using the training set data.

[0037] The anisotropic fraction feature extraction module is used to construct anisotropic fraction templates for children and perform template standardization, and extract anisotropic fraction skeletons and anisotropic fraction values.

[0038] The autism spectrum disorder recognition model training module is used to train the autism spectrum disorder recognition model using the anisotropy scores, select the best performing model, and save the parameters.

[0039] The anisotropic fractional image generation model testing module is used to input the test set into the trained anisotropic fractional image generation model to obtain anisotropic fractional generated images.

[0040] The autism spectrum disorder recognition module is used to extract anisotropic score features from the anisotropic score generated image, input the anisotropic score features into the trained autism spectrum disorder recognition model, calculate evaluation indicators, and output autism spectrum disorder recognition results based on the evaluation indicators.

[0041] A third objective of the present invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0042] A fourth objective of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention can synthesize FA images from conventional MRI images without increasing scan time or changing imaging protocols, extracting quantitative indicators reflecting differences in white matter development, and achieving accurate differentiation between autism spectrum disorder and global developmental delay. This method avoids the time-consuming DTI scan, is particularly suitable for children who are difficult to cooperate with, and generates images with good anatomical consistency and biological interpretability, aiding physicians in understanding and accepting the information. Compared to existing multimodal or EEG methods, this invention has a simple structure, a stable recognition process, and a wide range of applications. It can be rapidly deployed on conventional imaging equipment, significantly improving the efficiency and accessibility of early autism screening and assisted diagnosis.

[0045] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0047] Figure 1 This is a flowchart of an image-based autism spectrum disorder identification method.

[0048] Figure 2 Flowchart for PT-GAN model training;

[0049] Figure 3 To construct anisotropic fraction templates for children and perform template standardization, as well as to extract the anisotropic fraction skeleton and anisotropic fraction values ​​flowchart;

[0050] Figure 4 Flowchart for training a model to identify autism spectrum disorder;

[0051] Figure 5 Visualize the generated anisotropy score plot;

[0052] Figure 6 This is a schematic diagram of an image-based autism spectrum disorder recognition system.

[0053] Figure 7 This is a schematic diagram of a computer device.

[0054] Figure 8 This is a schematic diagram of a computer-readable storage medium. Detailed Implementation

[0055] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0056] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0057] The drawing numbers in this application are only used to distinguish the steps in the scheme and are not used to limit the execution order of the steps. The specific execution order is as described in the specification.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0059] This invention provides a method and system for identifying autism spectrum disorder (ASD) based on image generation. It utilizes generative adversarial networks (GANs) to synthesize fractional anisotropy (FA) images from conventional T2-weighted MRI images without requiring diffusion tensor imaging, and extracts relevant structural indices to aid in the identification of ASD and GDD. The specific scheme is as follows:

[0060] Example 1

[0061] An image-based method for identifying autism spectrum disorder, such as Figure 1 As shown, it includes the following steps:

[0062] S100. Preprocess the original T1WI, T2WI, T2-FLAIR and anisotropic fractional map DICOM data, and divide all subjects into training set and test set;

[0063] In some embodiments, the preprocessing step for the raw T1WI, T2WI, T2-FLAIR, and anisotropic fractional map DICOM data includes:

[0064] S110. Convert the original T1WI, T2WI, T2-FLAIR and anisotropic fractional plot DICOM data into NIfTI format;

[0065] S120. Register the anisotropic fractional map, T2WI, and T2-FLAIR images to the T1WI image and unify their dimensions;

[0066] S130: Remove the background, retaining the foreground brain portion;

[0067] S140. Normalize the z-score of the foreground brain region.

[0068] S150. Extract cross-sectional slices from T1WI, T2WI, T2-FLAIR, and anisotropic fractional images and perform preprocessing.

[0069] S200. Train the anisotropic fractional image generation model using the training set data;

[0070] In some embodiments, such as Figure 2 As shown, the step of training the anisotropic fractional image generation model using the training set data includes:

[0071] The training set data is input into the PT-GAN model for training. The generator learns the mapping relationship for generating anisotropic fractional images from regular sequences, and the discriminator is used to collaboratively guide and improve the realism and structural accuracy of the generated images.

[0072] S300, Construct anisotropic fraction templates for children and perform template standardization, and extract the anisotropic fraction skeleton and anisotropic fraction values;

[0073] In some embodiments, such as Figure 3 As shown, the steps of constructing a child anisotropic fraction template and performing template standardization include:

[0074] S310. Select target anisotropic fraction images from the subject images without head motion artifacts. Align all anisotropic fraction images to the target image after linear and nonlinear registration, and merge them to generate a four-dimensional image to construct a child anisotropic fraction template.

[0075] S320. The child anisotropic fraction template is registered to the anisotropic fraction template and brain atlas provided by JHU, and all anisotropic fraction images are further registered to the preschool template space using the FSL toolkit to achieve image standardization.

[0076] In some embodiments, such as Figure 3 As shown, the steps for extracting the anisotropic fraction skeleton and anisotropic fraction values ​​include:

[0077] S330. Average the registered anisotropic fraction images to generate an average anisotropic fraction image and extract the anisotropic fraction skeleton. Project the images of each subject onto the skeleton to form standardized 4D skeleton data.

[0078] S340. Use standardized JHU brain atlases to extract the true and generated anisotropy scores for each subject.

[0079] S400. Train the autism spectrum disorder identification model using the anisotropic scores, select the best performing model and save the parameters;

[0080] In some embodiments, such as Figure 4 As shown, the steps of training the autism spectrum disorder identification model using the anisotropic scores, selecting the best-performing model, and saving the parameters include:

[0081] S410. Input the obtained anisotropic scores into the AutoGluon classification framework for training.

[0082] S420. Use five-fold cross-validation to optimize the model hyperparameters on the training set, select the best-performing model, and save the parameters.

[0083] S500. Input the test set into the trained anisotropic fraction image generation model to obtain anisotropic fraction generated images; the visualization result of the generated anisotropic fraction images is as follows. Figure 5 As shown.

[0084] S600. Extract the anisotropic fraction features from the generated anisotropic fraction image, input the anisotropic fraction features into the trained autism spectrum disorder recognition model, calculate the evaluation index, and output the autism spectrum disorder recognition result based on the evaluation index.

[0085] Furthermore, the step of calculating the evaluation index includes:

[0086] Calculate the accuracy, sensitivity, precision, and area under the receiver operating characteristic curve.

[0087] This embodiment combines machine learning to evaluate the recognition efficacy of real and generated anisotropic score maps in differentiating between autism spectrum disorder and global developmental delay. The recognition efficacy results are shown in Table 1.

[0088] Table 1. Evaluation results of the combined machine learning approach to assess the identification efficacy of real and generated anisotropic score maps in differentiating between autism spectrum disorder and global developmental delay.

[0089]

[0090] This embodiment proposes an image-based autism spectrum disorder (ASD) identification method. It utilizes generative adversarial networks (such as PT-GAN) to generate FA images from conventional MRI sequences (T1WI, T2WI, T2-FLAIR). Based on the white matter features extracted from the synthesized images, an ASD identification model is constructed to achieve structural auxiliary diagnosis for children who have not undergone DTI scans, effectively improving the clinical applicability and identification accuracy of the method.

[0091] Example 2

[0092] An image-based autism spectrum disorder recognition system applies the method described above. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here. Figure 6 As shown, the system 700 includes a data preprocessing module 710, an anisotropic fractional image generation model training module 720, an anisotropic fractional feature extraction module 730, an autism spectrum disorder recognition model training module 740, an anisotropic fractional image generation model testing module 750, and an autism spectrum disorder recognition module 760; wherein,

[0093] The data preprocessing module is used to preprocess the original T1WI, T2WI, T2-FLAIR and anisotropic fractional graph DICOM data, and divide them into training set and test set.

[0094] The anisotropic fractional image generation model training module is used to train the anisotropic fractional image generation model using the training set data.

[0095] The anisotropic fraction feature extraction module is used to construct anisotropic fraction templates for children and perform template standardization, and extract anisotropic fraction skeletons and anisotropic fraction values.

[0096] The autism spectrum disorder recognition model training module is used to train the autism spectrum disorder recognition model using the anisotropy scores, select the best performing model, and save the parameters.

[0097] The anisotropic fractional image generation model testing module is used to input the test set into the trained anisotropic fractional image generation model to obtain anisotropic fractional generated images.

[0098] The autism spectrum disorder recognition module is used to extract anisotropic score features from the anisotropic score generated image, input the anisotropic score features into the trained autism spectrum disorder recognition model, calculate evaluation indicators, and output autism spectrum disorder recognition results based on the evaluation indicators.

[0099] Based on the technical solutions of the above embodiments, optionally, the preprocessing step for the original T1WI, T2WI, T2-FLAIR and anisotropic fractional map DICOM data includes:

[0100] Convert the original T1WI, T2WI, T2-FLAIR and anisotropic fractional plot DICOM data into NIfTI format;

[0101] Register the anisotropic fractional map, T2WI, and T2-FLAIR images to the T1WI image and unify their sizes;

[0102] Remove the background and keep the foreground brain portion;

[0103] z-score normalization was performed on the foreground brain region;

[0104] Cross-sectional slices of T1WI, T2WI, T2-FLAIR, and anisotropic fractional images were extracted and preprocessed.

[0105] Based on the technical solution of the above embodiments, optionally, the step of training the anisotropic fractional image generation model using the training set data includes:

[0106] The training set data is input into the PT-GAN model for training. The generator learns the mapping relationship for generating anisotropic fractional images from regular sequences, and the discriminator is used to collaboratively guide and improve the realism and structural accuracy of the generated images.

[0107] Based on the technical solution of the above embodiments, optionally, the step of constructing a child anisotropic fraction template and performing template standardization includes:

[0108] Target anisotropic fraction images are selected from the subject images without head motion artifacts. All anisotropic fraction images are aligned to the target image after linear and nonlinear registration and merged to generate a four-dimensional image to construct an anisotropic fraction template for children.

[0109] The anisotropic fraction templates for children were registered to the anisotropic fraction templates and brain atlases provided by JHU, and the FSL toolkit was used to further register all anisotropic fraction images to the preschool template space to achieve image standardization.

[0110] Based on the technical solution of the above embodiments, optionally, the step of extracting the anisotropic fraction skeleton and the anisotropic fraction values ​​includes:

[0111] The anisotropic fraction images after registration are averaged to generate an average anisotropic fraction image and an anisotropic fraction skeleton is extracted. The images of each subject are projected onto the skeleton to form standardized 4D skeleton data.

[0112] The true and generated anisotropy scores for each subject were extracted using a standardized JHU brain atlas.

[0113] Based on the technical solution of the above embodiments, optionally, the step of training the autism spectrum disorder identification model using the anisotropic scores, selecting the best-performing model, and saving the parameters includes:

[0114] The obtained anisotropic scores are input into the AutoGluon classification framework for training.

[0115] Five-fold cross-validation is used to optimize the model hyperparameters on the training set, and the best-performing model is selected and its parameters are saved.

[0116] Based on the technical solutions of the above embodiments, optionally, the step of calculating the evaluation index includes:

[0117] Calculate the accuracy, sensitivity, precision, and area under the receiver operating characteristic curve.

[0118] This embodiment proposes an image-based autism spectrum disorder (ASD) identification system. It utilizes generative adversarial networks (such as PT-GAN) to generate FA images from conventional MRI sequences (T1WI, T2WI, T2-FLAIR). Based on the white matter features extracted from the synthesized images, an ASD identification model is constructed to achieve structural auxiliary diagnosis for children who have not undergone DTI scans, effectively improving the clinical applicability and identification accuracy of the method.

[0119] Example 3

[0120] A computer device 800, such as Figure 7 As shown, the system includes a memory 810, a processor 820, and a computer program 830 stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of an image-based autism spectrum disorder identification method. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here.

[0121] Example 4

[0122] A computer-readable storage medium, such as Figure 8 As shown, a computer program is stored thereon. When executed by a processor, the computer program implements the steps of an image-based autism spectrum disorder identification method. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0123] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0124] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0125] The apparatus, computer device, and non-volatile computer storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, computer device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, computer device, and non-volatile computer storage medium will not be repeated here.

[0126] Those skilled in the art will also know that, besides implementing the controller in the form of purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller take the form of logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices included within it for implementing various functions can also be considered structures within that hardware component. Alternatively, the devices for implementing various functions can be considered as both software units implementing the method and structures within a hardware component.

[0127] The systems, apparatuses, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above apparatuses are described separately as various units based on their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0128] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0133] This specification may be described in the general context of computer-executable instructions, such as program units, that are executed by a computer. Generally, program units include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program units may reside in local and remote computer storage media, including storage devices.

[0134] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0135] The above description is merely an embodiment of this specification and is not intended to limit the scope of one or more embodiments of this specification. Various modifications and variations can be made to one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of one or more embodiments of this specification.

Claims

1. A method for identifying autism spectrum disorder based on image generation, characterized in that, Includes the following steps: The original T1WI, T2WI, T2-FLAIR and anisotropic fractional map DICOM data were preprocessed and divided into training and test sets. The anisotropic fractional image generation model is trained using the training set data. Construct anisotropic fraction templates for children and perform template standardization, then extract the anisotropic fraction skeleton and anisotropic fraction values; The autism spectrum disorder identification model is trained using the anisotropic scores, and the model with the best performance is selected and its parameters are saved. The test set is input into the trained anisotropic fraction image generation model to obtain anisotropic fraction generated images; Extract the anisotropic fraction features from the generated anisotropic fraction image, input the anisotropic fraction features into the trained autism spectrum disorder recognition model, calculate the evaluation index, and output the autism spectrum disorder recognition result based on the evaluation index; The step of training the anisotropic fractional image generation model using the training set data includes: The training set data is input into the PT-GAN model for training. The generator learns the mapping relationship from regular sequences to generate anisotropic fractional images, and the discriminator is used to collaboratively guide and improve the realism and structural accuracy of the generated images. The steps of constructing anisotropic fraction templates for children and performing template standardization include: Target anisotropic fraction images are selected from the subject images without head motion artifacts. All anisotropic fraction images are aligned to the target image after linear and nonlinear registration and merged to generate a four-dimensional image to construct an anisotropic fraction template for children. The anisotropic fraction templates for children were registered to the anisotropic fraction templates and brain atlases provided by JHU, and all anisotropic fraction images were further registered to the preschool template space using the FSL toolkit to achieve image standardization. The steps for extracting the anisotropic fraction skeleton and anisotropic fraction values ​​include: The anisotropic fraction images after registration are averaged to generate an average anisotropic fraction image and an anisotropic fraction skeleton is extracted. The images of each subject are projected onto the skeleton to form standardized 4D skeleton data. The true and generated anisotropy scores for each subject were extracted using a standardized JHU brain atlas.

2. The autism spectrum disorder identification method based on image generation as described in claim 1, characterized in that, The preprocessing steps for the original T1WI, T2WI, T2-FLAIR, and anisotropic fractional map DICOM data include: Convert the original T1WI, T2WI, T2-FLAIR and anisotropic fractional plot DICOM data into NIfTI format; Register the anisotropic fractional map, T2WI, and T2-FLAIR images to the T1WI image and unify their sizes; Remove the background and keep the foreground brain portion; z-score normalization was performed on the foreground brain region; Cross-sectional slices of T1WI, T2WI, T2-FLAIR, and anisotropic fractional images were extracted and preprocessed.

3. The autism spectrum disorder identification method based on image generation as described in claim 1, characterized in that, The steps of training the autism spectrum disorder identification model using the anisotropic scores, selecting the best-performing model, and saving the parameters include: The obtained anisotropic scores are input into the AutoGluon classification framework for training. Five-fold cross-validation is used to optimize the model hyperparameters on the training set, and the best-performing model is selected and its parameters are saved.

4. The autism spectrum disorder identification method based on image generation as described in claim 1, characterized in that, The steps for calculating the evaluation indicators include: Calculate the accuracy, sensitivity, precision, and area under the receiver operating characteristic curve.

5. An image-based autism spectrum disorder identification system, using the method as described in any one of claims 1 to 4, characterized in that: It includes a data preprocessing module, an anisotropic fractional image generation model training module, an anisotropic fractional feature extraction module, an autism spectrum disorder recognition model training module, an anisotropic fractional image generation model testing module, and an autism spectrum disorder recognition module; among which, The data preprocessing module is used to preprocess the original T1WI, T2WI, T2-FLAIR and anisotropic fractional graph DICOM data, and divide them into training set and test set. The anisotropic fractional image generation model training module is used to train the anisotropic fractional image generation model using the training set data. The anisotropic fraction feature extraction module is used to construct anisotropic fraction templates for children and perform template standardization, and extract anisotropic fraction skeletons and anisotropic fraction values. The autism spectrum disorder recognition model training module is used to train the autism spectrum disorder recognition model using the anisotropy scores, select the best performing model, and save the parameters. The anisotropic fractional image generation model testing module is used to input the test set into the trained anisotropic fractional image generation model to obtain anisotropic fractional generated images. The autism spectrum disorder recognition module is used to extract anisotropic fraction features from the anisotropic fraction generated image, input the anisotropic fraction features into the trained autism spectrum disorder recognition model, calculate the evaluation index, and output the autism spectrum disorder recognition result based on the evaluation index.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 4.

Images