WO2023104173A1 - Autism classifier construction method and system based on functional magnetic resonance images of human brains - Google Patents

Abstract

Disclosed in the present invention is an autism classifier construction method and system based on functional magnetic resonance images of human brains. The method comprises: performing pre-processing on functional magnetic resonance images of brains of a normal subject and an autistic subject, so as to obtain feature vectors; for each feature of the feature vectors, performing a feature selection method based on a gradient distribution curve difference, so as to obtain a significant feature, combining the obtained significant features into a new feature vector, and taking the new feature vectors of all the subjects as a training sample, the training sample comprising a training set and a validation set; and performing pre-training on a variational auto-encoder by using the training set, and after pre-training ends, migrating an encoder parameter of the variational auto-encoder to a multi-layer perceptron and performing supervised training on the multi-layer perceptron by using the training set, performing fine adjustment on parameters of the multi-layer perceptron, performing assessment by using the validation set after each round of training until a set number of rounds of training is reached or the assessment precision of the validation set reaches standards, and taking the multi-layer perceptron, the parameters of which have been subjected to fine adjustment, as an autism classifier.

Description

Method and system for constructing autism classifier based on functional magnetic resonance images of human brain technical field

The invention relates to the field of medical image classification, in particular to a method and system for constructing an autism classifier based on human brain functional magnetic resonance images.

Background technique

Autism spectrum disorder (ASD) is a common complex neurodevelopmental disorder that occurs in early childhood and is characterized by social interaction and restricted repetitive sensorimotor behaviors. Traditional symptom-based classification methods cannot reveal the underlying pathogenesis of ASD and are therefore often unreliable. With the development of neuroimaging, non-invasive brain imaging technology has become a powerful tool to study and reveal neurological diseases such as ASD. Among them, functional magnetic resonance (rs-fMRI) measurement of blood oxygen level-related change signals can help clinicians and neuroscientists to visually assess the functional characteristics or properties of the brain, and has become a powerful tool for early classification of ASD. In recent years, rs-fMRI combined with machine learning and deep learning techniques for ASD classification has achieved good results and has become one of the most promising imaging methods for ASD classification.

In recent years, machine learning (including deep learning) methods have been widely used in the classification of autism. In recent years, machine learning (including deep learning) methods have been widely used in the prediction and research of cell images. Plitt et al. ^[1] only used support vector machines to classify the autism samples and normal samples of the ABIDE dataset, and achieved a classification accuracy of 69%. Heinsfeld et al. ^[2] used stacked autoencoders (SAE) and fully connected neural networks to achieve the highest prediction accuracy of 70% on the public dataset ABIDE I at that time. Parisot et al. ^[3] proposed a general framework for brain analysis of large-scale populations using imaging and non-imaging information. The framework is based on graph convolutional neural network (GCN), and achieved a classification accuracy of 70.4% on the ABIDE dataset. . Zhi-An Huang et al. ^[4] achieved a prediction accuracy of 76.4% on the ABIDE I dataset using a deep belief network (DBN). In addition, feature selection methods are often combined with machine learning to achieve better classification performance.

However, the main challenge of the current functional magnetic resonance imaging of the brain is that the preprocessed data contains a large amount of redundant information, which will lead to the deterioration of the performance of the classification model. The accuracy of the machine learning classification model based on functional magnetic resonance imaging still needs to be improved. The current classification model cannot flexibly adjust the sensitivity and specificity, making it unable to adapt to some specific practical needs.

1. Plitt, M., Barnes, K.A., and Martin, A. (2015). Functional connectivity classification of autism identifies highly predictive brain features but falls short of biomarker standards. YNICL 7, 359–366. doi: 10.1016/j .nicl. 2014.12.013

2. Heinsfeld, A.S.; Franco, A.R.; Craddock, R.C.; Buchweitz, A.; Meneguzzi, F., Identification of autism spectrum disorder using deep learning and the ABIDE dataset. NeuroImage: Clinical 2018,17,16-23.

3. Parisot, S.; Ktena, S.I.; Ferrante, E.; Lee, M.; Guerrero, R.; Glocker, B.; Rueckert, D., Disease prediction using graph convolutional networks: application to autism spectrum disorder and Alzheimer's disease.Medical image analysis 2018,48,117-130.

4. Huang, Z.-A.; Zhu, Z.; Yau, C.H.; Tan, K.C., Identifying autism spectrum disorder from resting-state fMRI using deep belief network. IEEE Transactions on Neural Networks and Learning Systems 2020.

Contents of the invention

The purpose of the present invention is to provide a method and system for constructing an autism classifier based on functional magnetic resonance images of the human brain to overcome the defects of the prior art. The present invention can select useful features that are more refined than conventional methods, and It has high precision, sensitivity, specificity and training speed.

To achieve the above object, the present invention adopts the following technical solutions:

A method for constructing an autism classifier based on functional magnetic resonance images of the human brain, including:

The feature vectors are obtained by preprocessing the brain fMRI images of normal subjects and autistic subjects;

Perform feature selection based on the difference of the gradient distribution curve for each feature of the feature vector to obtain salient features, form the new feature vectors with the obtained salient features, and use the new feature vectors of all subjects as training samples, the training samples Including training set and validation set;

The training set is used to pre-train the variational autoencoder. After the pre-training, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the training set is used to perform supervised training on the multi-layer perceptron. The parameters of the multi-layer perceptron are fine-tuned, and after each round of training, the verification set is used for evaluation until the training reaches the set number of rounds, and the multi-layer perceptron with fine-tuned parameters is used as an autism classifier.

Further, the process of preprocessing the brain functional magnetic resonance images of normal subjects and autistic subjects to obtain feature vectors is specifically: based on the time series extracted from human brain functional magnetic resonance images, calculate the time series of interest Functional connections between regions, all functional connections form a feature vector, and each functional connection is used as a feature of the feature vector;

The specific calculation method of the functional connection is: calculating the Pearson correlation coefficients of the time series corresponding to all pairwise different regions of interest.

Further, the feature selection based on the gradient distribution curve difference is performed on each feature of the feature vector to obtain the salient features, specifically:

For each feature in the feature vector, calculate the DSDC score;

The features whose DSDC score is greater than the preset threshold are selected as salient features, and the features whose DSDC score is less than or equal to the preset threshold are discarded.

Further, the DSDC score is calculated for each feature in the feature vector, specifically:

Divide the range of features in all feature vectors into several equal-length subintervals;

The DSDC score of a feature is calculated by the following formula:

In the formula, b ₀ and b ₁ are the lower bound and upper bound of the value of the feature; δ is the length of the subinterval; i represents the i-th subinterval, and n _i ⁺ and n _i ^- are in the [i-δ, i) interval is the number of autistic subjects and normal subjects, N ⁺ and N ^– is the number of autistic subjects and normal subjects in the training set.

Further, the supervised training of the multi-layer perceptron by using the training set is specifically training without constraints, training with sensitivity constraints, or training with specificity constraints;

When using unconstrained training, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the parameters that have been trained in this round are saved, otherwise they are not saved;

When training with sensitivity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set increases, the difference between the evaluation sensitivity of the verification set and the evaluation specificity of the verification set increases, and the improvement value is less than the preset value , then save the parameters after this round of training, otherwise not save;

When training with specificity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the difference between the evaluation specificity of the verification set and the evaluation sensitivity of the verification set is increased, and the improvement value is less than the preset value, save the parameters after this round of training, otherwise do not save.

An autism classifier construction system based on functional magnetic resonance images of the human brain, including a preprocessing module, a training set acquisition module and a training module, in which:

Preprocessing module: used to preprocess the brain functional magnetic resonance images of normal subjects and autistic subjects to obtain feature vectors;

Training set acquisition module: used to perform feature selection based on the difference of the gradient distribution curve for each feature of the feature vector, to obtain salient features, to form new feature vectors with the obtained salient features, and to use the new feature vectors of all subjects as Training samples, the training samples include a training set and a verification set;

Training module: It is used to pre-train the variational autoencoder using the training set. After the pre-training is completed, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the multi-layer perceptron is effectively implemented using the training set. Supervised training, fine-tuning the parameters of the multi-layer perceptron, and using the verification set for evaluation after each round of training, until the training reaches the set number of rounds, the multi-layer perceptron with parameter fine-tuning is used as an autism classification device.

Further, the process of preprocessing the brain functional magnetic resonance images of normal subjects and autistic subjects to obtain feature vectors is specifically: based on the time series extracted from human brain functional magnetic resonance images, calculate the time series of interest Functional connections between regions, all functional connections form a feature vector, and each functional connection is used as a feature of the feature vector;

The specific calculation method of the functional connection is: calculating the Pearson correlation coefficients of the time series corresponding to all pairwise different regions of interest.

Further, the feature selection based on the gradient distribution curve difference is carried out to each feature of the feature vector to obtain the salient features, specifically:

For each feature in the feature vector, calculate the DSDC score;

The features whose DSDC score is greater than the preset threshold are selected as salient features, and the features whose DSDC score is less than or equal to the preset threshold are discarded.

Further, the DSDC score is calculated for each feature in the feature vector, specifically:

Divide the range of features in all feature vectors into several equal-length subintervals;

The DSDC score of a feature is calculated by the following formula:

In the formula, b ₀ and b ₁ are the lower bound and upper bound of the value of the feature; δ is the length of the subinterval; i represents the i-th subinterval, and n _i ⁺ and n _i ^- are in the [i-δ, i) interval is the number of autistic subjects and normal subjects, N ⁺ and N ^– is the number of autistic subjects and normal subjects in the training set.

Further, the supervised training of the multi-layer perceptron by using the training set is specifically training without constraints, training with sensitivity constraints, or training with specificity constraints;

When using unconstrained training, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the parameters that have been trained in this round are saved, otherwise they are not saved;

When training with sensitivity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set increases, the difference between the evaluation sensitivity of the verification set and the evaluation specificity of the verification set increases, and the improvement value is less than the preset value , then save the parameters after this round of training, otherwise not save;

When training with specificity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the difference between the evaluation specificity of the verification set and the evaluation sensitivity of the verification set is increased, and the improvement value is less than the preset value, save the parameters after this round of training, otherwise do not save.

Compared with the prior art, the present invention has the following beneficial technical effects:

First of all, the method of the present invention adopts the feature selection method based on the step distribution curve difference (DSDC) to select more refined useful features than the conventional methods; secondly, the accuracy, sensitivity, specificity and training speed of the classifier constructed by the present invention are all the same. It exceeds the state-of-the-art experimental results in previous studies based on the same data set; finally, the present invention can flexibly adjust the sensitivity and specificity of the classifier by imposing constraints during the training process, enabling the model to use different practical application requirements.

Description of drawings

The accompanying drawings in the description are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention.

Fig. 1 is a schematic flow chart of the method of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only It is an embodiment of a part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first" and "second" in the description and claims of the present invention and the above drawings are used to distinguish similar objects, but not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the invention described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having", as well as any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to the expressly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

The invention provides a method for constructing an autism classifier based on functional magnetic resonance images of the human brain, which specifically includes:

The first step is to preprocess the brain fMRI images of 455 normal subjects and 477 autistic subjects to obtain feature vectors, specifically: based on time series calculations extracted from human brain fMRI images Functional connections between regions of interest, each functional connection is used as a feature, and all functional connections form a feature vector. The specific calculation method of functional connectivity is to calculate the Pearson correlation coefficient of the time series corresponding to all two different regions of interest.

In the second step, based on the feature vector obtained in the first step, a feature selection method based on the difference of the gradient distribution curve is performed. For each feature, calculate the DSDC score of the feature: first divide the range of features in all feature vectors into 20 equal-length subintervals, then count the number of samples in each subinterval, and divide the number by Normalize with the total number of samples belonging to the category, and generate a ladder-shaped distribution curve according to the normalized value of the number of samples in each interval. This process is equivalent to a coarse-grained fitting of the original distribution curve of the feature, that is, the DSDC score of the feature is calculated by the following formula.

In the formula, b ₀ and b ₁ are the lower and upper bounds of the features; δ is the length of the sub-interval; i represents the i-th sub-interval, and the value of i is an integer within the range of [1,20]; n _i ⁺ and n _i ^- the number of autistic subjects and normal subjects falling in the [i-δ,i) interval, N ⁺ and N ^– the number of autistic subjects and normal subjects in the training set .

The present invention pre-sets a filtering threshold, selects features with DSDC scores greater than the threshold as salient features, discards features with DSDC scores less than or equal to the threshold, and uses new feature vectors of all subjects as training samples, the training samples Includes training set and validation set.

In the third step, the present invention simplifies the encoder structure of the variational autoencoder, uses the same neural network to generate two parameters of the potential space, and uses the training set to pre-train the variational autoencoder.

In the fourth step, after the pre-training is completed, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the training set is used for supervised training of the multi-layer perceptron, and the parameters of the multi-layer perceptron are fine-tuned. And after each round of training, the verification set is used for evaluation until the training reaches the set number of rounds, and the multi-layer perceptron with fine-tuned parameters is used as an autism classifier for autism classification. In the multi-layer perceptron training process, the present invention designs two constraint conditions (sensitivity constraint and specificity constraint) as options. In the training process of multi-layer perceptron, adding sensitivity constraints and specificity constraints can greatly improve sensitivity and specificity at the cost of a small amount of accuracy reduction.

The main function of sensitivity constraints and specificity constraints is to determine whether the parameters obtained by the classifier after each round of training are saved. In the present invention, one-tenth of the data in the training set is taken out as a verification set, and the verification set is mainly used to evaluate the training model of each round without participating in the training process. When there are no constraints, after a certain round of training, if the evaluation accuracy of the verification set is improved, the parameters after this round of training will be saved, otherwise they will not be saved. When using sensitivity constraints, after a certain round of training, if the evaluation accuracy of the verification set is improved, and the difference between the evaluation sensitivity of the verification set and the evaluation specificity of the verification set is improved and less than 0.3, the parameters after this round of training will be saved, otherwise not saved. When specificity constraints are used, after a certain round of training, if the evaluation accuracy of the verification set is improved, and the difference between the evaluation specificity of the verification set and the evaluation sensitivity of the verification set is improved and less than 0.3, the parameters after this round of training will be saved, otherwise not. Using sensitivity constraints can help improve model sensitivity; using specificity constraints can help improve model specificity.

The present invention also provides an autism classifier construction system based on functional magnetic resonance images of the human brain, including a preprocessing module, a training set acquisition module and a training module, wherein:

Preprocessing module: used to preprocess the brain functional magnetic resonance images of normal subjects and autistic subjects to obtain feature vectors;

Training set acquisition module: used to perform a feature selection method based on the difference of gradient distribution curves for each feature of the feature vector to obtain salient features, form the obtained salient features into a new feature vector, and combine the new feature vectors of all subjects As a training sample, the training sample includes a training set and a verification set;

Training module: It is used to pre-train the variational autoencoder using the training set. After the pre-training is completed, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the multi-layer perceptron is effectively implemented using the training set. Supervised training, fine-tuning the parameters of the multi-layer perceptron, and using the verification set for evaluation after each round of training, until the training reaches the set number of rounds, the multi-layer perceptron with parameter fine-tuning is used as an autism classification device.

Utilize the present invention to pass through experiments and tests on the ABIDE I data set, and the experimental results show that the accuracy (78.12%) of the classifier constructed by the present invention (without constraints), sensitivity (87.20%) (with sensitivity constraints), specificity Both accuracy (88.55%) (using specificity constraints) and training speed (10-fold cross-validation took 85 seconds) surpassed the state-of-the-art experimental results in previous studies based on the same dataset.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention rather than limiting its protection scope, although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: Those skilled in the art can still make various changes, modifications or equivalent replacements to the specific embodiments of the invention after reading the present invention, but these changes, modifications or equivalent replacements are all within the protection scope of the pending claims of the invention.

Claims (10)

The method for constructing an autism classifier based on human brain functional magnetic resonance images is characterized in that it includes: The feature vectors are obtained by preprocessing the brain fMRI images of normal subjects and autistic subjects; Perform feature selection based on the difference of the gradient distribution curve for each feature of the feature vector to obtain salient features, form the new feature vectors with the obtained salient features, and use the new feature vectors of all subjects as training samples, the training samples Including training set and validation set; The training set is used to pre-train the variational autoencoder. After the pre-training, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the training set is used to perform supervised training on the multi-layer perceptron. The parameters of the multi-layer perceptron are fine-tuned, and after each round of training, the verification set is used for evaluation until the training reaches the set number of rounds, and the multi-layer perceptron with fine-tuned parameters is used as an autism classifier. The method for constructing an autism classifier based on human brain functional magnetic resonance images according to claim 1, wherein the preprocessing of the brain functional magnetic resonance images of normal subjects and autistic subjects is obtained The process of eigenvectors is specifically: based on the time series extracted from human brain functional magnetic resonance images, the functional connections between regions of interest are calculated, all functional connections form a eigenvector, and each functional connection is used as a feature of the eigenvector; The specific calculation method of the functional connection is: calculating the Pearson correlation coefficients of the time series corresponding to all pairwise different regions of interest. The method for constructing an autism classifier based on functional magnetic resonance images of the human brain according to claim 1, wherein the feature selection based on gradient distribution curve differences is carried out to each feature of the feature vector to obtain salient features, Specifically: For each feature in the feature vector, calculate the DSDC score; The features whose DSDC score is greater than the preset threshold are selected as salient features, and the features whose DSDC score is less than or equal to the preset threshold are discarded. The autism classifier construction method based on human brain functional magnetic resonance image according to claim 3, is characterized in that, for each feature in the feature vector, calculate the DSDC score, specifically: Divide the range of features in all feature vectors into several equal-length subintervals; The DSDC score of a feature is calculated by the following formula:

In the formula, b ₀ and b ₁ are the lower bound and upper bound of the value of the feature; δ is the length of the subinterval; i represents the i-th subinterval, and n _i ⁺ and n _i ^- are in the [i-δ, i) interval is the number of autistic subjects and normal subjects, N ⁺ and N ^– is the number of autistic subjects and normal subjects in the training set. The method for constructing an autism classifier based on functional magnetic resonance images of the human brain according to claim 1, wherein the supervised training of the multi-layer perceptron using the training set is specifically unrestricted training, using Sensitivity constraint training or training with specificity constraints; When using unconstrained training, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the parameters that have been trained in this round are saved, otherwise they are not saved; When training with sensitivity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set increases, the difference between the evaluation sensitivity of the verification set and the evaluation specificity of the verification set increases, and the improvement value is less than the preset value , then save the parameters after this round of training, otherwise not save; When training with specificity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the difference between the evaluation specificity of the verification set and the evaluation sensitivity of the verification set is increased, and the improvement value is less than the preset value, save the parameters after this round of training, otherwise do not save. The autism classifier construction system based on human brain functional magnetic resonance images is characterized in that it includes a preprocessing module, a training set acquisition module and a training module, wherein: Preprocessing module: used to preprocess the brain functional magnetic resonance images of normal subjects and autistic subjects to obtain feature vectors; Training set acquisition module: used to perform feature selection based on the difference of the gradient distribution curve for each feature of the feature vector, to obtain salient features, to form new feature vectors with the obtained salient features, and to use the new feature vectors of all subjects as Training samples, the training samples include a training set and a verification set; Training module: It is used to pre-train the variational autoencoder using the training set. After the pre-training is completed, the encoder parameters of the variational autoencoder are transferred to the multi-layer perceptron, and the multi-layer perceptron is effectively implemented using the training set. Supervised training, fine-tuning the parameters of the multi-layer perceptron, and using the verification set for evaluation after each round of training, until the training reaches the set number of rounds, the multi-layer perceptron with parameter fine-tuning is used as an autism classification device. The autism classifier construction system based on human brain functional magnetic resonance images according to claim 6, wherein the preprocessing of the brain functional magnetic resonance images of normal subjects and autistic subjects is obtained The process of eigenvectors is specifically: based on the time series extracted from human brain functional magnetic resonance images, the functional connections between regions of interest are calculated, all functional connections form a eigenvector, and each functional connection is used as a feature of the eigenvector; The specific calculation method of the functional connection is: calculating the Pearson correlation coefficients of the time series corresponding to all pairwise different regions of interest. The autism classifier construction system based on functional magnetic resonance images of the human brain according to claim 6, wherein the feature selection based on gradient distribution curve differences is carried out to each feature of the feature vector to obtain salient features, Specifically: For each feature in the feature vector, calculate the DSDC score; The features whose DSDC score is greater than the preset threshold are selected as salient features, and the features whose DSDC score is less than or equal to the preset threshold are discarded. The autism classifier construction system based on functional magnetic resonance images of the human brain according to claim 8, wherein, for each feature in the feature vector, the DSDC score is calculated, specifically: Divide the range of features in all feature vectors into several equal-length subintervals; The DSDC score of a feature is calculated by the following formula:

In the formula, b ₀ and b ₁ are the lower bound and upper bound of the value of the feature; δ is the length of the subinterval; i represents the i-th subinterval, and n _i ⁺ and n _i ^- are in the [i-δ, i) interval is the number of autistic subjects and normal subjects, N ⁺ and N ^– is the number of autistic subjects and normal subjects in the training set. The autism classifier construction system based on human brain functional magnetic resonance images according to claim 6, wherein the supervised training of the multi-layer perceptron using the training set is specifically unrestricted training, using Sensitivity constraint training or training with specificity constraints; When using unconstrained training, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the parameters that have been trained in this round are saved, otherwise they are not saved; When training with sensitivity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set increases, the difference between the evaluation sensitivity of the verification set and the evaluation specificity of the verification set increases, and the improvement value is less than the preset value , then save the parameters after this round of training, otherwise not save; When training with specificity constraints, the verification set is used for evaluation after each round of training. If the evaluation accuracy of the verification set is improved, the difference between the evaluation specificity of the verification set and the evaluation sensitivity of the verification set is increased, and the improvement value is less than the preset value, save the parameters after this round of training, otherwise do not save.

Images