Fig 1.
The entire process of this experiment.
Details of the CNN for extracting features from static and dynamic FC parts are illustrated in Figs 3 and 4 separately. The left figure in the raw data part is republished from [26] under a CC BY license, with permission from Pixabay, original copyright 2017. The right figure in the raw data part is republished from [27] under a CC BY license, with permission from Pixabay, original copyright 2015.
Fig 2.
The distribution of subjects in the dataset.
Table 1.
Distribution of the data from rs-fMRI ABIDE database used in this study.
Fig 3.
The entire process of processing static functional connectivity.
Fig 4.
The entire process of processing dynamic functional connectivity.
Fig 5.
The performance of different classifiers.
Fig 6.
The performance of the CNN-SVM model on different gender.
Fig 7.
The performance of the CNN-SVM model on different age range.
Fig 8.
The performance of the CNN-SVM model on data from different sites.
Table 2.
Performance comparison of multiple papers based on ABIDE dataset.
DNN: Deep neural network; DBN: Deep Belief Network; CNNG:convolutional neural network and gate recurrent unit; BNC-DGHL:a brain network classification method based on deep graph hashing learning; C-GAN:Conditional Generative Adversarial Network; A-GCL: an adversarial self-supervised graph neural network based on graph contrastive learning; AWSO: the Adam war strategy optimization.
Fig 9.
The most discriminating brain areas related to ASD.
Table 3.
The discriminating brain areas of static FC and dynamic FC.
Fig 10.
The Shapley value corresponds to SRS scores, static FC and dynamic FC from random selected subjects.