Functional connectivity (FC) analysis, which measures the connection between different brain regions, has been widelyused to study brain function and development. However, FC-based analysis breaks the local structure in MRI images,resulting in a challenge for applying advanced deep learning models, e.g., convolutional neural networks (CNN). To fitthe data in a non-Euclidean domain, graph convolutional neural network (GCN) was proposed, which can work on graphsrather than raw images, making it a suitable model for brain FC study. The small sample size is another challenge.Compared with natural images, medical images are usually limited in data sample size. Moreover, labeling medical imagesrequires laborious annotation and is time-consuming. These limitations result in low accuracy and overfitting problem whentraining a conventional deep learning model on medical images. To address this problem, we employed a semi-supervisedGCN with a Laplacian regularization term. By exploiting the between-sample information, semi-supervised GCN canachieve better performance on data with limited sample size. We applied the semi-supervised GCN model to a brainimaging cohort to classify the groups with dierent Wide Range Achievement Test (WRAT) scores. Experimental resultsshowed semi-supervised GCN can improve classification accuracy, demonstrating the superior power of semi-supervisedGCN on small datasets.
展开▼
