We propose a deep learning-based DWI connectome (DWIC) analytic method, characterized by convolutional neural network combined with graph convolutional network. This method trained DWIC features to predict the severity of expressive and receptive language impairment, defined by the clinical evaluation of language fundamentals test. It outperformed other state-of-the-art deep learning approaches in predicting the expressive/receptive language scores in children with focal epilepsy. It also demonstrated the smallest prediction error without a noticeable variation in the random permutation test. Further investigation is warranted to determine the feasibility of a DWIC-based prognostic biomarker of language impairment in clinical practice.
This abstract and the presentation materials are available to members only; a login is required.