A convolutional autoencoder for identification of mild traumatic brain injury

Mild traumatic brain injury (mTBI) is the most common form of traumatic brain injury (TBI), yet its timely diagnosis has remained challenging due to lack of established criteria and biomarkers. Widefield optical imaging of neuronal populations over the cerebral cortex in animals provides a unique opportunity to study injury-induced alterations of the brain function. Using a convolutional autoencoder (CAE), this paper aims to develop a framework for detecting mTBI from calcium imaging data to identify the most informative injury-related features. A support vector machine is then trained to classify healthy and injured subjects, and an average classification accuracy of 96.47% is obtained for the best case scenario. Our results suggest that the spatial features obtained through CAE can discriminate the injured and healthy brains better than naive convolutional neural network (CNN).