Detection of Mild Cognitive Impairment by Facial Videos

In this study, we proposed a two-stream ConvNet model to detect the mild cognitive impairment (MCI) using facial videos. The image frame containing the facial spatial information and the stacked optical flow fields containing the motion information were extracted from facial videos. Both were input to the two-stream CovnNet model to predict MCI. The experimental results showed that the validation accuracy reaches 91%. This finding indicates that an automatic, non-invasive, and inexpensive MCI screening methods from facial videos is feasible.