F-Dit-V: An Automated Video Classification Tool For Facial Weakness Detection

Facial weakness is a common presenting sign of several neurological diseases including stroke, traumatic brain injury (TBI), and Bell's palsy. Tools to improve the accuracy of facial weakness detection can prompt quicker evaluation into these diseases possibly resulting in earlier diagnoses. In this study, we propose an automated video classification detection tool, Facial Deficit Identification Tool for Videos (F-DIT-V), for facial weakness detection. This tool exploits Histogram of Oriented Gradients (HOG) features to perform more accurate facial weakness detection for a given video. Using experimental data we demonstrate that F-DIT-V achieves a classification accuracy of 92.9%, precision of 93.6 %, recall of 92.8%, and specificity of 94.2%. F-DIT-V is able to achieve higher and more reliable performance compared to existing (e.g. LBP-TOP, RNN-based) methods which are widely used in previous and current studies for facial weakness video classification. As the proposed camera based analysis system requires no extra hardware, F-DIT-V could be implemented in a low-cost, portable, and easy to use format for generalizability to real world settings.