Application of ResLSTM in Hypomimia Video Detection for Parkinson's Disease

Hypomimia is one of the primary clinical manifestations of Parkinson's disease, characterized by a reduction in facial expressions and progressive loss of facial animation. In routine clinical diagnosis, the assessment of Hypomimia remains subjective and relies on manual scoring by doctors. With the development of artificial intelligence, digital healthcare has become an essential component of the medical ecosystem. In this paper, we explore a method for diagnosing Parkinson's disease Hypomimia symptoms by analyzing facial videos using deep learning. We collected facial expression videos from PD patients and matched control groups from open-source resources. After preprocessing the video frames, we employed deep learning algorithms to diagnose Hypomimia. Experimental results show that our model achieved an average accuracy of 81.73% in the Hypomimia classification task. This research contributes to the digitized diagnosis of Parkinson's disease based on facial videos and validates our PD Hypomimia diagnostic model through experiments. The research findings in this paper have potential applications in assisting with preliminary home-based symptom assessment.