The authors' approach integrated two streams of data: facial frame features and optical flow, processed through a transformer-based architecture. Various configurations of embedding dimensions, dense layer sizes, and attention heads were examined to enhance model performance. The final model, trained on 183 Parkinson disease patients, attained an accuracy of 86% in differentiating between ON- and OFF-medication state. Moreover, uniform classification performance was obtained across various stages of Parkinson disease severity, as expressed by the Hoehn and Yahr scale. These values highlight the potential of authors' model as a non-invasive, cost-effective instrument for clinicians to remotely and accurately detect patients' response to treatment from early to more advanced Parkinson disease stages.
