Building a Quantitative Telemedicine Platform for Myasthenia Gravis: Augmenting the Physical Examination

Myasthenia gravis (MG), a fluctuating autoimmune neuromuscular disease, presents unique challenges for remote assessment due to its reliance on detailed physical examination. To address this, we developed a quantitative telemedicine platform that augments traditional neurological assessments using computer vision, signal processing, and augmented intelligence. In response to the COVID-19 pandemic that motivated a shift to telemedicine evaluations, the Myasthenia Gravis Core Exam (MGCE) was developed by the rare disease consortium MGNet, to provide guidance in the performance of telemedicine clinical encounters. The MGCE includes eight sentinel tasks such as ptosis assessment, sit-to-stand, and speech-based respiratory measures, all amenable to digital capture. To assess the reproducibility and reliability of the MGCE, videos of 52 MG patients across six centers were performed with each having examinations performed twice. We utilized this unique resource to apply machine learning algorithms to extract clinically relevant features from video and audio data, enabling quantitation of continuous variation, in contrast to the categorical measures (mild, moderate, and severe) used in standard clinical assessments. Inter-rater variability prompted the development of a reproducibility of score metric and revealed that variations in examiner instructions and video quality significantly affect reliability. Our findings suggest that a digital examination framework can enhance MG assessment precision, reduce variability in physical examination evaluation, and support the telemedicine examination. This scalable approach has the potential to integrate digital biomarkers into neuromuscular disease care and clinical trials.