Looking into the Future of Myasthenia Gravis Management Using Artificial Intelligence


Clinical Management

Poster Number: M261


Shaweta Khosa, MD, Olive View-UCLA Medical Center, Namrata Shetty, BS, University of California, San Francisco, Gurveer Khosa, MBBS, Adesh Institute of Medical Sciences & Research

Artificial intelligence’s (AI) integration into medicine has revolutionized patient care, increased accessibility, and transformed health outcomes. Recent AI-based models in neurology have enabled a multifaceted approach to evaluating Myasthenia Gravis (MG):

(1)Ptosis, a common symptom, is often measured using margin-reflex distance 1 (MRD1) with a manual ruler. New AI-models have automated MRD1 evaluation via patient selfies on smartphones.
(2)Researchers trained a deep learning (DL) model to evaluate disease severity by quantifying facial weakness using facial recognition software.

To describe novel AI-based tools for Myasthenia Gravis management.

(1)In a three-month prospective observational study, 82 individuals with MG yielded 664 self-recorded smartphone videos during an eyelid fatigability exercise. Non-clinical annotators established ground truth for MRD1 and video frame quality. An artificial neural network, trained as an MRD1 measurement tool, analyzed the images.
(2)A cross-sectional study assessed video recordings of 69 healthy controls (HC) and 70 MG patients. Using FaceReader software, facial weakness was analyzed, quantifying six emotions. Then, a DL model was trained to classify MG diagnosis and severity using 50 HC and 50 MG videos.

(1)The MRD1 ground truth was determined from eye landmark annotations in video frames using the visible iris diameter. Findings showed a notable correlation between the ground truth MRD1 and predicted values. The reported mean absolute error was 0.822 mm.
(2)FaceReader findings indicated that MG individuals exhibited significantly lower expressions of anger (p=0.026), fear (p=0.003), and happiness (p<0.001) compared to HC. The DL model showed a 76% accuracy in diagnostic ability and 80% accuracy in categorizing disease severity. Conclusions: To mitigate myasthenic crisis risk, accessible tools for tracking symptoms are vital for patients and providers. The emergence of AI-driven models for automated symptom assessment holds promise for precise, patient-focused tools to enhance MG management in the future.