New AI Tool May Measure Myasthenia Gravis Ptosis on Smartphone Videos

Researchers developed an artificial intelligence (AI)-based tool that can measure ptosis, or eyelid drooping, in patients with myasthenia gravis (MG). The tool uses selfie video clips recorded on a smartphone and is able to predict margin reflex distance 1 (MRD1), which is an accepted clinical measure of ptosis that is normally assessed using a hand-held ruler, with a strong correlation between ground truth and predicted value.

“Our work demonstrates the feasibility of automated ptosis assessment from frames of video data collected remotely over a broad range of smartphones,” the researchers wrote in a report that they published in the journal Digital Biomarkers. “The model developed holds promise as a patient-centric tool for objective, remote measurement of this important MG symptom.”

Read more about the symptoms of MG

Ptosis is an MG symptom that is widely used to follow disease progression. The model presented here allows the MRD1 to be measured automatically in video clips recorded by patients themselves using their own smartphones.

The team, led by Francesca Rinaldo, MD, PhD, from Sharecare Inc in Atlanta, Georgia, asked participants to perform an eyelid fatiguability exercise, while filming the selfie videos on their smartphones.

A total of 664 images were obtained and analyzed. The results of the analysis showed there was a strong correlation between the ground truth MRD1, which was calculated from eye landmark annotations in the video frames using the horizontal visible iris diameter of the human eye, and the MRD1 predicted by the model. 

The team reported the mean absolute error was 0.822 mm, the mean of differences was −0.256 mm, and the 95% limits of agreement were −0.214 mm to 1.768 mm. 

“On data generated under highly challenging real-world conditions from a variety of different smartphone devices, the model predicts MRD1 with a strong correlation between ground truth and predicted MRD1,” the researchers wrote.

This article originally appeared on Rare Disease Advisor