AI Estimates Myoclonic Jerk Severity Using Video Footage

The severity of epilepsy-related muscle pricking symptoms can be reliably estimated with artificial intelligence, according to a recent study. An artificial intelligence-based tool was used to analyze video recordings of patients with myoclonic epilepsy.

Myoclonus, or sudden muscle spasms, is the most troublesome and drug-responsive symptom in the progressive type 1 myoclonic epilepsy (EPM1), which is part of the Finnish disease heritage. Myoclonus is sensitive to stimuli and varies in severity throughout the day. Lack of sleep and anxiety also contribute to the worsening of myoclonic symptoms. As a result, automatic tools have been sought by physicians and the pharmaceutical industry to improve the reliability of drug response and monitoring of disease progression.

The purpose of this study was to provide a rapid, objective and automated method for assessing myoclonus from video recording. Neuro Event Labs, a health technology company, has developed an algorithm that utilizes learning neural network architectures and pre-taught modeling to identify and track critical points in the body and to model their movements in myoclonia.

Currently, the most established method of assessing myoclonus is the UMRS test panel, in which a physician scores symptoms based on video recording. The study evaluated ten video-recorded UMRS test panels using a novel method based on automatic body part identification and modeling of myoclonic movements. The study showed that automated analysis was successful in detecting and tracking the movement of predetermined body points. In addition, the method successfully detected and scored myoclonic jerks by analyzing fluidity and velocity changes in movements. Automated scoring of myoclonic movements correlates with a patient's UMRS test scoring by a highly experienced medical researcher.

Based on the results, the severity of myoclonic jerks in EPM1 patients can be reliably estimated by the method used in the study. Automatic analysis correlates well with the physician's judgment. The algorithm was able to efficiently model motion fluency and also detected small and fast myoclones that were difficult to measure with sufficient sensitivity.

Reference

Hyppönen et al. (2020) Automatic assessment of the myoclonus severity from videos recorded according to standardized Unified Myoclonus Rating Scale protocol and using human pose and body movement analysis. Seizure. DOI: https://doi.org/10.1016/j.seizure.2020.01.014

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source.