Machine learning model may help improve care in Rett syndrome

A new machine learning model that incorporates cardiac activity and body movement may help estimate the severity of Rett syndrome symptoms, enabling researchers to identify effective treatments, according to findings published in PLOS ONE.

“As of 2022, no clinically meaningful disease-modifying treatments exist for patients with Rett syndrome,” Pradyumna Byappanahalli Suresha, MS, PhD, of Georgia Institute of Technology and Emory University at the time of the study, and colleagues wrote. “We instead rely on multiple therapeutics and symptomatic treatment strategies geared towards managing respiratory ailments, treating seizures, improving gastrointestinal function and improving motor skills. As new drugs and therapeutics are discovered, the need for objective measures that can be used in clinical trials increases.”

Suresha and colleagues enrolled 20 patients with Rett syndrome to wear a biosensor patch (BioStamp nPoint, MC10 Inc.) for a 48-hour period. Patients were required to wear the patch for each visit they attended, resulting in 32 visits with associated data.

The researchers used the BioStamp patch to evaluate heart rate variability with ECG and movement with three-axis acceleration data. These data — in combination with multiscale transfer entropy (MSTE) metrics and multiscale network representation (MSNR) — were used to categorize Rett syndrome symptoms as high or low severity on the Clinical Global Impression – Severity.

In total, there were 18 visits by 10 patients with high-severity symptoms and 14 visits by 11 patients with low-severity symptoms.

Analyses revealed that compared with using heart rate variability data alone (area under the curve = 0.76), incorporating MSTE and MSNR improved the estimation of symptom severity by 21% (area under the curve = 0.92).

“This algorithm provides an objective metric that could be used to automatically assess the effect of a medication or other intervention on the symptoms experienced by a Rett syndrome sufferer,” Gari D. Clifford, DPhil, chair and professor in the department of biomedical informatics at Emory University School of Medicine and professor of biomedical engineering at Georgia Institute of Technology, said in a related press release. “We are excited that these biomarkers could potentially allow for more personalized and effective treatment in this population, and perhaps others.”

Reference:

• New study shows how machine learning can improve care for people with Rett syndrome. https://news.emory.edu/stories/2023/03/hs_machine_learning_rett_syndrome_01-03-2023/story.html. Published March 1, 2023. Accessed March 7, 2023.