AI Effective at Predicting Disease Progression in Friedreich’s Ataxia

Artificial Intelligence (AI) has proven more effective than 2 gold-standard clinical scale assessments at precisely predicting gene expression and disease progression in Friedreich’s ataxia (FA), according to new research published in Nature Medicine. Full-body movement data gathered from wearable sensors were accurate at longitudinally predicting the clinical scores in 9 patients with FA.

Participants were divided into a cohort of 9 healthy individuals and a matching control cohort of healthy participants. Both cohorts wore a motion capture suit to record movement data during the 9-hole peg test (9 HPT) and the 8-minute walk (8 MW). The collected data were used to compare changes in movement at baseline, 3 weeks, 3 months, and 9 months. 

After capturing the kinesthetic data, researchers used AI to predict 2 commonly used clinical scale scores: 1) disease progression and 2) Frataxin expression levels by creating a digital avatar of each participant. The algorithm's prediction of disease progression was compared with 2 gold-standard clinical assessments: 1) the Scale for Assessment and Rating of Ataxia (SARA) and 2) the Spinocerebellar Ataxia Functional Index (SCAFI). 

The algorithm was able to produce a more accurate prediction of SARA and SCAFI scores 9 months into the future that was 1.7 times more accurate than SARA alone and 4 times more accurate than SCAFI alone. These scales are less sensitive to the slow progression of FA because they rely on subjective, observational data collected “by eye”.

According to researchers, “Data-derived wearable biomarkers can track personal disease trajectories and indicate the potential of such biomarkers for substantially reducing the duration or size of clinical trials testing disease-modifying therapies and for enabling behavioral transcriptomics.”