Objective: To develop a computer vision-based pipeline for extraction of digital biomarkers from hand drawing tasks and to evaluate their correlation with clinical ataxia severity.

Background: Scale for the Assessment and Rating of Ataxia (SARA) is the gold standard for quantifying ataxia severity, yet it relies on clinical observation and may not fully capture granular deficits. Kinematic analysis of hand drawing tasks, including Archimedes spirals and straight-line drawings, has shown promise as a sensitive and objective measure of motor control in various movement disorders, though, application to cerebellar ataxia remains limited.

Method: This study was approved by the Institutional Review Board at University of California, Irvine. Cerebellar ataxia patients underwent standardized paper-based drawing tasks consisting of 4 spiral tracings and 6 straight-line tasks. Clinical severity was assessed using the total SARA score and the upper-limb SARA subscore (finger chase, nose-finger, fast alternating hand movements). Images were processed using a custom computer vision pipeline. 11 total features representing spatial control, tremor characteristics, and spiral complexity measures were extracted. Spearman correlations with Benjamini-Hochberg false discovery rate (FDR) correction were computed against SARA score.

Results: We enrolled 11 patients (mean age 57.4±13.9) with diagnoses of SCA6 (n=3), SCA3 (n=2), MSA-C (n=1), GAD65-associated cerebellar ataxia (n=2), and others (n=3). Total SARA scores ranged from 5.5 to 26. Three line-derived features were significantly correlated with total SARA score after FDR correction: root mean square (RMS) deviation (rho=0.82, p-adj=0.024), coverage ratio (rho=-0.77, p-adj=0.031), and tremor RMS (rho=0.73, p-adj=0.037). No features survived FDR correction when correlated against the upper-limb SARA subscore.

Conclusion: In this preliminary study, digital biomarkers derived from hand drawing tasks demonstrated significant correlations with total SARA severity in ataxia patients, while no features correlated to upper-limb SARA subscores. Our study is limited by small sample size and require validation in larger cohorts. Nonetheless, a key contribution is the curation of a dedicated ataxia hand drawing dataset, and future analyses leveraging image-based machine learning models and sophisticated statistical analysis is warranted.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/archimedes-spiral-and-straight-line-drawings-as-digital-biomarkers-in-cerebellar-ataxia-a-preliminary-study/