Reliable Objective Assessment of Friedreich Ataxia Through Isolation Forest-Based Anomaly Detection

Objective assessment of motor impairments due to Friedreich Ataxia (FRDA) is vital for monitoring disease progression and for timely therapeutic interventions. The Ataxia Instrumented Measurement Cup (AIM-C), an IMU-based device, enables precise quantification of motor function during a simulated drinking task. However, anomalies in recorded data-caused by irregular task execution or external disturbances-can undermine the reliability of severity predictions. This study presents a robust anomaly detection framework using the concept of Isolation Forest to identify and exclude anomalous segments within the AIM-C data. Additionally, an adaptive median filtering approach, guided by anomaly scores, dynamically adjusts denoising filters to enhance the signal quality. The proposed method significantly improves the reliability of extracted kinematic and kinetic features, leading to higher inter-class correlation (ICC) values for severity predictions. These findings underscore the utility of AIM-C, augmented with anomaly detection and signal optimization techniques, as a reliable mechanism to enhance objective assessment capabilities in FRDA research and clinical practice.Clinical relevance- This study improves the reliability of objective Friedreich's ataxia assessments, providing clinicians with a more consistent and accurate tool for tracking disease progression and evaluating treatment effects.