PhD Dissertation Proposal Defense: Juhyeon Lee, Advancing Objective Motor Severity Assessment in Cerebellar Ataxias Using Wearable Sensors and Machine Learning
Content
Speaker
Abstract
Cerebellar ataxias are a group of etiologically diverse neurological diseases that cause dysfunction of the cerebellum and related pathways, leading to impaired motor coordination and balance. As most ataxias are progressive, frequent and objective assessment of motor severity is essential for monitoring disease progression and evaluating therapeutic interventions. The lack of disease-modifying treatments for most ataxias further highlights the need for sensitive and frequent assessment to support ongoing clinical trials. Current assessments rely on clinicians' visual observations, which limit the frequency, sensitivity, and objectivity of disease severity monitoring.
To address these limitations, wearable inertial sensors have emerged as a promising tool for continuous and objective assessment of motor impairments. This dissertation explores analytic pipelines and machine learning models for estimating ataxia severity using inertial sensor data collected during various motor tasks involving the upper and lower limbs.
The first part of this work presents an analytic pipeline to estimate overall motor severity in cerebellar ataxias using ankle-worn inertial sensors during a simple gait task. By analyzing sub-second movement profiles and extracting submovement features, conventional supervised learning models are trained to estimate overall motor severity. This approach is further applied to pediatric populations with ataxia-telangiectasia, disentangling developmental effects from estimating disease severity.
To reduce reliance on subjective clinical scores for training machine learning models, the dissertation introduces a contrastive learning framework for estimating ataxia severity using wrist-worn sensors during upper-limb motor tasks. By employing a pair-wise contrastive loss function, this method captures relative differences in ataxia severity, leveraging inherent data variations to enable bias-reduced and precise severity assessments.
Building on these prior works, the dissertation extends the contrastive learning framework to a multimodal approach, integrating data from multiple tasks to provide a comprehensive understanding of ataxia progression. This integrative model aims to advance the objective monitoring and assessment of cerebellar ataxias, ultimately facilitating more effective clinical trials and improving medical interventions.
Advisor
Sunghoon Ivan Lee