Vision-based Estimation of MDS-UPDRS Gait Scores for Assessing Parkinson's Disease Motor Severity (2007.08920v1)

Abstract: Parkinson's disease (PD) is a progressive neurological disorder primarily affecting motor function resulting in tremor at rest, rigidity, bradykinesia, and postural instability. The physical severity of PD impairments can be quantified through the Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS), a widely used clinical rating scale. Accurate and quantitative assessment of disease progression is critical to developing a treatment that slows or stops further advancement of the disease. Prior work has mainly focused on dopamine transport neuroimaging for diagnosis or costly and intrusive wearables evaluating motor impairments. For the first time, we propose a computer vision-based model that observes non-intrusive video recordings of individuals, extracts their 3D body skeletons, tracks them through time, and classifies the movements according to the MDS-UPDRS gait scores. Experimental results show that our proposed method performs significantly better than chance and competing methods with an F1-score of 0.83 and a balanced accuracy of 81%. This is the first benchmark for classifying PD patients based on MDS-UPDRS gait severity and could be an objective biomarker for disease severity. Our work demonstrates how computer-assisted technologies can be used to non-intrusively monitor patients and their motor impairments. The code is available at https://github.com/mlu355/PD-Motor-Severity-Estimation.

• The paper presents a novel vision-based approach that estimates Parkinson’s gait scores using computer vision and 3D pose extraction.
• It employs a three-step framework with SORT for tracking, SPIN for 3D mesh extraction, and DD-Net enhanced by a hybrid ordinal-focal loss.
• The method achieved an F1-score of 0.83 and 81% balanced accuracy, underscoring its potential for non-intrusive PD monitoring.

Vision-based Estimation of MDS-UPDRS Gait Scores for Assessing Parkinson's Disease Motor Severity

Overview

The paper presents a novel approach to estimating the severity of Parkinson's Disease (PD) motor impairments using a vision-based model. This paper marks a significant advancement in utilizing computer vision for clinical applications, specifically in evaluating motor symptoms of Parkinson’s Disease through gait analysis. Traditional methods for assessing PD, relying heavily on neuroimaging or intrusive wearable sensors, are often costly or burdensome for patients. Here, the authors propose a non-intrusive, scalable alternative: a computer vision approach leveraging video recordings to extract and analyze 3D body dynamics during movement.

Methodology

The research introduces a pioneering framework using monocular videos to analyze the MDS-UPDRS (Movement Disorder Society Unified Parkinson's Disease Rating Scale) gait scores. The process involves a three-step approach:

1. Participant Detection and Tracking: Employing the SORT algorithm to isolate the subject in each video frame.
2. 3D Pose Extraction: Utilizing the SPIN model to extract detailed 3D body mesh and pose data from video frames.
3. Gait Score Estimation: Implementing a novel Double-Features Double-Motion Network (DD-Net) architecture enhanced with a hybrid ordinal-focal loss function for classification. This combination is designed to account for class imbalance and the intrinsic ordinal nature of MDS-UPDRS scores.

Results

The proposed method demonstrated robust performance with an $F_1$-score of 0.83 and balanced accuracy of 81%, outperforming chance and several baseline methods. Notably, the combination of 3D joint extraction and the hybrid ordinal-focal loss function was pivotal. This approach not only fully utilizes the spatial and temporal features available in video data but also effectively manages imbalances typical in clinical datasets.

Implications and Future Directions

The approach sets a benchmark for video-based assessment of motor severity in PD, suggesting implications for non-intrusive patient monitoring and potential applications in remote health diagnostics and management of movement disorders. Such a system could alleviate the reliance on specialized equipment and trained personnel, particularly useful in low-resource settings.

Future research may explore scaling datasets to include more participants, addressing variability in human movement, and enhancing model robustness. Additionally, integrating multi-angle video captures and advanced tracking methods can refine pose estimation and improve classification accuracy.

The potential for expanding this framework to other neurological disorders is significant, given the increasing accessibility of high-resolution imaging and the growing capabilities of machine learning in understanding complex human movements. This alignment between computer science and medical diagnostics opens pathways for robust, scalable, and cost-efficient methods in healthcare technology.