Automated Shoulder Girdle Rigidity Assessment in Parkinson's Disease via an Integrated Model- and Data-Driven...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Wearable Technology

Background:

• Parkinson's disease (PD) diagnosis relies on subjective motor assessments, particularly for rigidity.
• Current methods for quantifying rigidity lack objectivity and can be influenced by rater variability.
• There is a need for objective, scalable, and reliable methods to assess PD-related rigidity.

Purpose of the Study:

• To develop and validate a novel hybrid framework for objective quantification of shoulder girdle rigidity in Parkinson's disease.
• To integrate model-driven biomechanical and data-driven statistical features from wearable sensor data.
• To enhance the accuracy and interpretability of rigidity assessment using weak supervision.

Main Methods:

• Utilized wearable sensor data during a modified pendulum test.
• Extracted model-driven biomechanical features (damping ratio, decay rate) and data-driven statistical features (maximum detail coefficient).
• Employed weak supervision to unify features and generate robust labels from limited data.

Main Results:

• Achieved a 10% improvement in Parkinson's disease/healthy control classification accuracy (0.71 vs. 0.64) compared to data-driven methods alone.
• The hybrid framework matched the performance of model-driven methods (0.70 accuracy).
• Identified velocity-dependent aspects of rigidity, challenging traditional clinical characterizations, and demonstrated strong correlation with UPDRS rigidity scores (r = 0.78, p < 0.001).

Conclusions:

• The proposed hybrid framework provides objective, interpretable, and scalable assessment of shoulder rigidity in Parkinson's disease.
• This sensor-based neurotechnology advances PD diagnosis and management through remote monitoring capabilities.
• The identified biomechanical biomarkers hold significant clinical utility for early diagnosis and telemedicine applications.