Objective:

To examine whether wrist-worn actigraphy can estimate MDS-UPDRS scores in people with Parkinson's disease.

Approach:

• Data Collection: Continuous accelerometer recordings at 25 Hz were collected over up to 28 days using GeneActiv devices.
• Feature Extraction: Three feature representations were derived: non-embedding actigraphy features, self-supervised accelerometer embeddings, and a combined feature set.
• Model Evaluation: A small set of regression models was evaluated using strict leave-one-participant-out cross-validation (LOPO-CV).

Key Findings:

• The strongest estimation performance was for MDS-UPDRS Part IV with a mean absolute error (MAE) of 1.6 and a correlation of 0.83.
• The combined feature set performed best for MDS-UPDRS Part I (MAE = 3.0, r = 0.60), Part III (MAE = 8.2, r = 0.47), and total MDS-UPDRS score (MAE = 13.3, r = 0.49).
• Non-embedding features performed best for MDS-UPDRS Part II (MAE = 2.7, r = 0.61).

Interpretation:

Wrist-worn actigraphy contains information related to PD severity in daily life, although estimation accuracy varies by MDS-UPDRS target.

Limitations:

• Estimation accuracy remains limited.
• Performance varies depending on the specific MDS-UPDRS target.

Conclusion:

Wearable-derived measures may provide complementary information to clinical assessments, particularly for motor complications.