Objective: To evaluate the accuracy of a machine learning model using smartphone-captured finger tapping videos to distinguish patients with Parkinson’s disease (PD) from healthy controls (HC) in both clinical and home environments.
Background: Clinical assessment of finger tapping, a key component of the MDS-UPDRS, is subjective. Existing video-based studies are often limited to controlled clinical settings, brief recording durations, and manual frame selection, reducing real-world applicability. An automated, accessible tool for objective assessment is needed.
Method: We recorded 40-second finger tapping videos via smartphone from 30 PD patients and 30 HC (60 participants, 120 hands). Videos were preprocessed by removing the first/last 5 seconds, resizing frames, and automatically rejecting low-quality frames. MediaPipe extracted thumb tip and index fingertip landmarks. Thirteen temporal features (amplitude, velocity, frequency, halts/hesitations) were derived from the distance time-series. A Random Forest classifier was trained and tested on unseen data (10 PD, 7 HC videos). Feature correlation was analyzed.
Results: The Random Forest model achieved 88.23% accuracy, 83.33% precision, 90.0% recall, and an F1-score of 0.90 on the test set. Correlation analysis revealed strong coupling between velocity and amplitude features (e.g., F_vel_var–F_amp_var r=0.89) and high internal consistency among frequency metrics, while also highlighting feature redundancy.
Conclusion: Our automated pipeline, utilizing longer-duration smartphone videos from varied environments and robust preprocessing, accurately classifies PD. This approach supports accessible, objective telemonitoring. Future work will focus on severity prediction and extracting explainable kinematic biomarkers.
References: D.Bose, A.Mukherjee, M.Acharya, S.Choudhury, and N.Ghosh, “Artificial Intelligence for Detection of Parkinson’s Disease From Speech Signals—A Comprehensive Review,” BioFactors51, no. 6 (2025): e70065, https://doi.org/10.1002/biof.70065.
Acharya, Mrinal; Banerjee, Subhadeep; Chatterjee, Apratim; Mukherjee, Adreesh; Biswas, Samar; Gangopadhyay, Goutam; Biswas, Atanu. Predicting Long-Term Outcome of Patients of Early Parkinsonism with Acute Levodopa Challenge Test. Neurology India 69(2):p 430-434, Mar–Apr 2021. | DOI: 10.4103/0028-3886.314539
Treatment Reconciliation in Parkinson’s Disease Patients with Particular Reference to Wearing-off and Motor fluctuations: A Registry-based, Prospective, Observational Study.
S SAHOO, A PAL, SM NASER, C BAGCHI, S MUNSHI, SK TRIPATHI, …
Journal of Clinical & Diagnostic Research 17 (6)
Acharya, M. (2024, September). Decoding non-IParkinson’s disease (PD) Parkinsonism-A Comparative Analysis of Atypical and Secondary Parkinsonism. In MOVEMENT DISORDERS (Vol. 39, pp. S32-S33). 111 RIVER ST, HOBOKEN 07030-5774, NJ USA: WILEY.
Antiparkinson’s Drug-Effects On Quality Of Life And Safety Among Parkinson Disease Patients-A Prospective Observational Study In A Tertiary Care Hospital Of West-Bengal; S Sahoo, A Pal, SM Naser, CH Bagchi, S Munshi, M Acharya, SH Dasgupta, SK Tripathi, S Samajdar, S Sarkar; 2024; Panacea Journal Of Medical Sciences; Volume 14; Issue 2; Pages- 409-414
P. Dutta, A. Eshore, P. Guha, M. Acharya. Electrochemical Sensing for Accurate L-Dopa Detection in Parkinson’s disease [abstract]. Mov Disord. 2025; 40 (suppl 1). https://www.mdsabstracts.org/abstract/electrochemical-sensing-for-accurate-l-dopa-detection-in-parkinsons-disease/. Accessed March 12, 2026.
Acharya, M., Biswas, A., Ganguly, G., & Das, S. (2018, October). Utility of Acute Levodopa Challenge Test In Early Parkinsonism. In MOVEMENT DISORDERS (Vol. 33, pp. S439-S441). 111 RIVER ST, HOBOKEN 07030-5774, NJ USA: WILEY.
Mandal, R., & Acharya, M. K. (2025, May). Parkinsonism And Related Disorders 134 (2025) 107588 Parkinson’s Disease: Fear Of Fall, Slow Cadence; Efficacy Of Spinal Extensors Strengthening Exercise. In Parkinsonism & Related Disorders (Vol. 134). 125 London Wall, London, England: Elsevier Sci Ltd.
To cite this abstract in AMA style:
M. Acharya, D. Bose, N. Ghosh, A. Mukherjee. Smartphone Video-Based Finger Tapping Analysis Using Machine Learning for Parkinson’s Disease Classification [abstract]. Mov Disord. 2026; 41 (suppl 1). https://www.mdsabstracts.org/abstract/smartphone-video-based-finger-tapping-analysis-using-machine-learning-for-parkinsons-disease-classification/. Accessed October 1, 2026.« Back to 2026 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/smartphone-video-based-finger-tapping-analysis-using-machine-learning-for-parkinsons-disease-classification/