Category: Tremor
Objective: To detect tremor syndromes from electrophysiological signatures using classical machine learning techniques.
Background: Electromyographic recordings combined with accelerometry data are widely used to measure and distinguish upper extremity tremors. This is done by calculating standard features such as peak frequency and power of the tremor spectrum at physiological frequencies. Here, we try to automate the process of distinguishing between patients and healthy controls via means of training classical machine learning algorithms on electrophysiological data.
Method: We recruited 25 healthy controls and 75 patients (25 Dystonic Tremor (DT), 25 Essential Tremor (ET), 25 Parkinsonian Tremor(PD)) with upper extremity tremor and carried out electrophysiological analysis on them using the combined tri-axial accelerometer-EMG system, in 8 positions . For each position, we derived Peak frequency (PF), Peak Power (PP) and Total Power (TP) from the power spectral density data (1-30Hz) by an automated algorithm. Further we applied classical machine learning classifiers such as Support Vector Classifier (SVC), Random Forest (RF) and K-Nearest Neighbour (KNN) to the derived electrophysiological parameters, in order to distinguish tremor patients from controls.
Results: Our trained models were able to classify between tremor patients and controls in the test data with considerable accuracy. Random Forest ( Accuracy = 0.927, F1 score = 0.947) and XGBoost ( Accuracy = 0.927, F1 score = 0.947) were the best performing models, followed by KNN classifier (Accuracy =0.852, F1 score =0.895).
Conclusion: Classical machine learning techniques applied to standard electrophysiological data can distinguish between tremor patients and healthy controls. These findings may help in enhancing the applicability of tremor analysis to wider clinical settings.
References: [1] Yao L, Brown P, Shoaran M. Resting tremor detection in Parkinson’s disease with machine learning and Kalman filtering. In2018 IEEE Biomedical Circuits and Systems Conference (BioCAS) 2018 Oct 17 (pp. 1-4). IEEE.
[2] de Araújo AC, Santos EG, De Sà KS, Furtado VK, Santos FA, De Lima RC, Krejcová LV, Santos-Lobato BL, Pinto GH, Cabral AD, Belgamo A. Hand resting tremor assessment of healthy and patients with Parkinson’s disease: an exploratory machine learning study. Frontiers in bioengineering and biotechnology. 2020:778.
[3] Myszczynska MA, Ojamies PN, Lacoste A, Neil D, Saffari A, Mead R, Hautbergue GM, Holbrook JD, Ferraiuolo L. Applications of machine learning to diagnosis and treatment of neurodegenerative diseases. Nature Reviews Neurology. 2020 Aug;16(8):440-56.
To cite this abstract in AMA style:
A. Vishnoi, R. Anandapadmanabhan, D. Biswas, A. Srivastava, D. Joshi, A. Mahabal. Classical Machine Learning approaches to automate tremor detection from electrophysiological data [abstract]. Mov Disord. 2022; 37 (suppl 2). https://www.mdsabstracts.org/abstract/classical-machine-learning-approaches-to-automate-tremor-detection-from-electrophysiological-data/. Accessed September 7, 2024.« Back to 2022 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/classical-machine-learning-approaches-to-automate-tremor-detection-from-electrophysiological-data/