AUTHOR=Xing Xupo , Luo Ningdi , Li Shun , Zhou Liche , Song Chengli , Liu Jun 

TITLE=Identification and Classification of Parkinsonian and Essential Tremors for Diagnosis Using Machine Learning Algorithms

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 16 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.701632

DOI=10.3389/fnins.2022.701632

ISSN=1662-453X

ABSTRACT=Due to the overlapping tremor features, the medical diagnosis of Parkinson’s disease (PD) and essential tremor (ET) mainly relies on the clinical experience of doctors, which can induce a common misdiagnosis between them. In order to better differentiate between PD and ET by using accessible demographics and tremor information of upper limbs, we evaluated several predictive models through machine learning and neural network algorithms. Tremor signals including tremor acceleration and surface electromyogram (sEMG) recorded from 398 patients (PD=257, ET=141) were used to train the established models to predict the diagnosis of PD and ET. Seven predictive models were evaluated and compared, including random forest, XGBoost, support vector machine, logistic regression, ridge classification, backpropagation neural network, and convolutional neural network. Evaluation indexes including accuracy, area under the curve, etc. were used to evaluate each of the models. The relative importance of sex, age, four postures, and five tremor features were analyzed and ranked. The ensemble learning models including random forest and XGBoost performed the best overall predictive ability (area under the curve above 0.90, precision above 0.84). Thus, we concluded that easily available demographics and tremor information of upper limbs can be used to effectively distinguish PD and ET. We also confirmed that the dominant frequency of sEMG of flexors, the average amplitude of sEMG of flexors, resting posture, and winging posture had a greater impact on the diagnosis, while sex and age were less important. These results provide reference for the intelligent diagnosis of Parkinson's disease and show promise for use in wearable tremor suppression devices.