Machine Learning-Based Real-Time Tremor Level Detection for Parkinson Disease

Parkinson's disease is one of the neurodegenerative diseases that affects neurons in the brain and causes motor functions to deteriorate. The most common symptom of this disease is involuntary tremor, especially in the hands and fingers, when the patient is in a resting position. In this study, a machine learning-based embedded system is proposed that can detect tremor and determine its level according to sensor data obtained from fingers. Subsequently, tremor data was obtained using Arduino UNO and MPU-6050 sensor, machine learning models were trained, and autonomous decision making have been performed. The study aims to evaluate tremor autonomously in real time, report it to the specialist, and assist in diagnosis and treatment. Unlike the studies in the literature, in this study, tremor signals were processed in real time with machine learning techniques instead of rule-based decision making. Tremor signals are digitally generated using sensors via the Internet of Things. Since mobility is crucial in the healthcare industry, the data was transferred wirelessly to the local server and evaluated for ease of use. As a result of this study, 96% accuracy was achieved using artificial neural networks in tremor level detection. By increasing the amount of data and the number of participants, the potential for the system to be developed and used in clinics is quite high.