An automated cloud-based tool for Screening of Parkinson's disease in Bangladesh

• Published in: 2021 2nd International Conference on Robotics, Electrical and Signal Processing Techniques (ICREST) pp. 664 - 668
• Main Authors: Ehsan, Md. Tanvir, Sajal, Md. Sakibur Rahman, Mamun, Khondaker A
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.01.2021
• Subjects: accelerometer; Accelerometers; Feature extraction; machine learning; Parkinson's disease; Random forests; rest tremor; Signal processing algorithms; smartphone-based app; Support vector machines; wavelet packets
• ISBN: 9781665415743; 1665415746
• DOI: 10.1109/ICREST51555.2021.9331233

Parkinson's disease is the second most severe neurodegenerative disorder globally, which affects the nerve cells of our brain. As a result, it causes symptoms like speech impairment, tremor, stiffness, or balance problems of our body. Since PD is not curable, regular monitoring plays a vital role in improving patients' quality of life. Modern countries provide a specialized program to address this disease, but there are no such activities in Bangladesh. As a result, most of the patients remain untreated, especially those living in remote areas. Our primary purpose of this study was to propose a cloud-based automated tool for screening Parkinson's disease using smartphone-based tri-axial accelerometer data. To evaluate this method, we obtained the resting tremor of both hands from 10 patients of PD and 7 Healthy individuals using our application. We applied the three ML algorithms (Random Forest, KNN, SVM) on the extracted features from accelerometer data, where the Random forest classifier performed the optimum result with 90% sensitivity and 85% specificity to effectively discriminate between PD patients and healthy individuals. Our research also showed that this tool could be a cost-effective solution for screening Parkinson's disease for the people of Bangladesh.