ISSN:1005-3026

INTEGRATED MACHINE LEARNING POWERED SYSTEM TO DETECT PRESENCE OF PARKINSON’S DISEASE

Kotta Snigdhasree

UG scholar, Bachelor of Engineering in Department of Artificial intelligence and Machine Learning (AIML), BNM Institute of Technology, Bangalore, Karnataka. Email: ksnigdhas02@gmail.com.

G.Chenna Keshava Reddy

Head of Department, Department of Artificial intelligence and Machine Learning (AIML), BNM Institute of Technology, Bangalore, Karnataka.

Harshitha K

UG scholar, Bachelor of Engineering in Department of Artificial intelligence and Machine Learning,  BNM Institute of Technology, Bangalore, Karnataka.

ABSTRACT

An effective method to slow down the progression of Parkinson’s disease (PD), a neurodegenerative ailment primarily marked by motor-related dysfunction, is an accurate, quantitative, and objective diagnosis. Based on the circumstances of mobility of the upper limbs in 100 people, 50 of whom have Parkinson’s disease and 50 of whom are healthy, this study builds an accessible auxiliary diagnostic method for the disease. Wearable sensors track the motion of the upper limbs, and the system’s graphic user interface (GUI) and host computer process and categorize the data. Classification of PD and normal states using the evolutionary algorithm optimized random forest classifier is introduced. The classifier’s performance is evaluated using 50 trials leave-one-out cross-validation, and it achieves the highest accuracy of 94.4%. Results demonstrate that tasks involving only alternation of hands also achieve satisfactory classification accuracy, and that sensors on both wrists outperform one sensor on a single wrist when comparing the classification accuracy across various upper limb movement tasks and with varying numbers of sensors. The suggested system’s usefulness is demonstrated by neurologists conducting clinical investigations using a deployed GUI, which opens up several potential uses in the auxiliary diagnosis of PD.

Index Terms—Parkinson’s disease, Machine Learning, , Accuracy