AN IMAGE BASED PARKINSON’S DISEASE DETECTION SYSTEM USING MACHINE LEARNING APPROACH

Abstract

Quite several people have been sent to untimely grave and some with part of their body system paralyzed resulting from Parkinson Disease (PD). PD is a neurodegenerative disorder that affect neuron in the brain called substantial nigra with central movement. In this research work, an attempt is being made to design and implement a system for early detection of (PD) making use of relevant data (drawing of waves) from patient of PD and non-patient of the disease. Machine Learning models which include random forest (RF), Decision Tree (DT) and K- Nearest Neighbor (KNN), Dummy Classifier, Support Vector Machine (SVM), Convolutional Neural Network (CNN) were used to develop the model and the model in turn were used to develop the system. Machine Learning (ML)
models helped in predicting whether the patient have PD or not using both spiral and wave drawing and hence, displays the result.
Performance metrics such as Precision, recall and f1-score were used to determine the accuracy of the model result. Confusion matrix on the various algorithms were also displayed. Random Forest Classifier is discovered to be the most accurate machine learning algorithm for both the wave and spiral classification with an accuracy of 100% and a precision score for healthy (100%) and Parkinson’s (100%) after the extracted features were trained for wave classification, and an accuracy of 80% with a precision score for healthy (75%) and Parkinson’s (86%) after the extracted features were trained for spiral classification. Logistic Regression and Dummy Classifier had the worst accuracy with wave classification with an accuracy of 47%. Logistic Regression produced the worst accuracy (47%) in term of spiral classification. The best model (Random Forest) with 100% accuracy was used to develop the system and hence, the system was found to be highly efficient and far better than the existing methods or techniques.