A novel multi-model neural network for Parkinson’s Disease detection

Abstract Parkinson’s Disease (PD) is categorized as a neurodegenerative progressive disease caused by the destruction of the cells in the substantia nigra. Roughly out of 1000 elderly people, 1 to 2 suffers from PD and its complication. Complications of PD disease involved tremors, rigidity, impaired balance, etc. These complications affect the routine life of Parkinson’s patients. Accurate PD detection helps the healthcare system treat these complications properly. This research proposes a multi-modal emotion recognition technique for PD detection. Since the proposed methods used more than one modality for detection , its architecture had composed of three main sections. These sections are: feature extracting, merging, and classifying. As feature extractors, a combination of Convolutional Neural Network (CNN) attention mechanisms is developed to extract features from pictures. To extract features from related electronic healthcare features combination between CNN and Long-Short Term Memory is used Finally, Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), Extreme Boot Classifier (XGB), and voting classifier are used as classification. A Parkinson’s disease handwriting database is used to evaluate the proposed model. The experimental result indicates 99.95% accuracy, 99.99% precision, 99.98% sensitivity, and 99.95% F1-score using the proposed fusion technique and voting classifier. The comparative result indicates state-of-the-art results for PD detection.