An optimized deep learning approach for the prediction of parkinson s disease

Abstract

Parkinson s Disease is a progressive neurological disorder which is newlinecharacterized by motor dysfunction such as tremors, difficulty in balancing and newlinecoordination issues. This issue is due to degeneration of dopamine producing brain newlinecells. Since there is no cure, only medication and therapy can significantly newlineimprove the quality of the life. After a detailed study on the limitations newlineof existing predictive methods, in initial phase, a novel Convolutional Neural newlineNetwork Long Short-Term Memory Unsupervised Fine-Tuned Deep newlineSelf-Organizing Map (CNN-LSTM-UFDSOM) is proposed. The gathered newlinedatasets were processed by the ODBN model to remove extraneous features and newlinereduce the dimensionality of the data. The extracted features are then classified newlineinto types of Parkinson s disease using a CNN-LSTM based classifier which is newlineaugmented by Unsupervised Fine-tuned Deep Self-Organizing Map (UFDSOM). newlineThis model is evaluated with the number performance criteria, including accuracy, newlineprecision, recall, F1-measure, Root Mean Square Error (RMSE), error rate, and newlineMean Absolute Error (MAE). In comparison to all other models, the ODBN with newlineCNN-LSTM-UFDSOM model has a mean accuracy of 98.12% and an error newlinerate of 3.0. newline