Enhancing parkinson disease detection through feature based deep learning with autoencoders and neural networks.

Parkinson's disease is a neurodegenerative disorder that is associated with aging, leading to the progressive deterioration of certain regions of the brain. Accurate and timely diagnosis plays a crucial role in facilitating optimal therapy and improving patient outcomes. This study presents an innovative approach to identify Parkinson's disease (PD) through the examination of audio waves using Feature Based - Deep Neural Network (FB-DNN) techniques. Autoencoder, a specific form of Artificial Neural Network (ANN) that is designed to excel in the task of feature extraction, is utilized in our study to effectively capture complex patterns present in audio data. Deep Neural Networks (DNNs) are utilized in the task of classification, using the capabilities of deep learning (DL) to differentiate between audio samples that exhibit Parkinson's disease (PD) and those that do not. The deep neural network (DNN) model is trained using the retrieved data, allowing it to effectively distinguish minor variations in voice characteristics that are linked to Parkinson's disease. The suggested methodology not only enhances the precision of diagnosis but also enables prompt identification, perhaps resulting in more efficacious treatment methodologies. The present study introduces a potentially effective approach for the automated and non-intrusive identification of Parkinson's disease through the analysis of audio data. The integration of Autoencoder-based feature extraction with Deep Neural Networks (DNN) presents a dependable and easily accessible solution for the early detection and continuous monitoring of Parkinson's disease. This approach has promise for significantly improving the quality of life for persons affected by this condition. The implementation in Python was conducted as part of our experimentation. Upon analyzing the accuracy, it became apparent that the Feature-Based Deep Neural Network (FB-DNN) exhibited superior performance compared to the other models. Notably, the FB-DNN achieved the highest accuracy score of 96.15%.