Multi-Omics System Based on Predictive Analysis with AI-Driven Models for Parkinson’s Disease (PD) Neurosurgery

Abstract

In addition to Alzheimer's disease, Bradykinesia, stiffness, tremor, and postural instability are symptoms of Parkinson's disease (PD), the second most prevalent neurological illness globally. The symptoms might overlap with those of other neurological diseases, making early identification difficult. This research investigates the possibilities of deep learning to detect PD through non-invasive voice analysis, which offers a practical and accessible diagnostic approach. Leveraging a biomedical voice dataset, propose to improve prediction accuracy and rectify the inherent class imbalance, a convolutional neural network (CNN) model can differentiate between healthy individuals and those with Parkinson's disease. SMOTE and feature selection strategies were employed. Experimental results demonstrate that the CNN model outperforms traditional classifiers, achieving a classification accuracy of 98.05%, as well as strong F1-score, precision, and recall. These results demonstrate how deep learning may help diagnose Parkinson's disease early and allow for quicker treatments. This study advances the development of voice-based, reasonably priced diagnostic tools for practical clinical applications.