Early Parkinson's Disease Diagnosis Using Multi-Modal CASENet CNN-LSTM

Abstract

By analyzing the deviation of features earlier stages can be segmented with subtle patterns in patients' handwriting dynamics and voice recordings, this innovative method showcases deep learning's potential to revolutionize medical diagnostics. By applying Casenet convolutional neural network framework, a hybrid architecture incorporating CNNs and improved long short-term memory networks is implemented using Kaggle datasets, which excels in spatial feature extraction from handwriting features with individual cases. while LSTM captures temporal patterns from voice recordings. Demonstrating a robust 94.6% accuracy rate, the model proves its effectiveness in Parkinson's disease prediction in earlier stages that can support complete diagnosis. Model assessment includes precision, recall, and F1-score evaluations using Principal Component Analysis (PCA) by integrating the Casenet CNN framework to enhance the diagnosis system and reliable accuracy that can predict early detection of Parkinson's disease from multimodal data.