Advanced AI solution to detect Parkinson's disease through voice data analysis

The affliction referred to as Parkinson's disease (PD) results in a continuous decline in neurological performance, evident through assorted motor and non-motor symptoms that are being more widely discovered internationally. The prompt and precise identification of PD is essential for facilitating timely interventions and enhancing patient well-being. This scholarly article delves into a multitude of innovative techniques and strategies for the early identification of Parkinson's disease, spotlighting the utilization of deep learning models harnessed from the UCI Parkinson's disease remote monitoring dataset. Evaluating the severity of Parkinson's disease relies heavily on the UPDRS as a significant resource. In facing the challenges of multicollinearity within the dataset and optimizing the dimensionality of the spatial data, methodologies like Principal Component Analysis (PCA) and its extensions, including kernelPCA, incremental PCA, and sparse PCA, alongside Independent Component Analysis (ICA), are carried out. The results of these dimensionality reductions were subsequently fed into Deep Neural Network (DNN) models, which were configured with diverse parameters to assess their efficacy in predicting mobility and Total-UPDRS scores during the manifestation of PD. The model's performance is scrutinized through evaluations, and the findings are juxtaposed with those from earlier models utilizing the same dataset. To determine the model's precision, we scrutinize performance measures including mean absolute error (MAE) and mean squared error (MSE) to gain insights into the dependability of the forecasts. We also tested several different model designs with different hyperparameters on the same dataset. Among all the variants of the DNN model, we obtained the best results using DNN variant 2 of the 8-component sparse PCA. The MAE, MSE and values for the motor UPDRS are 0.0622, 0.0126, respectively. For the total UPDRS, we obtained the values of 0.0935 and 0.0185, respectively.