Prediction of Parkinson’s disease based on artificial neural networks using speech datasets

Parkinson’s disease (PD) is a progressive disorder of the nervous system that affects movement. Early prediction of PD can increase the chances of earlier intervention and delay the onset of the disease. Vocal impairment is one of the most important signs in the early stages of PD. Therefore, PD detection based on speech analysis and vocal patterns has attracted significant attention recently. In this paper, we propose a vowel-based artificial neural network (ANN) model for PD prediction based on single vowel phonation. Firstly, we propose a novel multi-layer neural network based on speech features to predict PD. The speech samples from 48 PD patients and 20 healthy individuals are processed into four types: vowel, number, word, and short sentence. Secondly, we establish ANN models with single-type speech samples versus combinations of multi-type speech samples, respectively. Comparative experiments demonstrate that the single-type vowel model is superior to other single-type models as well as multi-type models. Finally, we build a vowel-based ANN model for PD prediction and evaluate its performance. Extensive experiments demonstrate that the proposed model has a prediction accuracy of 91%, sensitivity of 99%, specificity of 82%, and area under the receiver operating characteristic curve (AUC) of 91%, which is superior to the performance of previous methods. Overall, this study demonstrates that the proposed model can provide good classification accuracy for predicting PD and can improve the rate of early diagnosis.