Speech-based biomarkers for Parkinson’s disease detection and classification using AI Approach

Parkinson's disease (PD) is a progressive neurological condition that impairs motor and speech function. Early and precise detection is critical for prompt intervention and illness treatment. This work applies machine learning approaches to classify Parkinson's disease using speech biomarkers collected from voice recordings. The dataset includes a variety of acoustic parameters that capture speech anomalies often seen in people with Parkinson's disease. The Chi-Square (Chi2) approach was used to pick the most important predictors, which improved model performance and reduced computational complexity. The fine K-Nearest Neighbors (KNN) classifier was implemented, achieving a validation accuracy of 74.7%. The model demonstrated a moderate ability to distinguish between Parkinson’s and non-Parkinson’s cases, as indicated by an area under the curve (AUC) score of 0.7421. However, the confusion matrix revealed challenges in misclassification, with false positives leading to potential unnecessary medical evaluations and false negatives resulting in missed diagnoses. This study highlights the potential of machine learning in Parkinson’s detection while emphasizing the need for further refinement to enhance classification accuracy