This paper reports that KNN applied to combined neuroimaging and speech features achieved high diagnostic performance (accuracy 92.8%, F1 0.953) for classifying Alzheimer's, Parkinson's, and epilepsy.

The work has limited direct therapeutic discovery value—no mechanistic targets or interventions—but the multimodal diagnostic method could help early detection and patient stratification for Parkinson's clinical studies if rigorously validated on diverse cohorts.

This study investigates the application of machine learning (ML) techniques combined with neuroimaging and speech signal processing for the early detection of neurological disorders, including Alzheimer's disease, Parkinson's disease, and epilepsy. A multisource analysis dataset consisting of neuro-images and speech features was utilized to train and evaluate various ML classifiers, such as K-nearest neighbors (KNN), Support Vector Machines, Random Forest, Naive Bayes, Decision Trees, XGBoost, and ADABoost. Performance assessment was based on metrics like accuracy, precision, recall, F1 score, and AUC-ROC. Among all models, KNN demonstrated the highest diagnostic accuracy and overall performance, with an accuracy of 92.8% and an F1 score of 0.953. The results suggest that KNN is particularly well-suited for classifying neurological conditions using integrated biomedical data. Although these findings highlight the promise of artificial intelligence (AI)-driven approaches in neurological diagnostics, further validation with diverse datasets is recommended to improve generalizability and clinical relevance.