NeuroAid An Open-Data Multimodal Screening Framework for Parkinson's and Depression Risk Estimation

NeuroAid is an open-source AI screening system designed for early detection of Parkinson's disease and depression risk using non-invasive signals. The system analyzes three independent signal pathways: acoustic speech biomarkers, visual emotional cues from video, and clinical digital biomarkers. On the Figshare Parkinson's audio dataset, the system achieved balanced accuracy of 78.6% in cross-validation and 75% on test data, with F1-score of 80% and AUC-ROC of 82.8%. The depression vision branch reaches balanced accuracy of 59.7% on test data. NeuroAid is built exclusively on publicly available datasets with complete reproducibility. The system includes production-grade MLOps infrastructure including a Streamlit interface and Docker deployment.