Accurate and timely neurological decision support increasingly relies on the effective integration of heterogeneous clinical and biosignal data. In current practice, physiological signals and neuroimaging modalities are often analysed independently, limiting robustness, interpretability, and reliability in real-world clinical environments affected by noise, missing data, and modality-specific variability. This research addresses this limitation by proposing MedAI, a unified and reproducible multimodal medical intelligence framework for reliable neurological prediction and decision support. MedAI integrates electroencephalography (EEG), functional magnetic resonance imaging (fMRI), and associated clinical features through systematic preprocessing, feature extraction, dimensionality reduction, and supervised machine learning, with optional cognitive support using IBM-based analytics components. The framework is evaluated using standard classification metrics and benchmarked against conventional single-modality and baseline machine learning approaches. Experimental results demonstrate that MedAI achieves improved prediction stability and consistent performance while preserving transparency in feature contribution and decision logic.
Sivakumaran et al. (Thu,) studied this question.