Dimitris Askounis

Input-adaptive multimodal modeling can significantly enhance Alzheimer's diagnosis accuracy, achieving up to 95.47% in distinguishing between normal cognition and advanced disease stages.

A context-guided cross-modal attention mechanism boosts Parkinson's disease detection accuracy by integrating diverse speech representations, achieving over 91% accuracy.

Explicitly maximizing mutual information between speech and text embeddings unlocks more accurate dementia detection from patient interviews.