Wojciech Ciezobka; Joan Falcó-Roget; Cemal Koba; Alessandro Crimi
This study proposes a framework to construct directed graph representations of brain networks. By further integrating explainable AI techniques, the method reveals disruptions in brain connectivity associated with stroke. The pipeline also compares the performance of reservoir computing-based causality with Granger causality and transfer entropy, offering a comprehensive assessment of effective connectivity estimation methods. Explainable AI tools allowed insights into critical network alterations, clarifying the role of effective connectivity biomarkers in stroke. This transparent analytical approach highlights the potential of directed graph models for both improved diagnostic precision and understanding stroke mechanisms, with broader implications for brain disorder analysis.
DOI: 10.1109/ACCESS.2025.3529179
Autors: Wojciech Ciezobka; Joan Falcó-Roget; Cemal Koba; Alessandro Crimi
Keywords: Reservoir Computing, Brain Connectivity, Explainable Artificial Intelligence (XAI) Effective Connectivity, Neuroimaging Biomarkers, Magnetic Resonance Imaging (MRI), Causality Analysis, Machine Learning in Healthcare, Brain Network Disruption, Stroke Diagnosis
READ HERE