Portable RF brain scanner enables deep-learning-based stroke type detection and ordering of brain-event dielectric permittivity

Abstract Stroke remains a leading cause of mortality worldwide, demanding rapid, accurate diagnostic solutions; distinguishing among stroke types matters because treatments differ significantly. Using a recently developed, non-invasive brain scanner that integrates a 16-antenna radio-frequency array, we present a deep-learning model that reliably discerns the two main stroke types. We employ masked autoencoder-based self-supervised learning and supervised contrastive strategies to address small dataset constraints. Our system achieved 92% sensitivity and 85% specificity for hemorrhagic versus non-hemorrhagic detection and 95% sensitivity and 80% specificity for differentiating ischemic stroke from non-ischemic cases. Beyond classification, the model revealed a distinctive ordering of dielectric permittivity (hemorrhagic, ischemic, mimic, and healthy), challenging traditional assumptions about neural dielectric properties. This new insight into the relative dielectric permittivity emphasizes the scanner’s breakthrough potential in real-time stroke diagnostics. These findings pave the way for faster and more reliable differentiation between stroke subtypes, offering a potentially transformative impact on patient care.