A Healthcare Digital Twin for Diagnosis of Stroke

Neurological impairment is a common disorder observed in stroke population and Electroencephalography (EEG) monitoring is considered a significant marker for diagnostics stroke onset. This study aims to propose a proof-of-concept of a healthcare “digital twin” and utilize EEG data and machine-learning models to build a digital twin for the stroke patients. We examined 48 stroke patients admitted to a rehabilitation clinic and 75 healthy persons. Portable EEG devices were used to capture EEG using frontal, central, temporal, and occipital cortical electrodes. The statistical analysis revealed that the revised brain-symmetry index, theta, and delta activities are relevant characteristics for classifying stroke patients and healthy individuals in motor and cognitive states. Using the machine learning approach, Support vector machine (SVM) classified the EEG feature dataset with 76% accuracy (AUC: 0.84) for classifying the stroke and the control group. This healthcare digital twin framework may assist in clinical decision-making for stroke preventive measures and post-stroke treatment.