Enabling efficient stroke prediction by exploring sleep related features

Stroke is one typical chronic disease, which is caused by the degenerative disorder of the central nervous system and has a serious impact on the daily lives of human beings. Thereby, it is of great value to enable early diagnosis or prediction of stroke by monitoring peoples daily physiological data and designing useful stroke predictors when the symptoms are not apparent. Specifically, in this paper, we propose a novel approach for stroke prediction by exploring sleep related features. In the first step, we present a stroke prediction framework, which integrates common medical features with fine-grained sleep features for stroke risk prediction. In the second step, we propose a stroke risk prediction model, which consists of two key components to control the false negative rate of stroke prediction. We evaluate the framework using a real polysomnogram dataset that contains 66 patients and 159 healthy individuals. Experimental results demonstrate that the proposed model can predict stroke events effectively, and the Precision, Recall and AUC are 83.1%, 83.6%, and 0.782, respectively.