Building a Patient-Centered Virtual Hospital Ecosystem Using Both Access Control and CNN-Based Models

Abstract

Virtual hospitals empower traditional hospitals to deliver more accessible, affordable, and comprehensive patient-centered (PC) care services. However, the legacy information systems of traditional hospitals are ill-equipped to support the needs of virtual hospitals. We propose a holistic virtual hospital ecosystem design that addresses these issues. We have developed two models. The first is a VHealth-CNN model that extracts PC knowledge from multi-sourced biomedical big data by (1) extracting disease health-related features; (2) structuring the relevant health-related features as per the pre-identified factors; (3) training a convolutional neural network (CNN) double-layer structure, where we select significant health-related features in the first layer, and classify the positively and negatively correlated features in the second one; and (4) generating disease class outputs representing the PC knowledge. The second model is a granular VHealth-AC model that seamlessly grants healthcare practitioners at a hub hospital remote access to PC knowledge at the right point of care. We have deployed a granular 5-tier PC information classification scheme to enforce information security rules across hospitals. In addition, we examined the feasibility of the proposed design through a tele-monitoring service experimental case study for predicting obesity, hypertension, and diabetes. The experimental results show that the proposed model predicts obesity, hypertension, and diabetes diagnoses with 91.3%, 93.5%, and 95% accuracy, respectively. Finally, our ecosystem design should encourage the adoption of virtual hospitals and the adoption of virtual healthcare services as a new norm.