Title:Digital Twin approach to Clinical DSS with Explainable AI

Abstract:We propose a digital twin approach to improve healthcare decision support systems with a combination of domain knowledge and data. Domain knowledge helps build decision thresholds that doctors can use to determine a risk or recommend a treatment or test based on the specific patient condition. However, these assessments tend to be highly subjective and differ from doctor to doctor and from patient to patient. We propose a system where we collate this subjective risk by compiling data from different doctors treating different patients and build a machine learning model that learns from this knowledge. Then using state-of-the-art explainability concepts we derive explanations from this model. These explanations give us a summary of different doctor domain knowledge applied in different cases to give a more generic perspective. Also these explanations are specific to a particular patient and are customized for their condition. This is a form of a digital twin for the patient that can now be used to enhance decision boundaries for earlier defined decision tables that help in diagnosis. We will show an example of running this analysis for a liver disease risk diagnosis.