Using AI-based Plausible Reasoning for Inferring Missing Knowledge to Provide Automated Decision Support

Cancer treatment planning requires complete patient information, including all medical conditions and therapies. Varian Medical Systems has developed a cancer planning system (called 360 Oncology) which offers decision support to generate a holistic care path from initial cancer diagnosis to post-treatment follow-up. To gather a complete medical record on the patient's past and present diagnoses, the system integrates medical information from legacy information systems which leads to inaccurate and/or incomplete patient information that affects the accuracy of the cancer care plan. The company's current mechanisms to fill in the missing information are sub-optimal. This project aims to develop and use Artificial Intelligence (AI) methods to infer the missing clinical information within a medical record by reasoning over external evidence-based medical knowledge resources. The project will investigate a novel approach-called plausible reasoning that is similar to human reasoning processes when confronted with incomplete knowledge--to infer medical relationships between diagnoses and associated interventions. These inferred relationships will provide the missing information. We will evaluate the research output with real-world cases (provided by the company) to determine the accuracy and efficacy of the AI based solution to handling missing clinical information. It is anticipated that this AI based method will improve 360 Oncology's accuracy of automated cancer care planning, and, in turn, increase the company's competitiveness in the global marketplace. This novel Canadian based research to infer missing clinical information in medical records has benefits to the Canadian healthcare sector, as the technology can be applied to other patient support system to improve the quality of information in medical records.