Abstract

CDSS or Clinical Decision Support Systems have become a critical component in the sphere of contemporary healthcare, offering intelligence based feedback to clinicians to enhance the quality of their correct diagnoses, and subsequent treatments. Even though single applications in isolated machine learning (ML) methods and semantic reasoning through knowledge graphs (KGs) have advanced, they tend to fail to provide context-based factual, explainable and scaleable decision support. Currently, the proposed hybridization of knowledge graphs and machine learning models is a possible solution presented in this paper to improve the performance and explainability of CDSS. On combining the advantages of structured domain knowledge with the data-driven inference possibilities, namely hybrid architecture, is the possibility to provide even closer-to-clinical recommendations, more interesting reasoning possibilities, and more confidence among medical workers. The proposed study proposes an integrated approach, tests its performance on several sets of clinical data, and provides a discussion of its deployment in real-world practice.