Med-KGMA: A novel AI-driven medical support system leveraging knowledge graphs and medical advisors

Healthcare systems worldwide face a growing burden, struggling to provide timely diagnosis and personalized care due to resource constraints. The increasing demand for medical expertise often results in delayed interventions, making automated decision support crucial. However, existing medical question-answering (QnA) systems struggle with hallucinations, limited contextual understanding, and difficulty handling complex queries, which often results in unreliable responses. To address these challenges, this study proposes Med-KGMA, an artificial intelligence-driven medical QnA system that leverages the proposed SequentialRotatE, a knowledge graph embedding model, along with the Mixture-of-Medical-Advisors (MoMA) framework to enhance diagnostic accuracy and treatment recommendations. Unlike conventional methods, SequentialRotatE effectively captures contextual relationships between medical entities, improving the system's reasoning capabilities. Additionally, the MoMA framework, which dynamically routes queries to specialized advisors based on complexity and relevance, ensures more precise recommendations. Experimental results demonstrate that Med-KGMA achieves 91.32% accuracy, outperforming state-of-the-art baselines. This approach advances medical knowledge representation through optimized query processing, tailored knowledge graphs, and intelligent advisor selection, providing an efficient, scalable solution. By addressing initial symptom analysis and reducing healthcare load, Med-KGMA empowers users with reliable medical insights, bridging the gap between patients and timely care.