Enhancing Interoperability of Knowledge Graphs through Structural and Semantic Entity Alignment

Abstract

Autoimmune diseases represent a growing health concern globally, significantly impacting millions of people with chronic, life-threatening disabilities. These diseases trigger the immune system to attack healthy tissues and require the integration of vast amounts of complex medical information, including symptoms, treatments, and patient outcomes for effective diagnosis and treatment. However, the large volume and diversity of data, often presented in unstructured formats such as medical reports and research articles, provide challenges in extracting significant insights. The proposed approach for addressing this challenge is a knowledge graph (KG), which facilitates a structured presentation of medical knowledge. By capturing the semantic and contextual relationships between various medical entities, such as symptoms, diagnostic tests, and treatments, a KG can enhance the understanding and management of autoimmune diseases. A primary challenge in creating a Knowledge Graph (KG) is to guarantee the precise alignment of medical entities considering the extensive contextual and semantic information linked to each entity. Ensuring entity alignment is crucial for constructing an accurate and relevant knowledge graph in medical data, especially in autoimmune disorders, which exhibit variances in terminology and context. For example, a symptom such as "fatigue" can be associated with several autoimmune conditions, requiring accurate identification and alignment. This research work focuses on aligning entities from diverse sources such as PubMed articles, biomedical websites, and patient discharge summaries. By utilizing both contextual and semantic information, the entity alignment process enhances the accuracy and completeness of the knowledge graph, ensuring that related concepts are correctly linked. The results of the alignment process were promising, with hit@1 = 0.8823, hit@10 = 0.9090, and MRR = 0.8917, indicating strong performance in mapping relevant entities. This research presents a knowledge graph that integrates both contextual and semantic information through accurate entity alignment, offering a structured framework for advancing clinical decision support systems (CDSS) and improving the diagnosis and treatment of autoimmune diseases