Automated Mapping of Healthcare Concepts to a Standardized Healthcare Taxonomy

Sabbir Mollah, Rakib Mohammed, Mashrur Wasek, AKM Shahariar Azad Rabby, Fuad Rahman, Nabeel Mohammed

Keywords: EHR, SNOMED CT, Clinical Terminology, Sentence Encoding

TL;DR: Mapping medical phrases to paths in the SNOMED Clinical Terminology.

Abstract: SNOMED CT presents an opportunity for numerous research prospects to learn medical terminologies and effectively assist the diagnosis process. In this work, we propose mapping the information in medical records to paths extracted from the SNOMED CT knowledge graph. To achieve this, we have leveraged language models to extract feature embeddings from EHR and compare them against the paths found in SNOMED CT. Our motivation is that this method can significantly assist the diagnosis process, ultimately leading to better healthcare outcomes. We have used several strategies to score the paths and evaluated them with experts in this field who have provided encouraging feedback on the outcomes of this approach.