Unifying ICOP Knowledge Graph and Large Language Models for Clinical Reasoning in Orofacial Pain
Clinical or diagnostic reasoning is a dynamic thinking process between the observed clinical evidence and disease identification. Clinical reasoning involves integrating Electronic Health Records (EHRs), relevant medical knowledge, clinicians’ experience, and other contextual or situational factors. The ever-increasing complexity and verbosity of EHR clinical narratives, often laden with redundant information, present the risk of cognitive overload for healthcare providers, potentially culminating in diagnostic inaccuracies (Liu et al., 2022). Physicians often skip sections of lengthy and unstructured notes and rely on decisional shortcuts (i.e. decisional heuristics) that contribute to diagnostic errors (Croskerry, 2005). This study aims to leverage the dental narratives of the patients available in Axium to the factual knowledge in the International Classification of Orofacial Pain (ICOP) in order to develop a knowledge graph using LLMs. The proposed knowledge graph in this study can support dentists in their clinical decision making for patients with Orofacial pain.
Tahereh Firoozi
Dr. Firoozi currently holds the Alberta Innovates Postdoctoral Fellowship in the School of Dentistry at the University of Alberta. Her two-year fellowship is focused on dental analytics and applications of artificial intelligence and large language modelling to health science research.
She is implementing the most recent transformer-based language models—including LLaMA, LLaVA, and BERT—to develop AI systems for information retrieval and disease prediction and diagnosis. The information retrieval system is used to extract information from clinical notes, electronic health records, and medical reports. The disease predictive system uses text and image data to assist in early disease detection and diagnosis.
When
Sessions: April 2nd – 3rd 2024
Trainings: April 15th – 19th 2024