AAN 2024 | NLP algorithm to extract headache frequency from clinical notes

Chia-Chun Chiang, MD, from Mayo Clinic, Rochester, MN, discusses her recently published large language model-based natural language processing (NLP) framework, able to extract headache frequency from electronic health records. This is particularly important in the headache field as diagnosis and treatment outcome evaluations rely on human natural language descriptions, and not lab testing or imaging. The framework focuses on headache frequency recorded in free text clinical notes, the most commonly used measure to evaluate headache severity. The GPT-2 large language model was used which, when fine-tuned using clinical notes, was able to extract headache frequency with an accuracy close to 90%. Dr Chiang expresses her excitement at the potential of this model to facilitate research and extract key outcomes in headache medicine. She emphasizes that the model has been made open source for collaboration with other researchers. This interview was conducted at the American Academy of Neurology (AAN) Annual Meeting 2024 in Denver, CO.

These works are owned by Magdalen Medical Publishing (MMP) and are protected by copyright laws and treaties around the world.