Broadening Patient Treatment Options: Large Language Models Pipeline Improves Cancer Patient Matching to Clinical Trials in Day-To-Day Practice

Poster Number: P34
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Precision Medicine, Large Language Models (LLMs), Information Extraction, Data Standards, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Most modern cancer trials are targeted towards a patient’s molecular profile. These trials may provide longer survival and better quality of life than standard treatments, especially for patients whose cancer advances after initial treatment. Despite availability of these novel treatments, clinicians are poorly equipped to match patients whose cancer progresses to the available treatment options in a timely manner. One of the major challenges in automation of the matching process is recognition of cancer progression in clinical notes. Addition of Large Language Models to the arsenal of tools extracting information from text revolutionizes capabilities in this domain. At the NYU Perlmutter Cancer Center, we successfully implemented an LLM-based pipeline that alerts thoracic oncologists about available clinical trial options based on patient’s molecular profile one day after radiology reports cancer progression. In this manuscript, we intend to describe how combination of novel informatics methods such as AI and good informatics practices such as data standardization help make real impact in clinical care.

Speaker(s):
Rimma Belenkaya, MS, MA
NYU Langone

Author(s):
Ferris Hussein, Computer Science - NYU Langone Health; Abraham Chachoua, MD - NYU Langone; Vamsidhar Velcheti, MD - NYU Langone; William H. Moore, MD - NYU Langone; Kanan Shah, MD - NYU Langone;