2024 World Conference on Lung Cancer (WCLC) - Posters-P2.04A.02 Using an AI Transformer Model to Identify Incidental Pulmonary Nodules (IPN) For Lung Cancer Early Detection

P2.04A.02 Using an AI Transformer Model to Identify Incidental Pulmonary Nodules (IPN) For Lung Cancer Early Detection

Pdf Summary

The study "Using an AI Transformer Model to Identify Incidental Pulmonary Nodules (IPN) For Lung Cancer Early Detection" is part of the Identifying Early Lung Cancer In A Diverse Population (IDEAL) trial, focusing on the Vancouver site. It aims to improve the early detection of lung cancer through the identification and follow-up of incidental pulmonary nodules (IPNs).<br /><br />This project utilizes a Medical Named Entity Recognition Model, SapienNER v3a, to analyze 'lung-in-view' CT scans from three major hospitals in Greater Vancouver. The November 2023 to July 2024 timeline for scans targets identifying patients, particularly those aged 50 to 80, with incidental nodules of 6mm or larger.<br /><br />The necessity of such a model arises because many clinically significant nodules, which fall outside formal cancer screening programs, do not receive appropriate follow-up, potentially delaying critical lung cancer diagnoses. The AI model helps address this gap by automating the identification process, thus increasing the chances of early detection and intervention.<br /><br />Initial results from the first 8 months indicate that 4.4% of the examined CT scans identified significant incidental nodules in the target age range. Cardiac scans, in particular, were noted as significant contributors to the identified IPNs, suggesting a collaborative opportunity to improve nodule management and follow-up protocols.<br /><br />The study underscores the role of natural language processing (NLP) systems in healthcare, enabling large-scale, efficient screening processes. Furthermore, the research team is developing a Large Language Model to enhance the model's ability to compare prior imaging studies and extract relevant medical history, aiming to refine the accuracy of IPN identification and ensure necessary clinical follow-up. <br /><br />Overall, the study demonstrates the effective deployment of AI technology in an extensive healthcare setting for early lung cancer detection and highlights future development directions.

Asset Subtitle

Renelle Myers

Meta Tag

Speaker Renelle Myers

Topic Screening & Early Detection

Keywords

AI Transformer Model

Incidental Pulmonary Nodules

Lung Cancer Early Detection

IDEAL trial

SapienNER v3a

CT scans

Natural Language Processing

Large Language Model

Vancouver

Healthcare AI