WCLC 2025 - Posters & ePosters-EP.17.20 Natural Language Processing in Rapid Evaluation of Suspicious Lung Findings: Impact on Lung Cancer Outcomes

EP.17.20 Natural Language Processing in Rapid Evaluation of Suspicious Lung Findings: Impact on Lung Cancer Outcomes

Pdf Summary

This study evaluates the impact of a Natural Language Processing (NLP)-assisted Incidental Pulmonary Nodule (IPN) evaluation protocol on lung cancer (LC) outcomes at Assuta Medical Centers (AMC) in Israel. Early detection and prompt treatment initiation are critical for improving LC prognosis, yet current lung cancer screening (LCS) eligibility criteria based on age and smoking history leave many at-risk individuals unscreened. The NLP system analyzes chest CT reports to identify suspicious lung findings (SLFs) requiring further investigation, supplementing traditional referral pathways.<br /><br />A retrospective analysis was conducted on 200 consecutive LC patients diagnosed between 2019 and 2022, divided evenly into two groups: Group A (NLP-assisted evaluation) and Group B (standard care with community physician referral). Both groups were comparable in baseline demographics, smoking history, tumor histology, and mutation profiles. Notably, Group A showed a significant shift toward earlier stage diagnosis (48% at stage I versus 27% in Group B, p=0.013), indicating enhanced early detection. Treatment modalities and timelines suggested potentially faster diagnosis and management in the NLP group, although overall survival data were immature with no significant difference observed yet.<br /><br />The findings suggest that integrating NLP for automated identification of suspicious lung nodules from routine CT reports can improve early lung cancer detection beyond standard LCS eligibility constraints. This method enhances diagnostic efficiency and could expedite the care pathway for patients with incidental findings, potentially improving clinical outcomes. While longer follow-up is necessary to confirm survival benefits, the study highlights the promise of NLP-assisted protocols as complementary tools to existing lung cancer screening strategies, addressing gaps in current screening coverage and facilitating timely interventions.

Asset Subtitle

Elizabeth Dudnik

Meta Tag

Speaker Elizabeth Dudnik

Topic Global Health, Health Services, and Health Economics

Keywords

Natural Language Processing

Incidental Pulmonary Nodule

Lung Cancer

Early Detection

Lung Cancer Screening

Chest CT Reports

Suspicious Lung Findings

Assuta Medical Centers

Retrospective Analysis

Diagnostic Efficiency