2022 World Conference on Lung Cancer (ePosters)-EP11.02-001. Natural Language Processing to Abstract Preneoplastic and Incidental Pulmonary Lesions from Pathology Reports

EP11.02-001. Natural Language Processing to Abstract Preneoplastic and Incidental Pulmonary Lesions from Pathology Reports

Back to course

Pdf Summary

Researchers at the University of Toronto and Princess Margaret Cancer Centre have successfully utilized natural language processing (NLP) to extract preneoplastic and incidental pulmonary lesions from pathology reports. The study analyzed the pathology report text from 942 early stage non-small cell lung cancer (NSCLC) patients who underwent surgical resection at the cancer center between 2014 and 2019. A total of 279 unique pathology reports were analyzed, with 27% of patients having at least one term extracted through the NLP pipeline. Of the patients with positive terms, 65% were positive for precancerous lesions, 9% for squamous metaplasia, and 36% for idiopathic pulmonary fibrosis (IPF). However, there was a low positivity rate for carcinoma in situ (CIS) and diffuse idiopathic pulmonary neuroendocrine cell hyperplasia (DIPNECH). The researchers noted that pathologists may not consistently record the presence or absence of these lesions, leading to the lower positivity rates. The successful implementation of NLP has the potential to improve the efficiency of molecular precancer research by identifying incidentally-discovered preneoplastic lesions of the lung. The researchers plan to further improve the performance of their NLP method by incorporating machine learning methodologies and expanding the cohort to include all patients with resected lung tissue, not limited to NSCLC patients. Overall, this study demonstrates the ability of NLP to rapidly extract and evaluate lung pathology terms from unstructured pathology reports, reducing the need for time-consuming manual abstraction.

Asset Subtitle

Jessica Petricca

Meta Tag

Speaker Jessica Petricca

Topic Pathology - New Technology

Keywords

natural language processing

pathology reports

lung cancer

precancerous lesions

squamous metaplasia

idiopathic pulmonary fibrosis

carcinoma in situ

pathologists

molecular precancer research

machine learning methodologies