WCLC 2025 - Posters & ePosters-EP.04.02 Fully-Automated Classification of Parenchymal Low Attenuation-Clusters as Indicators of Pulmonary Emphysema in Lung Cancer Screening

EP.04.02 Fully-Automated Classification of Parenchymal Low Attenuation-Clusters as Indicators of Pulmonary Emphysema in Lung Cancer Screening

Back to course

Pdf Summary

This study addresses the automated detection and classification of pulmonary emphysema in lung cancer screening (LCS) populations, where emphysema prevalence ranges between 31% and 43%, and is recognized as an independent lung cancer risk factor. The researchers used the Nelcin B3 LCS dataset comprising 257 cases, each visually scored for emphysema severity by four readers based on the Fleischner criteria, focusing on cases without disagreement between normal/trace and at least mild emphysema severity.<br /><br />The methodology involved low-attenuation analysis (LAA) and low-attenuation cluster (LA-cluster) analysis performed on CT scans, with cluster sizes ranging from 5.0mm to 8.0mm (incremented by 0.5mm) evaluated at different Hounsfield Unit (HU) thresholds (-950HU, -960HU, -970HU). The goal was to optimize discrimination between normal/trace and mild or more severe emphysema cases.<br /><br />Results demonstrated that LA-cluster analysis achieved very good agreement (Cohen’s kappa 0.870–0.876) with visual scores, particularly at thresholds below -950HU and cluster sizes around 7.0mm, outperforming LAA analysis which showed moderate agreement (kappa 0.497–0.601). For example, the highest concordance was at -950HU with a 7.0mm cluster size cutoff resulting in a 96.1% agreement (κ=0.876). These findings indicate that LA-cluster size measurements align closely with visual assessments, suggesting greater potential for early emphysema detection in low-dose CT screening.<br /><br />In conclusion, fully automated LA-cluster size analysis offers superior reliability over traditional LAA quantification for CT-based emphysema evaluation in lung cancer screening cohorts. This advancement could improve identification of individuals at elevated lung cancer risk through more accurate emphysema assessment, facilitating early intervention and better patient management.

Asset Subtitle

Hailan Liu

Meta Tag

Speaker Hailan Liu

Topic Screening and Early Detection

Keywords

pulmonary emphysema

lung cancer screening

automated detection

Low-Attenuation Analysis (LAA)

Low-Attenuation Cluster (LA-cluster) analysis

CT scan

Fleischner criteria

Cohen's kappa

Hounsfield Unit thresholds

early emphysema detection