2022 World Conference on Lung Cancer (ePosters)-EP01.04-005. Quantitative Characteristics in Global CT Lung Cancer Screening Populations Using the ELIC Distributed Database and Computation Environment

EP01.04-005. Quantitative Characteristics in Global CT Lung Cancer Screening Populations Using the ELIC Distributed Database and Computation Environment

Back to course

Pdf Summary

This study aimed to analyze low-dose screening CT scans for potential lung cancer risk biomarkers. The researchers used the Early Lung Imaging Confederation (ELIC) database and a cloud-based infrastructure to perform a quantitative analysis of lung shape and density in 597 randomly selected screening CT scans. The scans were obtained from participants with and without lung cancer, including cases with lung nodules and cases without nodules.<br /><br />A deep learning AI algorithm was used to segment lung and lobe boundaries and measure cancer case lobe distribution. Quantitative measures such as Perc15 (the point below which 15% of voxels are distributed) and LAV950 (% of lung volume less than -950 HU) were calculated for each lung and lobe.<br /><br />The preliminary analysis demonstrated the potential of AI-based quantitative lung lobe segmentation and analysis to provide insights into early lung cancer. However, further analyses using the ELIC database and infrastructure are needed to determine the utility of these quantitative lung measures in globally distributed CT lung cancer screening populations.<br /><br />In terms of demographics, the study included sites from Gdansk, Poland; Milan, Italy; Ishikawa, Japan; Perth, Australia; Porto Alegre, Brazil; and Vancouver, Canada.<br /><br />The study also explored the relationship between emphysema metrics and lung cancer, as well as emphysema metrics and pack years of exposure. The initial analysis suggested that lung cancer cases had a lower standard deviation of HU intensities within lung regions compared to non-cancer cases. However, analysis of a larger dataset and removing outliers did not show any significant ability of quantitative metrics to separate benign from malignant cases.<br /><br />Overall, this study highlights the potential of AI-based quantitative analysis of CT lung scans in assessing lung cancer risk, but further research with a larger number of high-quality CT image data is needed to determine the extent to which these metrics can identify and characterize early lung cancer.

Asset Subtitle

Stephen Lam

Meta Tag

Speaker Stephen Lam

Topic Early Detection and Screening - Implementation Quality Control

Keywords

low-dose screening CT scans

lung cancer risk biomarkers

ELIC database

quantitative analysis

deep learning AI algorithm

lung lobe segmentation

emphysema metrics

benign and malignant cases

CT lung scans

early lung cancer