2022 World Conference on Lung Cancer (ePosters)-EP01.01-002. Challenges in the Use of NLST Image Data for Quantitative Algorithm Development

EP01.01-002. Challenges in the Use of NLST Image Data for Quantitative Algorithm Development

Pdf Summary

The National Lung Screening Trial (NLST) has created a public database of CT images from the trial, which is being used for the development of algorithms related to lung nodule measurement and detection. However, the CT scans in the NLST database are 15-20 years old, and there is a need for algorithms to be developed on more recent real-world data. The study analyzed the NLST images to determine how often they met the minimum quality guidelines for scan acquisition and reconstruction set by the Radiological Society of North America's Quantitative Imaging Biomarker Alliance (QIBA). The results showed that only 2.9% of the NLST images met the protocol recommendations of the QIBA Small Nodule Committee. The most common reason for non-compliance was the slice thickness of the CT images. The NLST database is limited in terms of its suitability for developing algorithms, as only a small number of images meet modern standards. The study concludes that there is a need for high-quality image databases that represent current CT images for lung cancer screening and for new databases and data distribution approaches to successfully develop AI algorithms. Various organizations are currently developing newer databases that will be made available for this purpose.

Asset Subtitle

Artit Jirapatnakul, United States

Meta Tag

Speaker Artit Jirapatnakul, United States

Topic Early Detection and Screening - Biomarkers

Keywords

National Lung Screening Trial

CT images

lung nodule measurement

lung nodule detection

algorithm development

minimum quality guidelines

Quantitative Imaging Biomarker Alliance

QIBA

slice thickness

image databases