CT offers opportunistic lung cancer risk assessment in heavy smokers

Sunday, November 29 | 1:00 p.m.- 2:00 p.m. | S4-SSCH02-7 | Room E451A

Automated mortality risk stratification from lung screening CT may enable opportunistic identification of high-risk individuals without additional imaging or manual body composition assessment, researchers have found.

Presenter Johannes Jahn, MD, of the University of Erlangen in Germany and colleagues developed a deep-learning (DL) model called CTchest-risk to predict mortality from low-dose lung screening CT images in heavy smokers and tested whether it could predict mortality independently of clinical or quantitative body composition measures such as skeletal muscle, subcutaneous adipose tissue, and adipose tissue. The team developed the model using data from 14,208 participants (equal to 39,304 individual CT scans) and tested it using data from 8,971 participants.

The investigators then conducted a study using data from the National Lung Screening Trial (NLST) divided into three categories: CT only, CT plus body composition segmentation results, and body composition segmentation results only. The study endpoint was six-year all-cause mortality.

Jahn's group reported that in the test cohort (mean age 61.3; 41% female), 462 deaths (5.1%) occurred over a median follow-up of 6.5 years. 

How did the DL model perform for each study category when it came to predicting lung cancer mortality? Attend this session to get the full details.

Page 1 of 2
Next Page