Online Program Home
My Program

Abstract Details

Activity Number: 382
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 11:15 AM
Sponsor: Section on Medical Devices and Diagnostics
Abstract #321637
Title: Prediction of Lung Cancer Risk in the CT Arm of the National Lung Screening Trial
Author(s): Menghan Hu* and Fenghai Duan
Companies: Brown University and Brown University
Keywords: NLST ; Lung Cancer ; Risk Prediction ; CT imaging ; ROC

Background: In NLST a 20% relative reduction in lung cancer mortality was observed using LDCT relative to chest-X-ray screening in older smokers. The study's aim is to determine risk factors associated with lung cancer diagnosis and assess if adding imaging features improves the prediction of risk. Methods: In 26,455 participants who underwent at least one CT screen, risk factors for lung cancer were determined through fitting the multivariate logistic regression models. The previous developed model (PLCOall2014) was used as reference and we added the imaging features at the participant level to improve the predictive accuracy. The outcome is lung cancer diagnosis by the end of the 6th year. Accuracy was measured by the area under ROC curves. Optimism was calculated via bootstrapping to assesse overfitting. Results: By adding imaging features, the area under the curve increases from 0.706 to 0.772, 0.696 to 0.754, and 0.704 to 0.787 for all participants, current smokers and former smokers, respectively. The optimisms show no overfitting for all models. Conclusions: Adding imaging features at the participant level can improve the accuracy of lung cancer risk prediction.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association