Imaging-Based Mortality Prediction in the National Lung Screening Trial

Principal Investigator

Name
Metin Gurcan

Degrees
Ph.D.

Institution
Wake Forest University Health Sciences

Position Title
Professor

Email
metin.gurcan@wfusm.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-1524

Initial CDAS Request Approval
Jul 27, 2026

Title
Imaging-Based Mortality Prediction in the National Lung Screening Trial

Summary
This project will use data from the National Lung Screening Trial (NLST) to study the relationship between low-dose chest CT imaging features and mortality outcomes in a lung cancer screening population. We already have access to publicly available NLST imaging data and limited clinical variables from TCIA. We request access to the NLST Participant dataset to obtain participant-level mortality and follow-up variables, including vital status, time from randomization to death or censoring, and lung cancer-specific mortality indicators.

The study design is retrospective. We will link imaging data with participant-level clinical outcomes using NLST participant identifiers. Imaging-derived features and/or machine learning representations will be evaluated for their association with all-cause mortality and lung cancer-specific mortality. Age and other available clinical covariates may be used for adjustment or stratified analysis. Planned analyses may include descriptive cohort summaries, survival analysis, risk modeling, and evaluation of imaging-based prediction models.

All analyses will be conducted using de-identified data under the terms of the CDAS data use agreement. Results may be used for scientific presentations, manuscripts, and future methodological development in imaging-based risk prediction.

Aims

- Aim 1: Link publicly available NLST low-dose CT imaging data with participant-level mortality and follow-up variables from the NLST Participant dataset using NLST participant identifiers. This will create an analysis cohort suitable for retrospective imaging-based survival and mortality research.

- Aim 2: Characterize the relationship between imaging-derived features and mortality outcomes in the NLST screening population. Outcomes of interest include all-cause mortality, time from randomization to death or censoring, and lung cancer-specific mortality where available.

- Aim 3: Develop and evaluate statistical and/or machine learning models for mortality and survival prediction using imaging features, age, and other available clinical covariates. Model performance will be assessed using appropriate validation procedures and survival/risk prediction metrics.

- Aim 4: Examine whether imaging-based models provide useful risk stratification beyond basic demographic and clinical variables, with the long-term goal of supporting research on personalized risk assessment in lung cancer screening.

Collaborators

Metin Gurcan Wake Forest University Health Sciences
ZHENGJIE ZHU Wake Forest University Health Sciences
Onur Koyun Wake Forest University Health Sciences