Study
NLST
(Learn more about this study)
Project ID
NLST-883
Initial CDAS Request Approval
Feb 16, 2022
Title
Optimization of lung cancer screening interval by deep learning-based risk prediction
Summary
The United States Preventive Services Task Force (USPSTF) recommends annual lung cancer screening with low dose computed tomography (LDCT) for certain at-risk patients. Considering about 15 million people in the US are recommended for screening each year, optimization in screening strategies could be highly beneficial as evidenced in the latest USPSTF report on “High-Priority Evidence Gaps” stating the need for research “to determine optimal screening frequencies and strategies.” Personalized screening frequency based on cancer risk offers the potential to minimize radiation exposure, maximize the benefits of early detection, and increase cost efficiency. It is unknown whether hand-crafted radiomic features or automated feature extraction with convolutional neural networks (CNNs) can outperform the prior models that only utilized radiologist-derived CT features. Furthermore, prior models used smoking history as one of the clinical predictors of cancer risk; however, this is an imprecise measure of tobacco exposure or its effect for several reasons: recall bias, reporting bias, depth of inhalation, brand/type of cigarettes, genetics, dietary modifiers etc. An “effective” smoking exposure metric does not currently exist. In short, LDCT screening is currently only used to detect lung nodules, despite the potential of these LDCT images to grade lung health and cancer risk. This study will be among the first to determine if LDCT images can guide precision screening in diverse populations.
Aims
Aim 1: To develop a computational risk model of lung cancer that incorporates pixel-level CT image
data, as well as clinical predictors.
AIM 2: To develop a nomogram that reliably facilitates assessment of effective tobacco exposure.
Collaborators
Christine Neslund-Dudas, PhD, Associate Research Scientist, Department of Public Health Sciences; Co-Leader, Cancer Epidemiology, Prevention and Control Program, Henry Ford Cancer Center
Kelly A. Hirko, PhD, MPH, Assistant Professor, Department of Epidemiology & Biostatistics, College of Human Medicine, Michigan State University
Snehal Patel, MD, PhD, Research Associate, Institute for Quantitative Health Science & Engineering, Michigan State University
Michael J. Simoff, MD, FACP, FCCP, Senior Staff, Pulmonary and Critical Care Medicine, Henry Ford, Director, Interventional Pulmonology and Lung Cancer Screening Programs
Thomas K. Song, MD, Senior Staff, Division Head, Thoracic Radiology, Henry Ford Health System
Yalei Chen, PhD, Assistant Research Scientist / Biostatistician, Department of Public Health Sciences, Henry Ford Health System
