Validation of Deep Learning Algorithm by using extended follow-up data in the National Lung Screening Trial
Principal Investigator
Name
Ping Hu
Degrees
ScD
Institution
NCI
Position Title
Mathematical Statistician
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-715
Initial CDAS Request Approval
Oct 22, 2020
Title
Validation of Deep Learning Algorithm by using extended follow-up data in the National Lung Screening Trial
Summary
The National Lung Screening Trial (NLST) compared two ways of detecting lung cancer: low-dose computed tomography (LDCT) and standard chest X-ray. Extended follow-up of participants in the NLST continued to show a lung cancer-related mortality benefit for LDCT.
A Deep Learning Algorithm (DeepLR, Dr. Huang, etc 2019) has been developed using NLST original data from participants who had received at least two CT screening scans up to 2 years apart in the NLST. In this project, we will apply this DeepLR to extended follow-up data in the NLST. We aim to compare accuracy of DeepLR scores to predict lung cancer incidence at 1 year, 2 years, and 3 years with the Lung CT Screening Reporting & Data System (Lung-RADS) and volume doubling time, using time-dependent area under the receiver operating characteristic curve (AUC) analysis. We will also modify DeepLR with more clinical variables.
A Deep Learning Algorithm (DeepLR, Dr. Huang, etc 2019) has been developed using NLST original data from participants who had received at least two CT screening scans up to 2 years apart in the NLST. In this project, we will apply this DeepLR to extended follow-up data in the NLST. We aim to compare accuracy of DeepLR scores to predict lung cancer incidence at 1 year, 2 years, and 3 years with the Lung CT Screening Reporting & Data System (Lung-RADS) and volume doubling time, using time-dependent area under the receiver operating characteristic curve (AUC) analysis. We will also modify DeepLR with more clinical variables.
Aims
The aims of this study are i) to compare accuracy of DeepLR scores to predict lung cancer incidence in extended NLST follow-up data, and 2) to modify DeepLR by adding more clinical variables.
Collaborators
Peng Huang, Johns Hopkins University