Development and validation of a candidate selection model for lung cancer screening and a nodule risk prediction model using CT scans

Principal Investigator

Name
Hyungjin Kim

Degrees
M.D., Ph.D.

Institution
Seoul National University Hospital

Position Title
Clinical assistant professor

Email
khj.snuh@gmail.com

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-538

Initial CDAS Request Approval
Jul 12, 2019

Title
Development and validation of a candidate selection model for lung cancer screening and a nodule risk prediction model using CT scans

Summary
Our study aims to develop and validate two prediction models using NLST data through deep learning. One is the LDCT screening candidate selection model using clinical variables and the other is the a malignancy prediction model for the LDCT-detected lung nodules.

Aims

1. Screening candidate selection model
- Development of a candidate selection model using variables that can be obtained prior to LDCT screening.
- Deep learning with end-to-end learning may enable an effective nonlinear modeling strategy
- Deep learning model validation and comparison with pre-established models including PLCOm2012, PANCAN, Bach, and etc.
- Pre-established models will be recalibrated and revised for the fair comparison.

2. Nodule risk prediction model
- Development of a risk prediction model per-nodule basis.
- Baseline CT scans will be used as inputs.
- Annotation will be performed by the radiologists.
- Risk prediction model will be developed using convolutional neural network and validated.
- Deep learning model will be compared with the Brock model.

Collaborators

Jin Mo Goo, Seoul National University College of Medicine
Chang Min Park, Seoul National University College of Medicine