Modeling lung cancer risk with Gradient-Boosted Decision-Trees

Principal Investigator

Name
Jean-Emmanuel Bibault

Degrees
MD, PhD

Institution
INSERM UMR 1138 Team 22: Information Sciences to support Personalized Medicine, Cordeliers Research Centre, Paris Descartes University, Paris, France

Position Title
Dr

Email
jbibault@stanford.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-814

Initial CDAS Request Approval
Jul 15, 2021

Title
Modeling lung cancer risk with Gradient-Boosted Decision-Trees

Summary
The aim of this project is to validate on NLST data a gradient-boosted model we created to predict 6-year lung cancer risk. Several models already exist: they are currently being used to guide patients and physicians on lung cancer screening. The state-of-the-art models, the PLCO m2012, the Lung Cancer Death Risk Assessment Tool and the Lung Cancer Risk Assessment Tool have been validated on PLCO, NLST and UK Biobank data.

Aims

- Calculate Accuracy, AUC, precision-recall of our XGBoost model on NLST data and validate our approach on an external dataset (the NLST dataset).
- Deploy the model online to guide lung screening strategies

Collaborators

Lei Xing (Stanford University)