This project has 3 specific aims

1) Convolutional neural networks can be used to identify those as greatest risk for the development of lung cancer. Prior investigation suggests that smokers with emphysema are amongst those most likely to develop lung cancer. We will leverage these prior observations along with all of the co-morbid conditions evident on thoracic CT to determine which NLST subjects are at greatest risk for the development of lung cancer (based upon their enrollment CT)

2) Convolutional neural networks can be used to discriminate benign from malignant nodules. We will use these neural networks to assess all of the enrollment and follow up NLST scans. Our goal will be to detect lung nodules and then determine if they are benign or malignant

3) Smokers who develop lung cancer are likely to have comorbid diseases such as sarcopenia, atherosclerosis, emphysema, bronchiectasis. Our goal is to build a model to assess the competing risk of death in smokers with and without lung cancer. These models will allow clinicians to further understand the implications of there therapeutic decisions. Survival may be limited by the presence of comorbid disease rather than lung cancer.

David Gierada - Washington University
Mark Dransfield - University of Alabama, Birmingham
Matthew Kinsey - University of Vermont
Jason Bates - University of Vermont
Raul San Jose Estepar - Brigham and Women's Hospital
German Serrano - Brigham and Women's Hospital
Samual Ash - Brigham and Women's Hospital
Gonzalo Vegas - Brigham and Women's Hospital
Pietro Nardelli - Brigham and Women's Hospital