Automatic Lung Nodule Detection
Principal Investigator
Name
Paul Sajda
Institution
Columbia University
Position Title
Professor of Biomedical Engineering
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-236
Initial CDAS Request Approval
Aug 23, 2016
Title
Automatic Lung Nodule Detection
Summary
Lung cancer screening is accomplished by examining chest CTs for lung nodules. Nodules are evaluated and tracked over time if they are considered to be potentially cancerous. Finding the nodules is done manually by the the radiologist who must scan through large CT images searching for small nodules. Some nodules are inevitably missed.
We seek to design a machine learning application to automatically detect lung nodules. We will utilize convolutional neural networks in combination with a variety of pre-processing approaches.
We seek to design a machine learning application to automatically detect lung nodules. We will utilize convolutional neural networks in combination with a variety of pre-processing approaches.
Aims
(1) Determine which preprocessing approaches are most effective for improving the accuracy of lung nodule detection.
(2) Identify the most effective deep learning approach to identify nodules.
(3) Build an optimal machine learner and characterize its accuracy.
Collaborators
Nick Waytowich - Columbia University/U.S Army Research Laboratory
Addison Bohannon - U.S. Army Research Laboratory
Vernon Lawhern - U.S. Army Research Laboratory
Brent Lance - U.S. Army Research Laboratory