Deep learning for lung nodule detection and cancer prediction
Principal Investigator
Name
Quan Chen
Degrees
Ph.D.
Institution
University of Kentucky
Position Title
Associate Professor
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-320
Initial CDAS Request Approval
Jun 22, 2017
Title
Deep learning for lung nodule detection and cancer prediction
Summary
It has been shown that the low-dose CT screening on the high-risk population can improve the early detection and improve the overall survival. However, the screening will generate large amount of data for radiologist to inspect. This generated a lot of work to the radiologist. In addition, human fatigue may occur which will impact the accuracy. Even with the most up-to-date Lung-RADS guideline, false-positive rate is still high. Those false-positive patients will have to go through biopsy unnecessarily. This further increases the healthcare cost.
The recent success in using Neural Network to detect lung nodules and to predict whether it is cancer from a single CT has shown the power of the deep learning technology. However, more information regarding patient’s medical history, smoke history could be combined to aid the artificial intelligence to make a better decision.
The recent success in using Neural Network to detect lung nodules and to predict whether it is cancer from a single CT has shown the power of the deep learning technology. However, more information regarding patient’s medical history, smoke history could be combined to aid the artificial intelligence to make a better decision.
Aims
Specific aim 1: Use deep learning technology to detect suspicious nodules in the Lung CT scans.
Specific aim 2: Use deep learning technology to predict whether subject will develop lung cancer based on CT image alone.
Specific aim 3: Add patient’s demographics, smoke history, and other information to the neural network to see how much improvement in prediction accuracy can be made.
Collaborators
NA