Detection and prediction analysis of lung cancer based on deep learning with low-dose CT

Principal Investigator

Name
Seyoun Park

Degrees
Ph.D.

Institution
Johns Hopkins University

Position Title
Research Associate

Email
spark139@jhmi.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-356

Initial CDAS Request Approval
Sep 26, 2017

Title
Detection and prediction analysis of lung cancer based on deep learning with low-dose CT

Summary
In this project, we are developing detection and prediction models for lung cancer on low-dose CT images and pathology images using deep learning methods. Recently deep convolutional neural networks (CNN) has been shown similar or outperforming accuracy for nodule detection and classification, especially false-positive reduction. However, prediction models including survival rates are still challenging issues requiring new learning models. For this purpose, first we will develop advanced detection algorithm including semantic segmentation for the whole lung, such as vascular structures, airway, and tumor regions as well using deep networks. Based on the detection result, the initial goal is to develop classification model for diagnosis based on CT images. The final goal of this project is to develop the prediction model for prognosis of lung tumor based on CT.

Aims

Our studies include to develop
- an advanced detection algorithm for lung tumors including semantic segmentation on CT.
- deep learning algorithm for diagnosis for small nodules on CT.
- deep learning model for prediction of lung cancer based on CT.

Collaborators

Alan Yuille (Johns Hopkins University)