CADx Development for Lung Nodule Detection and Tumor Malignancy Assessment
Principal Investigator
Name
Kunlin Cao
Degrees
PhD
Institution
Curacloud Corporation
Position Title
Principle Scientist
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-195
Initial CDAS Request Approval
Feb 22, 2016
Title
CADx Development for Lung Nodule Detection and Tumor Malignancy Assessment
Summary
Lung cancer is the major cause of cancer-related deaths in U.S. It accounted for ~27% of all cancer deaths in 2015. Traditional chest radiography is commonly used to screen lung diseases. Recent advances in low-dose helical computed tomography (CT) improve the sensitivity of lung cancer screening and help detect tumors at early stages, which may lead to a decrease in the number of advanced-stage cancers diagnoses. The National Lung Screening Trial (NLST) showed that annual screenings with low-dose CT reduced mortality from lung cancer as compared screenings with chest radiography.
Applying advanced computer vision technologies on the large-scale data collected in NLST (including nodule appearance characteristics in low-dose CT and patients' clinical information) may benefit the computer-aided system development towards more objective and efficient lung cancer prognosis. The goal of this project is to explore deep learning technologies to learn attractive image features and design classifiers for tumor detection/characterization. Based on the investigation results, we aim to develop an interactive computer-aided diagnosis (CADx) framework for CT images interpretation during lung cancer screening.
Applying advanced computer vision technologies on the large-scale data collected in NLST (including nodule appearance characteristics in low-dose CT and patients' clinical information) may benefit the computer-aided system development towards more objective and efficient lung cancer prognosis. The goal of this project is to explore deep learning technologies to learn attractive image features and design classifiers for tumor detection/characterization. Based on the investigation results, we aim to develop an interactive computer-aided diagnosis (CADx) framework for CT images interpretation during lung cancer screening.
Aims
1. Apply deep learning algorithms on NLST datasets to learn useful image features.
2. Design classifiers to detect and characterize lung nodules in low-dose CT scans.
3. Develop image segmentation algorithms for lung tumors and airway structures.
4. Design an interactive framework to provide potential tumor locations, segmented contours, and malignancy assessment. This framework will allow user interactions for reviewing and revising the computer-aided detection results.
Collaborators
Ding, Kai (Johns Hopkins University School of Medicine)
Song, Qi (Curacloud Corporation)
Yin, Youbing (Curacloud Corporation)