Skip to Main Content

An official website of the United States government

Government Funding Lapse

Because of a lapse in government funding, the information on this website may not be up to date, transactions submitted via the website may not be processed, and the agency may not be able to respond to inquiries until appropriations are enacted. The NIH Clinical Center (the research hospital of NIH) is open. For more details about its operating status, please visit  cc.nih.gov. Updates regarding government operating status and resumption of normal operations can be found at OPM.gov.

Image feature extraction with deep learning for mortality risk stratification on low-dose lung computed tomography

Principal Investigator
Name
Chang-Fu Kuo
Degrees
M.D., Ph.D.
Institution
Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Taoyuan, Taiwan
Position Title
Postdoctoral research fellow
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-572
Initial CDAS Request Approval
Sep 26, 2019
Title
Image feature extraction with deep learning for mortality risk stratification on low-dose lung computed tomography
Summary
The National Lung Screening Trial (NLST) is a multi-center trial to examine the value of low-dose CT (LDCT) scans in the reduction of mortality in high risk patients due to lung cancer. We plan to use the NLST data to construct a mortality risk prediction model using demographic and habitual data and automatically extracted image features from LDCT scans. The lungs will be segmented and be partitioned into nodule and non-nodule areas. Image features of lung nodules and non-nodule lung areas will be extracted automatically using neural network, which will be merged into feature vectors. In the second stage, we will use machine learning approach to construct prediction models using the clinical and histopathological information and image feature vectors to predict mortality risk. We hypothesize that image features extracted from LDCT images can improve predictive accuracy for mortality. The model may allow a better risk stratification for patients at risk for lung cancer.
Aims

Aims
-To build a deep-leaning model to extract image features of lung nodule and non-nodule areas using LDCT images from the National Lung Screening Trial (NLST).
Motivation: Hidden image features in lung nodules and non-nodule area inconceivable by human eyes may have predictive power for mortality. Therefore, we plan to use CNN to extract image features that may be relevant for mortality prediction.
-To use machine learning methods to construct a model to predict mortality, using clinical information and image features within and out of lung nodules.
Motivation: Current approaches to predict mortality generally rely on structured information such as clinical information or histopathological cancer staging. Addition of image feature extracted by CNN may add valuable information to the model to improve prediction accuracy.

Collaborators

- Chang-Fu Kuo, M.D., Ph.D., Principle investigators, Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Taiwan
- Chi-Hung Lin, Ph.D., Associate investigator, Chang Gung Memorial Hospital, Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Taiwan
- Jung-Sheng Chen MSc, Research engineer, Chang Gung Memorial Hospital, Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Taiwan
- Fu-Chi Chen Ph.D., Postdoctoral research fellow, Chang Gung Memorial Hospital, Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital, Taiwan