Using Big Data for Computer-Aided Diagnosis of Chest CTs
Principal Investigator
Name
Ronald Summers
Degrees
PhD,MD
Institution
National Institutes of Health Clinical Center
Position Title
Senior Investigator
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-319
Initial CDAS Request Approval
Jun 30, 2017
Title
Using Big Data for Computer-Aided Diagnosis of Chest CTs
Summary
Recent advances in machine learning, e.g., deep learning, have pushed forward the possibility of reliable tools for computer-aided diagnosis and detection from radiological scans. Nonetheless, many challenges remain. Because modern machine learning algorithms require very large amounts of data to train, one important facet is leveraging and exploiting big-data sources like the NLST images and associated data.
Given the prevalence of lung cancer, developing algorithms for automated screening represents an important societal aim. As such, the NLST provides a uniquely rich source, as NLST CTs are accompanied by structured annotations, risk factors, demographics, and outcomes.
The Imaging Biomarkers and Computer-Aided Diagnosis Laboratory is an internationally recognized group with extensive experience in medical imaging and machine learning to extract knowledge from complex multi-factorial data. We aim to use the NLST dataset, in concert with other data sources, to develop large-scale and reliable computer-aided diagnosis tools for chest x-rays. With this, we hope to develop clinically relevant tools that can be used for automated or computer-aided screening.
Given the prevalence of lung cancer, developing algorithms for automated screening represents an important societal aim. As such, the NLST provides a uniquely rich source, as NLST CTs are accompanied by structured annotations, risk factors, demographics, and outcomes.
The Imaging Biomarkers and Computer-Aided Diagnosis Laboratory is an internationally recognized group with extensive experience in medical imaging and machine learning to extract knowledge from complex multi-factorial data. We aim to use the NLST dataset, in concert with other data sources, to develop large-scale and reliable computer-aided diagnosis tools for chest x-rays. With this, we hope to develop clinically relevant tools that can be used for automated or computer-aided screening.
Aims
-Develop machine learning algorithms, trained on NLST data, to automatically screen for lung cancer and other diseases
-Develop machine learning algorithms to roughly localize nodules, masses, and other abnormalities
-Investigate how to best use NLST data in concert with other, less structured datasets, in order to further improve performance
Collaborators
Adam P. Harrison, NIH
Le Lu, NIH
Ke Yan, NIH
Xiaosong Wang, NIH
Yuxing Tang, NIH