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.

Application of deep learning to automate detection of lung nodules on single and serial chest radiographs

Principal Investigator
Name
Kartik Varadarajan
Degrees
PhD
Institution
Visionairy Health Inc.
Position Title
Chief Operating Officer
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
PLCOI-748
Initial CDAS Request Approval
Mar 15, 2021
Title
Application of deep learning to automate detection of lung nodules on single and serial chest radiographs
Summary
Chest radiographs are widely used for diagnosis of various respiratory and cardiac conditions. The interpretation of these images is performed by physicians with a wide range of training and specialization (e.g. family physicians, emergency physicians, pulmonologists, radiologists etc.). Thus, the risk of a critical clinical condition being missed or a significant delay in diagnosis occurring is dependent on the training and expertise of the interpreter. To address disparity in quality of care, we are developing deep learning-based algorithms to detect lung nodules/masses on chest radiographs. The first objective of the proposed project is to assess how well our current deep learning model generalizes on the PLCO dataset. Many deep learning models are currently designed to analyze images from a single timepoint in the disease process. However, comparing a given image to available historical images in standard clinical practice. This comparison helps the physician judge whether changes in previously detected findings have occurred or if new findings have developed. Therefore, a second goal of this project is to develop and test deep learning algorithms that use timeseries of radiographs as input for detection of nodules/masses.
Aims

1) Assess generalizability of our current single timepoint deep learning model for detection of lung nodules on the PLCO dataset
2) Develop deep learning model for detection of lung nodules on serial radiographs
3) Evaluate whether deep learning model trained to compare sequential radiographs is more sensitive in detecting lung nodules than models trained on images from a single timepoint

Collaborators

None