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.

Principal Investigator
Name
Shadab Khan
Degrees
PhD
Institution
Group 42
Position Title
Head of Applied Science
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-654
Initial CDAS Request Approval
Apr 6, 2020
Title
COVID-19 AI Screening Model Pretraining Using NLST Data - Is CT a clinically meaningful COVID screening tool?
Summary
Several recent studies have presented imaging features of COVID 19 subjects in CT. Guam et al (NEJM) state that 18% of symptomatic non-severe COVID subjects were asymptomatic on radiology scans. Others have similarly raised concerns about CT's ability to be used a useful screening tool for COVID 19 (ACR/RSNA expert consensus report in Radiology: Cardiothoracic Imaging). Despite this, several studies have reported features that were common and nearly unique to COVID, e.g. peripheral presentation of ground glass opacity and consolidation near the periphery.

Working further against CT is the fact that it takes much longer to clean a CT scanner post COVID-scan than it does to clean an X-ray unit. While no COVID-specific distinguishing features have been reported X-rays, it motivates us to assess whether any radiology tools, CT or X-ray, offer anything meaningful in screening or managing care of a COVID-19 subject.

To this end, we intend to assess the utility of CT imaging. We will use the NLST data to train an AI model that will be fine tuned using COVID data. The volume of NLST images will allow us to try sophisticated modeling approaches in order to address our questions.
Aims

1) Train a lung nodule detection and classification model using NLST data.
2) Fine tune the model using COVID-19 scans as well as bacterial and viral pneumonia scans, in addition to scans from other morbidities that result in radiology features similar to pneumonia (e.g. diffuse alveolar hemorrhage).
3) Assess the specificity and sensitivity of the model in correctly diagnosing COVID-19 from a hold out set comprising scans from bacterial and viral pneumonia.

Collaborators

This is an internal project, no external collaborators are expected to participate.