COVID-19 Compatible Chest Radiograph Pattern


Provide a likelihood of a diagnosis of COVID-19 compatible lung disease




Define-AI ID



Eric J. Stern


Eric J. Stern

Panel Chair

Eric J. Stern

Panel Reviewers

Thoracic Panel


Creative Commons 4.0


Public Comment

RadElement Set


Clinical Implementation

Value Proposition

Chest radiography is often the first-line imaging modality used for patients with suspected COVID-19 after positive testing with RT-PCR.  

Chest radiographs are often normal in early or mild disease (up to 60%). Of patients with COVID-19 requiring hospitalization, most have an abnormal chest radiograph at the initial time of admission, and approximately 80% have radiographic abnormalities sometime during hospitalization.  Radiographic findings are most evident at approximately 10-12 days after symptom onset and are relatively slow to resolve. The hallmarks of COVID-19 on chest radiographic imaging are bilateral and peripheral ground-glass and consolidative opacities, most prominent in the mid to lower lung zones.. There is notable absence of ancillary findings such as lymphadenopathy, pleural effusions, pulmonary nodules, and lung cavitation. Described patterns  of disease are  similar to that described in other coronavirus outbreaks such as SARS and MERS, and similar to the response to acute lung injury whereby an initial (often infectious or inflammatory) acute insult causes ground-glass opacities that may coalesce into dense consolidative lesions, and then progressively evolve and organize with predilection for the lung periphery. Chest radiography is limited for identifying specific viruses and distinguishing between viruses emphasizing the need for RT-PCR testing in a high-probability environment. 


A 82 year-old man who resides in a long term nursing care facility presents with fever (T > 101o F or 38.3o C) , rhinorrhea, cough, and new onset shortness of breath of duration two days.

A 49 year-old woman who initially presents with productive cough, new onset shortness of breath, muscle soreness, and malaise develops increasing fever (T > 101o F or 38.3o C)  and profound hypoxia requiring intensive care unit admission and mechanical ventilation,  without alternative explanatory diagnosis.

A 37 year-old woman who is a front-line healthcare employee at a local hospital who had close contact with a laboratory confirmed COVID-19+ patient,  developed new onset shortness of breath, cough, headache, and fatigue with fever (T > 101o F or 38.3o C) over the course of five days.

Workflow Description

Image obtained from modality and sent to PACS and the AI engine. Image analyzed by engine. System returns a likelihood score for COVID-19 compatible disease and returns prediction along with any relevant clinical information. An alert message is sent to PACS from the engine with the prediction and graphic highlighting the most important features the model drew upon for its prediction of COVID-19.

Considerations for Dataset Development


≥ 18 years old

Sex at birth

Male, female, other

Comorbid lung diseases

Emphysema, bronchitis, bronchiolitis, lung fibrosis, and other diffuse lung diseases, malignancy, immunosuppression, congestive heart failure, pulmonary embolism.

Pathologic Diagnosis

Negative STAT rapid influenza/RSV PCR tests. Positive Real-Time Reverse Transcriptase Polymerase Chain Reaction (rRT-PCR)


PA and Lateral chest radiograph. AP portable chest radiograph (For ease of decontamination, use of portable radiography units is preferred.)


Frontal and lateral, supine, semi-erect, erect


fever, cough, dyspnea and synonyms, exposure

Associated findings

presence or absence of pleural fluid, presence or absence of pulmonary edema, presence or absence of lymphadenopathy, presence or absence of pulmonary nodules, cardiomegaly.

Lung tissue involvement

segmental, patchy, lobar, multi-lobar, unilateral, bilateral, diffuse, cavitary, peripheral, central, ground glass, consolidation, cavitation, rounded morphology, bronchial wall thickening, interlobular septal thickening


Focal, Multi-focal

Technical Specifications




Chest radiograph


Supine, semi-upright, upright frontal. Upright lateral.

Data Type




Body Region


Anatomic Focus


Primary Outputs

COVID-19 Detection

RadElement ID



COVID-19 compatible disease

Data Type


Value Set


% likelihood of COVID19

Secondary Outputs

Extent of lung involvement

RadElement ID



Extent of lung involvement by percent area of ground glass or consolidation.

Data Type


Value Set


% area of the lungs

Future Development Ideas

  • Track serial/temporal examination changes.

  • Include identification of change from prior CXRs (in order to differentiate opacities due to chronic changes rather than COVID-19, or identify new opacities due to COVID-19 superimposed upon chronic changes)

Related Datasets

No known related public datasets at this time,  please alert us if you know of any.