Pneumonia

Purpose

Detection of pneumonia on chest radiograph in adults

Tag(s)

 

Panel

Thoracic

TOUCH-AI ID

TAI-08180002

Originator

Eric J. Stern

Panel Chair

Eric J. Stern

Panel Reviewers

Thoracic Panel

License

Creative Commons 4.0
Status Public Commenting
                               

Clinical Implementation


Value Proposition

Pneumonia is a potentially life-threatening, treatable lung disease. AI meeting this use case would help in accurately detecting, comparing, grading severity, and alerting both non-specialists and radiologists to suspected disease. In clinical settings where a radiologist is not readily available, such detection would be of value to non-radiologists. In clinical settings where radiologists are available, such detection could be used to alert the radiologist and prioritize its reporting and notification to the treating providers. For patients with pre-existing pulmonary opacities from other diseases such as interstitial lung disease, cystic fibrosis, or malignancy, detection of a new opacity is sometimes difficult, even for radiologists. Subtle pneumonias, such as those projecting below the dome of the diaphragms on front chest radiographs, can easily overlooked and lead to unnecessary CT scans, which AI could help reduce. One possible extension to this use case is to compare treated cases with previously detected pneumonia to assess for resolution, stability, deterioration, or complication.

Narrative(s)

Community Acquired Pneumonia
69-year-old presents to primary care provider with short duration fever, cough, and dyspnea. A chest radiograph (CXR) is obtained. Algorithm evaluates the image and categorizes as pneumonia present, absent, or undetermined. If a radiologist is not present at the time of imaging, an alert is provided to the ordering physician for positive exams. If a radiologist is present, a positive exam is prioritized in the radiologist worklist for urgent interpretation and reporting. Cases of multifocal or diffuse pneumonia, and those with pleural effusions or other co-morbidities, should take higher priority in the work queue compared to less severe disease. 

Hospital Acquired Pneumonia
A 55 year old admitted to ICU and placed on mechanical ventilation is at risk for ventilator-associated pneumonia. Algorithm evaluates image and categorizes the image as pneumonia present, absent, or unknown. If a radiologist is not present at the time of imaging, an alert is provided to the ordering provider. If a radiologist is present, a positive exam is prioritized in the radiologist worklist for urgent interpretation and reporting. 

Opportunistic Pneumonia
A 42 year old immunosuppressed patient has had a recent bone marrow transplant and receives multiple surveillance chest radiographs for early detection of opportunistic pneumonia.

Workflow Description

Image obtained from modality and sent to PACS and the AI engine. Image analyzed by engine. System detects and characterizes pneumonia. An alert message is sent to PACS from the engine with the information, and identification, and graphic highlighting the pneumonia.

 

Considerations for Dataset Development


Procedures(s): XRAY, Chest

View(s): {AP, PA/Lat, inclination e.g., upright, semi-upright, supine}

 

Sex at Birth: {Male, Female}

 

History: {fever, cough, dyspnea and synonyms, hypoxia, decreased oxygen saturation}

 

Comorbidities: {pleural fluid (including air/fluid levels), lung disease (eg. pneumonia/lung abscess, bullous emphysema, bronchiectasis), pneumomediastinum, other extrapleural air collections, atelectasis, lobar collapse, malignancy, left-sided congestive heart failure, pulmonary edema, prior chest surgery, prior lung injury, prior treated/healed lung disease}

Clinical Note: Distinguishing pneumonia from basilar atelectasis is very difficult on the basis of radiographic findings alone and so requires additional clinical data , eg. White blood cell count, elevated body temperature, abnormal sputum production and character, acute or chronic cough, etc.

 
Lung Tissue Involvement: {segmental, patchy, lobar, multilobar, diffuse, cavitary, nodular}
 
Presentation: Subtle pneumonia (include cases in which opacities project below the diaphragms on the frontal view)

Technical Specifications


Inputs

DICOM Study

Procedure

XRAY, Chest

Views

AP, PA/Lat, inclination e.g., upright, semi-upright, supine

Data Type

DICOM

Modality

XRAY

Body Region

Chest

Anatomic Focus

Lung

 

Primary Outputs

Pneumonia Detection

RadElement ID

RDE343

Definition

Detection of pneumonia in adult patient

Data Type

Categorical

Value Set

0–Unknown

1–Pneumonia present

2–Pneumonia absent

Units

N/A

Future Development Ideas


  • Saliency map to localize the region of suspected pneumonia on radiograph
  • Identify presence of possible tuberculosis

Public Commenting


Use cases are meant to be a primary vehicle for distributing clinical information to the developer community. They pinpoint precise scenarios within radiology workflows where potential automation could add noticeable value and establish standards for interpreting and passing corresponding common data elements. Implementing effective standards requires the perspective from all stakeholders. So to that end, we encourage your feedback on use cases.

To submit comments, please email DSIUseCases@acr.org with the use case title(s) and relevant comments by January 1, 2019. If more convenient, you may also download this use case and comment directly on the PDF. Just attach the PDF copy on the email.