Tuberculosis, Triage


Detection and characterization of tuberculosis on chest radiograph, in symptomatic adults





Define-AI ID



Eric J. Stern

Lead Eric J. Stern

Panel Chair

Eric J. Stern

Panel Reviewers

Thoracic Panel


Creative Commons 4.0


RadElement Set RDES79 

Clinical Implementation

Value Proposition

The value proposition is well stated by the World Health Organization publication: Chest radiography in tuberculosis detection – summary of current WHO recommendations and guidance on programmatic approaches:
“Triaging is defined as the processes of deciding the diagnostic and care pathways for people seeking healthcare, based on their symptoms, signs, risk markers and test results. Triaging involves assessing the likelihood of various differential diagnoses as a basis for making clinical decisions. It can follow more- or less-standardized protocols and algorithms and may be done in multiple steps. Effective triaging that helps to rapidly identify TB is important both for optimizing care for the individual and for ensuring good infection control” [1].

“Triaging is different from systematic screening in that it focuses on the clinical management of a person seeking healthcare for one or several unexplained complaints or concerns, while systematic screening normally is initiated by a provider and targets apparently healthy individuals with or without risk markers for a given disease” [1].

“The chest radiograph (CXR) has historically been one of the primary tools for detecting tuberculosis (TB), especially pulmonary TB. CXR has high sensitivity for pulmonary TB and thus is a valuable tool to identify TB as a differential diagnosis for patients, especially when the radiograph is interpreted to identify any abnormality that is consistent with TB” [1].

However, many CXR abnormalities that are consistent with pulmonary TB are seen also in several other lung diseases, often with significant intra- and interobserver variation in interpretations, which can lead to both overdiagnosis and underdiagnosis.

Digital radiography is a suitable alternative to conventional radiography for TB detection due to its lower running costs, portability, superior image quality and better safety (due to decreased radiation dose), as well as possibilities for its use in telemedicine.

Beyond identifying active TB disease, CXR also identifies one of the populations at highest risk of developing TB disease, namely those who have inactive or latent TB or fibrotic lesions without a history of TB treatment. Once active TB has been excluded, patients with fibrotic lesions should be followed-up, given their high risk for developing active disease.

CXR may have higher specificity for pulmonary TB than assessing clinical symptoms alone, depending on how the radiograph is interpreted. Therefore, triage using CXR can help reduce the number of persons who undergo bacteriological TB testing without decreasing the detection of true TB cases. CXR also improves the positive predictive value of subsequent bacteriological tests by increasing the pre-test probability of TB.


People with pulmonary TB who are seeking care often initially present with non-specific respiratory symptoms that need to be evaluated. Given that respiratory conditions are among the most common acute and chronic diseases worldwide, occurring in all societies and in all age groups, respiratory symptoms are among the most common reasons patients seek primary healthcare. Such respiratory complaints (including cough, sputum production, and shortness of breath) constitute approximately 20% of the symptoms that prompt a visit to a primary health center. Thus, respiratory symptoms are both common and non-specific. Most people with respiratory symptoms consistent with TB do not have TB, even in settings where TB is highly endemic. Therefore, it is important to identify in a sensitive and efficient manner those who have a high likelihood of TB among those with respiratory symptoms and to determine the underlying cause of disease for those who are not ultimately diagnosed with TB.

People with respiratory symptoms who are seeking care need to be evaluated not only for TB but for all relevant respiratory diseases, which may include, for example, non-TB lung infections, diseases of airflow obstruction (such as chronic obstructive pulmonary disease or asthma), neoplasms (such as lung cancer or pulmonary metastasis), occupational lung diseases (such as silicosis) or bronchiectasis.

Where it is available and feasible in the outpatient primary care setting, CXR can be used as an effective triage test for those seeking care for respiratory complaints. CXR is a sensitive tool for identifying TB, meaning that it identifies most people with a high likelihood of having the disease, while correctly ruling out TB in most persons when the the CXR is interpreted to identify any abnormality consistent with TB. In addition, CXR can help identify other pulmonary conditions, such as lung cancer and occupational lung diseases like silicosis, as well as other intrathoracic diseases that require further diagnostic evaluation. Therefore, CXR is a useful general triage test for pulmonary conditions because it helps identify which type of further diagnostic evaluation patients require to correctly diagnose the cause of their illness. A normal CXR helps rule out a number of pulmonary conditions and prompts diagnostic evaluation for conditions consistent with no radiological findings, while an abnormal CXR prompts evaluation for conditions consistent with radiographic changes, including but not limited to bacteriological evaluation for TB.

Workflow Description

Image obtained from modality and sent to PACS (or mini-PACS) and the AI engine. Image analyzed by engine. System detects and characterizes exam as either normal or positive for any lung abnormalities. An alert message is sent to PACS from the engine with the information, and identification, and graphic highlighting abnormalities.

Cases in which AI has detected active TB are flagged and prioritized for expedited reading in worklist.

Considerations for Dataset Development


Xray, Chest


primarily AP, PA only, PA/Lat (supplementary Apical Lordotic or decubitus views)

Sex at Birth

Male, Female




pleural fluid (including air/fluid levels), other lung disease (eg. bullous emphysema, malignancy, left-sided congestive heart failure, pulmonary edema, prior chest surgery, prior lung injury, sarcoid, nontuberculous mycobacterial infection, apical radiation fibrosis, chronic aspergillosis

Lung Tissue Involvement

right, left, bilateral} {Distribution: upper, middle, lower, diffuse, segmental or lobar} {cavitary, micronodular, nodular, calcifications, linear scarring

Other abnormalities

pleural effusion, pleural calcification, tracheal deviation, lymphadenopathy, lobar collapse, extrathoracic disease

Disease prevalence

population demographics and disease prevalence

Technical Specifications




X-ray, Chest


PA, AP, PA/LAT, apical lordotic, lateral decubitus

Data Type




Body Region


Anatomic Focus


Primary Outputs

Tuberculosis Detection

RadElement ID



Detection of Tuberculosis in symptomatic adult patient

Data Type


Value Set

  • Unknown

  • Absent

  • Present



Future Development Ideas

• Scaling up detection and treatment of latent TB infection


  1. Chest radiography in tuberculosis detection – summary of current WHO recommendations and guidance on programmatic approaches. I. World Health Organization. ISBN 978 92 4 151150 6

Related Datasets

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