Innovations for
bi-directional algorithms

The COVID-19 pandemic response pushed aside TB outreach and services, resulting in a 20% drop in diagnosis and treatment worldwide, according to WHO urgent recovery is needed. Data emerging from India and South Africa shows that people coinfected with TB and COVID-19 have three times higher mortality than people infected with TB alone. This makes case finding, contact tracing and bi-directional TB and COVID-19 screening and testing essential.

Strategies to actively find cases and ensure prompt effective treatment exist for high TB burden settings [1]. However, to scale up such strategies for bi-directional screening of TB and COVID-19 in key populations, rapid and more affordable diagnostic pathways are required. As symptoms of TB and COVID can coincide, bi-directional screening is in particular required for people displaying influenza-like illness and severe acute respiratory infections.

The choice for an algorithm to find the missing people with TB will be setting specific, but generally an algorithm composed of Chest X-ray (CXR) screening followed by confirmatory testing with Xpert® can achieve the lowest Number Needed to Screen and highest positive predictive value, and is the least amenable to setting-specific variation [2].

The use of computerized scoring of CXRs has the potential to extend the use of CXR in resource constrained settings which still face challenges of sufficient qualified staff to interpret CXR correctly, and thus has the potential to minimize the cost of TB diagnosis especially when used prior to more expensive new diagnostic tools. [3]

TB & COVID-19 bi-directional screening algorithms

As computerized scoring can on the same dCXR apply simultaneous identification and quantification of abnormalities suggestive of TB and COVID-19, the need for rapid bi-directional screening at senior radiologist accuracy levels can be served using AI through CE Certified CAD systems.

Some examples:

  • FIND and Swiss Tropical Institute started to perform operational research on diagnostic algorithms that include next to symptom screening AI for TB and COVID-19 using the same chest radiograph (see illustration below).

COVID-19 and TB/HIV screen, triage and treatment algorithm for LIC with community transmission [4]

  • In Ghana CAD4COVID facilitates early COVID-19 case detection and allows clinicians to take faster and better clinical decisions. The use of CAD makes the Ghanaian health system more resilient for this pandemic and as well as for the fight against the world’s deadliest infectious disease, TB.

    GHS using CAD4COVID as part of bi-directional COVID-19 and TB triaging.

Innovative algorithm

Digital Chest Images are instantly available and can be interpreted by CAD or by a human reader who does not need to be in the same location by using the internet or mobile phone network connections. This makes digital CXR a rapid and relatively low cost triage tool for Xpert® eligibility also fit for use in low resource areas. This enables NTP’s to boost diagnostic output at same budgetary input (online Economics of Screening model).

The above cost indications for dCXR/CAD and Xpert® include salary cost (LMIC), cartridge cost and provisions for maintenance for dCXR and GeneXpert® system.

TB Innovations Summit NYC 2018

The TB Innovation Summit was forward-looking and solutions-oriented with the tagline for this event being "Our Journey to End TB Begins Now”. At this summit preceding the historic UN General Assembly on TB in September 2018, Computer Aided Detection for TB (CAD) was selected and presented as one of the key innovations for early detection.



WHO guidelines on CXR

  • Important triaging tool for pulmonary conditions including TB
  • Useful to decide who should be tested with Gene Xpert® MTB/RIF
  • Most sensitive TB screening tool and therefore recommended as the first screening test in prevalence surveys
  • Optimal first screening tool for high risk groups.
  • Recommended tool to rule out TB before initiating LTBI treatment (in well-resourced countries with TB incidence less than 100/100.000)
  • Useful diagnostic tool for non-bacteriologically confirmed TB, for example in PLHIV and children.

Case detection in children

It is more difficult to diagnose TB in children than in adults as they are less likely to have the classical signs and symptoms of the disease, like couch. Children have usually less bacilli in the sputum and the younger children often cannot produce a specimen. Therefore CXR plays an important role in TB diagnosis in children. The intended use of CAD4TB as from the 2018 version 6 was enlarged to serve subjects as from 4 years of age.