How computers can help screen for TB

December 9, 2021 by Shibu Vijayan, Victoria Musonda, Pranati Jha, and Vaishnavi Jondhale

In low-resource settings, it can be extremely difficult to access effective TB screenings—but today, computer-aided detection software makes it easier.

A technician at a chest x-ray laboratory in Nagpur, India, updates patient details in a computer-aided detection software program before running a cloud-based artificial intelligence screening for tuberculosis. Photo: Stop TB Partnership.

A technician at a chest x-ray laboratory in Nagpur, India, updates patient details in a computer-aided detection software program before running a cloud-based artificial intelligence screening for tuberculosis. Photo: Stop TB Partnership.

In 2020, tuberculosis (TB) claimed 1.5 million lives globally, making it the leading cause of death from a single infectious agent. That same year, approximately 4.1 million cases were either not reported or not diagnosed. This was a sharp increase—largely attributable to the COVID-19 pandemic—from the previous year, when there were an estimated 2.9 million missing cases.

Despite the existence of advanced tools and technologies, diagnosing TB remains a challenge, especially in resource-limited settings in low- and middle-income countries.

In many cases, patients are not diagnosed despite having access to diagnostic services. This can occur if the tests are not sensitive enough to detect infection. For example, sputum smear microscopy is the most commonly used diagnostic test, particularly in countries with high rates of TB infection, but it is not especially sensitive.

The chest x-ray, on the other hand, is a highly sensitive tool for early TB screening. The downside of x-rays, however, is that they can have poor inter-reader variability—meaning there is subjective inconsistency from one reader to another—and only modest specificity—or ability to identify patients without disease. Furthermore, in many low-resource settings with high TB burdens, including India, there aren’t enough qualified radiologists who can interpret chest x-rays.

Around the world, public health officials continue to prioritize timely and high-quality case detection for TB—and now, a new tool could help make early screening and diagnosis easier, more affordable, and more effective.

Digital tools for early TB screening and more

Recently, the World Health Organization recommended computer-aided detection (CAD) software programs for interpreting digital chest x-rays in screening and triage for TB.

CAD software helps health care workers overcome the obstacles that come with chest x-ray interpretation, enabling more efficient screening of patients.

Studies found that CAD programs are an effective and suitable alternative to human readers—a huge advantage for countries seeking to increase TB diagnoses but lacking radiologists.

And the technology can diagnose other diseases. For example, CAD can classify common bone, heart, and lung abnormalities such as calcification, cardiomegaly, mass, nodule, and pleural effusion. CAD can also be used to detect COVID-19 and lung cancer.

Considering that most patients fall out of the care cascade during pre-diagnosis and diagnosis, uptake of CAD is necessary to begin closing the gap of the millions of unreported or undiagnosed TB cases.

There is also a suite of other, emerging digital applications that can be used alongside CAD. For example, digital applications can use artificial intelligence (AI) to evaluate the sound of a cough, for a high-quality, efficient, and standardized approach to screening for respiratory issues.

Additionally, ultra-portable digital x-ray systems—which are often small, wireless, and easily mobile—are excellent tools for providing care to communities in hard-to-reach geographies.

Once integrated in the TB care cascade, these technologies can strengthen case detection and treatment, while also improving patients’ care experiences.

Advancing patient-centered TB care

CAD and other digital tools contribute to value-based health care—a health care model that more and more countries are moving toward.

Value-based health care seeks to improve the patient experience and reduce the patient burden, from travel time and lost wages to the cost of care and medicines. This kind of patient-centered care requires a health system that communicates clearly, makes it easy to navigate and access services, and provides reasonable wait times—all areas that digital tools can strengthen.

For instance, CAD programs can help health care workers streamline and optimize their workflows. This enables them to provide higher quality care with shorter turnaround times for diagnosis and thus faster pathways to treatment when necessary.

Despite these benefits, health care workers and patients alike can be hesitant when it comes to new technologies such as CAD.

To manage hesitance and resistance, new technologies should be implemented with transparency and compassion. It is well documented that sharing knowledge and information about new technologies with patients and health care workers early on and throughout the implementation process can increase understanding and therefore acceptance of the tools.

Training programs for health care workers and other programmatic staff are good avenues for sharing knowledge about the technologies and increasing support for successful adoption and scale-up.

Sustainable implementation and a way forward

Digital screenings have the potential to deliver high-quality services to more people, at minimal cost. But to sustainably scale up the use of CAD and other digital innovations will require cross-sector collaboration. New diagnostic tools are often too expensive for national programs in low-resource, high TB burden countries. Technology partners should consider subsidizing the cost to encourage uptake and scale of their products.

Additionally, best practices for long-term success for CAD programs will vary in different regions and health care settings. The Stop TB Partnership’s practical guide on the implementation of CAD technology and ultra-portable x-ray systems shares insights and guidance from a range of implementation experiences, making it a useful tool for strengthening CAD rollout.

PATH is already supporting the use of CAD tools in the countries where we work. For instance, we have collaborated with partners in India, Myanmar, and Vietnam to test and demonstrate’s CAD triaging tool, qXR. Available at an affordable price, qXR’s AI algorithm for reading chest x-rays enhances the diagnostic accuracy of TB to greater than 95 percent.

And in Zambia, PATH deployed the mobile OneStopTB clinic in the Copperbelt Province in March 2020. The clinic uses a digital x-ray with AI-augmented picture reading and a GeneXpert machine. Together these tools helped diagnose more than 1,300 people, or 21 percent, of the more than 6,300 people screened in the last half of 2020.

PATH will continue to collaborate with governments and technology partners to implement and scale CAD technologies, with a focus on the Stop TB Partnership’s Global Drug Facility for increasing access to TB diagnostics and treatment.