A Ray Of Hope: AI Paves The Way For TB Detection In Underserved Regions

In a world where technology continues to reshape the way we live, the field of healthcare is no exception. One area that’s experiencing significant transformation is the use of artificial intelligence (AI) in identifying diseases through chest X-rays. These cutting-edge algorithms can analyze chest X-rays, identifying anomalies and various conditions such as pneumonia, tuberculosis, and lung cancer. The result? Faster, more accurate diagnoses that lead to improved patient care.

A recent study highlights the accuracy and life-saving potential of these technologies even in underserved communities with a shortage of detection tools and trained physicians, which usually are also those that bear the hardest tool of the infections.

Just think that in 2021 there were an estimated 1.6 million deaths from tuberculosis worldwide (making it the world’s most deadly infectious disease): 82% of these were in Africa and South East Asia.

Presented last week at the European Congress of Clinical Microbiology & Infectious Diseases (ECCMID), the “Comparison of chest radiograph interpretations by artificial intelligence vs radiologists in the diagnosis of pulmonary TB” study, led by Dr. Frauke Rudolf from Aarhus University, compared the performance of AI software qXR in correctly diagnosing tuberculosis (TB), with that of two Ethiopian radiologists with different levels of experience. It turned out that the software was able to match or improve the performance of the two professionals.

In a retrospective analysis of X-rays images of 498 patients, including 57 diagnosed with TB, the AI software demonstrated a “sensitivity” of 75% and a “specificity” of 85.7%, meaning that it correctly identified 75% of the positive cases and demonstrated an 85.7% accuracy rate in providing negative results for patients without TB.

These results outperformed the less experienced radiologist (62.5% and 91.7% respectively) and matched the experienced radiologist (sensitivity of 75% and specificity of 82%).

What makes the experiment even more remarkable is that it was conducted on mobile phone photographs of analogue chest X-rays, enhancing its applicability in underserved areas.

“In low resource areas with a high incidence of TB but a shortage of radiologists, chest X-rays could be photographed with a mobile phone and the image sent to be analyzed remotely by the AI,” Dr. Rudolf said in a statement.

This does not mean, of course, that doctors should rely only on the AI’s assessments, or that these technologies can be a silver bullet, a one-size-fits-all approach to identify the disease.

A study published in September 2021 in the Lancet Digital Health journal, while confirming the effectiveness of computer-aided detection systems in detecting tuberculosis in chest X-rays, also mentioned some limits of these systems.

Their performance may vary in certain subgroups of patients, such as those with HIV or other immunosuppressive conditions. Additionally, while AI systems can help identify potential TB cases, a full clinical examination and follow-up tests, such as the GeneXpert molecular test, are still recommended by experts.

The best results could be obtained by combining the two approaches: automated AI interpretation of chest X-rays for mass screening, and the more expensive GeneXpert molecular testing to confirm the diagnosis. This is what the U.S. Agency for International Develop (USAID) and the FHI 360 nonprofit are doing in Vietnam, in the Support to End Tuberculosis project.

Beginning in 2022, the initiative has employed AI for community outreach and busy district-level facilities, analyzing more than 100,000 X-ray images, which helps develop tailored strategies for at-risk groups. According to a report published by the nonprofit, roughly half of the individuals diagnosed with TB exhibited no symptoms and might have gone undetected without the use of chest X-rays.

Whatever the approach, it’s crucial to identify and treat TB infections as widely and as early as possible. Tuberculosis is preventable and curable, but undiagnosed disease and delayed medical care can lead not just to serious health issues (and potentially, death) for the patient, but also to the spread of the infection; according to the WHO, people ill with TB can infect up to 10-15 other people through close contact over the course of a year.