Artificial Intelligence in Tuberculosis Diagnostics: Opportunities and Emerging Challenges

Dear Editor, Tuberculosis (TB) continues to be a formidable global health challenge, particularly in countries of the Global South, with India alone contributing nearly one-third of the global burden.[1] Despite advances in diagnostic technologies such as GeneXpert, Line Probe Assays (LPA), and digital radiography, persistent gaps in timely diagnosis and treatment initiation have impeded progress toward the End TB targets. In this context, artificial intelligence (AI) has emerged as a transformative tool that promises to enhance diagnostic accuracy, expand access, and improve programmatic efficiency.[2] AI-powered computer-aided detection (CAD) tools for chest radiography are among the most widely deployed applications.[3,4] Platforms such as CAD4TB, qXR, and Lunit INSIGHT have demonstrated diagnostic performance comparable to that of radiologists in adult population, particularly in high-burden and resource-limited settings.[4,5] They have been integrated into mobile-van-based screening programs and are increasingly endorsed by global health agencies. However, it is important to note that current CAD models are trained largely on adult datasets, limiting their accuracy in children, where atypical radiological manifestations – such as hilar lymphadenopathy or miliary patterns - pose diagnostic challenges.[4] Ongoing efforts are directed toward developing pediatric-inclusive datasets to improve sensitivity in this vulnerable group. Beyond pulmonary disease, AI applications in extrapulmonary TB diagnosis are gaining traction. Research has begun to explore AI-assisted ultrasound for detecting abdominal and lymph node TB, automated computed tomography interpretation for spinal TB, and histopathology-based algorithms for tuberculous lymphadenitis. Given that extrapulmonary TB contributes significantly to morbidity in India and other high-burden countries, AI holds the potential to strengthen this underprioritized domain.[6] The diagnosis of TB in human immunodeficiency virus (HIV)-coinfected patients is another area where AI may provide critical support. HIV-positive individuals often present with atypical chest radiographs and reduced bacillary loads, which lower the sensitivity of conventional diagnostics.[7] AI models trained on diverse datasets may help detect subtle radiological or clinical patterns, thus improving case detection in this vulnerable subgroup. Molecular diagnostics are also being transformed through AI integration. For GeneXpert, algorithms are being developed to analyze raw amplification curves and flag borderline or invalid results, while in LPA, computer vision models are automating the interpretation of banding patterns.[8] These advances reduce turnaround time and interobserver variability, thereby supporting more reliable detection of drug resistance. Similarly, AI systems such as TBDx and BacterIA are automating sputum microscopy and digital pathology, minimizing dependence on highly trained human resources.[9,10] Importantly, the promise of AI is not confined to diagnosis alone. Treatment monitoring and follow-up are emerging frontiers. AI tools capable of automated scoring of radiographic improvement, smartphone-based cough frequency monitoring, and predictive analytics from digital adherence records are being piloted. These innovations may support personalized therapy and reduce treatment interruptions – key barriers in TB control.[11,12] Integration with national TB programs, such as India’s national tuberculosis elimination program (NTEP) and Nikshay platform, is critical for sustainability. Pilot projects have demonstrated that AI can be embedded within Nikshay to generate real-time diagnostic alerts, geotag new cases, and link screening outcomes directly to program databases.[13] Such integration not only strengthens surveillance but also enhances accountability and follow-up. While these advances are promising, caution is warranted. Data bias, infrastructural barriers, ethical concerns regarding patient privacy, and the absence of clear regulatory pathways continue to be important limitations.[14] In the future, equitable deployment, transparency in algorithm design, and context-specific validation are essential to prevent AI from reinforcing existing disparities.[15] In conclusion, AI offers a powerful opportunity to bridge long-standing diagnostic and programmatic gaps in TB care. Its thoughtful deployment, particularly in the Global South, could accelerate progress toward TB elimination. However, careful attention to inclusivity, ethics, and sustainability must guide this integration. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.