Researchers at Ateneo de Manila University have found that artificial intelligence could reduce the cost of tuberculosis screening in rural Philippine health units by roughly 23%, but only if the technology is carefully tailored to local conditions rather than simply imported as a high-tech solution. The study challenges the assumption that AI's primary value lies in outperforming human experts, instead focusing on whether the technology can deliver expert-level care to underserved communities at a price they can actually sustain.

Why Does TB Screening Cost Matter in the Philippines?

The Philippines carries a staggering share of the global tuberculosis burden. According to the World Health Organization, an estimated 739,000 people in the Philippines developed tuberculosis in 2024 alone, accounting for 6.8% of the 10.8 million TB cases worldwide. Yet many Filipinos lack access to timely diagnosis and care. In geographically isolated or disadvantaged communities and rural health units, even when patients manage to get a chest X-ray taken, waiting for a radiologist or teleradiology services to interpret it can take weeks or longer. That delay often means another trip to a health facility, additional out-of-pocket expenses, time away from work, or a missed opportunity for continued care.

Early detection is critical for TB. Finding the disease before it becomes severe can mean the difference between receiving treatment that prevents irreversible lung damage and developing advanced, harder-to-treat disease. For rural and economically disadvantaged communities, the bottleneck is not access to imaging technology itself, but access to the human expertise needed to interpret those images.

What Did the Ateneo Researchers Actually Find?

Researchers Dr. Harold Chiu, Dr. Bryan Lao, and Dr. Gloanne Adolor developed a cost-effectiveness model based on a theoretical annual cohort of 1,000 presumptive TB patients undergoing chest radiography in rural health units. The analysis tracked costs and health outcomes over five years, including AI software and operating expenses, radiologist reading fees, and confirmatory GeneXpert testing (a molecular test used to confirm TB diagnosis).

The numbers were striking. The AI-assisted strategy would cost an estimated 877,330 Philippine pesos annually, compared with 1.14 million pesos for manual interpretation by radiologists or teleradiology services. Divided across the 1,000 individuals screened, this translated to approximately 877 pesos per person with AI-assisted interpretation, versus approximately 1,142 pesos per person using manual interpretation. That represents a savings of about 265 pesos per person, or roughly 23% cost reduction.

But Cost Savings Aren't the Real Story. Here's Why?

The researchers emphasize that the significance of AI extends far beyond its ability to read an X-ray efficiently. The fundamental question for resource-constrained communities is not whether AI can match or exceed the performance of expert radiologists, but whether it can extend expert-level diagnostic support to places where such expertise is scarce in a way that is affordable, sustainable, and equitable.

"For resource-constrained communities, the most important question is therefore not whether AI can outperform or assist an expert reader, but whether it can extend expert-level support to places where expertise is scarce in a way that is affordable, sustainable, and equitable," the researchers stated.

Dr. Harold Chiu, Dr. Bryan Lao, and Dr. Gloanne Adolor, Ateneo de Manila University

This reframing is crucial. It shifts the conversation from "Is AI as good as a human radiologist?" to "Can AI help deliver healthcare to people who otherwise would have no access to radiologist expertise at all?" For rural communities where waiting weeks for a teleradiology reading is the current reality, even an imperfect AI system that provides immediate feedback could represent a dramatic improvement in care access.

What Conditions Would Make AI Work in Rural Health Units?

The researchers identified several practical requirements for successful implementation:

• Portable Digital X-ray Systems: The AI system must work with portable or mobile radiography equipment that can be deployed to remote health units, not just hospital-based imaging centers.
• Limited Connectivity Requirements: The AI system must function in areas with unreliable or limited internet connectivity, rather than requiring constant cloud connection or real-time data uploads.
• Integration with Existing TB Programs: Rather than introducing another standalone high-tech tool, the AI must be woven into existing tuberculosis screening and treatment workflows that communities already use.
• Affordability at Local Price Points: The cost structure must align with what rural health units and patients can actually pay, not just what the technology costs to develop.

What Could Go Wrong With This Plan?

The study's own findings reveal important caveats. When researchers modeled scenarios with lower manual or teleradiology reading fees, or when they applied diagnostic performance estimates based on Philippine patient populations rather than international benchmarks, the AI approach remained more effective but was no longer necessarily cost-saving. In other words, the economic case for AI depends heavily on local conditions. If a region already has access to affordable radiologist services, AI may not offer financial advantages. If diagnostic accuracy differs between populations, the model's projections may not hold.

Additionally, the study is based on theoretical assumptions about costs and diagnostic accuracy. Real-world implementation would require confirmatory testing for all AI-flagged cases, meaning the AI system would not replace radiologists entirely but rather serve as a first-line screening tool.

What Do the Researchers Recommend Instead of Immediate Nationwide Rollout?

Rather than rushing to deploy AI across the entire country, the researchers recommend a measured, evidence-based approach. They suggest starting with targeted pilot implementation in carefully selected underserved rural health units, alongside local validation of the AI system's accuracy in Philippine populations, quality assurance protocols, ongoing monitoring of outcomes, and detailed budget assessment. This phased approach would allow policymakers and health systems to learn what works in their specific contexts before committing to larger investments.

The broader lesson is that technology alone does not solve healthcare access problems. The Philippines carries a significant share of the world's tuberculosis burden and struggles to provide universal healthcare to all its citizens. For a country facing these challenges, the question is not simply "Can we bring the newest technology into healthcare?" but rather "Where can that technology help deliver expertise to meet the realities of people with the least access to it?" For TB screening in rural communities, AI may be part of the answer, but only if it is designed and deployed with local conditions, costs, and community needs at the center.