The doctor in the box

Imagine a nurse in a rural health unit in some far-flung province. She has a basic X-ray machine, a basic smartphone, no internet access, and a patient in front of her with a cough that has not gone away in three months. The nearest radiologist is four hours away. The nearest specialist is in Manila.

That gap between the patient who needs a diagnosis and the specialist who can give one is a key concern of our public healthcare system. The Universal Health Care Act of 2019 gave us the legal basis to fix it. What we have not yet done is reach for the tools that could actually address the concern.

One of those tools arrived quietly last year as the result of research by the US National Institutes of Health, through its National Institute of Biomedical Imaging and Bioengineering.

The tool is called Merlin, aptly named after the legendary, powerful wizard and prophet in Arthurian mythology.

A doctor can feed Merlin a 3D scan of a patients body and Merlin reads that scan the way an experienced radiologist would, looking for signs of disease across hundreds of conditions at once. This is what makes Merlin unique as an Artificial Intelligence (AI) tool for medical science.

Older medical AI programs were narrow: one program for tuberculosis, a separate one for lung tumors, another for broken bones. If a patient had all three problems, the doctor had to run three separate programs. Merlin looks at everything in a single pass, predicts five-year health risks with 75% accuracy, and flags the most urgent cases first — strokes, tumors, and blocked arteries.

Merlin learned to do this by studying 1.3 million scans alongside the written reports that radiologists made about those scans. Over time, it learned to connect what a scan looks like with what a doctor would say about it. A clinician can now ask it a question in plain language, and it will look, even for things it was never specifically programmed to find.

For a well-equipped hospital in an urban center like Makati City, this is a useful upgrade. But for the rest of the country, where there are not enough specialists and where internet connections outside cities are unreliable, it is something bigger. It is a chance to improve diagnostic capability.

This is already happening in other countries with similar challenges. In India, for instance, a company has deployed AI software on ordinary smartphones and low-cost devices to screen chest X-rays for tuberculosis and other conditions like pneumonia, operating offline without internet.

Available information indicate that the AI runs locally on small devices or phones at clinics, giving rapid imaging analysis without distant servers, and providing specialist-level opinions on medical images regardless of connectivity, like a compact diagnostic expert.

In China, various AI-assisted systems have also been deployed in rural clinics to enhance diagnostics, and provide offline diagnosis recommendations, similar case retrieval, and medical info lookups without internet. Other initiatives include smart village doctor systems with AI for exams like blood pressure and glucose, plus tools using data for prescription references.

And in Nigeria and Uganda, AI tools reportedly enable faster TB and fracture diagnosis in remote areas with minimal prior support, processing images offline on local devices or phones. These act as compact, on-site “diagnostic experts” independent of servers or connectivity.

In all these situations, the AI does not live in a distant server somewhere. It lives in a small box on-site, doing its job whether or not the internet is working. The “doctor” is a “jack in the box,” offering informed opinions on medical imaging data provided by the rural health unit or the patient.

This is the model the Philippines needs to follow. And the law we already have, the Universal Health Care Act, gives us the legal foundation to do it. One section of that law requires hospitals and clinics to use compatible digital health records, the data AI needs to work properly. Another section of the law created a government body responsible for approving new medical tools for public use. The pieces are in place. What is missing is the connection between them.

The approval process for new medical tools is currently designed for physical equipment like an X-ray machine, or a blood pressure monitor. AI is different. It improves as it processes more data. A tool approved today will be more capable in two years. What we need is an approval process that accounts for that.

We should also settle the issue of legal responsibility. If an AI tool clears a scan as healthy, the doctor agrees, and the patient later turns out to be seriously ill, who is at fault? The Department of Health needs to formally recognize AI as a support tool, something that helps the doctor but does not replace the doctor’s judgment. The doctor makes the final call. The law needs to say that clearly. Without that protection, most doctors will simply refuse to use AI tools, and no one can blame them.

Also, a tool like Merlin was trained mostly on scans from patients in Western countries. A 70-year-old Filipino patient has a different medical history from the American patients whose scans shaped Merlin’s training. Different childhood diseases, different diet, different baseline health conditions, etc.

A tool that reads scans accurately in the US may miss things here in the Philippines. So, before any foreign AI tool is used widely here, it must be tested and adjusted using a representative sample of Filipino patients. This is not red tape. This is basic patient safety.

Moreover, clinics will not adopt new technology if it only adds to their costs. PhilHealth can change that by paying clinics more when they use AI to catch tuberculosis and other diseases early. In short, the government should turn a public health goal into a financial incentive.

None of this has to happen all at once. It can start small: AI on smartphones in rural health units, the way India began. It can grow to regional centers with better equipment. It can eventually connect to a national system with full predictive capability. But the groundwork has to be laid now, in law, in regulation, and in budget.

The Universal Health Care Act gave every Filipino the right to health. Merlin, and AI tools like it, give the government the means to deliver on that promise without waiting for enough specialists to be trained and posted to every far-flung clinic in the country. We should start digitizing healthcare and make it real.

Marvin Tort is a former managing editor of BusinessWorld, and a former chairman of the Philippine Press Council.

matort@yahoo.com