Leapfrogging cancer diagnostics with AI: a mixed-methods assessment of implementation determinants in low- and middle-income countries

Cancer is a relentless disease that strikes people everywhere, but the ability to diagnose it quickly and accurately is not shared equally across the globe. In many low- and middle-income countries, the path to a cure is often blocked long before treatment even begins. The first step in fighting cancer is finding it, which usually requires a pathologist—a doctor who examines tiny slices of tissue under a microscope to identify disease. These specialists are in short supply in many parts of the world, and the laboratories they work in often lack the advanced tools needed to see the smallest details. Without a clear diagnosis, patients wait too long for care, and their chances of survival drop.

In recent years, a new tool has emerged that promises to change this picture: artificial intelligence. This is computer software trained to recognize patterns, capable of looking at the same tissue slides a human doctor would and spotting signs of cancer with great speed. The idea is that if these smart programs can be built into the diagnostic systems of developing nations, they could help doctors work faster and more accurately, potentially allowing these countries to skip the slow, expensive steps of building traditional pathology labs and jump straight to a modern, digital system. But for this to work, the technology must fit the reality of the places where it is needed. It cannot simply be dropped in from the outside; it must be understood how local doctors, budgets, and rules interact with the software.

A team of researchers set out to understand exactly how this might work in practice. They did not just build a computer program and test it; instead, they went to the source, asking the people who would actually use the technology what they needed and what stood in their way. They spoke with twelve doctors from twelve different countries, including ten low- and middle-income nations and two high-income nations for comparison. These doctors were experts in pathology and gynecology, the fields most involved in cancer diagnosis. The researchers used a mix of methods, starting with surveys to map out the current state of laboratories, followed by group discussions to dig deeper into the challenges, and finally a second survey to measure how strongly the doctors felt about specific issues.

The researchers found that the situation on the ground was far more complicated than a simple list of rich and poor countries would suggest. While every single location had access to the basic method of staining tissue slides to make them visible, the ability to perform more advanced tests varied wildly. In seven out of ten of the lower-income sites, the labs did not have enough capacity to run the specialized chemical tests often required to confirm a cancer diagnosis. In six of those ten sites, molecular testing—which looks at the genetic makeup of the tumor—was either completely missing or barely available. This meant that even if a computer could spot a problem, the human doctors often lacked the tools to confirm it or understand exactly what kind of cancer it was.

The availability of the digital tools needed for artificial intelligence was also a mixed bag. Most of the sites had scanners that could turn physical slides into digital images, a necessary first step for the computer to "see" them. However, the powerful computers required to run the artificial intelligence software were missing in more than half of the locations. This created a strange disconnect: some places had the scanners to create the images but not the computers to analyze them, while others had the computers but not the scanners. The researchers noted that this variation did not follow the usual pattern of national wealth. A country's income level was not a reliable way to guess whether a specific hospital was ready to use artificial intelligence. Some institutions in lower-income countries were surprisingly well-equipped, while others in wealthier nations struggled with basic gaps.

When the doctors were asked what they wanted the artificial intelligence to do, their answers were practical and focused on immediate needs. They did not ask for the software to replace them or to predict the future with perfect certainty. Instead, they wanted the computer to help them find cancer earlier, to double-check their own diagnoses, and to predict certain biological markers from the standard stained slides. They imagined a system where the computer could sort through thousands of slides, flagging the ones that looked suspicious so a human doctor could focus their time on those specific cases. This would be like having a tireless assistant who never misses a detail, allowing the overworked doctors to see more patients without sacrificing accuracy.

However, the path to making this vision a reality is blocked by a series of interconnected hurdles. The researchers identified six main factors that determine whether artificial intelligence can actually be used in these settings. The first is literacy, meaning that doctors and patients need to understand what the technology is and trust it. The second is evidence, or the need to prove that the software works well for the specific populations in these countries, rather than just for people in wealthy nations. The third is governance, which involves creating rules and laws to guide how the technology is used and who is responsible if something goes wrong. The fourth is clinical deployment, which covers the physical steps of getting the software into the hospital workflow. The fifth is funding, and the sixth is collaboration between different groups.

Among all these challenges, money emerged as the single biggest obstacle. The doctors reported that financial constraints were the dominant barrier across every single area. They explained that without a steady source of funding, they could not buy the necessary computers, could not pay for the internet connections needed to send images, and could not afford to train their staff. While international grants and research money were available to buy the initial equipment, the doctors worried that these funds would run out, leaving them with expensive machines they could not maintain or repair. They emphasized that for artificial intelligence to last, it needed to be supported by the government's regular health budget, not just by temporary donations.

Collaboration was identified as the key to unlocking the other solutions. The doctors realized that no single person or hospital could solve these problems alone. They needed to work together with governments to create policies, with universities to train the next generation of doctors, and with private companies to share the cost of equipment. They saw that the current efforts were often fragmented, with different groups working in isolation. The researchers found that the most successful path forward would be a coordinated effort where these different stakeholders joined forces, sharing resources and knowledge to build a system that could sustain itself over the long term.

The study concluded that while artificial intelligence holds great promise for helping low- and middle-income countries leapfrog over the traditional, slow development of pathology labs, it is not a magic solution that can be applied everywhere in the same way. The readiness to use this technology varies greatly from one hospital to the next, regardless of the country's overall wealth. Success depends on addressing the specific local needs, securing sustainable funding, and building strong partnerships. Without these foundations, the technology risks becoming another piece of unused equipment gathering dust in a laboratory, rather than a tool that saves lives. The researchers suggest that before rolling out these systems, each institution must be carefully assessed to ensure it has the people, the rules, and the money to make the technology work.