How AI is rewriting future of diagnostics

This room at Makerere University’s College of Sciences – the Artificial Intelligence (AI) Lab – is buzzing—not with machinery, but with the hum of minds at work.

Screens glow with datasets of specimen, while a whiteboard stands battle-scarred from endless equations and breakthroughs and a few researchers are seen across huddling over a microscope, adjusting settings with the careful precision of surgeons.

So what’s happening here? A mix of exciting projects, but one really stands out—turning hours of waiting for disease diagnoses into mere seconds. How? By leveraging large language models. Right now, it's zeroing in on tuberculosis (TB), cervical cancer, and malaria, but they’re already plotting to scale this magic trick far beyond those.

Here’s how it works: A microscopist receives a sample—let’s say a blood smear to test for malaria. Now, anyone with a medical background knows that malaria has a very distinct look under the microscope.

So, the sample is examined, and here’s where the magic happens—the data –imagery or some text or both – is fed into the AI model that Dr Rose Nakasi and her team are developing with thousands of datasets from hospitals here in Uganda.

The AI then takes over, quickly scanning the sample and making a call—like, ‘There’s a 95 percent chance this is malaria’ and gives a reason why. What used to take about 30 minutes in a lab now takes mere seconds. In fact, Dr Nakasi says, it’s just ‘two seconds’.

I’m here on the sixth floor of the College of Computer Sciences, deep in the lab, and across, the team behind the Ocular project is leading a training session for medical professionals from Mulago hospital on how to annotate microscope images.

The process is fascinating: when they receive digital images from the microscope, they must carefully examine them for key indicators. This includes identifying artefacts like debris or acid-fast bacteria, as well as focusing on critical elements such as bacterial structures, thread-like artefacts, white blood cells,and monocytes.

This step is especially important when test results are negative—ensuring that even subtle, often overlooked details don’t go unnoticed. I’m now facing Dr Nakasi,a trailblazing AI researcher turning computer vision and machine learning into game-changers for disease diagnosis and spatial modelling in low-resource countries. I ask her about this ‘Ocular project’ and she does give me an analogy.

"Imagine this," she says, her hands slicing through the air like a conductor leading an invisible orchestra. "Thirty minutes per patient, maybe 30 patients a day, tops. With Ocular? We’re talking about 1,000 patients in a single day." Her voice is now taking on a measured intensity.

"And let’s talk about strain. Imagine spending your entire day peering into a microscope, straining your eyes, battling fatigue. That’s the reality for lab technicians. Ocular eliminates that burden.

Now, instead of scanning slides all day, they can focus on what matters—patient care precision, decision-making." Her hands then come to rest on the table, fingers spread, as if settling a case before the jury.

The cost equation

Across the table, I raise an eyebrow. "But doesn’t AI make this more expensive?" She grins. “Good question. And here’s the reality—Ocular doesn’t require a $10,000 (Shs36.4m) lab setup. You need a smartphone,a 3D-printed adapter (costs less than $10 (Shs36,000)), and a microscope. The entire system? Under $3,000 (Shs10.9m).”

She leans forward. “Now, compare that to the cost of traditional diagnostics—the equipment, the time lost,the human resources stretched thin. Ocular isn’t adding costs. It’s shifting them where they matter—toward efficiency, toward impact.”

The impact

Africa’s healthcare system is grappling

with a perfect storm of challenges—limited technical expertise and a shortage of specialised equipment are crippling its diagnostic capabilities.

The result? Delays, misdiagnoses, and a cascade of consequences: improper treatments, late-stage interventions,and escalating morbidity and mortality rates.

Without reliable diagnostics, patients are often handed unnecessary or wrong treatments, tying up precious resources and forcing the system to deal with repeat visits, longer treatment regimens, and the looming threat of drug resistance. And the punchline? A lack of timely data keeps disease surveillance in the dark, complicating efforts to track outbreaks and allocate resources effectively.

Now, when we talk about diseases like malaria, TB, and cervical cancer—each presents its own monster-sized public health challenge in Uganda.

Malaria alone is responsible for 30 to 40 percent of outpatient visits and 20 percent of hospital admissions, making Uganda one of the world's top hotspots for transmission, public health data shows.

TB,with 200-250 cases per 100,000 people, coupled with the rise of drug-resistant TB (MDR-TB), paints an urgent picture of the need for better diagnostic methods. Then there’s cervical cancer—the silent killer, making up 40 percent of all cancer cases in Ugandan women, largely due to inadequate screening and late detection.

For effective management of these diseases, this means we need both skilled professionals and cutting-edge diagnostic tools.

Microscopy is crucial in disease diagnosis, but while microscopes are common in Uganda and many developing nations, the real bottleneck is the shortage of trained laboratory personnel. This lack of expertise limits access to high-quality diagnostics, especially in areas battling endemic diseases like malaria and cervical cancer. Enter transformative technology, the game-changer that’s shifting the paradigm.

“By automating pathogen detection for diseases like malaria and TB, this technology is reducing diagnostic errors and speeding up treatment decisions. It’s already making waves in Uganda, revolutionizing medical diagnostics, and setting the stage for broader regional applications,”Dr Nakasi says. The fear of the unknown But like anything new in society, this innovation faces skepticism because—well, change is hard.

People question if the tech can truly deliver, some worry about the learning curve, and others just prefer the comfort of the status quo. Plus, let’s face it, the fear of the unknown is always a tough pill to swallow—especially when it’s shaking up a deeply entrenched system.

“You tell a clinician AI is diagnosing patients, and their first reaction? ‘Over my dead body.’” she laughs, shaking his head.

“In Africa especially, AI in healthcare was a foreign concept. Clinicians asked, ‘Why should I trust a machine over my own expertise?’” she pauses, then points to the team’s solution. “So, we didn’t just build a tool—we built trust. We worked side by side with hospital staff, demonstrated results, let them experience the difference firsthand. The shift was gradual, but it happened.” And the other fear? ‘Will AI take my job?’

Dr Nakasi raises a reassuring hand. “AI is not replacing humans—it’s augment- ing them. We made that clear from the start – Ocular doesn’t remove experts from the loop; it enhances their capability. Instead of spending hours on a single slide, they get fast, accurate results, then make informed decisions. That’s the future—not automation, but collaboration.”

A key feature is a teaching aid system designed to provide real-time guidance and feedback, significantly improving diagnostic accuracy while providing continuous training for microscopists and other laboratory professionals.

Scaling

This project is currently building a machine-learning framework that plugs into existing disease surveillance, offering real-time data analysis and outbreak predictions. Over three years, it’ll be piloted across Uganda’s four regions, with Dr Nakasi focusing on perfecting the AI, testing it in clinics, and measuring its impact on both patients and the healthcare system.

The project has secured both administrative and institutional review board (IRB) approvals for malaria and cervical cancer, ensuring it meets all ethical standards and health research regulations in Uganda.

With more than 30,000 data sets from Ministry of Health-approved hospitals, it’s fueling the AI’s need for quality data, ensuring evidence-based diagnostics that pack a punch.

The malaria model hit a solid Map50 of 0.685, while the cervical cancer mod-el boasts an impressive 94 percent accuracy. With 150 technicians and 10 pathologists trained, the tool is now being used across six health facilities turning data into diagnosis, one breakthrough at a time. "This is bigger than just Makerere," she says, hands clasped, voice steady. “This is about changing global diagnostics. Scaling this technology beyond Uganda, beyond Africa. Making AI-powered healthcare the norm,not the exception.” She exhales, a quiet moment of reflection. Then, with a small grin, she adds—“And we’re just getting started.”

Obstacles

This project is sizzling with innovation but not without its hurdles. The TBIRB approval is still in limbo, which has slowed down TB data collection and AI model development. But the team is locked in with the Mulago Research Ethics Committee, hopeful to get the green light soon.

And then there’s the classic Ugandan innovator’s headache: hardware. The Ocular project hit roadblocks with 3D printers that needed constant recalibration. But no worries, early hiccups were smoothed out with rapid redesigns—because who doesn’t love a challenge?

But the real kicker? Funding. Dr Nakasi is optimistic that if Uganda’s government sees the immense potential of this solution and backs it financially, the project could go from ground-breaking to game-changing, scaling across diseases and health centres nationwide—giving the country a much-needed boost in health innovation.

Interventions

The Science, Technology & Innovation Secretariat (STI-OP), the government-backed powerhouse under the Ministry of ICT, is on board with AI’s transformative potential for sectors like agriculture, health, national security, tourism,and banking. But here's the catch: the data is scattered, digital and non-digital resources are fragmented, and access to computing is limited—hello, innovation roadblock!

Add to that a dearth of local research, reliance on imported architectures, and a workforce largely focused on fine-tuning existing tools,not building them.

But fear not! The STI-OP is pushing for a game plan with initiatives like the Uganda Data Exchange Platform to centralise and monetise datasets, AI for Business Process Outsourcing (BPO) to equip talent for global AI gigs, and an AI Studio to fuel research and adoption.

Their call to action: developers, build on the infrastructure; academia, create the algorithms,data,and talent; and government, unlock free access to datasets and AI tools for all. Dr Nakasi’s Ocular project is not just a game-changer; it’s rewriting the playbook for healthcare diagnostics.

In a world where waiting for test results feels like waiting for the next season of your favourite show, Ocular fasttracks answers with AI-powered precision, proving that innovation doesn’t need to break the bank to be transformative.

Sure,there are hurdles—like the classic “we need more funding and IRB approvals”—but with the drive and vision that Dr Nakasi and her team bring to the table, the future of healthcare doesn’t just look bright—it looks lightning fast. And as she says,“We’re just getting started.” Let the AI revolution begin!