Doctors in white coats walk quickly past tired patients clothed in bright, patterned kitenge fabrics.
Upstairs, lab technicians tackle a mounting pile of slides for microscopic examination - staring through the eyepiece at blood samples suspected of containing parasites of malaria or tuberculosis. It's a time consuming process, with each slide finely adjusted by hand around 100 times before a confident diagnosis can be given.
But this is changing. Uganda's first Artificial Intelligence (AI) lab, at Makerere University, has developed a way to diagnose the blood samples using a cell phone.
The program learns to create its own criteria based on a set of images that have been presented to it previously. It learns to recognize the common features of the parasites.
"Microscopists usually have a problem with their eyes because of over straining," says Martha Nakaya - an experienced lab technologist of 11 years.
Lab technicians should process no more than 25 slides each day, but a lack of qualified workers lead some to process four times as many.
"We have so many patients who may require malaria and TB tests, and we have one technician looking at all these slides," agrees doctor Alfred Andama, standing in the busy lab. "Apart from affecting their eyes, this also compromises their ability to report correctly what they see."
Andama is among a team of healthcare workers and coders trialling the prototype device that could put an end to this painstaking process - diagnosing patients more quickly, cheaply and accurately.
So how does it work?
Clamped in place over one microscope eyepiece, a basic smartphone brings to light a detailed image of the blood sample below - each malaria parasite circled in red by artificially intelligent software.
Nakaya verifies that the computer is correct, pointing to the tell-tale dot and comma shape of the parasite.
PhD researcher, Rose Nakasi, 31, is the brains behind this technology. "Almost everyone in Uganda, including me, has had malaria" says Nakasi, who is a researcher in computer science. "It affects me as a person, and it affects Uganda. So I feel attached and want to contribute in any way that I can to its proper diagnosis."
In 2016, Uganda's Ministry of Health found that the disease is the leading cause of death in the country - accounting for 27 per cent of deaths.
Mortality rates are particularly high in rural areas, where the lack of doctors and nurses is acute. Nursing assistants are often taught to read slides instead, but inadequate training can lead to misdiagnosis.
"You have cases where someone goes to the hospital and is diagnosed negative, but after a few days they come back and there is malaria," says Nakasi.
With her technology, pathogens are counted and mapped out quickly, ready to be confirmed by a health worker. Diagnosis times could be slashed from 30 minutes to as little as two minutes.
The idea is not to take away technicians jobs, but to make them easier. The technology has to "work hand in hand with the lab technicians," Nakasi explains. Their expertise is needed to train the device, and moving forward will make their work more efficient.
The AI software is built on deep learning algorithms that use an annotated library of microscope images to learn the common features of plasmodium parasites that cause malaria and the bacteria called Mycobacterium tuberculosis that is responsible for tuberculosis.
The device is yet to be rolled out beyond small-scale trials in Kampala's hospitals, but the biggest challenge may lie ahead as the technology is taken to remote areas.
'Phone to detect viral diseases'
Convincing patients of the value in this technology over conventional methods will be difficult, says Daniel Mutembesa, another researcher at the Makerere AI Lab. "People know a phone for calling and sending messages, they don't know that a phone could do your diagnosis," he explains, adding that ethically, patients can choose to have their diagnosis made by a technician.
By offering the smartphone diagnosis to patients for a fraction of the price, Mutembesa hopes that the technology will be adopted, and trusted, quickly.
Like all projects out of the Makerere AI lab, this device makes use of the falling cost and increasing power of smartphones. Another app developed by the lab, called mCROPS, also harness the smartphone's inbuilt camera. Farmers across Uganda are using mCROPS to detect viral diseases in cassava crops, and to monitor their spread across the country.
The technology used in the mCROPS programme uses a basic spectrometer to analyze the chemicals in the plant leaves, and imaging software that can be used to count the number of white flies on plant leaves.
Looking out across Kampala's urban sprawl, Nakasi says she is a strong believer in developing local tech solutions to local problems. Intermittent internet connections and under-resourced Ugandan laboratories present challenges that imported technology may not be equipped to handle she explains.
"Silicon Valley develop technology that fits that environment, we want to find solutions that can fit our environment," she says.
No comments:
Post a Comment