NAIROBI, Kenya- Kenya’s medical research community is venturing into uncharted territory by exploring the potential of artificial intelligence (AI) to diagnose tuberculosis (TB) and other respiratory diseases—using nothing more than your smartphone.
At the forefront of this effort is Dr. Videlis Nduba and his team at the Kenya Medical Research Institute (KEMRI).
With AI technology advancing at a rapid pace, the dream of diagnosing TB by simply recording a cough may soon be a reality.
Inside a quiet, specially designed room at KEMRI, Dr. Nduba and his team are working to fine-tune a mobile application that listens to coughs and, using AI, distinguishes between those caused by TB and those due to other respiratory illnesses or even healthy individuals.
The secret weapon? The ResNet 18 system—a sophisticated software from the University of Washington that uses AI to analyze cough sounds.
“We collect coughs from individuals, both healthy and those with respiratory diseases like TB, using three microphones—a cheap one, a high-definition one, and even a smartphone mic,” explains Dr. Nduba.
“The software then creates something called cough spectral grams, mathematical images of coughs, to determine if there’s a difference between someone with TB and someone without,” Dr. Nduba added.
While this tech is impressive, it’s still in the trial phase. The goal? To drastically reduce the time between symptom onset and diagnosis, slashing the potential for community spread.
According to Nduba, “The biggest achievement here is reduced time to diagnosis. Today, the average time from symptom development to diagnosis can stretch anywhere from three months to a year. That’s a long window where patients are actively spreading the disease.”
Though the potential is massive, the software needs some refining. Currently, Nduba’s AI diagnostic tool can detect TB with about 80pc accuracy, while it identifies non-TB cases at around 70pc.
The World Health Organization (WHO) requires a minimum 90pc accuracy in diagnosing TB and 80pc in confirming non-infections before granting its stamp of approval.
Nduba is optimistic, though. “We’re not far from WHO’s target, just a bit of fine-tuning is needed,” he says.
Former TB patient Johnson Munori, who participated in the trials, shares his excitement for the technology.
“Before, I didn’t know I had TB until a doctor diagnosed me. With this app, I recorded my cough, and they confirmed my TB status, which was really convenient. I think this new tech will help a lot of people,” says Munori.
AI’s role in healthcare is expanding rapidly, and experts like Jarim Omogi, a public health specialist at Amref International University, are optimistic about its potential.
“AI is already being used in various ways in healthcare. What’s key here is how cost-effective and fast it is. It’s not just about diagnosis; it’s about how quickly patients can get help,” Omogi explains.
With the backing of the U.S. National Institutes of Health (NIH) but no regulatory approval yet, this mobile diagnostic tool still has hurdles to clear.
However, if it succeeds, the application could transform TB diagnostics in resource-limited areas, helping curb one of the world’s deadliest infectious diseases.
