In the startup hall at MWC 2026, a booth for the South African startup AI Diagnostics showcases its new medical accessory: a wireless stethoscope tuned to listen to a person's breathing to screen for tuberculosis.
In the US and other countries with widespread childhood vaccination, a single early-life shot helps protect infants and children against severe tuberculosis. However, it isn't recommended in adulthood due to limited efficacy, leaving large populations worldwide vulnerable to the disease. Tuberculosis is among the top 10 causes of death worldwide; it infected 10 million people and killed 1.23 million in 2024, according to the World Health Organization, so increasing screening with digital gadgets is a modern way to expand preventative care.Â
AI Diagnostics' tuberculosis monitor works like a digital stethoscope, using two microphones -- one facing the body and the other facing externally for noise cancellation -- to listen to people's breathing to detect the condition early. The device is designed to be easy to use without medical expertise, so that communities can field them for large-scale prescreenings. The monitor aims to filter out patients at under $1 per test, in a way that could be easier to do at scale than more complex screening methods, such as chest X-rays, and cheaper than other methods, such as sputum tests, which cost $7 per examination.Â
"The goal is that the device is usable by lower levels of health care professionals in order to bring their clinical skills up to a doctor, who use AI to [span] that gap," said Jamie Arkin, business development and partnerships lead at AI Diagnostics.
AI Diagnostics leases its monitors as part of a service and doesn't sell them to private clients, though it plans to let governments purchase them directly. The company has a sliding scale of rates from $75 to $500 based on the volume of screenings a customer wants to do.Â
The AI Diagnostics monitor sends patient lung recordings to a paired device, which compares them to a TB audio database.
The monitor sends those recordings to a paired device (I saw it used with a tablet at the company's booth), where AI Diagnostics' artificial intelligence software compares it to an audio database of confirmed TB patients -- and if the AI algorithm determines that it's a close enough match, the person would want to seek an official diagnosis and medical treatment, if needed. It's important to note that the gold standard for a tuberculosis diagnosis is a sputum culture or molecular test; this AI-powered device is a triage tool.Â
AI Diagnostics' audio dataset comprises patient recordings from 15 studies across 10 countries. The company believes it has the largest such TB audio database in the world, and its next goal is to build additional subsets within populations to confirm that the device works with demographics such as pregnant women and older diabetic men, said Arkin.
The paired tablet app instructs where to place the TB monitor on the patient's chest to get lung audio recordings.
AI Diagnostics leases devices to communities in South Africa for a monthly fee, and they've been certified for use by the country's SAHPRA regulatory board. However, the company plans to expand it to other nearby countries in Africa and Asia, such as Namibia, Botswana, Zimbabwe, Bangladesh, Indonesia and the Philippines, once the device is medically certified in each nation in which it seeks to operate.Â
The device is accredited under ISO 13485, a global standard for medical device accreditation, and the company will seek WHO recommendation for its device to help enter additional African and Asian markets.Â
Beyond expanding into other countries, AI Diagnostics also intends to use its monitor device to screen for other lung-based conditions, such as bronchitis, pneumonia, silicosis and asthma, as well as heart conditions, such as arrhythmias, murmurs and congenital heart failure. Given how AI has been used in the tech industry for voice recognition, speech-to-text and other natural language processing, it's encouraging to see it used in the medical field as well.


