Dr. Jaishree Naidoo was in charge of pediatric radiology at a South African hospital in 2014, when she had a moment of epiphany after coming across a news story on the usage of AI recognition pattern in distinguishing animals.
As a radiologist with 20 years of experience, Naidoo was already familiar with pattern recognition, and she could immediately see how AI could be used in the industry to transform the access to diagnostic imaging. The flame had been lit, and in 2019, together with her husband, Terence Naidu, and Andrei Migatchev, Naidoo launched Envisionit Deep AI, a healthtech startup that uses AI to bridge demand for diagnostic imaging.
Envisionit Deep AI is now on the path to growth backed by a $1.65 million investment from New GX Ventures SA, a joint venture between New GX Capital, RMB Ventures and GIIG Africa. This was after the startup emerged the Southern Africa regional winner at the African Startup Awards.
“We have this exciting goal to combine revolutionary technology like artificial intelligence with radiology and we want to transform the way radiologists look, interpret imaging and make diagnosis,” said Naidoo, who doubles up as the company’s CEO.
Hybrid solution
The startup has a suite of products that it plans to scale beyond South Africa, including the Radify AI platform, which they claim ensures speedy, accurate, quality and affordable medical imaging diagnosis, factors that are critical in early diagnosis and treatment of disease.
“Radify AI has received approval from the South African Health Products Regulatory Association. But we want to go global and that is why we’re busy seeking approval from FDA and European Medical Agency,” said Naidoo.
Naidoo says Envisionit Deep AI’s ultimate goal is to reduce the burdens in the healthcare system, especially in Africa, where infrastructure and human resource investments remain dismal.
Data shows that the doctor to patient ratio across sub-Saharan Africa is one of the lowest in the world, a proportion that is even worse for specialty doctors. In radiology, the scarcity of human resources is so bad that the radiologist to population ratio in Kenya, for instance, is 1: 389,255, while in Nigeria it’s 1:566,000.
This dearth in radiologists is the driving force for Naidoo to make Radify AI accessible to all, especially in peri-urban and rural regions, and what informed the startup to build a hybrid solution.
She says Radify AI can be deployed anywhere, “whether in a first world setting or a rural clinic that may not have great broadband capacity or the best infrastructure… because the whole idea for our product is to democratize access to diagnostic imaging, and we will not be democratizing if we lack solutions that can go out into rural settings.”
The startup’s on-site product can be integrated with devices such as X-ray machines to deliver diagnosis and treatment at the point of care. They also offer tele-radiology for patients that require radiology reports.
“Typically a radiographer takes X-rays, then patients would go back home and collect the results later, sometimes after several months. Delayed diagnosis means diseases could advance further. We are taking that lag away, because when you know what to treat, you treat it quickly,” said Naidoo, adding they are planning on introducing its solutions to South Africa’s mining sector, where employees have a higher risk of getting tuberculosis.
The startup started out by building models for interpreting chest X-rays, with the capacity to detect 25 different pathologies, including tuberculosis, breast cancer, and pneumonia, the number-one killer of children under five in Africa.
This platform, Naidoo said, proved beneficial especially during the Covid pandemic, when Envisionit Deep AI unveiled a product that could detect Covid pneumonia from chest X-rays in under 25 milliseconds. This was deployed to drive efficiency at a 700-bed hospital in South Africa’s Northern Cape province, which only had one radiologist. Naidoo says it was also used in several ICUs to triage, especially during the pandemic’s second peak.
The startup says that while the volume of data it processes is integral, it ensures that its models are trained using quality de-identified data drawn globally, and from diverse ethnic groups.
The data can also be reviewed by radiologists through a validation tool that gives them some reassurance that the product is working accordingly; and to get their input and feedback, which has enabled the startup to improve the accuracy of its models.
Envisionit Deep AI has recently rolled out a computer-assisted training model (an edtech tool) for medical practitioners that are keen on gaining radiology skills.