In a small clinic on the edge of the village of Ga-Maja in rural Limpopo, a nurse examines the young woman before her. She’s experiencing pain in her left breast, the woman explains, and there seems to be a lump of sorts. It’s new; she doesn’t remember it being there before. The nurse maintains her composure, keeps her mental diagnosis to herself, even as she hopes she is wrong – she’s never seen the symptoms of breast cancer in a woman so young before. But the future is here for accurate breast cancer diagnosis.

Some 350km south, at the Charlotte Maxeke Academic Hospital in Johannesburg, Professor Qing-Guo Wang from the University of Johannesburg (UJ) is developing technology that will radically change the work of this nurse and countless others. Professor Wang joined UJ’s Institute for Intelligent Systems at the beginning of 2016 and, since his arrival, has been making use of artificial intelligence (AI) to improve systems and processes in South Africa’s healthcare system. His goal: nothing short of saving lives.

When artificial intelligence becomes real and relevant

Originally from Suzhou, China, Professor Wang holds a PhD in industrial automation and is an internationally renowned researcher in multiple engineering-related fields. He is also a leading expert on the fourth industrial revolution (4IR) and AI, knowledge and skills he has now brought to South Africa.

“I’m researching how AI can be applied in real life,” he says, his arms animated, his voice fervent. “I want to explore how it can advance technology and improve services so that people can access a better quality of life.” It is for these reasons that Professor Wang regularly finds himself in the halls of Charlotte Maxeke.

For months, Professor Wang and his team – comprising experts in the fields of theoretical and experimental physics and computer science – have been collecting open data on 20,000 breast cancer cases from the hospital’s archives. They are digitally uploading this data to their unique AI model, a software platform that is becoming increasingly “intelligent” as more and more information is added to it. The cases the team are using are massively varied, and include both positive and negative diagnoses. Once this work is complete, the software will be able to automatically compare a brand new X-ray against the vast repository of X-rays it has in its database.

The result is an affordable and accurate diagnostic system that will be made widely available in health institutions across the country.

Improved access, improved accuracy

The nurse in Ga-Maja conducts a physical exam before breaking the news to her patient. Even if she were to refer her to nearby Lebowakgomo Hospital for an X-ray, she doesn’t have the necessary expertise to read the X-ray once she has it in hand. She’s not a radiologist, not a doctor, not an oncologist. The young woman will have to make an appointment with a specialist, and that could take weeks, maybe months.

“Receiving an accurate breast cancer diagnosis in South Africa is difficult,” says Professor Wang. “Most South Africans can’t afford private medical aid, and those who are dependent on the public system battle to speak to the doctors they need. The system is simply too overrun, specialists are in short supply and demand is significant.” In some instances, doctors who are either overcommitted or poorly qualified are also more likely to give inaccurate results, threatening patients’ lives.

With the arrival of this new AI-modelled software, however, staff at even the most basic hospitals and clinics will require only rudimentary training and an ordinary PC to reap the benefits. The software will demand no specialist expertise and, once installed, will incur no running costs. Once an ordinary X-ray is inserted into the software, it will be able to offer a free and accurate breast cancer diagnosis, the results of which any medical practitioner will be able to read.

This technology means that patients in remote areas like Ga-Maja won’t need to travel to see a radiologist or an oncologist – they’ll be able to be diagnosed by the nurses at their disposal. This increased accessibility has massive advantages in a country as vast, as rural and as under-resourced as South Africa. And the software will also have more knowledge than any real-life oncologist, purely by virtue of the number of cases to which it has virtual access. The sheer volume of data it has to draw on increases its level of accuracy to unprecedented levels.

Interest is one thing; action is another

Professor Wang and his team have a long-term view of this and other projects. For one, making the software available virtually will improve remote access and ensure a wider reach. The team is also looking into developing AI technology that can diagnose other common diseases. “At the moment, we’re exploring five to 10 conditions, which we will work on one at a time,” he says.

In order for projects such as this to develop, however, there needs to be ongoing and concerted national interest, commitment, investment and action, Professor Wang explains. “4IR is about technology. AI is about technology. It’s about people finding new and innovative solutions to context-specific problems and delivering on these ideas. These new services have the potential to improve people’s lives – that’s what matters.”

Eight months from now, our hypothetical nurse is visited by another patient with similar symptoms and an X-ray in hand. She uploads the X-ray into UJ’s system, as she has been taught to do, and runs the analysis. The process is revolutionary. It has made her a better nurse. And it can be depended upon to guide the course of her patients’ lives.

“I want to explore how it can advance technology and improve services so that people can access a better quality of life.”

“Most South Africans can’t afford private medical aid, and those who are dependent on the public system battle to speak to the doctors they need. The system is simply too overrun, specialists are in short supply and demand is significant.”

Eight months from now, our hypothetical nurse is visited by another patient with similar symptoms and an X-ray in hand. She uploads the X-ray into UJ’s system, as she has been taught to do, and runs the analysis. The process is revolutionary.