AI will cut medical errors, improve cancer diagnosis

“About a third of the $3.2 trillion spent on healthcare in the US is because of medical errors,” says Rick Mammone, a professor at Rutgers University in New Jersey and CEO of Clearview Diagnostics. Artificial intelligence, he adds, can address this but it’s not just a question of finance. While overall costs could certainly be reduced, Mammone’s main aim through start-up Clearview is to “change the way the world screens and diagnoses cancer.”

It’s a noble goal but one that is not entirely built on altruism. Clearview of course is a business operating in a sector which research firm MarketsandMarkets recently claimed is expected to grow from $667m in 2016 to almost $8bn by 2022. Healthcare technology is big business but Mammone, a near lifelong academic, says he started the business because of the impact that AI could have, not just on cancer diagnoses but on a number of conditions, and especially in places where there is a limited number of doctors, or none, to diagnose.

“Women die of breast cancer in places like India and Africa because they have no mammograms… nothing,” says Mammone, whose ultimate aim is for a smartphone-based diagnosis “so women can self-examine”.

That’s still a little way off but the work Mammone and his co-founder Christine Podilchuk are undertaking is already starting to bear fruit. Early trials have shown significant reductions in “unnecessary biopsies” which Mammone says also go some way to reducing unnecessary US medical spend.

“We are seeing additional imaging and testing and more complicated therapies for cancer, so healthcare costs have been rising steadily,” says Mammone.

“Seventeen percent of US GDP is going to healthcare. How do we reduce this? The thing about Obamacare is that it doesn't matter who pays, whether government pays or the employer pays for the bill; it's just too big of a bill and no one is concentrating on the real problem, which is that we have never ending increased costs.”



Artificial assistants

Mammone says the idea is to “use AI to make better decisions” as currently around 40 million women a year have a mammogram and out of that, around 10 to 15 per cent get an ultrasound or further diagnostic test. Out of those, around 1.7 million get a biopsy and out of the 1.7 million about 200,000 have cancer. “So, 1.5 million you could say were unnecessary,” he says.

It’s a process he believes that AI can improve.

“We have developed over the last several years an AI decision maker that looks at the images [of known tumours and cells] and determines whether or not it is definitely cancer and whether a biopsy is required,” says Mammone. “Biopsies cost around $2,000, plus you have the patient anxiety and subsequent scar tissue. It’s about eliminating the unnecessary biopsies and we think we can reduce this by 50 per cent at the moment.”

The Clearview AI engine uses neural networks and deep learning to scan images and compare with known examples of cancer but it also uses additional data provided by doctors and radiologists to “personalise the data and increase accuracy”, although Mammone is more than aware that measuring accuracy is not straightforward. To overcome this problem he uses a ROC Curve to help plot performance.

“We take the best radiologists and we hold the detection rate constant – a detection rate that will have less false detections by 50-70 percent. If we hold the false detections constant we'll find, say, a few more cancers - there's usually four to seven cancers per 1000 women.”

While the aim is to improve the decision making of all radiologists based on the best understanding of the patient, Mammone wants to make it clear that the technology will not replace radiologists.

“We want to help them, with the routine,” he says. “They have to look at thousands of these images and this helps them do it faster, and hopefully frees them up for more human interaction with patients.”



An army against cancer

Mammone is not alone in his goal. A team of researchers in the UK, at the University of Warwick claimed this month that “cancer cells are to be detected and classified more efficiently and accurately, using ground-breaking artificial intelligence.”

The University of Warwick’s Tissue Image Analytics (TIA) Laboratory has teamed-up with Intel to develop a large, digital repository of a variety of tumour and immune cells found in thousands of human tissue samples, and are developing algorithms to recognise these cells automatically.

Mammone is a little ahead in the game here and plans to extend the AI platform to other forms of cancer diagnoses in the near future.

“The AI won’t replace doctors either,” he adds, almost determined to douse the flames of Professor Stephen Hawking’s forthcoming documentary on AI domination. “Doctors and radiologists are overworked, so we want to help them identify whether palpable breast masses are benign or suspicious. The platform, via cloud-based delivery also makes cancer screening more widely available and accessible in regions even without an established healthcare infrastructure.”

In places like Mexico, where it could take up to 18 months for a woman to get an appointment for typical breast cancer screening, this technology could indeed be life-saving. For Mammone, it’s not ‘if’ but ‘when’ as AI technology develops rapidly and increases the capabilities of humans to solve complex medical problems. In the short term, AI certainly has a huge role to play in keeping us all alive. In the long term? That’s another matter.


« Millennials talk careers: Roxanne Abercrombie


Q&A with Flare: An ambulance app for Nairobi »
Marc Ambasna-Jones

Marc Ambasna-Jones is a UK-based freelance writer and media consultant and has been writing about business and technology since 1989.

  • Mail


Do you think your smartphone is making you a workaholic?