Every year around 3500 Australian men die of prostate cancer, according to the Prostate Cancer foundation of Australia. It is the most commonly diagnosed cancer in men, and by the age of 75 one in seven men will develop prostate cancer; by the age of 85, that increases to one in five, according to PCFA.
The good news is that the survival rate is relatively high and improving, research from the Australian Institute of Health and Welfare reveals: Around nine out of 10 men diagnosed with the cancer will survive five years from diagnosis. The five-year relative survival for men diagnosed with prostate cancer increased from 59 per cent in 1986 to 90 per cent in 2007, according to a 2013 report from the AIHW.
However, early detection can have a significant impact on the chances of survival. Brisbane-based startup Maxwell Plus is using artificial intelligence (AI) in an effort to increase early diagnosis of the cancer.
The business is named after physicist James Maxwell: Maxwell’s equations are fundamental for electrical engineering and helped made MRI scans possible. “It was named after him in a similar way that Tesla Motors was named after our good friend Nikola Tesla,” founder Dr Elliot Smith told Computerworld.
Before starting Maxwell Plus, Smith was an academic, doing a PhD focused on designing medical equipment. “Specifically, MRI scanners, that paired together with radiation therapy in order to target radiation therapy in real time, and change the way we do cancer treatments and make sure that radiation hits the cancer but doesn't hit healthy tissue,” he explained.
After completing his PhD, however, he decided the life of an academic wasn’t for him. “But this idea of using technology to really add this extra layer of precision to the medical field was something that really intrigued me,” he said.
“I felt that what we were doing in treatment with the MRI-guided radiation therapy was a wonderful piece of technology. But to me, I think, the key application was earlier on: Making sure that we used all the technology we could to get people diagnosed. Diagnosed accurately and early, and in a way that scales with technology.”
“I spent some time exploring that idea,” he said. In 2016 when the company was founded — then named Maxwell MRI — AI was in the midst of a renaissance, Smith said.
“Having skills in medical imaging and numerical optimisation, which at its core is what AI is, I felt there was a good marrying of those two ideas,” he said. Smith said there was an opportunity to combine his knowledge of medical imaging with AI to help “supercharge” doctors.
Early on, the focus was on MRI scans. Smith said the shift to the name ‘Maxwell Plus’ was a recognition that diagnostic processes aren’t just about capturing a single point in time.
“Images are a very powerful way to dig deep about a patient but any diagnosis is far more complex than a single image,” he said. “Today, we use AI to look at blood test data in combination with things like medical imaging, some genomics and other factors, which help us give us help to give us an overall picture about that patient’s risk, in this case, of prostate cancer.”
By the end of this year the system is expected to have processed around half a million patient cases; a tenfold increase on the 50,000 it processed in 2018.
Maxwell Plus has two portals: One for clinicians and another for men seeking prostate cancer testing.
The clinician portal synchronises data about a patient and presents a doctor with test results for approval. However, it also provides them with “a second read or a second pair of eyes over all that data,” Smith said.
“It analyses the data that we have on hand, gives them a prediction of risk and a recommendation what step to take next: What test is going to help us get the information that we need,” Smith said. “Not only do we want an accurate diagnosis, but we want to minimise unnecessary testing.”
When it comes to analysing images, it can help pick up areas that look suspicious and help determine whether a suspicious area is cancer or something else. “And, again, what the next best step is: Should we monitor this one and see them again in a year? Or should we take them for a biopsy get some final confirmation on what looks highly suspicious?” Smith said.
“It fits in with the clinician’s workflow, providing them with a deeper analysis, and really trying to get a standardised workflow enhanced by AI.”
Earlier this year Maxwell Plus launched its direct patient access portal. “We're partnering with clinicians who have existing patient bases as well as employers, to bring their employees on board to give them a portal to start getting tested for prostate cancer,” Smith said.
“It is the number one cancer in men and we know that about 75 per cent of men don't get tested. We've built the system to allow men through an app on their phones to sign up to start the testing process. They can join, get connected with specialist clinicians that use our software, and go through the testing — so order those blood tests, have those results go into the platform and reviewed by clinician — all from the convenience of their phone.”
Smith said he believes that the convenience of the process will help make men more likely to get tested, as well as make it easier for people living in rural parts of Australia to get diagnosed.
“We know that a lot of the men that have come on board this year live two to four hours away from a highly trained specialist doctor,” Smith said. “Giving them a portal to access not only a specialist, but a specialist using the latest technology, means that the outcomes and the access to medical care is now the same whether you live in a CBD or whether you live six hours out of a CBD on a large rural property.”
“If you get diagnosed at stage one, the five year survival rate is about 98 per cent. If you get diagnosed later, that drops to 26 per cent,” Smith said. “The very best thing we can do for prostate cancer patients is make sure they have an early diagnosis. And that's really what we're all about using this technology to enable large scale, early diagnosis and prostate cancer.”
Maxwell Plus claims that the system can significantly cut the time it takes a doctor to review test data.
“Our system will review the data, and clinicians now can review the output of that, check it off, and move on to the next patient really, really quickly,” he said. Smith said that potentially allows a doctor to see more patients as well as free up opportunities for longer consultations for those cases that need it, as well as help reduce the time men spend waiting for test results.
The system, including both basic infrastructure services and the AI engine, is underpinned by Google Cloud Platform, Smith said. “It's come a long way from where we started,” he explained. “Back in the early days, we ran most of our AI on a single computer that we have in our office, but we quickly outpaced the ability for that computer to do everything we needed it to do.”
“Any startup these days is evaluating cloud providers,” Smith said. “I don't think you find a lot of people running their own gear much anymore.”
He said that what appealed about Google’s cloud offering in particular was its AI infrastructure and associated services, including using Cloud Machine Learning Engine to run TensorFlow-based models.
“It was quite surprising to me, having been through academia and the old ways of spinning up your own server and provisioning everything, how quickly you can train large-scale AI models using the managed services that are out there today,” Smith said.
The company shifted to Google Cloud Platform in early 2017. Google Kubernetes Engine is used for orchestration, with Maxwell Plus able to scale up containerised workloads in response to demand. When medical data is uploaded by a doctor, it is cleaned and standardised. Images are then sent to Cloud Storage, while structured data is sent to Cloud Datastore and Cloud Spanner.
Although the key focus of Maxwell Plus has been prostate cancer, Smith said it is also working to commercialise CSIRO technology and develop a similar workflow to aid early detection of Alzheimer's and dementia.
“We were awarded a government grant to partner with CSIRO, I-MED, and Austin Health on commercialising that technology and bringing the same benefits that we're seeing in prostate cancer to what is becoming just as large a problem: The growing burden of Alzheimer's and dementia,” Smith said.
That project will employ Cloud Healthcare API, unveiled by Google in 2018. The API can be used to connect clinical systems to applications running on Google Cloud Platform. The API supports de-identification and data can be fed into services such as AutoML and Cloud ML Engine.
“While today there’s no cure available, all the research is currently saying that the best thing that we can do is have early diagnosis,” Smith said. “A lot of these methods that are coming up to treat Alzheimer's and dementia tend to inhibit its progress, so if we know it's progressing we can intervene early and stop it progressing further.”
Breast cancer and lung cancer are also potential areas where the company is looking at employing AI to augment diagnosis, he said.