Diagnosing prostate cancer is challenging and there is currently no single definitive test for it.
A GP will typically take urine and blood samples, and of course there’s the ‘digital rectal examination’, but the results of these can only highlight if a man is at risk. Those who are will be sent to hospital to discuss options for further tests.
In December, the UK led the way in improving diagnosis when the National Institute for Health and Care Excellence (NICE) recommended that all men at risk of prostate cancer receive an MRI scan ahead of a biopsy.
This revolutionary approach to using multiparametric MRI (mpMRI) to identify cancer was pioneered over the last decade.
Two clinical trials led by UCL and University College London Hospitals showed that the scan results in fewer important cancers being missed and fewer unnecessary invasive biopsies for men who do not have the disease.
The new clinical pathway, however, a new challenge for interpreting these complex scans, is painstaking and requires great skill.
Step forward Amiri Health, a Cambridge start-up led by artificial intelligence (AI) entrepreneur Dr Antony Rix, who has been working over the last year to solve the problem.
“There is an urgent need for something now to make the new pathway scalable,” Dr Rix tells the Cambridge Independent.
“We have developed an AI technology to support the diagnosis of prostate cancer. We are already getting better performance results than the published literature and we aim to be the best in the world.”
The company is seeking investment to complete the clinical studies needed to validate its system, and apply for a CE mark and FDA approval so it can sell the technology in Europe and the US respectively.
Beyond the healthcare benefits in prostate cancer – and chance to tap into a market estimated by Grand View Research last year to be worth $5.5billion by 2025 – the company’s technology could also prove valuable in detecting other hard-to-diagnose cancers, where there is growing evidence supporting the use of MRI.
“It’s widely recognised that if you catch the disease early, it is easier to treat with less invasive and lower-cost methods than the traditional approach of surgery, drugs or radiotherapy,” says Dr Rix.
“What is saddening is that of the prostate cancers diagnosed in England in 2016, 16 per cent of patients were found with metastatic cancer, which had already spread. A further 35 per cent were found with locally advanced disease that also required intensive treatment.
“It’s all because the disease is being found too late.”
The blood test for prostate cancer, measuring prostate-specific antigen (PSA), is not always accurate, which has prevented a national screening programme being introduced.
Once a patient is identified as being at risk, biopsies are then used to try to confirm a cancer, with needles inserted into the prostate through the skin behind the scrotum – a transperineal biopsy – or through the back passage, a transrectal biopsy.
The latter carries a greater risk of infection and in both cases it is possible for a tumour to be missed, although a transperineal biopsy can be more precise. Biopsies can also lead to unnecessary treatment if they identify small, low-risk cancers that cause anxiety for the patient.
Teaming up with University of Cambridge physicist Dr Chris Doran and Nuada Medical Prostate Experts founder Brian Lynch and director Clare Delmar, Dr Rix founded Amiri Health to address the need to improve the process of diagnosis.
“Brian and Clare have been working in the prostate cancer community for eight years to provide technology and expertise to open up access to this completely new pathway, so they have a fantastic network of clinical contacts and customers both across the UK and internationally,” he says.
“The work they have supported is really exciting. It dates back to the 2000s when a new MRI technique, called diffusion weighted imaging, was introduced.
“Effectively, this new way of imaging adds another two colours. Over the 10 years or so since, clinical evidence has shown you can use diffusion weighted imaging to diagnose disease, non-invasively, with higher accuracy compared to transrectal biopsy.
“The trials show that MRI can also lead to fewer biopsies with healthy patients, which can cause harm and are expensive in their own right.”
While patients are already benefiting from this new MRI scan, there is much more to do.
“The challenge is that although the new images give us a lot more information, they are still not easy to read,” explains Dr Rix.
“You don’t see a big red light that says ‘This is cancer’. You see fuzzy signals, with poor signal-to-noise ratio. There are often distortions because of the way scanning systems work. So consultant radiologists who have to read these images are worried about making mistakes and generally err on the conservative side.
“As a consequence the biopsy rates are still higher than we would like. So we have a system that works in theory in the context of a good teaching hospital but is challenging to scale out across a whole healthcare system like the NHS, as NICE now wants.”
Amiri Health’s plan is to use its proprietary computer vision-like technology to help radiologists interpret the scans.
“What we want to do is distil the experience of the elite radiologists who pioneered this new method and build on the very high-quality data we have,” says Dr Rix.
“We already have one of the largest datasets in the world and are looking to partner with clinical institutions to provide further validation data, and put this into a system that will support radiologists to answer simple but critical questions about whether a patient has clinically significant cancer. If they don’t, then they can be sent home straight away with their mind put at rest. If they do, the MRI shows where to biopsy, as it is usually a biopsy that gives you the positive diagnosis.
“We’ve developed a proof-of- concept version of our platform. The performance results we are getting so far are tremendous.
“The concept of an artificial intelligence system that encapsulates this clinical expertise, and can predict whether a patient has cancer or not, is working.
“We are currently not missing any patients with significant disease in more than 100 unseen cases kept back for testing. We need more data to develop this further but that’s really promising.”
MRI images offer radiologists a number of views in 3D of different properties of the tissue.
“Typically a radiologist looking for prostate cancer looks at four of these,” says Dr Rix. “One of them also has a time axis, which makes it more complicated.
“The first view is called T2 and gives a view of the hydrogen density of that tissue. Water shows up brightly, while muscle tissue shows up dark.
“The second view is called contrast and this is where a patient is injected with an agent that shows strongly and flows through the blood system. By making repeated views over time, you can see how this is picked up.
“Tumours tend to absorb the contrast agent more quickly than healthy tissue. One of the things we’re interested in clinically is whether we can reduce the use of this contrast agent, based on the rare earth metal gadolinium, as some people may develop a reaction to it and there are concerns that it can build up in the skin or brain. The other two views look at different magnetic properties of tissue. They give us an indication of how easily water diffuses through tissue.”
The rich data has enabled Dr Rix and his colleagues to train the AI system to identify abnormalities.
“We had to go some way beyond the standard techniques,” says Dr Rix, who wrote much of the code himself. “Computer vision is now a relatively mature market if you are trying to answer consumer-style questions like ‘What breed of dog is in this image?’ or ‘Who is the person whose face I can see?’ But those AI algorithms are really not designed for the clinical questions we face so we’ve had to develop our own system.”
This new type of AI and the training system to teach it are expected to form much of the intellectual property and value in Amiri Health, which has identified three areas where it could apply for patents.
As well as identifying abnormal areas for a possible biopsy, the system outputs a probability score, which provides an estimate of how likely it is that patient would have a clinically significant cancer.
“Depending on a patient’s history, other factors and what country you are in, different thresholds might be set,” notes Dr Rix.
Some abnormalities do not actually spread and grow, and therefore do not represent a risk to patients. By helping work out who needs a biopsy, and where to target, the company wants to support clinical teams to make the best decision for each patient.
Amiri Health would deploy its software in the cloud, and most NHS hospitals already have the infrastructure needed.
“When a patient goes into the scanner, their data will be automatically uploaded to a secure location where the AI system will process it and return the information the radiologist needs. They then use their clinical judgment,” says Dr Rix.
While the UK is leading the way in the use of MRI in prostate cancer diagnosis, other European countries are beginning to follow suit, and the US is not far behind, opening up a substantial market for the company.
Investment is now key for Amiri Health, which has bases at Barclays Eagle Labs in Chesterton Road and on the Cambridge Biomedical Campus, and is looking to add staff.
Longer-term, once it has developed its prostate cancer technology, it would aim to explore other areas.
“As we build clinical data and partnerships, we would like to be able to use the same technology – both the AI and MRI – to screen for other types of cancer as well,” says Dr Rix. “It opens a much larger market and it means we can help a lot more people.”
Two decades of expertise in artificial intelligence
“The genesis of this is 20 years ago when I co-founded my first AI start-up, measuring the quality of telephone calls, which was a big deal if you were BT or a mobile operator,” says Dr Antony Rix.
Psytechnics beat the likes of Ericsson and Deutsche Telecom in an international competition in 2001, and was acquired by one of its customers, Netscout, in 2011. The technology is still in use today.
“We were in some ways ahead of our time and the climate in the early 2000s was quite difficult, but we successfully built the company up and won international revenue. For me, it was a fantastic route into AI,” said Dr Rix, who had studied at Cambridge with world-leading AI researchers Peter Rayner, Bill Fitzgerald, Steve Young and Tony Robinson. He gained his PhD from the work he did at Psytechnics.
Dr Rix then spent 12 years consulting for Melbourn-based TTP, learning much about software and medical device development, before rejoining the world of start-ups by co-founding 8Power, which is creating energy harvesting technology for sensors.
After leaving 8Power in early 2018, Dr Rix formed AI consultancy Granta Innovation.
He linked up with University of Cambridge physicist Dr Chris Doran, a fellow and director of studies at Sidney Sussex College, who had previously co-founded computer game graphics technology company Geomerics, which was acquired by Arm.
Together with Brian Lynch and Clare Delmar, who supported clinicians in and around London to develop the new method of prostate MRI scanning, they formed Amiri Health.
Prof Anwar Padhani, professor of cancer imaging, Institute of Cancer Research, London, said: “The universal roll out of prostate MRI for diagnosis is imminent, representing a large implementation challenge. The delivery of the MRI pathway benefits requires high quality of the entire diagnostic process with multidisciplinary working. Success will require artificial intelligence solutions for diagnosis and biopsy guidance, and the integration with clinical biomarkers. The Amiri prostate AI development is world leading in performance, and its development plan is pertinent for delivering real world patient benefits.”
The company has already earned acclaim, as one of three winners of the CW Discovery Startups competition in 2018.
The name of the company is a play on MRI and AI.