Cancer is one of the UK’s biggest killers. Could artificial intelligence put a stop to this? Yessi Bello Perez investigates.
By the time you finish reading this article, five people in the UK will have been diagnosed with cancer*. Here’s another scary thought: by 2020, 47% of people – that’s almost one in two – will get the disease at some point in their lives.
Depressingly, 356,860 new cases of cancer were diagnosed in the UK in 2014 alone, and although survival rates have doubled over the past four decades, 160,000 people still perish from the disease every year.
Tech giants such as Microsoft are already looking for ways to cure the disease and although the firm undoubtedly has the necessary resources to research a cure and potentially disrupt the diagnosis process, the question is whether technology startups are also well-placed to achieve the same objective.
AI and melanoma
Take skin cancer, for example. It’s the most commonly diagnosed cancer in Britain, with more than 100,000 new cases of non-melanoma cancer and 13,000 new melanoma diagnoses being made per annum.
One of the main challenges doctors face when trying to diagnose non-communicable diseases such as melanoma is the fact people don’t always seek help when they’re concerned, in this case, by the appearance of a skin lesion (or mole).
Why virtual reality is being used for training purposes
Referred to as ‘patient delay’, this reluctance, coupled with the time it takes for people to be seen by specialists, can lead to irreversible disease progression.
Skin Analytics, a UK HealthTech startup founded by Australian Neil Daly, hopes to combat this by enabling patients to track matters of concern through an AI-powered smartphone image recognition app.
“There has been research around patient delay and as far as I can surmise, it’s down to a mixture of not being able to interpret or recognise the early symptoms and embarrassment about seeing the doctor with few tangible symptoms in case they ‘waste the doctor’s time’,” Daly, the founder.
“Having grown up with sun safety messages, skin cancer immediately popped into my head as an area worth exploring. The more I looked into it and understood what a huge and complex problem it is, the more I realised AI had the potential to really change the way in which we identify the disease,” he added.
Meet the AdTech/MarTech startups pitching at Elevator Pitch LIVE 2017
Although not fully launched, Skin Analytics has received support from professors at Oxbridge and Imperial College and the team, consisting of post-doctorate AI researchers and machine learning experts, is seeking to validate a system which could screen photographs for risk of melanoma.
If successful, Daly’s startup could help reduce associated costs for the NHS, while ensuring skin melanoma sufferers are efficiently diagnosed. “The system is designed to triage cases, so the output is a validated decision about whether the patient needs to have a physical appointment with a dermatologist,” noted the founder.
Competing with giants
Skin Analytics may be making headway, but it’s also competing against the likes of IBM, whose researchers have developed a computer system to help with skin cancer diagnosis.
IBM’s technology is in its infant stages, but it works by analysing photographs of skin lesions and cross-referencing these against its database of skin cancer photos. Skin Analytics’ approach, Daly claimed, is different.
Meet the HealthTech startups pitching at Elevator Pitch LIVE 2017
“The difference is that we’re very focused on how to build this technology into the patient pathway and that requires evaluating the technology in a prospective clinical study. Or put another way, where the data is collected in hospitals according to a study protocol for the express intent of evaluating the system,“ he explained.
Daly went on to argue that Skin Analytics is not a research project and the company is focused on achieving better health outcomes for patients.
“We have been talking with providers from the earliest stage of the patient pathway, through to when the diagnosis is confirmed by biopsy. So we have a clear idea of how we want our technology to build on the diagnostic performance to improve outcomes,” he added.
Bristol-based Micrima is a HealthTech startup seeking to disrupt the way in which doctors screen for breast cancer – responsible for killing 11,433 women in the UK in 2014 alone.
The company raised £2.6m from a mixture of investors earlier this year and is working on a CE Mark-approved breast imaging system, known as MARIA technology.
Micrima claims its screening tech will be safer, more comfortable and more accessible for a larger proportion of the world’s female population and, with Cancer Research UK data showing breast screening uptake in the UK has declined in the past six years, there’s probably never been a better time to make the process more appealing.
The technology, originally used to detect buried landmines, spun out from Bristol University after the founders recognised the impact that it could potentially have on breast cancer detection and mortality rates.
“Breast cancer is one of the most common causes of death in women between the ages of 35 and 55 in Europe, and the leading cause of death in many countries. Early diagnosis dramatically improves survival rates, yet today’s screening based on X-rays detects only the minority of all tumours. Most tumours are being discovered by the women themselves, often at a much later stage than is ideal for the most effective treatment,” said Micrima CEO Roy Johnson.
The problem is that X-ray mammography, a widely used breast imaging technique, struggles to discriminate between cancers and dense breast tissue. In fact, breast cancer (which looks like white gland tissue on an X-ray) is easier to see on a mammogram if it’s surrounded by fatty tissue (which looks dark).
To overcome this, MARIA uses ‘harmless’ radio-waves for breast imaging, meaning the system can be used more frequently and result in more cancers being detected at an earlier stage.
MARIA’s imaging array has 60 antennas placed beneath a form-fitting ceramic bowl which is placed against the patient’s breast. Once in action, the resultant emitted radio signals are combined to produce a high-res 3D image.
“This imaging process takes less than five minutes and detects tumours as small as 4mm under laboratory condition,” explained the CEO.
MARIA is currently being trialed by three UK hospitals in an attempt to expand its clinical database but Johnson said the company will begin additional trials in Europe throughout this year.
Meanwhile, a Swiss company with $33m in funding under its belt, is also making progress in the field.
Six year old Sophia Genetics, founded by Dr Jurgi Camblong, Dr Pierre Hutter and Professor Lars Steinmetz, is using AI to spot alterations in humans’ genomic makeup to diagnose cancer and other rare disorders. Its data-driven platform, Sophia DDM, is being used by 215 hospitals and private labs across the world to diagnose 200 patients on a daily basis.
The firm’s AI, known as SOPHiA, is tasked with processing and analysing raw genomic data produced by hospitals to help physician diagnose patients better and faster, and eventually, recommend personalised care.
Camblong, CEO and co-founder, said: “Sophia Genetics was founded with the ambition to break down information silos in healthcare and pool genomic data to give clinicians relevant insights to better diagnose and treat patients.
“From the start, the goal has been to put a sustainable process in place to gather and analyse different types of data from various sources to systematically improve diagnostics and ensure the genomic information of a patient in Madrid could help better diagnose another patient in London, for instance,” Camblong added.
The process, which takes about two days, starts with the extraction of the patient’s DNA via a blood draw or biopsy. The hospital then uses molecular biology processes to prepare the samples and subsequently digitises them using a DNA sequencer. The resulting genomic data is then submitted to the company’s AI system which must then dig around to pinpoint the patient’s genomic mutations.
Camblong is confident the use of AI will eventually speed up the process of cancer diagnosis, inevitably resulting in better outcomes for patients across the globe.
“This technology is paving the way for the democratisation of data-driven medicine, and is improving both the accuracy of diagnoses and the care that patients receive. AI can help address the limits of traditional approaches to interpreting genetics information, which rely on human input, and are not suited for high-volume, routine clinical testing,” added the doctor.
It may be too early to tell whether AI, or any other form of technology, could help doctors diagnose cancer – and potentially even treat it – but it certainly seems the industry is heading in that direction. By its sheer nature, the healthcare sector is characterised by the abundance of data and it’s now up to innovators to make use of it and to do so for the greater good in hope of fighting the disease.
*Based on NHS statistics showing every two minutes someone in the UK is diagnosed with cancer, plus the assumption this article will take 10 minutes to read.