The cancer tech revolution: ‘we’re not far from a diagnosis being no big deal’
At the Institute of Cancer Research in South Kensington there are many labs. Most are exactly what you’d imagine — scientists in white coats doing clever things with microscopes and petri dishes. But one is full of large, super-powerful computers, which are working round the clock to generate masses of data using AI. This is the Centre for Evolution and Cancer, where Professor Trevor Graham leads teams of computational biologists who are using AI modelling to track how cells change over time and predict how close a person is to getting cancer.
“When I was doing my PhD, it took us weeks to analyse one tiny stretch of DNA — these computers can do that in milliseconds,” he explains. At the moment Graham and his team are looking at cells from patients who have Inflammatory Bowel Disease. “We know that people with IBD have a higher risk of bowel cancer and the NHS will give them regular endoscopies but obviously no one likes having a camera stuck up their bum a lot,” he explains. “So instead we can now use AI to analyse their cells and get more information out of a sample, and start predicting patterns of whether and when a tumour is likely to develop. AI like this is hugely exciting because it’s much more efficient which can mean we get in there earlier and can save more lives. We know that if you catch bowel cancer at an early stage there’s a 97% chance of survival, but it drops to 7% later down the line.”
Dr Sam Godfrey, the Research Information Team Lead at Cancer Research UK, says we’re currently in a “golden age” of cancer research. Part of his job is picking which groundbreaking research the charity will fund. “Thanks to technology we’re understanding things we never knew before and there are some really quirky new tools and cutting-edge methods being trialled,” he says.
One of those quirky tools might be the iKnife. Developed at Imperial College, London, this surgical scalpel can “smell tumours” and diagnose womb cancer within minutes. “The iKnife uses electrical currents to differentiate between cancerous and healthy tissue by analysing the smoke that is emitted when the biopsy tissue is vaporised,” explains Athena Lamnisos, the chief executive of the Eve Appeal, the cancer charity which funded the research. “This tool has the potential to create a step change in faster diagnosis, and for the 90% of women with postmenopausal bleeding that isn’t cancer, it’s a really effective way to put their minds at ease. We know how important this is for patients.”
Imperial are also trialling a breath test for lung and upper respiratory system cancer, which is so simple it could even be done in the GP’s surgery. “We know that cancer cells leak molecules, a bit like an engine leaks on a crap car,” explains Dr Samuel Godfrey. “So a variety of different molecules detected in the breath might indicate you have cancer. The patient would breathe into a plastic bag type instrument like a breathalyser and you could get results really quickly.”
Dr Godfrey says there are amazing advances at every stage of cancer care, from diagnosis to remission. “We have microscopes such as Cryo-EM which are so powerful they can take a slice of a cell that’s the equivalent of a cross-section of the human hair and enlarge it to the size of the London Eye,” he says. “There’s work being done on cancer vaccines which can teach the immune system to recognise mutated cells and attack them. There are also some amazing experimental new drugs which act like Sellotape for a cell’s self-destruct switch, meaning that the cancerous selves can just blast themselves. Tasks that would take expert scientists months and months can now be done in minutes by AI.”
AI-based computer programs have been used to help doctors interpret mammograms and other scans for a while, but research in this area is quickly evolving. AI is now being taught not just what to look for in a scan, but how to study the eye movements of a radiologist to mimic exactly what they do to make decisions. At the moment two separate radiographers are required to look at every scan, but trials are being done to see if AI could be used in parallel with a human radiographer. And AI algorithms are even being used to see if cancer can be predicted from someone’s shopping habits.
“The AI is being trained to look for patterns in behaviour which might be early indicators for cancer, so for example how often they visited the GP, what they bought from the pharmacy,” says Dr Samuel Godfrey. “Maybe people who go on to develop esophageal cancer bought more antacids for example. There’s a huge amount of data and we’re seeing patterns which no one predicted.”
There are so many glimmers of hope in every area, we’re not far from being at the point where getting a cancer diagnosis is no big deal.
And rather than just blasting a tumour with a one-size-fits-all drug and hoping for the best, at Queen Mary University, east London they’re using machine learning to create bespoke treatments. The Drug Ranking Using Machine Learning (DRUML) method tested thousands of liver cancer cells and their response to 412 different drugs. “Patients who are diagnosed with primary liver cancer often have a very poor prognosis,” says Professor Pedro Cutillas, researcher at Queen Mary. “Cancers of the bile duct, in particular, exhibit great variation from patient to patient and a one-size-fits-all approach to treatment is not the most effective way to reduce cancer cell growth. But by utilising this new technology we can predict individual patient responses to therapies and prescribe the most effective treatment plan.”
And soon drugs may not be needed at all, if you can get into the cell and stop the cancer forming in the first place. That’s the kind of work being done at Crispr, which uses revolutionary gene editing technology to change the code of living cells. Think of it as a pair of tiny molecular scissors which can figure out which gene might be causing the cancer, snip it out and replace it with another gene. Dr Edward Stadtmauer led the first trial to test a CRISPR-made cancer therapy at the University of Pennsylvania, in which patients’ own immune cells were genetically modified to better “see” and kill their cancer. “This [trial] was really a proof-of-principle, feasibility, and safety thing that now opens up the whole world of CRISPR editing and other techniques of editing to hopefully make the next generation of therapies,” says Dr. Stadtmauer.
Back at the ICR, Dr James O’Connor is staring at a fuzzy colourful shape on a screen. It is a tumour in a patient’s neck and it changes colour as they breathe in pure oxygen. O’Connor is Professor of Quantitative Biomedical Imaging at the ICR and is working to develop a new type of scan which could help doctors deliver radiotherapy more efficiently. The machine is called the MR-Linac and it can measure oxygen levels in tumours.
“Most cancers are still detected using MRI or CT scans and all we can see is whether there’s a tumour and how big it is,” he explains. “But the MR-Linac can show us how the tumour functions, what its metabolism is like, how much oxygen it has, how it’s responding to treatment in real time. We can measure the whole tumour spatially in 3D and colour map it and find its weak point which opens up a whole new load of treatment options. We know that if we can get great images of a tumour and track it across time there’s a proven survival benefit.” At the moment there are only two MR-Linac machines in the UK, in London and Manchester, but more will be rolled out soon.
Although debate about the ethics and limitations of AI continues, this field is moving incredibly fast and new applications of technology are constantly popping up. “There are so many glimmers of hope in every area,” says Dr Godfrey. “We’re not far from being at the point where getting a cancer diagnosis is no big deal.”