Use of AI in the NHS is leading to a 30-times improvement in early cancer detection rates
Cancer may be the world’s second largest killer, but there is lots of innovative work going on in the realm of AI software to improve early diagnosis and, therefore, boost survival rates. Two companies doing just that are UK-based C the Signs and Panakeia Technologies.
Cancer, which is the world's second biggest killer behind heart disease, leads to the demise of about 9.6 million people each year. This means it is responsible for about one in six deaths around the globe, 70% of which occur in low- and middle-income countries.
Within this context, the importance of early diagnosis cannot be overemphasized, points out Bea Bakshi, a UK-based general practitioner (GP) and founder of C the Signs, a start-up that has created an artificial intelligence-based tool to help identify at-risk patients and is a member of the TechNation network for tech entrepreneurs. She explains:
In the US and UK, about half of people are diagnosed when they're in the later stages of cancer. But early diagnosis would mean that 80% survived for 10 years or more, whereas late diagnosis means that 80% die within five years instead. So the impact of early diagnosis on survival rates is monumental.
The problem is there are more than 200 types of cancer that can present anywhere in a patient's body at any time. To make matters worse, they often either mimic benign diseases or generate only vague symptoms that could be linked to a range of different conditions.
Another challenge is the fact that GPs are trained as generalists rather than specialists, which means they do not always recognise the signs. In the US, for example, Bakshi says, research indicates that patients need to visit their doctors on average five times before they are placed in the at-risk category.
A further issue is that many GPs do not have access to the information they require to enable them to refer patients to the most appropriate consultant or to obtain the most effective test, particularly if symptoms remain ambiguous - hence Bakshi's decision to build a tool to help.
The idea came to her a number of years ago following an encounter with a man who had come to the accident and emergency ward in which she was working. He had late stage pancreatic cancer that had been misdiagnosed and ended up dying three weeks later. Bakshi explains:
He didn't ask why he'd got cancer but why he'd been diagnosed so late. He was quite young - and his words stayed with me.
First step towards early cancer diagnosis
The C the Signs AI-based platform, which is already fully integrated with the UK's electronic patient record system (EPRS), has come about due to the impact of those words. The platform uses information stored in the EPRS together with the structured and unstructured data, which includes academic research and information from biobanks, held in its own database, to predict from what kind of cancer a patient is at most risk.
It then recommends the swiftest and most cost-effective action doctors can take to confirm this initial diagnosis. A dashboard also makes it possible to track a given patient's treatment journey in order to understand the impact of their actions and investment. Bakshi says:
We see ourselves as being the first step in the pipeline to early cancer diagnosis. If someone presents who could be at risk, the doctor is prompted by the system to ask if they have x, y or z symptoms, and is then shown recommendations for action based on risk. End-to-end, the process takes about 35 seconds, so it's about speeding up decision-making. But it also makes it easier and safer to take a decision. The system is 100% evidence-based but the processes involved are very automated, which means doctors need to do fewer things than they would normally, making it the path of least resistance.
Another benefit of the platform is that it enables health authorities to save significant amounts of money. As Bakshi points out:
The cost of treating cancer is rising hugely and if about half of patients are being diagnosed in the late stages, much of this cost is going on the chemotherapy and drug end of things. But if people are diagnosed early, tumours are generally localised and so more amenable to surgery, which is much cheaper overall.
The annual subscription-based system has already been adopted by about 20% of NHS Trusts, following a year-long pilot that started in mid-2017, which demonstrated a 30-times improvement in early cancer detection rates. Adoption has also been given a boost following a recommendation by NHS England and NHS Improvement in a national policy paper - as well as the arrival of the pandemic. Bakshi explains:
More recently, the biggest accelerator to growth has been Covid. Cancer referrals in the UK have dropped by 75%, which means about 30,000 fewer people are being referred each week. So there's been in surge in interest in using the system for remote risk assessment.
Saving both money and lives
But another market that Bakshi expects to move into as soon as she can secure a pilot site for a six-to-12 month trial is the US. She explains the rationale:
The US has a disproportionately higher spend on cancer compared with the rest of the world, so the ability to diagnose early could save billions for insurance companies but also have a high impact on patient outcomes. We're currently speaking to some insurance companies, and colleagues to broker discussions, but we believe our impact could be even more significant there than in the UK.
Another TechNation member, whose software uses AI-based algorithms to help pathologists analyse images of tissue samples for biopsy, is Panakeia Technologies. The company, which was set up in late 2018, is officially due to launch next year.
Currently in the pre-clinical phase of deployment, it is partnering with a range of universities and hospitals in Scotland, Wales and the north of England to undertake software testing. Co-founder Pahini Pandya, who decided to get involved in the field after suffering a cancer scare herself, describes how the prototype system works:
We use machine learning to analyse patient data and provide the same information as lab-based tests. So it eliminates the need for lab-based tests and makes the whole process quicker and cheaper.
Patient turnaround can drop to as little as 24 hours rather than the more common couple of days or weeks, leading to cost savings of up to 85% through a mixture of workflow efficiency and freeing up specialists' time, she claims. As for the company itself, it is named after the Greek goddess of medicines and cures. Pandya explains:
Panakeia is the Greek goddess from which the word ‘panacea' derives, and it means ‘universal'. So the name seemed appropriate as our ultimate goal is to help bring precision medicine to where patients need it most.
In an AI world that tends to be very male-dominated, it is refreshing to see women employing their skills and expertise to work with technology and use it for social good.