The latest ground-breaking development in artificial intelligence (AI) demonstrates how it can detect eye disease as well as leading experts can.
A study by Moorfields Eye Hospital in London, University College London, and the company Google DeepMind found a machine could learn to read complex eye scans and detect more than 50 eye disorders. It was published this week in Nature Medicine .
This is just one example of how AI has the potential to transform the NHS.
“There is a revolution going on but you have to distinguish the substance from the hype,” according to Professor Tony Young the national clinical lead for innovation at NHS England.
“We can’t afford NOT to embrace the latest and greatest technology to improve quality of care and to make sure it’s cost effective,” he adds.
The Government has said it is firmly committed to developing AI projects. It believes that AI will be used to transform the prevention, early diagnosis, and treatment of diseases by 2030.
However, there are challenges surrounding training of staff in new technologies and the use of patient data.
What is AI?
AI learns about patterns from huge databases like NHS health records that can in turn be used to diagnose conditions.
It uses algorithms and software to analyse complex medical data.
“The NHS is the largest health care dataset on the planet which we can access to gain insights into patient care,” explains Prof Young.
Prime Minister Theresa May in May 2018 estimated that AI will help to prevent 22,000 cancer deaths each year by 2033. It would be done by using new technologies to cross reference people’s genetics, habits, and medical records, with national data to spot those at an early stage of cancer.
An independent review covering developments in AI, robotics, and smart phone technology is currently being carried out for the NHS in England.
It’s being led by Dr Eric Topol, the US academic, geneticist, and cardiologist. He is also editor-in-chief of Medscape. It’s due to deliver its final report at the end of 2018.
Dr Topol delivered his interim report in June 2018. He said: “We desperately need innovation in healthcare. Artificial intelligence is already in every aspect of our lives – from navigation to voice recognition – and will now be applied to healthcare, the next frontier.”
What Are the Current AI Projects?
A lot of government money and investment from health tech companies is being spent on a range of AI projects in different areas of the NHS.
In the Moorfields project AI is used to analyse retinal scans. It can recommend the correct referral decision for over 50 eye diseases with 94% accuracy, which matches world-leading eye experts. The system could help health professionals to spot serious conditions earlier.
“Collaboration is still at a very early stage. This won’t be ready to be rolled out before it’s been tested and trialled appropriately. It needs a safe assessment and review process first,” explains Prof Young.
Another project the NHS is trialling is called HeartFlow. It uses AI software to create a 3D model of coronary arteries and analyses the impact the blockages have on blood flow to rapidly diagnose patients with coronary heart disease.
Another project which is being used at Addenbroke’s Hospital in Cambridge is Microsoft’s InnerEye system. It marks up scans of prostate patients automatically and creates a 3D image. It learned to mark up the scans by training on the database of past patients’ records which had been seen and marked by senior consultants. It saves time, and potentially lives.
In another study published in 2017 in the journal Nature AI was shown to be capable of classifying skin cancer as effectively as dermatologists.
The UK company Babylon Health is behind the NHS GP at Hand service in London. It allows people to book a video GP appointment via an app on their smartphone 24 hours a day.
It’s also used AI to produce a medical chatbot. Patients symptoms and details are entered into the chatbot and it gives a diagnosis. It caused a degree of controversy in June 2018 after tests appeared to show the bot could perform in medical exams as well as human doctors.
It’s not just diagnoses that AI can help with it can be used to monitor the health of patients through apps and wearable technology.
The Changing Role of Doctors
With the rise of AI in health care there’ll be changes to the role of doctors.
Royal College of Physicians President-elect Dr Andrew Goddard says: “It is becoming clear that new technologies can support and enhance the role of doctors, for example in freeing up time to focus on the doctor-patient relationship, care planning and personalisation in addition to addressing existing challenges in medicine.”
Doctors are more than just the sum of their medical knowledge. There’s much more to being a doctor than that.
“From what I have seen, artificial intelligence undoubtedly can offer a really important breakthrough, particularly by handling and quickly interpreting huge amounts of data. The power and speed of computers can be harnessed by artificial intelligence to make big improvements to some types of NHS care,” says Andrew Foster, chief executive of Wrightington, Wigan and Leigh NHS Foundation Trust.
“However, we must never forget the fundamental importance of human care, compassion, empathy, and even the importance of a gentle, physical, human touch. For it to be welcome, health services will need to sensitively blend new technologies with old-fashioned care,” adds Foster.
Artificial Intelligence may even lead to a different breed of doctor.
“It’s almost as though a new speciality in medicine is emerging,” Prof Young speculates. “Like surgery and anaesthesia in the past, similarly with technology and AI you wonder if there’s a new medical technology specialism, someone with clinical interest and a specialist understanding of AI,” he adds.
The Challenges of AI
Like all emerging technologies AI does pose some challenges, including how best to integrate it into hospitals and other clinical settings. Staff who will be using the new systems will need training. That’s part of the remit of the Topol review managed by Health Education England. The education and training of medical students will also need to recognise AI based technologies.
“The landscape is changing so quickly. When you think that technology and science is evolving so rapidly, it’s important to work out how to prepare the workforce,” says Prof Young.
There are ethical considerations too. For example, if medics rely on AI to diagnose and something goes wrong, who’s to blame? Also overreliance on algorithm based diagnosis may have other consequences.
“We need to understand the impact on the knowledge of medical professionals. Could there be a cognitive deficit from medics should they increasingly rely on algorithms?” suggests privacy and data protection professional Ivana Bartoletti.
“Also how can we make algorithms sufficiently sensitive to the context, including the cultural one? This is very important especially when it comes to support for the elderly,” adds Bartoletti.
There are big issues surrounding data security and communication with patients, and transparency will be crucial. A deal to share 1.6 million NHS patient records with Google DeepMind as part of a trial of an app called Streams, which is an alert, diagnosis and detection system for acute kidney injury, inadvertently breached data protection law.
“The legal basis for processing patients’ data will have to be clearly identified, security around data handling will have to be paramount, and when partnerships are established between public and private sectors, clear governance will have to be in place and NHS organisations will need to be fully accountable,” suggests Bartoletti.
AI has the potential to be transformative. The medical profession needs to be adequately trained to make the best use of it. In essence, AI has the potential to help doctors help patients.
“The priority first and foremost must be patient benefits and safety. So close collaboration between industry, regulators, patients and clinicians and the same rigorous standards of testing and review that we expect from any new advancement in medicine is absolutely key,” says Dr Goddard.
Doctors aren’t just diagnostic tools. The idea is that AI will empower doctors, not be a threat to their existence.
“The possibilities are only as great as your imagination. AI is an amazing area. However, there’s a lot of hype, that’s why there needs to be testing and trialling before systems are proven to be effective. The next 10 years are going to be exciting,” predicts Prof Young.
Clinically applicable deep learning for diagnosis and referral in retinal disease. Nature Medicine (2018). Abstract .
Dermatologist-level classification of skin cancer with deep neural networks. Nature volume 542, pages 115–118 (02 February 2017). Abstract .