Artificial intelligence technologies have permeated many aspects of the medical world, enhancing the speed of disease diagnoses while also introducing chatbots that interact with patients. Since 2017, the scale of AI technologies has grown exponentially, and they have been incorporated to modernise all of the 20th century’s top five disruptive innovations, including telemedicine and robotics, m-health, basic laboratory biology, improved clinical environments, and health informatics, such as electronic patient records.
The potential of an AI algorithm to offer pertinent advice in human decision-making is exciting. But we must be mindful not to oversimplify AI or to characterise it as it has been in much of science fiction – as confusingly futuristic or even dystopian. There is a vast and growing ecology of AI algorithms that have distinct capabilities, including classical supervised learning, unsupervised learning, machine learning and deep learning.
Each so-called species of algorithm has its own place in the digital ecosystem. For example, 90 per cent of US Food and Drug Administration-approved AI innovations are currently in the field of diagnostics, including the use of mammograms to screen for breast cancer. Beyond screening, the role of point-of-care diagnostics is also particularly amenable to these algorithms, either in the diagnosis of urinary tract infections from an automated read of urine dipsticks or even algorithms to predict heart arrythmia through smart watches.
The next generation of algorithm may be more focused on the perspective of people’s needs in healthcare – anywhere from the routine prescribing of drugs to surgical robots’ closure of wounds. There is some speculation as to whether doctors could be replaced by robots for certain procedures or aspects of care.
The increasing adoption of these and the next generation of AI-based solutions will clearly change the face of the NHS and day-to-day tasks, though is unlikely to dislodge or replace staff. There have been a multitude of technologies brought in to facilitate care, and all had the potential to threaten staff, but actually ended up working in their favour – cross-sectional imaging (such as CT scans and MRIs), those of understanding our DNA plus the technical abilities of genomics, and the more recent influence of telehealth, have not put jobs at risk, but rather have acted to reduce mortality and enhance quality of life.
The AI-led shift in care will mean that our system adopts an increasingly pre-emptiveapproach, moving away from diagnosis of disease to predicting disease long before the early molecular or behavioural steps of pathology present themselves. The next generation of AI algorithms will offer this in a new way moving forward, and if the NHS applies this wisely it can reboot the care of the nation and transform the many pathways of care that can embrace the benefits of these new technologies. The goal is to target the pathways, with empowered patients at the centre, rather than just focus on single therapeutics or technologies as add-ons to current treatment plans.
The next-generation genomic tests with machine learning analysing countless nucleotide sources in parallel could rejuvenate that approach, predicting cancers well before they occur.
Noubar Afeyan’s concept of “emergent evolution” has helped teams at the entrepreneur’s Boston-based company generate an AI algorithm called Octavia to predict the next generation of highly virulent Covid variants, to give vaccine companies and health systems a major heads-up for their boosters. Such a paradigm not only offers a safety net of information but also a route to connectivity between our health systems and ultimately our health security.
AI is also key to any strategies pertaining to sustainable healthcare. Indeed, how can we make medicine more energy-efficient? AI optimisation can help with that, identifying how we might replace classical drugs with greener alternatives, or how we might guide patients towards more natural food solutions.
The adoption of AI technologies relies on greater patient engagement. The “digital divide” issue – the difference in patient skill levels when it comes to tech – is important to overcome. A joined-up approach to digital education is vital, and this will require collaboration between government, academia and industry. Social media may offer one way of spreading public health education effectively.
Ultimately, our progress with AI-based health will require a revamp of our relationship with our health providers, and a new social contract with the future of medicine is key.
The use of these AI technologies needs to take data ethics and people’s privacy into account. Patients must be consulted, transparently, in how their personal information is used and where it is stored. If we are to derive the greatest benefits of these AI technologies in our health and care systems, then patients must be actively involved in that journey. It’s not only a case of opening the eyes of future health specialists, but also a case of opening the eyes of the whole of society.
Ara Darzi is co-director of the Institute of Global Health Innovation at Imperial College London and the Paul Hamlyn chair of surgery. He is a Consultant Surgeon at Imperial College NHS Trust and the Royal Marsden NHS Foundation Trust. Hutan Ashrafian is an honorary clinical lecturer in surgery, also based at Imperial College London.