Artificial Intelligence in healthcare, an intelligent way forward?
We may be a victim of our own success with regards to healthcare. Improved scientific understanding and ever-diligent, we are extremely proficient in the treatment and management of acute conditions, including infectious diseases. However, whilst this success has increased life expectancy, the perfect storm of an ageing population, sedentary lifestyles and a diet of processed foods resulted in increased incidents of chronic health conditions, including heart disease and cancer. Unlike acute conditions, these often require care and management.
This takes place in the midst of severe financial constraints in the NHS, in the grip of staff shortages, with as many as one in 12 positions currently unfilled. Technology has played a part in significant productivity gains in other disciplines with the right expertise; it should be an integral part of the solution in the future of the NHS.
Early in the decade, I had the privilege of working with a wonderful mathematician on the use of artificial neural networks algorithms to automatically qualify the suitability of candidate protein crystals for structural analysis. The computational technique is now more commonly known as “deep learning”, a key underpinning of artificial intelligence (AI) systems. Similar to biological neural systems, a deep learning system is able to learn from past ‘experiences’, or training data, to make the appropriate abstraction and recognition when presented with a new set of information. Unlike humans, computer algorithms are more scalable. Since then, AI has matured and is proving itself to be a practical tool in clinical medicine.
The delivery of modern healthcare is intrinsically entwined with technology; an especially close partnership between clinicians and technology can be found in the field of diagnostics. The diagnosis can be tricky as patients often present a plethora of symptoms; experience is key, whereby the clinicians rely heavily on their experience of past presentations of similar sets of symptoms. The difficulty is that the numbers of such experienced clinicians do not match the demand. Massive scale cloning of suitable medical professionals, including their memory banks of experiences, is still not feasible in today’s world to match the real demand.
The NHS spends £2.2 billion a year on pathology diagnostics alone, excluding spends on inappropriate treatments; a solution which can reduce this figure would support a system under considerable strain. Indeed, within the past year, AI systems have demonstrated effectiveness at the diagnosis of breast, colorectal, lung and skin cancers. Separately, researchers at the John Radcliffe Hospital in Oxford have developed an AI system which has shown to be far superior to cardiologists at diagnosing heart disease, the misdiagnosis of which alone is estimated to cost the NHS £600 million a year, half of which could be saved by the AI system.
In addition to financial savings, AI systems could also relieve time pressures on clinicians, so that they can devote more actual patient care. A positive attitude from clinical professionals is known to affect patient outcomes; indeed, it is standard clinical practice to present even dire results in a positive manner so that patients can be in a resilient frame of mind. By the same token, if clinicians are stretched and stressed, the prognosis of their patients could be impacted. Within a system which is clearly stretched and under constraint, artificial intelligence could be just the medicine that the NHS needs. I have no doubt that AI technology will permeate into healthcare within the next five years. Could a solution borne out of necessity allow the beloved NHS become a pioneer of AI in healthcare?
Dr Wendy Ng, CISSP, CCNP; 15th May 2018