The rapid pace of innovation has made us both accept and expect continuous incremental improvements in our daily life, ranging from better appliances to lighter and more advanced phones. Progressive automation has also been welcomed, having liberated us from repetitive and dangerous tasks. However, up to now, we have kept a tight leash on our machines and have not really allowed them to encroach on environments where higher order decision-making is necessary.
The ascendance of artificial intelligence (AI) is now set to challenge this paradigm, and we are now approaching the technical and regulatory thresholds which could permit the application of autonomous intelligence to new applications such as driving. As exciting as this next chapter in our civilisation's progress may be, science fiction has long warned us of the perils of losing control to the machines. These mixed emotions of both excitement and fear are also being felt by the medical community.
As a field of science, healthcare has been slow to adopt digitalisation, lagging behind other industries. Cautious regulatory resistance has long been applied to all innovations entering clinical care, based on the painful lessons of the past, with Thalidomide being a frequently quoted example. Other industries have not been so cautious. Premature deployment of AI in the US judicial system executed racial based bias, as these algorithms had not been trained on a broad enough population. Of course, delaying the deployment of these technologies does also have its potential pitfalls. These innovations can enhance our clinical care, increase access to communities where expertise is limited, and deliver much-needed cost benefits in taxpayer funded systems.
As a cancer surgeon, I too saw both the promise of these innovations to better medical care, as well as the challenges of applying them safely. Having long had an interest in digital technologies, half a decade ago, I embarked on my doctoral research at the Centre for Precision Surgery, at University College Dublin in Ireland, seeking to better the diagnosis and treatment of cancer through the application of novel technologies.
My starting point was to use commercially available surgical cameras, and special dyes injected into the patient's veins. Using this optical equipment, it is already possible for surgeons to visualise blood flow as it traverses the body and display this on a screen as video. However, these rich recordings offer far more information than a surgeon can visually interpret. Here lied an opportunity to use computers to analyse this data in greater depth and capture more detail than a human possibly could. Now being able to discern these detailed fluctuations in blood flow, I sought to better understand how this related both to the development of cancer and, it's impact on healing following operations.
Looking at colon cancer first, I analysed recordings of normal and cancerous bowel, seeking to mathematical differentiate their blood flow patterns. Subsequently equipped with this understanding, I explored AI-based differentiation to develop a video-based diagnosis of colon cancer. This 'digital biopsy' portends to offer a new paradigm in care where the diagnosis of disease no longer requires doctors to remove pieces of tissue for laboratory analysis, promising to expedite the patient journey from diagnosis to surgical removal of these tumours.
Following that, I directed my attention to the healing phase which follows surgery. This is also dependent on a good blood supply as where this is lacking, poor healing can result in leaks from the bowel, with subsequent illness. Again, by deploying computational analysis, I was able to develop a mathematics-based aid, which provides surgeons with a recommendation on the ideal location to cut during the operation to give the patient with the best chance of healing from the procedure.
Armed with a better appreciation of the underpinning science and the available technologies I started to think about other parts of the body where a deeper understanding of wound healing might benefit cancer care. I turned my attention to breast cancer where, the reconstruction of the operated breast requires filling defects using synthetic materials (e.g. implants) or tissue from other parts of the body. The healing process following these operations, is also dependent on a good blood supply. Here, I was able to develop artificial intelligence-based decision support, where software could indeed take on-table surgical decisions on whether proceeding with surgery was safe (based on a computational prediction from blood flow patterns). Where the computer predicted poor healing, the recommendation would be to change the operative course.
What happens next? This is now where the regulatory caution comes into play. With all new ideas or applications of technology in healthcare we need to demonstrate robust scientific evidence and AI should not be gifted any exemptions to this. This research generated a body of literature for scientific scrutiny and dissemination. One component of this work has progressed to CLASSICA, an international European clinical trial where these tools are being tested on patients.
However, there are also pending ethical and legal questions on who is to take responsibility for decisions made by software, especially in a case like this where no surgeon can truly mentally replicate this level of computational decision-making and thus needs to have full faith in the machine. Of course, my work has not developed in isolation and other comparable tools in later stages of development are indeed now entering clinical availability.
One field of medicine where progress has been more rapid is the interpretation of x-rays, where software has matched and surpassed doctors. While this novel digital workstream is currently invisible to the patient, there will be a point where these tools will claim space from the doctor-patient interaction, thus diminishing contact with the physician. Having participated in discussions with patient groups, the underlying message remains that their trust to do what is right for them lies in the doctor's hands. While it appears almost inevitable that machines will supersede clinicians' technical abilities to diagnose and treat disease, we need to ensure that novel technological care pathways do not create new gaps in care, by failing to provide sufficient access to holistic emotional support, empathy and compassion.
Dr Jeffrey Dalli is a Consultant Oncoplastic (Cancer and reconstructive) Breast Surgeon at the Wrightington, Wigan and Leigh NHS Foundation Trust, UK, where he is also clinical lead and divisional research lead. Dr Dalli is a recipient of the Tertiary Education Scholarship Scheme (TESS), Malta. This scholarship provided funds for his Doctoral research at the University College, Dublin, Ireland. This research has also been supported by the Disruptive Technologies Innovation Fund, Ireland.