As the applications for AI in the medical field become more widespread, the healthcare industry is on the verge of a revolutionary transformation that promises both important opportunities as well as ethical challenges. Concern about the speed and unpredictability of change has led to urgent calls for regulation, particularly regarding the preservation of patient privacy and the incorporation of data equity within new computer models. Perhaps surprisingly, however, in certain circumstances AI may offer opportunities for the elimination of both privacy concerns and human bias from the medical domain.

The transforming role of AI in healthcare 

As healthcare professionals, we routinely measure a wide range of biomarkers, including blood pressure, heart rate, lipid profile, glucose and hormone levels – among others. We also perform physiologic testing of cardiac, pulmonary, cognitive, and other critical functions. In fact, the field of medicine is awash in data. 

AI models are ideally suited for processing vast amounts of new data, identifying patterns, and discerning relationships that may elude the human mind. However, the use of health information for research purposes requires informed consent from patients and data de-identification to safeguard patient confidentiality. 

When the focus shifts to the study of the human face, especially in the case of a child affected by a major congenital deformity such as a cleft lip, obtaining parental permission becomes challenging and de-identifying the data becomes impossible. This is problematic for the field of Reconstructive Craniofacial Surgery where there is a pressing need for a universal, objective, and feasible method of measuring the extent of facial normality and abnormality. If achieved, a uniform approach to measurement would greatly facilitate: (i) impartial self-assessment of clinical outcome on the part of the surgeon, (ii) meaningful discussion with patient or parent, (iii) outcome comparison between different surgical techniques and surgeons, (iv) enhanced surgical planning and education, (v) research on factors associated with cleft severity, and (vi) explicit characterisation of the clinical needs and benefits of surgery for third-party payers. 

To build such a new facial measurement system, I partnered over the past several years with the Computer Engineering team at Texas A&M University – Qatar under the leadership of Dr. Erchin Serpedin. To overcome the patient privacy issue, we developed an innovative AI model called CleftGAN that can generate an almost endless supply of fake (but highly realistic) children’s faces of different ethnicities depicting various types of cleft lip deformity. 

AI’s continued support across the healthcare sector

To avoid the risk of introducing bias when trying to establish a population ‘norm’ to compare an individual’s face, we developed a novel AI method whereby each patient is compared only to his or her own computer-normalised counterpart. Together, these new concepts are being developed into a new smartphone application that we hope will come to represent a new, universal standard with which facial form and facial surgical intervention can be measured.

While human judgement can be influenced by factors such as fatigue, distraction, bias and other types of cognitive heuristics, properly designed AI systems largely alleviate these concerns. Likewise, the ability of computer models to process vast amounts of data in ways well beyond the capacity of the human mind suggests that AI will soon penetrate all areas of medicine. 

It is my belief that with careful and continuing attention to ethical considerations, AI will bring great, and even unanticipated, benefits to patients and their providers in the years to come. 

Written by Dr. Mitchell Stotland (MD, MS, FRCSE) Vice Chair, Department of Surgery and Division Chief, Plastic, Craniofacial, and Hand Surgery at Sidra Medicine (member, Qatar Foundation).