Artificial Intelligence in Medicine and Dentistry (With a Special Focus on Medical Imaging)

Artificial Intelligence (AI) is definitely in the spotlight at the moment, where potential breakthroughs in medical practice (and research) are constantly being touted by the media as examples of where AI can potentially lead to major breakthroughs. Here I present some ways in which AI might lead to improvements medical and dental practice and research, with a particular focus on imaging. I then give a very brief overview of what AI is and I describe typical tasks in image processing that are mapped to specific software. Recent projects from the School of Dentistry in Cardiff are presented, namely: initial attempts to apply convolutional neural networks (CNNs) to “segment” teeth and to annotate panoramic dental radiographic images; and, machine learning models of 3D facial surface shape. Results of a recent systematic review of ours showed that academic articles describing the use of AI applied to dental radiographs to predict the presence of periodontitis were generally of “intermediate to poor” quality only. Particularly low scoring items in these papers (echoing “FAIR” principles in AI) were: transparency, specifying the data source, carrying out a bias assessment, error analysis, and description of a baseline model for comparison. Some of the practical issues involved in exploiting this technology are listed (i.e., sourcing data, ethics and governance, and routes for translation into practice) and results of a recent ad hoc survey of dental staff are given. In conclusion, there probably are many gains to be made in medical and dental practice and research, but that this might be hindered by: technical and reporting issues such as a lack of transparency (etc.); integration of these methods into a complex health system; and a lack of knowledge, skills, and time in health professionals, as well as some fear of the technical aspects of AI.

Short Bio

After completing his PhD on quantum magnetism at the University of Manchester Institute for Science and Technology (UMIST), Dr. Farnell spent a year working as a research software engineer for Logica UK Ltd. and then was a post-doctoral researcher in theoretical physics. Thereafter, his primary research and teaching focus therefore moved to medicine and dentistry, although he managed to keep his work in theoretical magnetism going as a “hobby” (see, e.g., his high-order CCM computer code).  He has worked in academia departments at three hospitals in the UK previously (i.e., St. Paul’s Eye Unit at the Royal Liverpool University Hospital, the Christie Hospital Manchester, and the Dental Hospital in Cardiff). During his time as a lecturer in Liverpool, he was the principal researcher for the Automated Retinal Image Analysis (ARIA) project, including retinal imaging and the creation of the ARIA archive of retinal images. Recent work has focussed on methods of Artificial Intelligence applied to medical images and 3D biological surface shapes. He has also worked as a statistician on a number of research studies that span from the bench-side to the bedside and on into the community. He is currently a Reader of Applied Mathematics in the School of Dentistry at Cardiff University and is the chair of the Dental School Research Ethics Committee (DSREC).