A multidisciplinary journey from model development to clinical implementation of medical computer vision using AI
Prof. Miguele Bernabeu Llinares
University of Edinburgh | Medical InformaticsRecent advances in Computer Vision and Machine Learning promise to revolutionise medical image interpretation. In the words of Prof. Geoff Hinton, a Turing Awardee in 2018 and Nobel Laureate in 2024, it is 'quite obvious that we should stop training radiologists', radiologists are 'the coyote already over the edge of the cliff who hasn’t yet looked down'. Unsurprisingly, radiologists, amongst others, have strongly disputed such statements and current data does not support any reduction in radiology job openings. In this talk, I will argue that while Computer Vision has taken a strong foothold in the area of medical image interpretation, successful translation of recent advances to clinical practice require a sophisticated interplay of multiple disciplines. Considerations around data quality, implementation, and evaluation are as relevant as model development, typically the area making the headlines. To illustrate challenges and approaches, I will present work undertaken at The University of Edinburgh around these three domains involving large multi-disciplinary teams.
