Computer Vision News - April 2023

40 Computational Medicine group, Usher, Edinburgh is a further aspect, using interviews with hospital clinicians and IT managers to understand the enablers and barriers to adoption and with patients to understand their perspectives. “ This type of work is pioneering because we carry out a holistic interpretation of the problem, ” Miguel shares. “ It’s not only the computer vision task; it’s everything around it. Providing a paper reporting on a computer vision task’s accuracy is very different from demonstrating that it’s improving patients’ lives. ” Finally, with such a comprehensive portfolio of work, is there one thing Miguel is most proud of? “ The project we’re getting off the ground on retinal image data curation at scale is really innovative, ” he responds. “ It’s going to make a real impact on the field! ” system that detects lung nodules on chest CTs , which has already been rolled out in several hospitals across the UK and other countries. “ This is the final piece of the puzzle – you develop the models, put them into practice, and then demonstrate that they’re ready, ” Miguel asserts. “ We’re doing a clinical study to understand whether, with the help of these tools, radiologists make more consistent decisions in line with the guidelines they receive. ” Evaluating computer vision systems in the real world is a multifaceted process. One part of the INPACT study is the clinical evaluation, which checks how accurately radiologists interpret an image against the ground truth with or without AI. Another aspect is health economics, demonstrating that the solution can be cost-effective for the health system. Qualitative research