Computer Vision News - June 2020

2 Summary The Lab of the Month 2 Ester Bonmati is a senior research fellow at the Wellcome / EPSRC Centre for Interventional and Surgical Sciences (WEISS). As part of our visit to the WEISS laboratory at University College London ( see part 1 here) , we speak to Ester about her work on computer-assisted endoscopic ultrasound. Thanks again to awesome Su-Lin Lee for organizing this fascinating visit! Ester’s journey in this area began when her supervisor Dean Barratt received a grant to develop an image-guided system for endoscopic ultrasound interventions in cases of pancreatic cancer . A computer scientist by training, she was employed as a research associate on the project. “Our group has extensive experience on ultrasound,” she tells us. “Several groups around the world have tried to implement guidance systems like this, but it is really challenging. The endoscope goes through the mouth all the way to the stomach, so obviously you can’t see where it is, which is difficult for clinicians.” In spite of the challenges, the team managed to create a demo system using electromagnetically tracked sensors to show the shape and position of the endoscope. In the next two years, their aim is to bring the technology closer to the endoscopy room and to have clinicians, either via guidance or training, performing the endoscopic ultrasound interventions themselves. At WEISS, engineers and clinicians work closely together, interacting and sharing expertise. “I believe the amount of time I have spent in theatre is really valuable,” Ester tells us. “ You can’t see how to help the clinicians until you work with them! The initial ideas for this project were completely different from the ones that we finally developed. We want to create a technology that allows clinicians to perform the procedure as they normally do, but with extra help from software. In this case, it’s guidance, but it could also be decision support.” Thinking more generally about future progress in her field, Ester says deep learning is now widely used in clinical applications and solving many problems, but points to surgical data science as a potential game changer. Surgical data science is an