Computer Vision News - June 2021

32 Fetoscopy Registration “With surgical vision, we are looking at multiple anatomies. Each imaging modality that we look at has a different anatomy,” she explains. “For example, if we are talking about upper GI, it has a totally different anatomy to colonoscopy, fetoscopy, or laparoscopy. For each of these, there is data available, which may or may not be annotated, but for which we want to apply computer-assisted intervention. Every year, EndoVis has a number of sub-challenges based on different modalities and procedures with new problems for the participants have to solve.” This year, the FetReg sub-challenge explores placental vessel segmentation and registration for mosaicking in clinical fetoscopy for the treatment of Twin-to-Twin Transfusion Syndrome (TTTS) . The team hope it will advance work to enhance surgeons’ vision by expanding the fetoscopic field of view and providing better visualization of the vessel map. Alessandro is a PhD student at the Italian Institute of Technology and Politecnico di Milano . Computer- assisted surgery and surgical data science are his main interests, and his PhD research is focused on computer vision and deep learning techniques for computer-assisted fetal surgery, so he has already done some work in this area. “My first work on TTTS was the identification of an anatomical structure that could be useful for a surgeon to orientate and navigate inside the placenta, ” he tells us. “Fetal surgery is very poorly explored. We hope with this work that we can push things forward so that in a few years we could have a complete framework for TTTS and be able to extend that to all fetal surgery practice.” The research dataset for the challenge is not currently available online, but the team will release it publicly afterwards