Computer Vision News - January 2020

3 Summary Endoscopic Vision 19 This year in the workflow challenge we had data sets from three different hospitals, and I think the field has to evolve in that direction. Also, federated learning is important when you deal with medical data. Hospitals want to resolve security and privacy issues and having a distributed learning environment could be a way to offer that.” Endoscopic vision’s role in surgical procedures currently being Karl Storz that provides the signals together with the endoscopic images for them to analyze. During an operation, she explains that a surgeon needs to have a recommendation based onmeaningful analysisbythecomputationalmethods. Human-machine interaction plays an important role, as no assistance system will be used if it’s not well integrated into the surgical workflow. p e r f o r m e d today – such as laparoscopy, c o l o n o s c o p y and colorectal surgery – includes e.g. w o r k f l o w analysis, instrument segmentation, 3D reconstruction, and classification of lesions. Stefanie’s group is currently working on computer and robotic-assisted surgery, including soft-tissue navigation in laparoscopic surgery, sensor-based training system to analyze the skill level of the surgeon as well as synthetic data generation for laparoscopic applications. They are also working on surgical workflow analysis with sensor data from the operating room (OR) – using not only endoscopic images, but also sensor signals. The team has access to data from an integrated OR by The challenge is to perform online analysis in real-time as to provide context-aware assistance in the OR. In terms of where the data processing takes place, Stefanie says that it depends on the application: “For some applications, you have to be real-time capable. For example, if you take visualization with augmented reality, latency is a factor you have to consider, and you need a high-speed network and swift processing. This could be in a local cloud.” Thinking ahead, she says they are working on 5G applications , which have the potential to be enablers for different kinds of assistance functions.