What to learn in instrument pose estimation

Watch the video here

Abstract: New Deep Learning Methods push the limits what computers are able to do to assist experts on a daily basis. With these developments an increasing gap is forming between those methods and the application and dataset realities, particular in Computer Assisted Interventions. In temporal bone surgery, sub-millimeter accuracy is needed to guide instruments around the ear. This talk discusses, how the gap can be closed for instrument pose estimation, which is needed so robotic control can be used in future high-precision minimally invasive surgeries. The i3PosNet method introduces an iterative scheme to achieve precision on x-rays significantly better than a pixel. Given this scheme, we develop AutoSNAP – a learning-centric method – to automatically design a Neural Architecture, which improves accuracies even further.

https://i3posnet.david-kuegler.de/

https://autosnap.david-kuegler.de/

Keywords: Deep Learning, Problem Design for Learning, Neural Architecture Search

Currently, David Kügler works as a researcher at the German Center for Neurodegenerative Diseases (DZNE) in Bonn, Germany. There he works on Deep Learning for Neuroimaging ranging from anatomical segmentation to reconstruction tasks to improve MRI acquisitions. In his PhD, David addressed the translation of AI on Computer-Assisted Interventions closing the gap between Deep Learning Method development and the reality of CAI applications and datasets.