I am currently working on Deep Learning for Neuro-Imaging. However, I can not share any publications, projects or results yet, since this is still work in progress.
Making CAI AI-ready
During my PhD at the Interactive Graphics Systems Group at the Technical University of Darmstadt, I have been working on making CAI AI-ready. I worked on Fluoroscopy-Guided Tracking as well as on Electromagnetic tracking.
Centered around the idea of estimating the pose of instruments using x-ray imaging, the core work culminated in i3PosNet, a dataset and a Deep-Learning setup to determine the pose of surgical instruments.
- AutoSNAP: Automatically Learning Architectures for Instrument Pose Estimation
- i3PosNet: Instrument Pose Estimation from X-Ray in temporal bone surgery
When no line-of-sight is available and 5- or 6-dimensional pose estimates are required, in addition to fluoroscopy, only electromagnetic tracking remains. However, for high-precision applications, the distortion of the electromagnetic field leads to metal artifacts that significantly limit the usability of Electromagnetic Tracking. To this point, all methods addressing the compensation of such artifacts fail when facing the dynamic environment of the surgical theatre.
Our work addresses these limitations and lays the foundation for Deep Learning to be applied to this topic while recognizing the dynamic nature of the operating theatre.