Annotation-efficient learning of surgical instrument activity in neurosurgery

Philipp M, Alperovich A, Lisogorov A, Gutt-Will M, Mathis A, Saur S, Raabe A, Mathis-Ullrich F (2022)

Publication Type: Journal article

Publication year: 2022

Journal

Current Directions in Biomedical Engineering De Gruyter

Book Volume: 8

Pages Range: 30-33

Journal Issue: 1

DOI: 10.1515/cdbme-2022-0008

Abstract

Machine learning-based solutions rely heavily on the quality and quantity of the training data. In the medical domain, the main challenge is to acquire rich and diverse annotated datasets for training. We propose to decrease the annotation efforts and further diversify the dataset by introducing an annotation-efficient learning workflow. Instead of costly pixel-level annotation, we require only image-level labels as the remainder is covered by simulation. Thus, we obtain a large-scale dataset with realistic images and accurate ground truth annotations. We use this dataset for the instrument localization activity task together with a studentteacher approach. We demonstrate the benefits of our workflow compared to state-of-the-art methods in instrument localization that are trained only on clinical datasets, which are fully annotated by human experts.

Authors with CRIS profile

Franziska Mathis-Ullrich

Involved external institutions

Karlsruhe Institute of Technology (KIT) DE Germany (DE) Carl Zeiss AG DE Germany (DE) Inselspital, Universitätsspital Bern CH Switzerland (CH)

How to cite

APA:

Philipp, M., Alperovich, A., Lisogorov, A., Gutt-Will, M., Mathis, A., Saur, S.,... Mathis-Ullrich, F. (2022). Annotation-efficient learning of surgical instrument activity in neurosurgery. Current Directions in Biomedical Engineering, 8(1), 30-33. https://doi.org/10.1515/cdbme-2022-0008

MLA:

Philipp, Markus, et al. "Annotation-efficient learning of surgical instrument activity in neurosurgery." Current Directions in Biomedical Engineering 8.1 (2022): 30-33.

BibTeX: Download