Keywords: Anatomy Segmentation, Real-Time, Surgical Computer Vision, Knowledge Distillation
Abstract: Automatic segmentation of anatomical structures in laparoscopic images or videos is an
important prerequisite for visual assistance tools which are designed to increase efficiency
and safety during an intervention. In order to be used in a realistic clinical scenario,
both high accuracy and real-time capability are required. Current deep learning networks
for anatomy segmentation show high accuracy, but are not suitable for real-time clinical
application due to their large size. As smaller, real-time capable deep learning networks
show lower segmentation performance, we propose a multi-teacher knowledge distillation
approach. We leverage the knowledge of multiple anatomy-specific, high-accuracy teacher
networks to improve the segmentation performance of a single and efficient student network
capable of segmenting multiple anatomies simultaneously. To do so, we minimize the
Kullback-Leibler divergence between the normalized anatomy-specific teacher logits and
the respective normalized logits of the student. We conduct experiments on the Dresden
Surgical Anatomy Dataset, the largest public data set of laparoscopic images to date.
Experiments demonstrate that our approach increases the overall Dice score of real-time
capable networks for anatomy segmentation from 60.5% to 64.9%.
Submission Number: 320
Loading