Uncertainty-Based Visual Guidance for Interactive Medical Volume Segmentation Editing

Sheng Shi,Bowei Zhou,Yibo Song,Li Chen

DOI: https://doi.org/10.1007/978-3-030-87358-5_64

2021-01-01

Abstract:The existing automatic medical volume segmentation methods are usually not accurate enough, and manually editing the segmentation result is still very tedious. The paper proposes a novel editing framework for medical volume data segmentation by providing users simple and convenient 3D interactions under the visual guidance of segmentation uncertainty measurements. In the system, the volume data will be visualized in a combination of volume rendering and surface rendering. Since it is hard for users to obtain volume data information as context from surface rendering, we have designed uncertainty metrics based on mesh shape descriptors and volume image structure descriptors. The measures are mapped and rendered on the segmentation boundary mesh to indicate the uncertainty of the current segmentation boundary, which can guide users in finding a optimal editing region to start editing. Existing interactive volume segmentation editing methods usually have a demand of per-slice 2D interactions which are inconvenient and time-consuming. Compared to these methods, our system allows users to use sketch-based interactions to generate editing slices of custom shapes, and refine the contours on 2D slices to edit the current medical segmentation result. The edit results will be propagated to the 3D volume using Laplacian mesh editing techniques, which makes only a few slice interactions required. The experiments have shown that our method can effectively improve the accuracy of segmentation results. Compared with the similar methods, the number of interactions is also effectively reduced.