Medical image segmentation with imperfect 3D bounding boxes

TL;DR

This paper introduces a bounding box correction framework that enhances weakly-supervised 3D medical image segmentation by improving bounding box tightness, leading to segmentation results comparable to fully-supervised methods.

Contribution

It proposes a novel bounding box correction method trained on limited pixel-level annotations to improve weakly-supervised segmentation accuracy in 3D medical images.

Findings

Bounding box correction improves segmentation accuracy.

Results approach fully-supervised performance.

Effective with limited pixel-level annotations.

Abstract

The development of high quality medical image segmentation algorithms depends on the availability of large datasets with pixel-level labels. The challenges of collecting such datasets, especially in case of 3D volumes, motivate to develop approaches that can learn from other types of labels that are cheap to obtain, e.g. bounding boxes. We focus on 3D medical images with their corresponding 3D bounding boxes which are considered as series of per-slice non-tight 2D bounding boxes. While current weakly-supervised approaches that use 2D bounding boxes as weak labels can be applied to medical image segmentation, we show that their success is limited in cases when the assumption about the tightness of the bounding boxes breaks. We propose a new bounding box correction framework which is trained on a small set of pixel-level annotations to improve the tightness of a larger set of non-tight bounding box annotations. The effectiveness of our solution is demonstrated by evaluating a known weakly-supervised segmentation approach with and without the proposed bounding box correction algorithm. When the tightness is improved by our solution, the results of the weakly-supervised segmentation become much closer to those of the fully-supervised one.