Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BoundarySeg:An Embarrassingly Simple Method To Boost Medical Image Segmentation Performance for Low Data Regimes

Tushar Kataria, Shireen Y. Elhabian

TL;DR

The paper tackles the challenge of limited annotated medical imagery under strict privacy by proposing BoundarySeg, a boundary-aware multitask framework that jointly learns organ segmentation and boundary prediction to provide extra supervision without unannotated data. It introduces a plug‑and‑play architecture with boundary labels generated via erosion and a boundary-consistency loss that ties the two tasks together, along with a forward-pass-only semi-supervised extension for unlabeled volumes. Empirical results on 3D LA MRI data show BoundarySeg surpasses state-of-the-art semi‑supervised methods when labeled data are scarce, and that even minimal unlabeled data used only in the forward pass can boost performance further. The work offers a simple, computationally efficient approach that improves low-data segmentation performance and can be integrated with existing architectures, with planned expansion to other modalities and organs.

Abstract

Obtaining large-scale medical data, annotated or unannotated, is challenging due to stringent privacy regulations and data protection policies. In addition, annotating medical images requires that domain experts manually delineate anatomical structures, making the process both time-consuming and costly. As a result, semi-supervised methods have gained popularity for reducing annotation costs. However, the performance of semi-supervised methods is heavily dependent on the availability of unannotated data, and their effectiveness declines when such data are scarce or absent. To overcome this limitation, we propose a simple, yet effective and computationally efficient approach for medical image segmentation that leverages only existing annotations. We propose BoundarySeg , a multi-task framework that incorporates organ boundary prediction as an auxiliary task to full organ segmentation, leveraging consistency between the two task predictions to provide additional supervision. This strategy improves segmentation accuracy, especially in low data regimes, allowing our method to achieve performance comparable to or exceeding state-of-the-art semi supervised approaches all without relying on unannotated data or increasing computational demands. Code will be released upon acceptance.

BoundarySeg:An Embarrassingly Simple Method To Boost Medical Image Segmentation Performance for Low Data Regimes

TL;DR

The paper tackles the challenge of limited annotated medical imagery under strict privacy by proposing BoundarySeg, a boundary-aware multitask framework that jointly learns organ segmentation and boundary prediction to provide extra supervision without unannotated data. It introduces a plug‑and‑play architecture with boundary labels generated via erosion and a boundary-consistency loss that ties the two tasks together, along with a forward-pass-only semi-supervised extension for unlabeled volumes. Empirical results on 3D LA MRI data show BoundarySeg surpasses state-of-the-art semi‑supervised methods when labeled data are scarce, and that even minimal unlabeled data used only in the forward pass can boost performance further. The work offers a simple, computationally efficient approach that improves low-data segmentation performance and can be integrated with existing architectures, with planned expansion to other modalities and organs.

Abstract

Obtaining large-scale medical data, annotated or unannotated, is challenging due to stringent privacy regulations and data protection policies. In addition, annotating medical images requires that domain experts manually delineate anatomical structures, making the process both time-consuming and costly. As a result, semi-supervised methods have gained popularity for reducing annotation costs. However, the performance of semi-supervised methods is heavily dependent on the availability of unannotated data, and their effectiveness declines when such data are scarce or absent. To overcome this limitation, we propose a simple, yet effective and computationally efficient approach for medical image segmentation that leverages only existing annotations. We propose BoundarySeg , a multi-task framework that incorporates organ boundary prediction as an auxiliary task to full organ segmentation, leveraging consistency between the two task predictions to provide additional supervision. This strategy improves segmentation accuracy, especially in low data regimes, allowing our method to achieve performance comparable to or exceeding state-of-the-art semi supervised approaches all without relying on unannotated data or increasing computational demands. Code will be released upon acceptance.
Paper Structure (4 sections, 6 equations, 1 figure, 2 tables)

This paper contains 4 sections, 6 equations, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. BoundarySeg Method
  3. Experimental Details and Discussion
  4. Conclusion and Future Work

Figures (1)

  • Figure 1: BoundarySeg Pipeline. (A) The proposed multi-task segmentation network takes the full 3D volume as input, uses a V-Net Architecture, and produces two outputs (i) segmentation of the full anatomy and (ii) boundary segmentation, which is obtained by using morphological operation. (B) Shows the pipeline to obtain the boundary labels using erosion and XOR operation.