Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clore: Interactive Pathology Image Segmentation with Click-based Local Refinement

Tiantong Wang, Minfan Zhao, Jun Shi, Hannan Wang, Yue Dai

Abstract

Recent advancements in deep learning-based interactive segmentation methods have significantly improved pathology image segmentation. Most existing approaches utilize user-provided positive and negative clicks to guide the segmentation process. However, these methods primarily rely on iterative global updates for refinement, which lead to redundant re-prediction and often fail to capture fine-grained structures or correct subtle errors during localized adjustments. To address this limitation, we propose the Click-based Local Refinement (Clore) pipeline, a simple yet efficient method designed to enhance interactive segmentation. The key innovation of Clore lies in its hierarchical interaction paradigm: the initial clicks drive global segmentation to rapidly outline large target regions, while subsequent clicks progressively refine local details to achieve precise boundaries. This approach not only improves the ability to handle fine-grained segmentation tasks but also achieves high-quality results with fewer interactions. Experimental results on four datasets demonstrate that Clore achieves the best balance between segmentation accuracy and interaction cost, making it an effective solution for efficient and accurate interactive pathology image segmentation.

Clore: Interactive Pathology Image Segmentation with Click-based Local Refinement

Abstract

Recent advancements in deep learning-based interactive segmentation methods have significantly improved pathology image segmentation. Most existing approaches utilize user-provided positive and negative clicks to guide the segmentation process. However, these methods primarily rely on iterative global updates for refinement, which lead to redundant re-prediction and often fail to capture fine-grained structures or correct subtle errors during localized adjustments. To address this limitation, we propose the Click-based Local Refinement (Clore) pipeline, a simple yet efficient method designed to enhance interactive segmentation. The key innovation of Clore lies in its hierarchical interaction paradigm: the initial clicks drive global segmentation to rapidly outline large target regions, while subsequent clicks progressively refine local details to achieve precise boundaries. This approach not only improves the ability to handle fine-grained segmentation tasks but also achieves high-quality results with fewer interactions. Experimental results on four datasets demonstrate that Clore achieves the best balance between segmentation accuracy and interaction cost, making it an effective solution for efficient and accurate interactive pathology image segmentation.

Paper Structure

This paper contains 17 sections, 2 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Methodology
  3. Overview
  4. Clore: Interactive Pathology Image Segmentation
  5. Click Simulation and Encoding
  6. Local Patch Selection
  7. Local Refinement
  8. Shared Segmentation Network
  9. Model Training
  10. Loss Functions
  11. Experiments
  12. Data Preparation
  13. Implementation Details and Evaluation Metrics
  14. Results
  15. Ablation Study
  16. ...and 2 more sections

Figures (4)

  • Figure 1: An overview of the proposed method
  • Figure 2: mDice performance on NuCLs and GlaS datasets at fixed number of clicks.
  • Figure 3: The visualized segmentation results of Glas dataset, where the upper row representing few click situation and the lower row representing many click situation. The yellow box indicates the better segmentation area of our method compared to the comparison method.
  • Figure 4: Result of the ablation study