Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fuzzy Rank-based Late Fusion Technique for Cytology image Segmentation

Soumyajyoti Dey, Sukanta Chakraborty, Utso Guha Roy, Nibaran Das

TL;DR

This paper tackles the challenging task of cytology image segmentation under limited annotated data by proposing a fuzzy rank-based voting late fusion method that combines three semantic segmentation models—UNet, SegNet, and PSPNet. The approach trains each base model independently to produce per-pixel class probabilities, then fuses these via a two-stage fuzzy ranking scheme that yields per-pixel class scores and final predictions. Evaluation on two cytology datasets, HErlev and JUCYT-v1, shows mean IoU improvements to 84.27% and 83.79%, respectively, outperforming traditional fusion rules such as arithmetic and geometric means, median, max/min, and Borda Count. The method demonstrates that fuzzy ensemble fusion can leverage complementary strengths of diverse architectures, with code available on GitHub to support reproducibility and further exploration in digital pathology segmentation.

Abstract

Cytology image segmentation is quite challenging due to its complex cellular structure and multiple overlapping regions. On the other hand, for supervised machine learning techniques, we need a large amount of annotated data, which is costly. In recent years, late fusion techniques have given some promising performances in the field of image classification. In this paper, we have explored a fuzzy-based late fusion techniques for cytology image segmentation. This fusion rule integrates three traditional semantic segmentation models UNet, SegNet, and PSPNet. The technique is applied on two cytology image datasets, i.e., cervical cytology(HErlev) and breast cytology(JUCYT-v1) image datasets. We have achieved maximum MeanIoU score 84.27% and 83.79% on the HErlev dataset and JUCYT-v1 dataset after the proposed late fusion technique, respectively which are better than that of the traditional fusion rules such as average probability, geometric mean, Borda Count, etc. The codes of the proposed model are available on GitHub.

Fuzzy Rank-based Late Fusion Technique for Cytology image Segmentation

TL;DR

This paper tackles the challenging task of cytology image segmentation under limited annotated data by proposing a fuzzy rank-based voting late fusion method that combines three semantic segmentation models—UNet, SegNet, and PSPNet. The approach trains each base model independently to produce per-pixel class probabilities, then fuses these via a two-stage fuzzy ranking scheme that yields per-pixel class scores and final predictions. Evaluation on two cytology datasets, HErlev and JUCYT-v1, shows mean IoU improvements to 84.27% and 83.79%, respectively, outperforming traditional fusion rules such as arithmetic and geometric means, median, max/min, and Borda Count. The method demonstrates that fuzzy ensemble fusion can leverage complementary strengths of diverse architectures, with code available on GitHub to support reproducibility and further exploration in digital pathology segmentation.

Abstract

Cytology image segmentation is quite challenging due to its complex cellular structure and multiple overlapping regions. On the other hand, for supervised machine learning techniques, we need a large amount of annotated data, which is costly. In recent years, late fusion techniques have given some promising performances in the field of image classification. In this paper, we have explored a fuzzy-based late fusion techniques for cytology image segmentation. This fusion rule integrates three traditional semantic segmentation models UNet, SegNet, and PSPNet. The technique is applied on two cytology image datasets, i.e., cervical cytology(HErlev) and breast cytology(JUCYT-v1) image datasets. We have achieved maximum MeanIoU score 84.27% and 83.79% on the HErlev dataset and JUCYT-v1 dataset after the proposed late fusion technique, respectively which are better than that of the traditional fusion rules such as average probability, geometric mean, Borda Count, etc. The codes of the proposed model are available on GitHub.
Paper Structure (17 sections, 7 figures, 3 tables)

This paper contains 17 sections, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Previous Work
  3. Dataset Description:
  4. Proposed Methodology
  5. Semantic Segmentation by Traditional Deep Learning Models
  6. Fuzzy based Late Fusion Rule:
  7. Fuzzy Rank based Voting Rule:
  8. Results and Discussion
  9. Comparative study
  10. Arithmetic Average rule:
  11. Geometric Average rule:
  12. Median value rule:
  13. Max Rule
  14. Min Rule
  15. Borda Count(BC) rule:
  16. ...and 2 more sections

Figures (7)

  • Figure 1: Ground truth preparation of JUCYT-v1 dataset
  • Figure 2: Cervical Cytology Data(HErlev).Top row indicated the original images, Bottom row indicates the corresponding ground truth images
  • Figure 3: Breast Cytology Data(JUCYT-v1).Top row indicates the original images, Bottom row indicates the corresponding ground truth images
  • Figure 4: Outline of proposed Fuzzy based Late fusion rule for cytology image segmentation
  • Figure 5: Output segmentation masks evaluating by base semantic Segmentation models
  • ...and 2 more figures