Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Segmenting Medical Images: From UNet to Res-UNet and nnUNet

Lina Huang, Alina Miron, Kate Hone, Yongmin Li

TL;DR

This paper tackles the problem of selecting effective deep learning models for diverse medical image segmentation tasks. It systematically compares UNet, Res-UNet, Attention Res-UNet, and nnUNet on brain-tumour, polyp, and multi-class heart segmentation, using metrics such as precision, recall, accuracy, DSC, and IoU. The main finding is that nnUNet generally delivers the best overall performance, especially in recall and boundary accuracy for tumours and polyps, while Res-UNet provides strong boundary delineation in brain tumours; but nnUNet's robustness across tasks makes it the preferred choice despite higher computational costs. The study highlights the importance of cross-task benchmarking and notes limitations like 2D data usage and data leakage, suggesting future work in 3D extensions and stricter data handling.

Abstract

This study provides a comparative analysis of deep learning models including UNet, Res-UNet, Attention Res-UNet, and nnUNet, and evaluates their performance in brain tumour, polyp, and multi-class heart segmentation tasks. The analysis focuses on precision, accuracy, recall, Dice Similarity Coefficient (DSC), and Intersection over Union (IoU) to assess their clinical applicability. In brain tumour segmentation, Res-UNet and nnUNet significantly outperformed UNet, with Res-UNet leading in DSC and IoU scores, indicating superior accuracy in tumour delineation. Meanwhile, nnUNet excelled in recall and accuracy, which are crucial for reliable tumour detection in clinical diagnosis and planning. In polyp detection, nnUNet was the most effective, achieving the highest metrics across all categories and proving itself as a reliable diagnostic tool in endoscopy. In the complex task of heart segmentation, Res-UNet and Attention Res-UNet were outstanding in delineating the left ventricle, with Res-UNet also leading in right ventricle segmentation. nnUNet was unmatched in myocardium segmentation, achieving top scores in precision, recall, DSC, and IoU. The conclusion notes that although Res-UNet occasionally outperforms nnUNet in specific metrics, the differences are quite small. Moreover, nnUNet consistently shows superior overall performance across the experiments. Particularly noted for its high recall and accuracy, which are crucial in clinical settings to minimize misdiagnosis and ensure timely treatment, nnUNet's robust performance in crucial metrics across all tested categories establishes it as the most effective model for these varied and complex segmentation tasks.

Segmenting Medical Images: From UNet to Res-UNet and nnUNet

TL;DR

This paper tackles the problem of selecting effective deep learning models for diverse medical image segmentation tasks. It systematically compares UNet, Res-UNet, Attention Res-UNet, and nnUNet on brain-tumour, polyp, and multi-class heart segmentation, using metrics such as precision, recall, accuracy, DSC, and IoU. The main finding is that nnUNet generally delivers the best overall performance, especially in recall and boundary accuracy for tumours and polyps, while Res-UNet provides strong boundary delineation in brain tumours; but nnUNet's robustness across tasks makes it the preferred choice despite higher computational costs. The study highlights the importance of cross-task benchmarking and notes limitations like 2D data usage and data leakage, suggesting future work in 3D extensions and stricter data handling.

Abstract

This study provides a comparative analysis of deep learning models including UNet, Res-UNet, Attention Res-UNet, and nnUNet, and evaluates their performance in brain tumour, polyp, and multi-class heart segmentation tasks. The analysis focuses on precision, accuracy, recall, Dice Similarity Coefficient (DSC), and Intersection over Union (IoU) to assess their clinical applicability. In brain tumour segmentation, Res-UNet and nnUNet significantly outperformed UNet, with Res-UNet leading in DSC and IoU scores, indicating superior accuracy in tumour delineation. Meanwhile, nnUNet excelled in recall and accuracy, which are crucial for reliable tumour detection in clinical diagnosis and planning. In polyp detection, nnUNet was the most effective, achieving the highest metrics across all categories and proving itself as a reliable diagnostic tool in endoscopy. In the complex task of heart segmentation, Res-UNet and Attention Res-UNet were outstanding in delineating the left ventricle, with Res-UNet also leading in right ventricle segmentation. nnUNet was unmatched in myocardium segmentation, achieving top scores in precision, recall, DSC, and IoU. The conclusion notes that although Res-UNet occasionally outperforms nnUNet in specific metrics, the differences are quite small. Moreover, nnUNet consistently shows superior overall performance across the experiments. Particularly noted for its high recall and accuracy, which are crucial in clinical settings to minimize misdiagnosis and ensure timely treatment, nnUNet's robust performance in crucial metrics across all tested categories establishes it as the most effective model for these varied and complex segmentation tasks.
Paper Structure (18 sections, 3 figures, 3 tables)

This paper contains 18 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Methods
  4. UNet
  5. Res-UNet
  6. Attention Res-UNet
  7. nnUNet
  8. Brain Tumour Segmentation
  9. Dataset
  10. Result
  11. Polyp Segmentation
  12. Dataset
  13. Result
  14. Heart Segmentation
  15. Dataset
  16. ...and 3 more sections

Figures (3)

  • Figure 1: Brain tumour segmentation: Four examples showing segmentation by four models. From left to right: input images, ground-truth, and segmentation by UNet, Res-UNet, Attention Res-UNet, nnUNet, in order.
  • Figure 2: Polyp segmentation: Four examples showing segmentation by four models. From left to right: input images, ground-truth, and segmentation by UNet, Res-UNet, Attention Res-UNet, nnUNet, in order.
  • Figure 3: Heart segmentation: Four examples showing segmentation by four models. From left to right: input images, ground-truth, and segmentation by UNet, Res-UNet, Attention Res-UNet, nnUNet, in order.