Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CMU-Net: A Strong ConvMixer-based Medical Ultrasound Image Segmentation Network

Fenghe Tang, Lingtao Wang, Chunping Ning, Min Xian, Jianrui Ding

TL;DR

A fully convolutional segmentation network (CMU-Net) which incorporates hybrid convolutions and multi-scale attention gate, and the ConvMixer module extracts global context information by mixing features at distant spatial locations.

Abstract

U-Net and its extensions have achieved great success in medical image segmentation. However, due to the inherent local characteristics of ordinary convolution operations, U-Net encoder cannot effectively extract global context information. In addition, simple skip connections cannot capture salient features. In this work, we propose a fully convolutional segmentation network (CMU-Net) which incorporates hybrid convolutions and multi-scale attention gate. The ConvMixer module extracts global context information by mixing features at distant spatial locations. Moreover, the multi-scale attention gate emphasizes valuable features and achieves efficient skip connections. We evaluate the proposed method using both breast ultrasound datasets and a thyroid ultrasound image dataset; and CMU-Net achieves average Intersection over Union (IoU) values of 73.27% and 84.75%, and F1 scores of 84.81% and 91.71%. The code is available at https://github.com/FengheTan9/CMU-Net.

CMU-Net: A Strong ConvMixer-based Medical Ultrasound Image Segmentation Network

TL;DR

A fully convolutional segmentation network (CMU-Net) which incorporates hybrid convolutions and multi-scale attention gate, and the ConvMixer module extracts global context information by mixing features at distant spatial locations.

Abstract

U-Net and its extensions have achieved great success in medical image segmentation. However, due to the inherent local characteristics of ordinary convolution operations, U-Net encoder cannot effectively extract global context information. In addition, simple skip connections cannot capture salient features. In this work, we propose a fully convolutional segmentation network (CMU-Net) which incorporates hybrid convolutions and multi-scale attention gate. The ConvMixer module extracts global context information by mixing features at distant spatial locations. Moreover, the multi-scale attention gate emphasizes valuable features and achieves efficient skip connections. We evaluate the proposed method using both breast ultrasound datasets and a thyroid ultrasound image dataset; and CMU-Net achieves average Intersection over Union (IoU) values of 73.27% and 84.75%, and F1 scores of 84.81% and 91.71%. The code is available at https://github.com/FengheTan9/CMU-Net.
Paper Structure (10 sections, 5 equations, 4 figures, 3 tables)

This paper contains 10 sections, 5 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. METHOD
  3. Encoder stage
  4. Decoder stage with skip connection
  5. RESULTS AND DISCUSSION
  6. Datasets, evaluation and implementation details
  7. Results
  8. CONCLUSION AND PERSPECTIVES
  9. Compliance with ethical standards
  10. Acknowledgments

Figures (4)

  • Figure 1: Examples of ultrasound image segmentation. The pink contours denote lesion boundaries.
  • Figure 2: Overview of the proposed CMU-Net architecture. Note that the channel numbers are adopted from U-Net, i.e., C1 = 64, C2 = 128, C3 = 256, C4 = 512, and C5 = 1024.
  • Figure 3: Multi-scale attention gate.
  • Figure 4: Segmentation result. Row 1 - BUSI dataset, Row 2 – TUS dataset. (a) Input and (b) Ground Truth. Predictions of (c) CMU-Net, (d) Attention U-Net, (e) TransUnet, (f) U-Net, (g) U-Net++ (h), U-Net3+, and (i) UNeXt.