/TriConvUNeXt

Exploring Dilated CNN, Deformable CNN, and Depthwise CNN for medical image segmentation.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

TriConvUNeXt

Crafting a Lightweight and Robust Symmetrical Network for Enhanced Biomedical Image Segmentation.

Motivation

Exploring Dilated CNN, Deformable CNN, and Depthwise CNN for medical image segmentation.

Requirements

  • Pytorch
  • Some basic python packages such as Numpy, Scikit-image, SimpleITK, Scipy ......
  • More information about environment development can be found in 'environment.yml' file.

DataSets

We use the Glas dataset originating from Gland Segmentation in Colon Histology Images Challenge from MICCAI Challenge 2015, which you can find more information here Official.

Models

We provide many baseline methods as well including UNet, AttentionUNet, LinkNet, ConvUNext, TransUNet, SwinUNet, and UnetLite, which you can be found in '/nn' folder.

We also provide evaluation metrics including Dice-Coefficient, IoU, Accuracy, Precision, Sensitivity, and Specificity, which you can be found in 'metrics.py' file.

Usage

  1. Clone the repo:
git clone https://github.com/ziyangwang007/TriConvUNeXt.git
cd TriConvUNeXt
  1. Train the model
python train.py
  1. Test the model
python val.py

Reference

Chao Ma, Yuan Gu, Ziyang Wang. "TriConvUNeXt: A Pure CNN-based Lightweight Symmetrical Network for Biomedical Image Segmentation." Journal of Imaging Informatics in Medicine (2024).