This repository contains the implementation of a novel attention based network (CaraNet) to segment the polyp (CVC-T, CVC-ClinicDB, CVC-ColonDB, ETIS and Kvasir) and brain tumor (BraTS). The CaraNet show great overall segmentation performance (mean dice) on polyp and brain tumor, but also show great performance on small medical objects (small polyps and brain tumors) segmentation.
:fire: NEWS :fire: The full paper is available: CaraNet
The journal version is available: CaraNet
We use Res2Net as our backbone.
We choose our CFP module as context module, and choose the dilation rate is 8. For the details of CFP module you can find here: CFPNet. The architecture of CFP module as shown in following figure:
As shown in architecture of CaraNet, the Axial Reverse Attention (A-RA) module contains two routes: 1) Reverse attention; 2) Axial-attention (The code of axial attention is applied from UACANET)
conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch
conda install opencv-python pillow numpy matplotlib
git clone https://github.com/AngeLouCN/CaraNet
Train.py
|-- TestDataset
| |-- CVC-300
| | |-- images
| | |-- masks
| |-- CVC-ClinicDB
| | |-- images
| | |-- masks
| |-- CVC-ColonDB
| | |-- images
| | |-- masks
| |-- ETIS-LaribPolypDB
| | |-- images
| | |-- masks
| |-- Kvasir
| |-- images
| |-- masks
dice_average.m
is to compute the averaged dice values according to sizes of objects, for small area analysis.Polyp Segmentation Results
Conditions of test datasets:
The x-axis is the proportion size (%) of polyp; y-axis is the average mean dice coefficient.
BraTS input | Segmentation truth |
---|---|
Results
Small tumor analysis
For very small areas (<1%):
The difference between results of CaraNet and PraNet:
If you think our work is helpful, please cite both conference and journal version.
@inproceedings{lou2021caranet,
author = {Ange Lou and Shuyue Guan and Hanseok Ko and Murray H. Loew},
title = {{CaraNet: context axial reverse attention network for segmentation of small medical objects}},
volume = {12032},
booktitle = {Medical Imaging 2022: Image Processing},
organization = {International Society for Optics and Photonics},
publisher = {SPIE},
pages = {81 -- 92},
year = {2022},
doi = {10.1117/12.2611802}}
@inproceedings{9506485,
author={Lou, Ange and Loew, Murray},
booktitle={2021 IEEE International Conference on Image Processing (ICIP)},
title={CFPNET: Channel-Wise Feature Pyramid For Real-Time Semantic Segmentation},
year={2021},
volume={},
number={},
pages={1894-1898},
doi={10.1109/ICIP42928.2021.9506485}}
@article{lou2023caranet,
title={CaraNet: context axial reverse attention network for segmentation of small medical objects},
author={Lou, Ange and Guan, Shuyue and Loew, Murray},
journal={Journal of Medical Imaging},
volume={10},
number={1},
pages={014005},
year={2023},
publisher={SPIE}
}