Dilation-supervised

This repository is the PyTorch code for the paper 'Dilation-supervised Learning: A Novel Strategy for Scale Difference in Retinal Vessel Segmentation' [Paper]

Abstract

Retinal fundus image segmentation based on deep learning is an important method for auxiliary diagnosis of ophthalmic diseases. Due to the scale difference of the blood vessels and the imbalance between foreground and background pixels in the fundus image, the deep learning network will inevitably ignore thin vessels when down-sampling and feature learning. For the scale difference problem, this paper aims to tackle its limitation from two perspectives: changing the supervised approach and adapting the feature learning. Correspondingly, a dilation-supervised learning method and an adaptive scale dimensional attention mechanism which are used to construct a two-stage segmentation model is proposed. Moreover, we introduce a quantitative approach to evaluate the scale difference of the blood vessels. With the help of the proposed weighted loss function, the segmentation results are refined, and the class imbalance problem between foreground and background pixels is resolved. Finally, the proposed adaptive threshold selection method is used in the post-processing of segmentation results. The experiments on DRIVE, STARE, CHASE DB1, and HRF datasets show that the proposed method achieves better segmentation performance compared with other state-of-the-art methods, and has good generalization ability and robustness.

Training on DRIVE, CHASE_DB1, SATRE, and HRF

Stage 1: Downloading Datasets

Please download the retina image datasets(DRIVE, STARE, CHASE_DB1, and HRF) to dilation_supervised/dataset1

Stage 2: data enhancement

cd root/dilation_supervised/data_process
python read_dateset_crop.py

Stage 3: train

cd root/dilation_supervised/train_test
python train_vessel.py

Testing on DRIVE, CHASE_DB1, SATRE, and HRF

cd root/dilation_supervised/train_test
python eval_test.py

Citation

If you think this repo and the manuscript helpful, please consider citing us.

@inproceedings{
  title={Dilation-supervised Learning: A Novel Strategy for Scale Difference in Retinal Vessel Segmentation},
  author={Wang Huadeng, Zuo Wenbin, Lan Rushi},
  booktitle={IEEE Transactions on Artificial Intelligence},
  year={2023},
}