Benchmarking Polyp Segmentation Methods in Narrow-Band Imaging Colonoscopy Images (Accepted for IEEE JBHI)

Abstract

In recent years, there has been significant progress in polyp segmentation in white-light imaging (WLI) colonoscopy images, particularly with methods based on deep learning (DL). However, little attention has been paid to the reliability of these methods in narrow-band imaging (NBI) data. NBI improves visibility of blood vessels and helps physicians observe complex polyps more easily than WLI, but NBI images often include polyps with small/flat appearances, background interference, and camouflage properties, making polyp segmentation a challenging task. This paper proposes a new polyp segmentation dataset (PS-NBI2K) consisting of 2,000 NBI colonoscopy images with pixel-wise annotations, and presents benchmarking results and analyses for 24 recently reported DL-based polyp segmentation methods on PS-NBI2K. The results show that existing methods struggle to locate polyps with smaller sizes and stronger interference, and that extracting both local and global features improves performance. There is also a trade-off between effectiveness and efficiency, and most methods cannot achieve the best results in both areas simultaneously. This work highlights potential directions for designing DL-based polyp segmentation methods in NBI colonoscopy images, and the release of PS-NBI2K aims to drive further development in this field.

Links to the DL methods that are compared in this paper

Checkpoints of the models

The checkpoints of the models above can be downloaded via BaiduNetDisk, with an extraction code of JrzX. For each method, we have uploaded the checkpoints of all 5 trials.

Prediction maps of the compared methods

The prediction maps of the methods above can also be downloaded via BaiduNetDisk, with an extraction code of Hy2z. For each method, we have uploaded the prediction maps of all 5 trials.

Evaluation of the models

For all methods, we use the same evaluation tool from the PraNet project to evaluate their performance. Please refer to the eval folder and carefully read the README on the github repository.

Access to the PS-NBI2K dataset

The dataset can now be accessed via OneDrive. To prevent non-scientific use of the dataset, the dataset archive is encrypted with a password. Please send an email to cooperzhuo@proton.me to get the password.

Citation

If you want to use the database, please cite our paper

@article{psnbi2k,
    author={Yue, Guanghui and Zhuo, Guibin and Li, Siying and Zhou, Tianwei and Du, Jingfeng and Yan, Weiqing and Hou, Jingwen and Liu, Weide and Wang, Tianfu},
    journal={IEEE Journal of Biomedical and Health Informatics}, 
    title={Benchmarking Polyp Segmentation Methods in Narrow-Band Imaging Colonoscopy Images}, 
    year={2023}
}