ADD

Holistic White-light Polyp Classification via Alignment-free Dense Distillation of Auxiliary Optical Chromoendoscopy

Qiang Hu^{1, *}, Qimei Wang^{1, *}, Jia Chen², Xuantao Ji², Qiang Li^{1, †}, Zhiwei Wang^{1, †}

¹ WNLO, HUST, ² UIH
(*: equal contribution, †: corresponding author)

1. Overview

This work implements a novel framework for enhancing holistic WLI polyp classification through cross-domain (NBI->WLI) knowledge distillation, without requiring any additinal labeling cost. The core innovation is Alignment-free Dense Distillation (ADD) module, which establishes dense distillation pathways between misaligned cross-domain features guided by learned affinities. Additionally, we capture the semantic relations to ensure distillation is restricted to semantically consistent regions. Extensive experiments demonstrate that our method achieves the state-of-the-art performance in WLI image classification on both the public CPC-Paired and our in-house datasets.

2. Checkpoints

Model	CPC-Paired (AUC)	In-house (AUC)	Weights (5-folds)
Ours	0.936	0.826	ckpts
CIC variant	0.801	0.603	ckpts
w/o ADD & SRG	0.857	0.683	ckpts
w/o SRG	0.925	0.775	ckpts
w/o Bi-A	0.918	0.762	ckpts
w/o PSR	0.928	0.786	ckpts

3. Visulization of Results

3.1 ROC Curve:

In-house Dataset

Public Dataset (CPC-Paried)

3.2 CAM Maps:

4. Getting Started

4.1 Recommended Environment:

• Python 3.8+
• PyTorch 2.1+
• TorchVision corresponding to the PyTorch version
• NVIDIA GPU + CUDA
• Install other dependent packages:

cd ADD
pip install -r requirements.txt

4.2 Data Preparation

• Downloading the CPC-Paired dataset (public WLI-NBI paired polyp classification dataset). The file paths should be arranged as follows:

ADD
├── dataset
├── ├── White_light
├── ├── ├── adenomas
├── ├── ├── ├── ├── 01-1.png
├── ├── ├── ├── ├── 02-1.png
├── ├── ├── ├── ├── ......
├── ├── ├── hyperplastic_lesions
├── ├── ├── ├── ├── 011-1.png
├── ├── ├── ├── ├── 011-2.png
├── ├── ├── ├── ├── ......
├── ├── NBI
├── ├── ├── adenomas
├── ├── ├── ├── ├── 01-1.png
├── ├── ├── ├── ├── 02-1.png
├── ├── ├── ├── ├── ......
├── ├── ├── hyperplastic_lesions
├── ├── ├── ├── ├── 011-1.png
├── ├── ├── ├── ├── 011-2.png
├── ├── ├── ├── ├── ......

• Note that the details of dataset splitation in the 5-fold experiment can be downloaded in here. You need to download these .txt files and put them into a newly created folder split and the file paths should be arranged as follows:

ADD
├── split
├── ├── xxx.txt
├── ├── ......

4.3 Training:

Stage 1: pre-traning the NBI classifier:

python train_teacher.py

Stage 2: training the WLI classifier:

python train.py

4.4 Testing and Evaluation:

python test.py

You can also directly download the well-trained model from Google Drive, and predict the results by test.py.

Citation

If you find our paper and code useful in your research, please consider giving us a star ⭐ and citing ADD by the following BibTeX entry.

@inproceedings{hu2025holistic,
  title={Holistic White-Light Polyp Classification via Alignment-Free Dense Distillation of Auxiliary Optical Chromoendoscopy},
  author={Hu, Qiang and Wang, Qimei and Chen, Jia and Ji, Xuantao and Liu, Mei and Li, Qiang and Wang, Zhiwei},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={251--261},
  year={2025},
  organization={Springer}
}