PDBL: Improving Histopathological Tissue Classification with Plug-and-Play Pyramidal Deep-Broad Learning

outline

Introduction

The iplementation of:

PDBL: Improving Histopathological Tissue Classification with Plug-and-Play Pyramidal Deep-Broad Learning

You can also download the repository from https://github.com/linjiatai/PDBL.git.

Abtract

Histopathological tissue classification is a simpler way to achieve semantic segmentation for the whole slide images, which can alleviate the requirement of pixel-level dense annotations. Existing works mostly leverage the popular CNN classification backbones in computer vision to achieve histopathological tissue classification. In this paper, we propose a super lightweight plug-and-play module, named Pyramidal Deep-Broad Learning (PDBL), for any well-trained classification backbone to improve the classification performance without a re-training burden. For each patch, we construct a multi-resolution image pyramid to obtain the pyramidal contextual information. For each level in the pyramid, we extract the multi-scale deep-broad features by our proposed Deep-Broad block (DB-block). We equip PDBL in three popular classification backbones, ShuffLeNetV2, EfficientNetb0, and ResNet50 to evaluate the effectiveness and efficiency of our proposed module on two datasets (Kather Multiclass Dataset and the LC25000 Dataset). Experimental results demonstrate the proposed PDBL can steadily improve the tissue-level classification performance for any CNN backbones, especially for the lightweight models when given a small among of training samples (less than 10%). It greatly saves the computational resources and annotation efforts.

Requirements

CUDA
1×GPU
2×CPU
Python 3.7
numpy==1.21.5
pytorch==1.11.0
torchvision==0.12.0
scikit-learn==0.24.2
Usage

Installation

Download the repository.

git clone https://github.com/linjiatai/PDBL.git

Install python dependencies.
```
pip install -r requirements.txt
```

PDBL

We provide an example to create and train a PDBL on 1% KMI set without CNN re-training burden. To use this example code, you must download the data set, you must download the 1% KMI set and KME set (Baidu Netdisk with code wfzk) and unpacked them in datset folder. And you can train and test the PDBL by the command:

python main.py --device 0 --save_dir save/ --traindir dataset/KMI_001/ --valdir dataset/KME --batch_size 20 -- n_class --n_workers 8

Moreover, the original Kather Dataset and LC25000 Dataset can be download at the links Kather2019 and LC25000.

Citation

If you find the code useful, please consider citing our paper using the following BibTeX entry.

@ARTICLE{lin2022pdbl,
  author={Lin, Jiatai and Han, Guoqiang and Pan, Xipeng and Liu, Zaiyi and Chen, Hao and Li, Danyi and Jia, Xiping and Shi, Zhenwei and Wang, Zhizhen and Cui, Yanfen and Li, Haiming and Liang, Changhong and Liang, Li and Wang, Ying and Han, Chu},
  journal={IEEE Transactions on Medical Imaging}, 
  title={PDBL: Improving Histopathological Tissue Classification with Plug-and-Play Pyramidal Deep-Broad Learning}, 
  year={2022},
  pages={1-1},
  doi={10.1109/TMI.2022.3161787}}

linjiatai / PDBL

readme