mldl / cancer

Autonomous Extraction of Gleason Patterns for Grading Prostate Cancer using Multi-Gigapixel Whole Slide Images

Introduction

This repository contains the implementation of our paper titled "Autonomous Extraction of Gleason Patterns for Grading Prostate Cancer using Multi-Gigapixel Whole Slide Images" currently under review in IEEE International Symposium on Biomedical Imaging (ISBI) 2021

Block Diagram of the Proposed Framework

The proposed framework is developed using TensorFlow 2.3.1 and Keras APIs with Python 3.7.8. Moreover, some preprocessing steps and result compilation is performed through MATLAB R2020a as well. The detailed steps for installing and running the code are presented below:

Installation

To run the codebase, following libraries are required. Although, the framework is developed using Anaconda. But it should be compatable with other platforms.

1. TensorFlow 2.3.1
2. Keras 2.3.1
3. OpenCV 4.4.0
4. imgaug 0.2.9 or above
5. tqdm
6. Matplotlib

Alternatively, we also provide a yml file that contains all of these packages.

Dataset

Please download the desired dataset and follow the below-mentioned hierarchy to train and test the proposed framework:

├── trainingDataset
│   ├── train_images
│   │   └── tr_image_1.png
│   │   └── tr_image_2.png
│   │   ...
│   │   └── tr_image_n.png
│   ├── train_annotations
│   │   └── tr_image_1.png
│   │   └── tr_image_2.png
│   │   ...
│   │   └── tr_image_n.png
│   ├── val_images
│   │   └── va_image_1.png
│   │   └── va_image_2.png
│   │   ...
│   │   └── va_image_m.png
│   ├── val_annotations
│   │   └── va_image_1.png
│   │   └── va_image_2.png
│   │   ...
│   │   └── va_image_m.png
├── testingDataset
│   ├── test_images
│   │   └── te_image_1.png
│   │   └── te_image_2.png
│   │   ...
│   │   └── te_image_k.png
│   ├── test_annotations
│   │   └── te_image_1.png
│   │   └── te_image_2.png
│   │   ...
│   │   └── te_image_k.png
│   ├── segmentation_results
│   │   └── te_image_1.png
│   │   └── te_image_2.png
│   │   ...
│   │   └── te_image_k.png

If you are using whole slide image dataset, then the following training and testing folders should contain WSI patches only. These patches can be obtained using 'patchGenerator.m'.

Steps

1. Download the desired dataset
2. Generate WSI patches using 'patchGenerator.m'
3. Resize patches (to make them compatible with the proposed framework) using the script 'resizer.m' within 'training_utils' folder
4. Put the resized training patches in '…\trainingDataset\train_images' folder
5. Put the resized training annotations (ground truth patches) in '…\trainingDataset\train_annotations' folder
6. Put resized validation patches in '…\trainingDataset\val_images' folder
7. Put resized validation ground truth patches in '…\trainingDataset\val_annotations' folder. Note: the patches and their annotations should have same name and extension (preferably .png).
8. Put resized test patches in '…\testingDataset\test_images' folder and their annotations in '…\testingDataset\test_annotations' folder
9. Use 'trainer.py' file to train the proposed framework and then evaluate it on the test scans. The results on the test scans are saved in ‘…\testingDataset\segmentation_results’ folder. This script also saves the trained model in 'model.h5' file
10. The trained models can also be ported to MATLAB using ‘kerasConverter.m’ (this step is optional and only designed to facilitate MATLAB users if they want to avoid Python analysis)

Citation

If you use the proposed framework (or any part of this code in your research), please cite the following paper:

@article{Hassan2021Cancer,
  title   = {Autonomous Extraction of Gleason Patterns for Grading Prostate Cancer using Multi-Gigapixel Whole Slide Images},
  author  = {Taimur Hassan and Ayman El-Baz and Naoufel Werghi},
  journal = {Submitted in IEEE International Symposium on Biomedical Imaging (ISBI)},
  year = {2021}
}

Contact

If you have any query, please feel free to contact us at: [email protected].