Codes for our Medical & Biological Engineering & Computing paper "MaskMitosis: a deep learning framework for fully supervised, weakly supervised, and unsupervised mitosis detection in histopathology images". Please see the paper for more details.

System Overview

Results

Method	Evaluation set	F-score
MaskMitosis12	2012 ICPR MITOSIS test set	0.863
MaskMitosis14 (weakly supervised)	2014 ICPR MITOSIS validation set	0.608
MaskMitosis14 (weakly supervised)	2014 ICPR MITOSIS test set	0.475
MaskMitosis14 (unsupervised)	2014 ICPR MITOSIS validation set	0.504
MaskMitosis14 (unsupervised)	2014 ICPR MITOSIS test set	0.395

Citing MaskMitosis

If you find MaskMitosis useful in your research, please consider citing:

@article{sebai2020maskmitosis,
title={MaskMitosis: a deep learning framework for fully supervised, weakly supervised, and unsupervised mitosis detection in histopathology images},
author={Sebai, Meriem and Wang, Xinggang and Wang, Tianjiang},
journal={Medical \& Biological Engineering \& Computing},
year={2020},
publisher={Springer}
}

Content

1. Hardware Requirements
2. Software Requirements
3. Installation

Hardware Requirements

In our experiments, we used a Nvidia Quadro P5000 GPU with 16 GB of memory.

Software Requirements

Python 3.4, TensorFlow 1.3, Keras 2.0.8 and other common packages are required.

Installation

Our mitosis detection and instance segmentation model is based on Mask RCNN model. We use the Matterport Inc implementation of Mask-RCNN. For more details see Mask RCNN

1. Clone the repository.
2. Install dependencies: pip3 install -r requirements.txt
3. Run setup from the repository root directory: python3 setup.py install

The trained MaskMitosis12 and MaskMitosis14 models are available in MaskMitosis12 model and MaskMitosis14 model, respectively.