Classification of Histopathologic Scans of Lymph Node Sections Using Machine Learning

When analyzing histopathological images, there is a need for an automated system that could help doctors in image analysis and diagnostics. Such a system could increase the accuracy and speed of analysis and diagnostics. Within this paper, an overview of the histopathological image analysis area is provided and a program implementation for the classification of lymph nodes based on machine learning has been developed. PatchCamelyon dataset has been used for training and testing of chosen machine learning models. The results of the following deep learning models have been studied: AlexNet, ResNet, DenseNet, Inception-v3. Also, the influence of different data augmentation methods on the model performance was investigated. Finally, the Inception-v3 model proved to be the best, which reached the 89% accuracy on the test set.

This repository represent Master Thesis at Faculty of Electrical Engineering and Computing, University of Zagreb. Original task statement can be found near the end of this readme.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

For building this project system needs to have installed the following:

• git
• virtualenv
• python3.6
• pip

Building

Guide for setting up the project.

Clone repository to local machine.

git clone https://github.com/domi385/Histo.git
cd Histo

It is recommended to work in a virtual environment and keep a list of required dependencies in a requirements.txt file.

Commands to setup virtual environment with requirements.

virtualenv -p python3.6 env
source env/bin/activate
pip install -r requirements.txt

In requirements pytorch and torchvision are not listed because of their dependence on system configuration. Please find more details on installation of torch at https://pytorch.org/

python setup.py

If you intend to develop part of this project you should use following command instead of last one.

python setup.py develop

Windows specifics

1. install python 3.6 64 bit with pip (if needed update pip python3 -m pip install --upgrade pip)
2. install and create virtual environment

pip3 install virtualenv
virtualenv -p python3 env

3. Activate environment

\path\to\env\Scripts\activate.bat -- using CMD
\path\to\env\Scripts\activate.ps1 -- using PowerShell

Note: To create a virtualenv under a path with spaces in it on Windows, you’ll need the win32api library installed.

3. Install requirements Install pytorch and torchvision. Detailed instructions are available at https://pytorch.org/

pip3 install https://download.pytorch.org/whl/cu90/torch-1.1.0-cp36-cp36m-win_amd64.whl
pip3 install https://download.pytorch.org/whl/cu90/torchvision-0.3.0-cp36-cp36m-win_amd64.whl

Install other requirements.

pip install -r requirements.txt
python setup.py install

4. Deactivate environment when needed

.\env\Scripts\deactivate.bat

Data

In this project PatchCamelyon dataset was used available at https://github.com/basveeling/pcam
Code for downloading the dataset is available in download_dataset
Current code expects a "data/pcam" directory in project root folder with all data unziped.

Usage examples

For usage examples see examples in histo/examples

Code style standards

In this repository we use numpydoc as a standard for documentation and Flake8 for code sytle. Code style references are Flake8 and PEP8.

Commands to check flake8 compliance for written code

flake8 histo

HR: Klasifikacija histopatoloških snimaka dijelova limfnih čvorova pomoću strojnog učenja

Analiza i generiranje slike spadaju pod tipične probleme računalnog vida i strojnog učenja. S nedavnim razvojem dubokih modela pomaknute su granice izvedivoga i računala često uspijevaju dosenguti ljudsku točnost u analizi slika. Jedan od problema koji su uključeni u ovo područje jest klasifikacija slika. Tema ovog rada je napraviti prototipno programsko rješenje na temelju slika histopatoloških skenova područja limfnih čvorova obavlja binarnu klasifikaciju slika na temelju prisutnosti tkiva s metastazama. U okviru rada potrebno je proučiti i opisati postojeće pristupe ovome problemu, ostvatiri prototipnu implementaciju sustava koji izvodi ovu zadaću te prikazati i ocijeniti dobivene rezultate. Radu priložiti izvorni kod razvijenih postupaka uz potrebna objašnjenja i dokumentaciju. Predložiti pravce budućeg razvoja. Citirati korištenu literaturu i navesti dobivenu pomoć.

Author

• Author: Domagoj Pluščec
• Project represents Master Thesis at Faculty of Electrical Engineering and Computing, University of Zagreb
• Academic year 2018/2019
• Faculty website: https://www.fer.unizg.hr/en/

License

This project is licensed under the MIT License - see the LICENSE file for details

List of References

Detailed list of references can be found in histo/doc