ChelovekHe / cardiac-segmentation

Right Ventricle Cardiac MRI Segmentation：2D U-net，2D Dilated u-nets, 2D Dilated densenets: multiple scales at once

Cardiac MRI Segmentation

This repository contains code and models to identify the right ventricle in cardiac MRI images. The project is one of the open calls for research from Francois Chollet's AI Open Network.

For the problem description, models and results, please see the blog post here.

Installation

The main code is written as a Python package named `rvseg'. After cloning this repository to your machine, install with:

cd cloned/path
pip install .

You should then be able to use the package in Python:

import matplotlib.pyplot as plt
from rvseg import patient, models

p = patient.PatientData("RVSC-data/TrainingSet/patient01")
# Explore p.images, p.endocardium_masks, etc.

Running models

Scripts for model training and evaluation are located under /scripts/.

python -u scripts/train.py defaults.config

Note: this package is written with the Tensorflow backend in mind -- (batch, height, width, channels) ordered is assumed and is not portable to Theano.