Three-stage-Curriculum-Training-for-Tumor-Segmentation

0. Introduction

This repository contains Pytorch code for the paper entitled with"A New Curriculum Learning Approach to Deep Network Based Liver Tumor Segmentation" . This paper was initially described in arXiv (https://arxiv.org/abs/1910.07895).

1. Getting Started

Clone the repo: https://github.com/Huiyu-Li/Three-stage-Curriculum-Learning.git

Requirements

python>=3.6
torch>=0.4.0
torchvision
csv
pandas
json
scipy
SimpleITK
medpy
numpy
time
shutil
sys
os

2. Data Prepare

You need to have downloaded at least the LiTS 2017 training dataset. First, you are supposed to make a dataset directory. Second, you may need to preprocess the data by https://github.com/Huiyu-Li/Preprocess-of-CT-data Third, change the file path in the hyperparameters part in the Main.py

3. Usage

To train the model:

• Stage 1:

Step1: split the data into training and valid dataset, respectively. LiTS_TumorNet_without_Source _on_wholeData>split_data.py Step2: Training

##########hyperparameters##########
if_test = False
if_resume = False# changed as True if you have saved model
##########hyperparameters##########

• Stage 2:

Step1: Extract tumor patches form the whole input GetTumorPathes>LiTSGetNegtiveTumorPatches.py and LiTSGetPositiveTumorPatches.py Step2: split the data into training and valid dataset, respectively. LiTS_TumorNet_without_Source_on_tumorPatches>split_datawithNegtive.py Step3: Training

##########hyperparameters##########
if_test = False
if_resume = True
##########hyperparameters##########

• Stage 3:

Just like the Stage 1.

##########hyperparameters##########
if_test = False
if_resume = True
##########hyperparameters##########

To Test and evaluate model:

Step1:

##########hyperparameters##########
if_test = True
if_resume = True
##########hyperparameters##########

Step2: LiTS_Evaluation>evaluator1.py