# Automated-Cardiac-Segmentation-and-Disease-Diagnosis
## Introduction
This repository contains the reference implementation for automated cardiac segmentation and diasease classification introduced in the following paper: "Fully Convolutional Multi-scale Residual DenseNets for Cardiac Segmentation and Automated Cardiac Diagnosis using Ensemble of Classifiers" https://doi.org/10.1016/j.media.2018.10.004
### Citation
If you find this reference implementation useful in your research, please consider citing:
```
@article{khened2019fully,
title={Fully convolutional multi-scale residual DenseNets for cardiac segmentation and automated cardiac diagnosis using ensemble of classifiers},
author={Khened, Mahendra and Kollerathu, Varghese Alex and Krishnamurthi, Ganapathy},
journal={Medical image analysis},
volume={51},
pages={21--45},
year={2019},
publisher={Elsevier}
}
```
## Usage
### ACDC Data Preparation
1. Register and download ACDC-2017 dataset from https://www.creatis.insa-lyon.fr/Challenge/acdc/index.html
2. Create a folder outside the project with name **ACDC_DataSet** and copy the dataset.
3. From the project folder open file data_preprocess/acdc_data_preparation.py.
4. In the file, set the path to ACDC training dataset is pointed as: ```complete_data_path = '../../ACDC_DataSet/training' ```.
5. Run the script acdc_data_preparation.py.
6. The processed data for training is generated outside the project folder named *processed_acdc_dataset*.
### Steps to train the model:
1. From the project folder open file estimators/train.py and configure the network hyper-parameters.
2. From the project folder open file estimators/config.py and configure the training hyper-parameters.
3. Run the script train.py.
4. Outside the project in the folder named *trained_models/ACDC/* the model weights and tensorboard summary are saved.
5. While training the training summary can be accessed running: ```tensorboard --logdir='path_to/trained_models/ACDC/FCRD_ACDC/summary' ```.
### Steps to test the model:
1. From the project folder open file estimators/test.py and configure the testing hyper-parameters, path to trained model weights and ACDC-2017 testing dataset.
2. Run the script test.py.
3. The predictions are saved in the path *trained_models/ACDC/FCRD_ACDC/predictions_TIMESTAMP*
### Cardiac Diagnosis
1. Extract Features from the training dataset by running: generate_cardiac_features_train.py
2. Extract Features from the testing dataset by running: generate_cardiac_features_test.py
3. Run the scripts stage_1_diagnosis.py and stage_2_diagnosis.py in sequence
4. The final cardiac disease prediction results on the test set are generated in the *prediction* folder
没有合适的资源?快使用搜索试试~ 我知道了~
温馨提示
本项目旨在利用深度学习方法实现心脏分割和心脏疾病的诊断。心脏分割是心脏疾病诊断的重要步骤,而心脏疾病的早期诊断对于患者的治疗和生活质量至关重要。 我们采用深度学习算法,通过分析心脏的医学影像数据,如心脏CT、MRI等,实现对心脏的自动分割。同时,我们利用分割后的数据,建立心脏疾病诊断模型。项目使用的数据集包括公开的心脏医学影像数据集,如Human Connectome Project、ABIDE等,并进行了预处理,包括图像裁剪、大小调整和归一化等。 在运行环境方面,我们使用Python编程语言,基于TensorFlow、PyTorch等深度学习框架进行开发。为了提高计算效率,我们还使用了GPU加速计算。此外,我们还采用了Docker容器技术,确保实验结果的可重复性。 项目完成后,将实现对心脏的快速、准确分割和心脏疾病的早期、准确诊断,为心脏疾病的诊断和治疗提供有力支持。同时,项目成果也可应用于其他医学影像分析领域。
资源推荐
资源详情
资源评论
收起资源包目录
基于深度学习的心脏分割和心脏疾病诊断内含数据集和环境运行说明.zip (29个子文件)
pretrained_models_weight
FCRD_ACDC
latest.ckpt.index 14KB
checkpoint 321B
latest.ckpt.meta 5.51MB
latest.ckpt.data-00000-of-00001 7.48MB
helpers
utils.py 6KB
__init__.py 0B
data_loaders
__init__.py 0B
data_augmentation.py 49KB
hdf5_loader.py 13KB
data_preprocess
__init__.py 0B
acdc_data_preparation.py 25KB
estimators
test_utils.py 32KB
estimator.py 14KB
train.py 3KB
test.py 8KB
config.py 2KB
models
__init__.py 0B
network.py 18KB
network_ops.py 7KB
ACDC_Diagnosis
prediction_data
Cardiac_parameters_prediction.csv 15KB
generate_cardiac_features_train.py 5KB
stage_2_diagnosis.py 14KB
training_data
Cardiac_parameters_training.csv 29KB
Cardiac_parameters_validation.csv 9KB
Cardiac_parameters_train.csv 21KB
utils_heart.py 5KB
generate_cardiac_features_test.py 5KB
stage_1_diagnosis.py 14KB
README.md 3KB
共 29 条
- 1
资源评论
小码蚁.
- 粉丝: 2633
- 资源: 4453
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功