Low-Dose CT via Transfer Learning from a 2D Trained Network

This repository contains the code for CPCE-3D network introduced in the following paper

3D Convolutional Encoder-Decoder Network for Low-Dose CT via Transfer Learning from a 2D Trained Network (IEEE TMI)

Installation

Make sure you have Python 2.7 installed, then install TensorFlow and Scikit-learn on your system.

Usage

Prepare the training data

In order to start the training process, please prepare your training data in the following form:

• data: N x D x W x H
• label: N x W x H

Here N, D, W, and H are number, depth, width, and height of the input data, respectively. Each label corresponds to the central slice of input data. Then data and label are stored in a hdh5 file.

Pre-trained VGG model

Please also download the pre-trained VGG model from here. Link updated on Jan 23, 2019.

Training network

python main.py

If you want to use the transfer learning from 2D to 3D, please train a 2D model first. The CPCE-3D model here can automatically deal with 2D input and 3D input with various depths (3, 5, 7, and 9), relying on the input size. A simple 2D model CPCE-2D and its shortcut connection version are added for only 2D case.

Citation

If you found this code or our work useful please cite us

@article{shan20183d,
  title={3-D Convolutional Encoder-Decoder Network for Low-Dose CT via Transfer Learning from a 2-D Trained Network},
  author={Shan, Hongming and Zhang, Yi and Yang, Qingsong and Kruger, Uwe and Kalra, Mannudeep K and Sun, Ling and Cong, Wenxiang and Wang, Ge},
  journal={IEEE Transactions on Medical Imaging},
  volume={37},
  number={6},
  pages={1522--1534},
  year={2018},
  publisher={IEEE}
}

Contact

shanh at rpi dot edu

Any discussions, suggestions and questions are welcome!