minghaobao / mtmr-net

Code for 《Multi-Task Deep Learning with Margin Ranking Loss for Lung Nodule Analysis》

MTMR-Net

Code for 《Multi-Task Deep Learning with Margin Ranking Loss for Lung Nodule Analysis》 Project.

---

Author	LIU Lihao
E-mail	[email protected]

---

Introduction

This is the pytorch implementation for 《MTMR-Net: Multi-Task Deep Learning with Margin Ranking Loss for Lung Nodule Analysis》

We can output a more robust benign-malignant classification result with persuasive semantic feature scores compared to other CAD techniques which can only output classification results, as shown in the figures.

Requirement

Python 2.7.13

PyTorch == 0.3.0

tensorboardX == 0.9

numpy == 1.14.3

Installation

pytorch: http://pytorch.org/

tensorboardX: https://github.com/lanpa/tensorboard-pytorch

Download and unzip this project: MTMR-net-master.zip.

Download and unzip preprocessed data into "./data/" folder：
https://drive.google.com/open?id=1xFRQBzuQLv4fO5ecsyKnPc5u2D2N9e76

Download resnet50 model into "./logs/middle_result_logs/imagenet/" folder from pytorch website:
https://download.pytorch.org/models/resnet50-19c8e357.pth

Dataset:

1.Original Dataset:

Original LIDC-IDRI dataset can be found in the official website:
https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI

The preprocessing methods can be found in below 2 links:
https://github.com/zhwhong/lidc_nodule_detection
https://github.com/jcausey-astate/NoduleX_code

2.Preprocessed Data:

Find the preprocessed data(2d slices) which can be used directly in the code from Installation section.

Todos

• Modify the args.yaml, add the parameters your deep learning model need under the "running_params" item. Details are shown in another project: https://github.com/CaptainWilliam/Deep-Learning-Model-Saving-Helper
• Pass the running_params (a python dict which contains the running parameters) to you own model.
• The first parameter "is_training" is True for training mode, "is_training" is False for test mode.
• Finish you mode(training or test), and run it.

$ cd MTMR-NET-master
$ python main.py

Acknowledgement

😙Thanks my dearest brother yong for this beautiful figure.