Home

Awesome

LSSANet

This repository includes codes, models, and test results for our paper: "LSSANet: A Long Short Slice-Aware Network for Pulmonary Nodule Detection", MICCAI 2022 (Early Accept). This project is licensed for non-commerical research purpose only.

Method

LSSANet

Results and Models

Method0.1250.250.51.02.04.08.0AvgPre-trained Model
LSSANet51.5951.5958.1866.8877.3385.3589.8768.69model & res

Requirements

The code is built with the following libraries:

Besides, you need to install a custom module for bounding box NMS and overlap calculation.

cd build/box
python setup.py install

Data

Please refer to PN9 for data preparation and then add the data information to single_config.py.

Testing

Run the following scripts to evaluate the model and obtain the results of FROC analysis.

python single_test.py --ckpt='./results/model/model.ckpt' --out_dir='./results/'

Training

This implementation supports multi-gpu, data_parallel training.

Change training configuration and data configuration in single_config.py, especially the path to preprocessed data.

Run the training script:

python single_train.py

Citations

If you are using the code/model/data provided here in a publication, please consider citing:

@inproceedings{DBLP:conf/miccai/XuLDKY22,
author    = {Rui Xu and Yong Luo and Bo Du and Kaiming Kuang and Jiancheng Yang},
editor    = {Linwei Wang and Qi Dou and P. Thomas Fletcher and Stefanie Speidel and Shuo Li},
title     = {LSSANet: {A} Long Short Slice-Aware Network for Pulmonary Nodule Detection},
booktitle = {MICCAI 2022},
series    = {LNCS},
volume    = {13431},
pages     = {664--674},
publisher = {Springer, Cham},
year      = {2022},
doi       = {10.1007/978-3-031-16431-6\_63}
}

Contact

For any questions, please contact: rui.xu AT whu.edu.cn.

Acknowledgment

This code is based on the SANet and NoduleNet.