Multi-View Vertebra Localization and Identification from CT Images

by Han Wu, Jiadong Zhang, Yu Fang, Zhentao Liu, Nizhuan Wang, Zhiming Cui and Dinggang Shen.

arXiv link: https://arxiv.org/abs/2307.12845 paper link: https://link.springer.com/chapter/10.1007/978-3-031-43904-9_14

Introduction

This repository is the reference code for our paper 'Multi-View Vertebra Localization and Identification from CT Images' in MICCAI 2023.

Get Started

This repository is based on PyTorch 1.12.1 + Plastimatch 1.9.4.

Installation

For installation of the Plastimatch, you can refer to Plastimatch.

Training WorkFlow

1. Preprocess the data

preprocess/preprocess.py

2. DRR Generation

preprocess/generate_drr.py
preprocess/generate_drr_heatmap.py

3. Multi-View Contrastive Learning

contrastive_learning/train_multi_view.py

4. Localization/ Identification network training

train/train_localization.py
train/train_id_as_seg.py

Inference WorkFlow

1. Preprocess the data

preprocess/preprocess.py

2. DRR Generation

preprocess/generate_drr.py

3. Single-View Localization & Identification
4. Multi-View Fusion
5. Evaluation

# 3-5 are all in eval_all.py
eavl/eval_all.py

Data Link

Public dataset:

• (VerSe'19 train): https://s3.bonescreen.de/public/VerSe-complete/dataset-verse19training.zip
• (VerSe'19 validation): https://s3.bonescreen.de/public/VerSe-complete/dataset-verse19validation.zip
• (VerSe'19 test): https://s3.bonescreen.de/public/VerSe-complete/dataset-verse19test.zip

Citation

@inproceedings{wu2023multi,
  title={Multi-view vertebra localization and identification from ct images},
  author={Wu, Han and Zhang, Jiadong and Fang, Yu and Liu, Zhentao and Wang, Nizhuan and Cui, Zhiming and Shen, Dinggang},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={136--145},
  year={2023},
  organization={Springer}
}