https://github.com/tarolangner/ukb_segmentation

PyTorch implementation for 2.5D U-Net segmentation of UK Biobank neck-to-knee body MRI
https://github.com/tarolangner/ukb_segmentation

deep-learning kidney mri neural-networks parenchyma pytorch segmentation semantic-segmentation u-net uk-biobank

Last synced: 2 months ago
JSON representation

PyTorch implementation for 2.5D U-Net segmentation of UK Biobank neck-to-knee body MRI

• Host: GitHub
• URL: https://github.com/tarolangner/ukb_segmentation
• Owner: tarolangner
• License: mit
• Created: 2020-10-12T09:29:03.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2023-02-10T16:42:07.000Z (almost 2 years ago)
• Last Synced: 2024-08-02T16:45:58.888Z (5 months ago)
• Topics: deep-learning, kidney, mri, neural-networks, parenchyma, pytorch, segmentation, semantic-segmentation, u-net, uk-biobank
• Language: Python
• Homepage:
• Size: 439 KB
• Stars: 18
• Watchers: 4
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-uk-biobank - ukb_segmentation - Net segmentation of UK Biobank neck-to-knee body MRI | (Imaging phenotypes / Abdominal MRI)

README

        
# Neural networks for semantic segmentation of UK Biobank neck-to-knee body MRI

![title](figures/header.png)

This repository contains PyTorch code for cross-validation and inference with neural networks for kidney segmentation on UK Biobank neck-to-knee body MRI, as described in:  

[_"Kidney segmentation in neck-to-knee body MRI of 40,000 UK Biobank participants"_](https://arxiv.org/abs/2006.06996) [1]

The included inference pipeline and trained snapshot enables measurements of left and right parenchymal kidney volumes (excluding cysts and vessels) from these images.

Contents:

- 2.5D U-Net architecture with residual connections (based on [TernausNet, Iglovikov et al. 2018](https://arxiv.org/pdf/1801.05746.pdf))

- Infrastructure for training and *cross-validation*

- Pipeline for *inference* on neck-to-knee body MRI DICOMs

- Code for *quality_controls* based on numerical metrics

- A [_trained snapshot for parenchymal kidney tissue can be found here_](https://uppsala.box.com/s/lan3d807uqz3rhk6vf7o8u5jjnex4jyp)

For any questions and suggestions, feel free to reach out!

# Notes

Access to the underlying image data can only be granted by [the UK Biobank Study](https://www.ukbiobank.ac.uk/register-apply/). Annotations from the quality controls used in our work [1] are available under return data ID 2345 for our application 14237. Our measurements and annotations for the kidneys will eventually be made available as well.

# Citation

If you use this code for any derived work, please consider citing [1] and linking this GitHub.

# References

[1] [_T. Langner, A. Östling, L. Maldonis, A. Karlsson, D. Olmo, D. Lindgren, A. Wallin, L. Lundin, R. Strand, H. Ahlström, J. Kullberg, “Kidney segmentation in neck-to-knee body MRI of 40,000 UK Biobank participants,” Scientific reports 10.1 (2020): 1-10_](https://www.nature.com/articles/s41598-020-77981-4)\