Deep learning based domain adaptation for mitochondria segmentation on EM volumes

This repository contains the code to reproduce the methods described in the publication entitled "Deep learning based domain adaptation for mitochondria segmentation on EM volumes".

The code and detailed instructions of the four different implemented strategies are available in the following links:

• Style-transfer based domain adaptation.
• Self-supervised learning (SSL) based domain adpation.
• DAMT-Net.
• Attention_Y-Net.

Qualitative results of all methods are summarized in the next figure:

Datasets

The EM datasets used are available here:

• Lucchi++
• Kasthuri++
• VNC
• Histogram matched datasets

Citation

This repository is the base of the following work:

@article{franco2022domain,
    title = {Deep learning based domain adaptation for mitochondria segmentation on EM volumes},
    journal = {Computer Methods and Programs in Biomedicine},
    volume = {222},
    pages = {106949},
    year = {2022},
    publisher={Elsevier}
    issn = {0169-2607},
    doi = {https://doi.org/10.1016/j.cmpb.2022.106949},
    url = {https://www.sciencedirect.com/science/article/pii/S0169260722003315},
    author={Franco-Barranco, Daniel and Pastor-Tronch, Julio and Gonz{\'a}lez-Marfil, Aitor and Mu{\~n}oz-Barrutia, Arrate and Arganda-Carreras, Ignacio},
}