imadtoubal / mip-supervision

An implementation of: Tracing in 2D to reduce the annotation effort for 3D deep delineation of linear structures by Kozinsky M. et. al. https://arxiv.org/abs/1811.10508

MIP Supervision

A simple PyTorch implementation of: Koziński M, Mosinska A, Salzmann M, Fua P. Tracing in 2D to reduce the annotation effort for 3D deep delineation of linear structures. Medical Image Analysis. 2020 Feb;60:101590. DOI: 10.1016/j.media.2019.101590.

Install Environement

• Using conda:

  conda env create

• Using pip (TODO)

Dataset

demo.ipynb uses a public segmentation dataset (CHAOS [1]) as an example. It is a good palce to start in this repository.

If you wish to use your own dataset, you can use dataset.SegDataset3D class as follows:

from dataset import SegDataset3D

trainset = SegDataset3D(X, Y)

Where X and Y are numpy arrays of shape (N, W, H, D), with:

• N: number of samples
• W: width of each sample
• H: height of each sample
• D: depth of each sample

Network

The network is a U-Net style of network [2] that uses two max-pooling operations instead of four. The following code shows how to instantiate and use the network:

from model import UNet

in_channels = 1
out_channels = 3
net = UNet(in_channels, out_channels)

In the above example, the input is 1 channel volume, and the output is 3-class (3 channel) volumetric probability map.

Training

TODO

References

[1] A.E. Kavur, M. A. Selver, O. Dicle, M. Barış, N.S. Gezer. CHAOS - Combined (CT-MR) Healthy Abdominal Organ Segmentation Challenge Data (Version v1.03) [Data set]. Apr. 2019. Zenodo. http://doi.org/10.5281/zenodo.3362844

[2] O. Ronneberger, P. Fischer, and T. Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation. In MICCAI, 2015.

[3] Koziński M, Mosinska A, Salzmann M, Fua P. Tracing in 2D to reduce the annotation effort for 3D deep delineation of linear structures. Medical Image Analysis. 2020 Feb;60:101590. DOI: 10.1016/j.media.2019.101590.