anisotropic_nets_brats_challenge_model_zoo.md 2.74 KB
Newer Older
1 2 3 4 5 6 7 8 9 10
# Automatic brain tumor segmentation on BRATS with anisotropic nets

This page describes how to acquire and use the whole tumor segmentation network
 as a part of the pipeline described in:

Wang et al., Automatic Brain Tumor Segmentation using
Cascaded Anisotropic Convolutional Neural Networks, MICCAI BRATS 2017

[https://arxiv.org/abs/1709.00382](https://arxiv.org/abs/1709.00382)

Wenqi Li's avatar
Wenqi Li committed
11
* This implementation ranked the first (in terms of averaged Dice score 0.90499) according
12 13
to the online validation leaderboard of [BRATS challenge 2017](https://www.cbica.upenn.edu/BraTS17/lboardValidation.html).*

Wenqi Li's avatar
Wenqi Li committed
14 15 16
* For a full implementation of the method described in this paper with three stages of the cascaded CNNs,
please see: https://github.com/taigw/brats17

17 18
## Downloading model zoo files

Wenqi Li's avatar
Wenqi Li committed
19
The network weights and examples data can be downloaded with the command
20
```bash
Wenqi Li's avatar
Wenqi Li committed
21
net_download anisotropic_nets_brats_challenge_model_zoo
22 23
```

Wenqi Li's avatar
Wenqi Li committed
24 25
(Replace `net_download` with `python net_download.py` if you cloned the NiftyNet repository.)

26 27
## Generating segmentations for example data

Wenqi Li's avatar
Wenqi Li committed
28 29 30 31 32 33 34 35 36 37 38 39 40
Generate segmentations for the included example image with the command,

For the network operates in axial view:
```bash
net_run inference -a anisotropic_nets_brats_challenge.brats_seg_app.BRATSApp \
                  -c ~/niftynet/extensions/anisotropic_nets_brats_challenge/whole_tumor_axial.ini
```
For the network operates in coronal view:
```bash
net_run inference -a anisotropic_nets_brats_challenge.brats_seg_app.BRATSApp \
                  -c ~/niftynet/extensions/anisotropic_nets_brats_challenge/whole_tumor_coronal.ini
```
For the network operates in sagittal view:
41
```bash
Wenqi Li's avatar
Wenqi Li committed
42 43
net_run inference -a anisotropic_nets_brats_challenge.brats_seg_app.BRATSApp \
                  -c ~/niftynet/extensions/anisotropic_nets_brats_challenge/whole_tumor_sagittal.ini
44 45
```

Wenqi Li's avatar
Wenqi Li committed
46
(Replace `net_run` with `python net_run.py` if you cloned the NiftyNet repository.)
47

Wenqi Li's avatar
Wenqi Li committed
48 49 50 51 52 53 54 55
## Generating averaged volume from the outcomes of the previous step

A script has been created to compute the averaged volumes and
 the Dice coefficients from the probabilistic outputs of the segmentation step.

```bash
python ~/niftynet/extensions/anisotropic_nets_brats_challenge/average_volume.py
```
Wenqi Li's avatar
Wenqi Li committed
56 57 58

---

Wenqi Li's avatar
Wenqi Li committed
59
Example data used in this model zoo entry are taken from
Wenqi Li's avatar
Wenqi Li committed
60 61 62 63
[Multimodal Brain Tumor Segmentation Challenge 2017](http://braintumorsegmentation.org/).

Data references:

Wenqi Li's avatar
Wenqi Li committed
64 65
> Menze, Bjoern H., et al.
> "The multimodal brain tumor image segmentation benchmark (BRATS)."
Wenqi Li's avatar
Wenqi Li committed
66 67 68 69 70 71 72
> IEEE transactions on medical imaging 34.10 (2015): 1993-2024.
> DOI: 10.1109/TMI.2014.2377694


> Bakas, S., et al.
> "Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features."
> Nature Scientific Data 4:170117 (2017).
Wenqi Li's avatar
Wenqi Li committed
73
> DOI: 10.1038/sdata.2017.117