https://github.com/fepegar/vesseg
Brain vessel segmentation using 3D convolutional neural networks
https://github.com/fepegar/vesseg
brain-vessel-segmentation cnn convolutional-neural-networks deep-learning segmentation
Last synced: over 1 year ago
JSON representation
Brain vessel segmentation using 3D convolutional neural networks
- Host: GitHub
- URL: https://github.com/fepegar/vesseg
- Owner: fepegar
- License: mit
- Archived: true
- Created: 2018-11-26T23:36:52.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2022-06-23T17:55:20.000Z (about 4 years ago)
- Last Synced: 2025-02-12T15:18:15.562Z (over 1 year ago)
- Topics: brain-vessel-segmentation, cnn, convolutional-neural-networks, deep-learning, segmentation
- Language: Python
- Homepage:
- Size: 2.46 MB
- Stars: 42
- Watchers: 5
- Forks: 10
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Citation: CITATION.cff
Awesome Lists containing this project
README
# Vesseg
Brain vessel segmentation from digital subtraction angiography (DSA) using a 3D convolutional neural network (CNN).
(Brain parcellation performed using [GIF](https://spiral.imperial.ac.uk/bitstream/10044/1/30755/4/07086081.pdf), not included in this repository).
## Installation
### GPU support
This package uses [NiftyNet](https://github.com/NifTK/NiftyNet), which is built on top of [TensorFlow](https://www.tensorflow.org/), so first of all you need to follow the [instructions to setup your NVIDIA GPU](https://www.tensorflow.org/install/gpu).
While it is possible to perform a segmentation without a GPU, inference of one large volume (512 × 512 × 394 voxels) using the default parameters takes one hour using CPU only and around 90 seconds using a GPU.
### `vesseg` package
Using [`conda`](https://conda.io/docs/) is recommended:
```shell
$ conda create -n "vesseg" python=3.6 -y # tensorflow doesn't support python 3.7 yet
$ conda activate "vesseg"
(vesseg) $ git clone "https://github.com/fepegar/vesseg.git" --depth 1
(vesseg) $ pip install --editable "./vesseg"
```
## Usage
```shell
$ conda activate vesseg
(vesseg) $ vesseg dsa_image.nii.gz dsa_vessels_seg.nii.gz
(vesseg) $ bin2mesh dsa_vessels_seg.nii.gz dsa_vessels_mesh.vtp
```
Tested on Linux and macOS.