Implementation of the Keras U-Net for ".mat" file of MATLAB.
Image pattern extraction using U-Net
(Windows) Python 3.5.2 ver.
Keras==2.0.4
tensorflow-gpu==1.6.0
scipy==1.0.0
numpy==1.14.2
matplolib==2.1.1
image augmentation with rotation 90 degree, up-down filp, left-right flip randomly.
![augmentation](https://raw.githubusercontent.com/wlstyql/keras_with_mat_for_unet/master/./images/augmentation.png)
U-Net-based model for pattern extraction
![model](https://raw.githubusercontent.com/wlstyql/keras_with_mat_for_unet/master/./images/model.png)
U-Net: Convolutional Networks for Biomedical Image Segmentation https://arxiv.org/abs/1505.04597
https://github.com/jocicmarko/ultrasound-nerve-segmentation/
A few slices of Input image, Label image, Result image comparison
![result](https://raw.githubusercontent.com/wlstyql/keras_with_mat_for_unet/master/./images/result.png)
(a) Input image (with field inhomogeneity artifact in MRI, out-of-phase angle image)
(b) Label image (with field inhomogeneity artifact removal "SUPER" method in MRI)
"SUPER" method https://synapse.koreamed.org/DOIx.php?id=10.13104/imri.2018.22.1.37
(c) Deep-Learning result image
Water-Fat seperation result
![wf_result](https://raw.githubusercontent.com/wlstyql/keras_with_mat_for_unet/master/./images/wf_result.png)
(a) Before artifact removal
(b) After artifact removal with "SUPER" method
(c) After artifact removal with trained "SUPER" method Deep-Learning
![feature](https://raw.githubusercontent.com/wlstyql/keras_with_mat_for_unet/master/./images/feature.png)