In-vivo optoacoustic sparse reconstruction artefact removal with U-Net

This repository contains the code of sparse reconstruction artefact removal with convolutional neural network that was employed in our work:

Requirements

Running the code

Training

Testing

You can downloed a trained model by running the script "download_pretrained_model_32.sh"

sh download_pretrained_model_32.sh

which will add the model to the created directory. Then by running "test.py"

python test.py

you will use the downloded trained model and provided sample test data to test the network. Sample test data includes the network input as artefactual sparse recostruction images, "test_32.mat", and ground truth artefact-free full reconstruction, "test_GT.mat", to be compared with network output for performance evaluation.

Some results from our paper