Improved Photoacoustic Imaging of Numerical Bone Model Based on Attention Block U-Net Deep Learning Network
Source code for ''Improved Photoacoustic Imaging of Numerical Bone Model Based on Attention Block U-Net Deep Learning Network''.
We designed an Attention Block U-Net (AB U-Net) Network from the standard U-Net by integrating the attention blocks in the feature extraction part, aiming to be more adaptive for imaging bone samples with complex structure.
The attention blocks originated from Convolutional block attention module (CBAM).
U-Net and Attention U-Net models are also contained in this repository.
python main.py
Run test.py
The curves of BCE loss, PSNR and SSIM with iterations are realized by TensorBoard.
# runs tensorboard --logdir runs
AB U-Net successfully removes artifacts and restores the high-frequency information, such as the micro-structure of the trabecular bone. Compared with Time Reversal method, the CNN-based network provides significant improvement in PSNR and SSIM, i.e., SSIM of sample 1 increases from 0.62 to 0.88, indicating an accurate modeling of the initial pressure.
