Frequancy-Domain-Optoacoustic-LSTM-FD-PA-LSTM

This is a novel LSTM based deep learning approach for the frequency domain photoacoustic reconstruction using LSTM.

Parameters

Set the path and other parameters in the beginning of the FD-PA LSTM.ipynb file.

Step 1. Generate forward model dataset for training

Generate slices using the three 3D numeric phantoms in the folder Breast Phantom 3D. Then generate the forward model dataset for training using the Generate Data Final.ipynb by setting correct path for source and output. To be done separately for 150 degree FOV and 300 degree FOV. The original source of the 3D phantom is

Step 2. Train the model

Train the model using FD-PA LSTM.ipynb after setting the appropriate parameters. Both the models are trained for data with 40dB SNR in the data and tested with data containg 20dB, 30dB and 40dB SNR.

Step 3. Test the performance

Validate the model using the test data and also compare PSNR and SSIM.

Files for training DL models based on ResNet, UNet, AUTOMAP, and Recon + UNet is also included