This project uses parameter optimization to select the optimal transducer locations to maximize the FSIM image quality metric measured between the ground truth and time-reversal reconstructed image.

• An NVIDIA GPU with CUDA support
• Python 3.11

1. python -m venv .venv
2. source .venv/bin/activate
3. pip install -r requirements.txt

• optimize.py: Runs optimization for the parameters defined in the file.
• run_simulation.py: Plots the reconstruction for the parameters defined in the file.
• get_optimized_parameters.py: Extracts the parameters corresponding to the sample with the highest FSIM from the checkpoint.

• overrides/*.py: Contains various patched files from the k-Wave Python library to allow time-reversal reconstruction to function properly.
• generateP0.py: Generates the branch geometry target for use as the initial pressure mask.
• treeToImage.py: A modified version of a similar function from the GenProcTree library that renders the branches as desired (such as without leaves).
• trials.py: Defines how to perform each reconstruction trial and how to create an elliptical sensor mask.

When running optimize.py, some of the trials will result in errors. This is expected in order to force an early exit for parameters that result in incorrect numbers of transducers being placed.

Fixed transducers are placed at the start and end of the arc. When the number of transducers is odd, a third fixed transducer is placed in the middle of the arc.