wgrissom/zebrography

Code for implementation and validation of optical focused ultrasound beam mapping using CW background oriented schlieren imaging (aka zebrography)

CW-BOS imaging of high intensity ultrasonic pressure fields.

In this study, there are three parts to quantitatively map focused-ultrasound (FUS) pressure fields based on CW-BOS imaging.

1. CW-BOS simulations: Perform numerical FUS beam simulations and generate the training data.

• demo_Simulations.m
  • Workflow to generate simulated data, perform SVD on the dictionary, and construct training set.
• wave_prop_simu.m
  • Generate simulated pressure of different levels and transducers. "ablvec.p", "march_asr.p" "precalculate_abl", "precalculate_ad.p" and "precalculate_mas.p" need to be called.
• forward_model.m
  • Calculate the projected pressure and displacements in both two dimensions on the background pattern.
• accum_d.m
  • Calculate displacements of each slice, need to be called in forward_model_dxdz.m

2. CW-BOS acquisition and hardware: Acquire FUS-photo by CW-BOS system.

• testacq
  • Workflow to acquire FUS and non-FUS photos.
  • Need to call "BOSTomoController.py" to communicate with iPad.
• BOSTomoController.py
  • On the experiment computer to switch the background pattern.
• BOSTomoDisplay_app_IPad.py
  • On an APP named Pythonista of the iPad to display background patterns and IP address of iPad.
• ShutterController_Arduino.ino
  • Code to be uploaded to the Arduino board, which allows Arduino to control waveform generator and DLSR camera.
• Modified_camera_shutter_design.zip
  • Modified camera shutter with an analog switch.

3. Reconstruction: Train deep neural network, process the acquired actual photo by CW-BOS system and reconstruct root-mean-square(RMS) projected pressure by pre-trained neural network.

• demo_trainingdata_writer.py
  • Write training data with HDF5 format.
• svd_trainDNN.py
  • Train a multi-layer deep neural network.
• process_photo.m
  • Segment actual photos acquired by DSLR camera in to small patches of rectangular histograms.
• demo_predict.py
  • Reconstruct RMS projected pressure from actual photos.
• "model116.h5" and "model225.h5" are pre-trained model for two transducers (1.16MHz and 2.25MHz).