optic diagnosis final project

purpose : naive idea to using high photons Monte Carlo result as target, and low photons Monte Carlo target as train set.

---

Simulation Steps

1. Put the segmented model and position/direction of the probes in the models folder.
   

2. Use the S1_make_the_sim_setting.m to make the setting, including:

   1. how many SDS are there to simulate.
   2. which mus combination to simulation.
   3. choose ground / train mode to match each simulation photon number

   We only need to set the mus to simulate, because the forward is using WMC, so the mua can be any combination.
   For example, the mus for each layer are setting as below:
   
   Then there will be 13X9X4X6=2808 sets of mus combinations. And the program will auto generate these combinations for you.

   %% param
   model_folder='models'; % the folder of the models
   target_name_arr={'ZJ'}; % the name of the model to simulate
   num_SDS=7; % number of detectors
   layer_mus={[50:25:350],[50:37.5:350],[10 19 28 37],[50:60:350]}; % mus for each layer, 1/cm
   mode = ["ground","train"];  % [ground/train]
   num_photon=[1E9,1E8]; % the number of photon
   n=1.4;
   g=0.9;
   sim_version=4.41;

3. Use 'thisPC_sim_wl_index.txt' to set the beginning and endding index for wavelength to simulate by this PC

   % The beginning and endding index for wavelength to simulate by this PC
   1 2808
   

4. Use S2_run_script.m to run the simulation.
   You can set it to run many simulation one-by-one.
   
   If you hany more than one GPU on your computer, you should set the GPU_setting.txt to determine which GPU is used and the load for each GPU.
   

5. The result folder for each subject will containe many [sim_ + index] folders
   
   In each folder is the WMC simulation result of the given mus combination. It records the pathlength in each layer for each detected photon.
   There will also be some files containing the information of the lookup table, e.g., the mus_table.txt containing the mus for each set of combination.
   

6. After the simulation, use S3_exam_the_simed_table.m to check if there is any error in the simulation result.

7. After checking simulation is done, use S4_gen_dataset.m to generate both ANN_ground data and ANN_train data.

8. Run S5_ANN_train.m to train our ANN model.

---

TODO

• Using more mua set to get make model more robustness.