Medical Bodies

• Taking in circular meshes (medical bodies), getting voltage data from the electrodes, and simulating the medical body back using AI/ML techniques with the voltages. Aka the sig_data_coarse values.

• Combines physical modeling and machine learning to solve an inverse imaging problem.

Results

NOTE: This model as you can see is

Data gathering

1. Run MATLAB
2. Run ML_EIT.m
3. Run e1_t1_e1_save.mat

Then we transform the data into csv format for python:

4. Open terminal and use these two commands to turn the .mat files into .csv files:

cd scripts/
python3 sig_data_mat_To_CSV.py --input data/sig_data_course.mat
python3 voltage_data_mat_to_CSV.py --input data/voltage_data.mat  

5. Then the data is ready for you to train with!

How to run the training/predictions

1. Create a conda environment

conda env create -f environment.yml

2. Change directory into Scripts

cd scripts/

3. Training/Prediction

This will train from the csv inside of the dataset csv files

python3 trainModel.py --input dataset/voltageDataset.csv --output dataset/datasetTrainSigData.csv --model_path data/model1.pth

This will do an evaluation by loading the model and placing it in evaluation mode

python3 predictModel.py --input dataset/voltageDataset.csv --output predictions/sig_data_predict.mat --model model.pth

4. Validate the predicted run and visualize it using the MATLAB code predictedexrun.m file.

Plot the circle mesh (course)

figure;pdemesh(p, e, t)

Plot the circle mesh (fine)

figure;pdemesh(p1, e1, t1)