Diffuse Optical Imaging with CAFNet

This repository contains the core implementation of the CAFNet model proposed in our manuscript "[Diffuse optical Imaging with CAFNet (channel attention fusion network)]" (JBO), an annotated simulated dataset, experimental phantom dataset as well as requirement file for readers.

---

Table of Contents

• Installation
• Dataset
• Training
• Plotting Reconstructed Optical Properties
• Usage Notes

---

Installation

1. Make sure you have Python 3.7.12 installed.

2. Install the required dependencies:

pip install -r py-37.yaml

Dataset

Ensure your dataset is in the correct folder structure expected by the scripts.

For test samples, the plotting utility uses a sample ID key (e.g., 'id_test_20_2d_exp') to select which data to visualize.

Training

To train the CAFNet model, run:

python helper1516.py

This script will:

1. Load the dataset

2. Save the dataset in the directory 'trainingtata'

For generating plots after training, run:

python CAFNet_Model_Plot.py

Plotting Reconstructed Optical Properties

To visualize the reconstructed optical-property images:

Open the plotting script or a Python session.

Change the key value to your desired test sample ID (Line #323):

python
for i in ['id_test_20_2d_exp']:  # Replace with your test sample ID
    plot_result(i, 'trainingtata', None)

Explanation of columns in the output plot:

Column 1: Ground truth optical properties

Column 2: Reconstructed results using Tikhonov Regularization

'trainingtata' refers to the directory containing the trained model outputs.

Usage Notes

Make sure your dataset paths match the paths expected in the script