KAT

KAT is a Matlab-based toolbox that implement the Katsevich algorithm for helical CT reconstruction.

Most of the codes in the toolbox are vectorized and so the computations can be transfered from CPU and GPU. In the toolbox, we provide both the CPU and GPU implementations and the switching is controlled by setting the value of parameter 'usegpu' to be 0 or 1. We provide two examples here to demonstrate how to use our toolbox.

Dependencies

Test on Matlab R2023a.
If you use GPU, about 24Gb GPU memory is needed.

Example 1

##Exapmle 1: Reconstruct CT images from the data subset 'L109' of the '2016 NIH-AAPM-Mayo Clinic Low Dose CT Grand Challenge'.

Step 1: Download the dataset of the challenge from https://aapm.app.box.com/s/eaw4jddb53keg1bptavvvd1sf4x3pe9h.

Step 2: Copy the 'L109' directory of the 'Training_Projection_Data' of the challenge dataset to our resposity folder and extract the compressed file 'DICOM-CT-PD_FD.zip' in the 'L109' folder. The file structure should looks like this:

.           
├── example                                          # Example script
├── helical_curve                                    # Source files
├── images                                           # Some reconstructed images
├── L109                                             # Dataset folder
│   ├── DICOM-CT-PD_FD          
│   │   ├── L109_4M_100kv_fulldose1.00001.dcm   
│   │   ├── L109_4M_100kv_fulldose1.00002.dcm  
                             ...
│   │   ├── L109_4M_100kv_fulldose1.29226.dcm 
│   │   ├── L109_4M_100kv_fulldose1.txt    
├── LICENSE
└── README.md

step 3: run the 'rec_L109.m' script in the 'example' folder.

Results. Some reconstructed images look like this:

Example 2

##Exapmle 2: Reconstruct CT images from the data subset 'dcm000' of the 'Truth-Based CT (TrueCT) Reconstruction Challenge'.

Step 1: Download the dataset of the challenge (This may needs to contact the challenge organizer https://www.aapm.org/GrandChallenge/TrueCT/).

Step 2: Copy the 'dcmproj_copd' directory of the challenge dataset to our resposity folder. The file structure should looks like this:

.           
├── example                                          # Example script
├── helical_curve                                    # Source files
├── images                                           # Some reconstructed images
├── dcmproj_copd                                     # Dataset folder
│   ├── dcm_000          
│   │   ├── mAs_vector_9000.bin
│   │   ├── proj_0001.dcm
│   │   ├── proj_0002.dcm
                 ...
│   │   ├── proj_9000.dcm   
├── LICENSE
└── README.md

step 3: run the 'rec_dcm000.m' script in the 'example' folder.

Results. Some reconstructed images look like this: