DumbICOM

This program can load CT data in DICOM format (e.g., from the Visible-Human Project), and render a 3D visualization from it.

If you really need a DICOM viewer use AlizaMS instead, because this is just a dumb student project and will not be maintained in any way.

Build

The project can be built with CMake.

I tried to use vcpkg and conan, but that way lay madness, so I stopped. C++ dependency management is a bigger dumpster fire than Python 🤡

Install boost-devel, dcmtk-devel, opencv-devel, jsoncpp-devel, and vtk-devel. If on a Debian based distro, replace devel with dev and hope for the best.

In the project folder:

cmake -DCMAKE_BUILD_TYPE=Release -S . -B build/
cmake --build build/

The dumbicom executable file will then be located in the build/ folder.

The following libraries were used:

• Boost v1.78.0
• OpenGL v4.6
• DICOM Toolkit v3.6.7
• OpenCV v4.6.0
• Visualization Toolkit v9.1.0

The version numbers refer to the ones I used initially. Other versions may also work fine.

Usage

Usage and possible parameters can be displayed with the --help flag.

Usage: ./dumbicom [options] <input>

Program options:
    -h [ --help ]          print help message and exit
    -t [ --threshold ] arg threshold for binarization of cleaning mask (default
                           is 250)
    -b [ --brush ] arg     size of brush for cleaning with morphological
                           operations (default is 25)
    -l [ --lower ] arg     comma separated pair of integer numbers "<row,col>", 
                           for defining region of interest
    -u [ --upper ] arg     comma separated pair of integer numbers "<row,col>", 
                           for defining region of interest

The positional <input> argument must be a folder containing DICOM files. Slices need to be in the correct alphabetical order.

The other arguments are optional.

Data Preparation

If the relevant parameters are provided, data preparation may take place. This preparation proceeds as follows:

1. A mask is generated from a copy of the input data.
2. The mask is binarized with --threshold.
3. All values outside a region of interest defined by --lower and --upper are zeroed out.
4. Each layer of the mask is first opened with a disk of radius --brush, then closed and dilated with 2x --brush.
5. The input data is then bitwise ANDed together with the mask.

For the example data in female_head, the parameters --lower 70,120 --upper 452,380 work particularly well.

Interactive Control

The animation is interactive and can be controlled.

• The camera can be controlled via drag-and-drop or with the camera orientation widget in the upper-right corner.
• Transparency can be adjusted using the arrow keys:
  • Up/Down adjust the width of the trasparency window.
  • Left/Right adjust the center of the trasparency window.
• The P key adjusts the projection mode. The following options are available:
  • "Composite" mode
  • "Maximum Intensity" mode
  • "Iso Surface" mode
  • "Additive" mode
• The camera perspective can be reset with R.
• Pressing Q (quit) or closing the window will terminate the program.

Legend

The legend in the lower-left corner displays the following information:

• Patient's name
• Date of birth
• Gender
• Date of acquisition
• Center of the transparency window
• Width of the transparency window
• Current projection mode

Example Data

Example data in DICOM format is provided in the data directory. These were used to test the functionality.

The data in male_head comes from the National Library of Medicine and are part of the aforementioned Visible-Human-Project. The data in female_head also comes from the VHP and the NLM®, but was made available by the University of Iowa.

This project, which mapped the human body in unprecedented detail and provided access to high-quality image data for students like me, was made possible only through body donations. Thank you "Adam" and "Eve".

License

My code (everything in the src folder) is available under the MIT license. See LICENSE for the exact license terms.

All other contents are unaffected by this and subject to their respective license terms.

The sample data is made available under these terms, and provided courtesy of the NLM® and UI Carver College of Medicine. This project is neither affiliated with, nor endorsed by any of these parties