Extraction of Abdominal Topology and Simulation
Over the term of Summer 2019, I collaborated with the Computer Graphics department and surgeons from the Hospital of the University of Pennsylvania to 3D visualize specific internal organs based off of CT scans.
EATS Team Project
Our team was presented with a problem from Penn’s Surgery Department regarding the treatment of hiatal hernias, a condition where the stomach slides up through the hole in the diaphragm where the esophagus passes through. There is a lack of tools to determine the best method of repairing this hole in the diaphragm, as there is high variance in its condition between patients, and treatment methods vary in risk and effectiveness. As such, the Surgery Department wanted a dynamic 3D model of hiatal hernia anatomy made from provided CT scans of the abdomen region in order to aid in surgical planning and repair strategy, allowing for prediction of the effect of certain surgical methods on different kinds of patients.
We took CT scans provided to us from Penn’s Hospital and brought them into different programs to try viewing them, and finding a way to separate specific organs from the rest of the scans -- namely the stomach, esophagus, and diaphragm. Some of us wrote scripts to edit the values of the CT scans to better differentiate certain regions from one another. We ended up doing this with a program called ITK-Snap, which allowed us to trace slices of the CT scans and convert them to a 3D model. We then cleaned up this 3D model in the 3D computer graphics software Maya, then brought it into the 3D animation software Houdini to simulate physics on the stomach, esophagus, and diaphragm.
The following is a video documenting our process, as well as the poster we presented at the Center for Undergraduate Research & Fellowships (CURF) Research Expo.
DICOM File Parser
DICOM files are images saved in a Digital Imaging and Communications in Medicine format, and these can store information from medical scans such as MRI scans or CT scans. I created a Dicom Parser in MATLAB that is able to read in DICOM files that can be viewed with any DICOM viewer software such as OsiriX or ITK-Snap, and it can edit Hounsfield values and write to new DICOM files. Hounsfield values affect how each pixel of the CT scan is visualized on DICOM viewer software, and editing these values can allow for easier segmentation of organs if certain values are isolated.
For example, we may have a CT scan that starts out like this:
Through the MATLAB command line, the user gets prompts and answers the prompts with the appropriate response:
In this example, the code multiplies the Hounsfield values by -1 and thus inverts the colors of the CT scan in the CT scan viewer. The result is:
Here is the link to the GitHub repository for this project.