University of Washington

Home

About Me

My Mentor

Outside UW

Summer Research

Final Report


Distributed Mentor Project



Final Report





Reconstruction of Blood Vessel Trees from Visible Human Data

Arati Kurani

Abstract:

VHD (Visible Human Data) Reconstruction is a program that was able to successfully reconstruct the blood vessel trees in the right lung of using Visible Human Data – sets of highly detailed cross-sectional images of the human body. Modifications were made, documentation was created, and tests were performed in order to establish the limitations of the program.

Introduction:

VHD Reconstruction is a program that was developed in Java by a previous Masters student, Zhenrong Qian. The goal of this project was to create a system that could effectively recover and reconstruct blood vessel tree structures using Visible Human Data [1]. VHD Reconstruction contributes to the development of the “canonical” human, which can then be linked with the Foundational Model of Anatomy [2]. It also sheds light on interesting topics related to the reconstruction of organ models from medical images including: semi-automation, low-level image processing, and model construction. At the time of creation, this program was able to successfully reconstruct the blood vessel trees in the right lung using data from the Visible Human Male. Unfortunately, at the time, further testing using this program was not completed and the project was set aside when Zhenrong graduated about two years ago. My contributions to the project this summer included: getting Zhenrong’s program up and running, creating a thorough documentation of the program, and testing the program on data from the right and left lung of the Visible Human Male and organs that are similar in shape such as the esophagus.

Contributions:

VHD Reconstruction is a program that lacked sufficient documentation needed in order to run it and understand how it works. After receiving the code and test data, I was able to begin my work. After making some modifications to the code, I was able to get the code to compile and run. What came next was the difficult part: understanding the code by reading through it and testing different parts of the interface. A few weeks later I had a pretty good idea of how VHD Reconstruction worked, and I was then able to confirm my thoughts after meeting with Zhenrong. It was at this point that I began creating a thorough documentation of the program. The documentation includes everything from a list of the programs that need to be downloaded to sample results from testing on the data: documentation.

Experiments:

I began testing on the right lung in order to see if I could duplicate the results Zhenrong had produced when she had performed testing on the right lung nearly two years prior. I found that I was able to successfully recreate and display various artery/vein trees that had previously been created. Testing involves finding an artery/vein (a darkened spot) that you would like to trace and click the center button on the mouse. A pop-up screen asking for the name of this artery/vein will appear. After entering in the name (can be determined by speaking to an anatomist), you can right click on that same artery/vein and click on start. The program at this point will begin to trace the vein and artery. For any single artery/vein, this process can take up to several hours.


Once this phase of testing has been completed, the images/contours need to be resampled and the resampled branch needs to be created. Resampling is used when the axis is not vertical. The axis is tracked through the center points of the contours that were found. A spline curve is then fit to the axis and the data is then resampled perpendicular to the spline curve. These resampled contours are then used for model creation of the artery/vein. This is done by running the set of images containing the found contours through another part of the program.

Once several of these arteries and veins have been created, they can be displayed together as an artery/vein tree. The following are some of the results that I have obtained through testing.

Results:

Results from the Right Lung



Early Results from the Left Lung



The Esophagus



Based on the testing results it is clear that the VHD Reconstruction is not confined to data from the right lung, nor is it restricted to tracing blood vessel trees. Now that we know the program runs successfully on both lungs as well as tubular organs, the next step is trace the complete artery/vein trees in the left lung. After that, testing should be extended to blood vessels and tubular organs in other areas of the body.

Throughout the testing process, I came across a few limitations. If the size of the contour becomes a certain size, it is no longer recognized as a blood vessel. Hopefully, this limitation can be eliminated by changing the values of certain parameters. Also, if the color of the contour becomes too close to that of the background, it is no longer recognized as a blood vessel.

With continued work, this program will become increasingly robust, in which case it can be used for organ modeling, as well as be linked with the Foundational Model of Anatomy.

References:

[1] Visible Human Project: http://www.nlm.nih.gov/research/visible/visible_human.html.

[2] Rosse, C., Shapiro, L.G., and Brinkley, JF. 1998. “The Digital Anatomist Foundational Model: Principles for Defining and Structuring its Concept Domain”; JAMIA Symposium Suppl. 820-824.

[3] Qian, Zhenrong. "Reconstruction of Blood Vessel Trees Using Visible Human Data." Quals Paper, Dept of Computer Science and Engineering, University of Washington, 2001.




Home   •   About Me   •   My Mentor   •   Outside UW   •   Summer Research   •   Final Report