HemoVis
A 2D Interactive Hemodynamic Visualization Tool

Created by Michelle Borkin

Click on above image to launch the visualization in a new window!
(Be patient - the program takes a little while to start-up!)

Note: If you get a Java Plugin error, make sure you have Java installed on your computer.

Overview

HemoVis was created to visualize and aid in the analysis of data from the "Multiscale Hemodynamics" Project in which Lattice-Boltzmann fluid flow simulations are being developed for studying coronary artery disease in humans. More specifically, the geometries and input characteristics for the simulation are all derived from actual patient data. With the large amounts of complex data being produced, having effective visualizations for the doctors to understand the output and answer questions like "where are the most likely places in an artery for disease to form?" are very important. Since doctors are used to viewing only one artery at a time, it is of importance to develop new effective ways of displaying a whole artery tree's worth of information. HemoVis was designed to address these aspects and effectively allow a doctor to navigate the simulation output for a patient's left coronary artery.

Background & Data

The data displayed is the simulation output from the Hemodynamic's code. Specifically, I have the geometry of a patient's left coronary artery derived from high resolution CT scan data (courtesy of the AISL lab at BWH) which has been used in the simulation. The fluid dynamics simulation models blood flow through the artery system and calculates values such as velocity and shear stress at each point (data courtesy of Simone Melchionna). This endothelial (inner wall of the artery) shear stress (ESS) is of great interest to cardiologists since it has been directly linked to sites of lipid deposits. Studies have shown a correlation between these low shear stress areas where the fluid is more stagnant and lipid deposits which can build-up to clog an artery thus causing a heart attack in a patient (e.g. Chatzizisis, et al. 2007).

A current convention used in many medical journal publications is to display a single artery as a cylindrical projection thus no data is occluded (see image below). Our group has recently devised a version of this cylindrical projection where the height of the box at any point along the artery represents the circumference of the artery. This gives the viewer a sense of where the artery is more constricted and can easily see how this geometry effects the shear stress. The data I am directly dealing with for this project's visualization are a series of these pseudo-cylindrical projections, one for each major artery in the system, where the value of each pixel represents the value of the ESS in Pascals (Pa). I also have the lengths of each artery, and the circumference of the artery as a function of its length. I also know the anatomy of the heart thus I am able to label each of the arteries in the system and identify where each one bifurcates. To see an example of what this looks like in 3D, see the image below.

Traditional cylindrical projection representation.

Sample 3D visualization with the shear stress mapped in color on artery "tree".

The Design

HemoVis displays the branched artery system as a tree diagram where each "node" is a 2D representation of the artery. In order to optimize screen real-estate and make text easy to read I chose to go with a landscape workspace. (Note: the convention for displaying artery trees is in portrait mode, but when I presented my sketches to cardiologists they really liked the landscape mode!) In the "tree" mode, a tree diagram of the artery is presented: each artery is labeled with its anatomical name, the color is mapped to the shear stress (ESS) value, and the color and size scales are on the left. The upper right of the screen displays the relevant metadata for the particular data set.

The user is able to view additional simulation data sets (same geometry, but different input flow rates) by selecting the little triangles. This was key for my design since not only do the doctor's care about how all the different arteries relate to each other in space and size, but they also greatly benefit from seeing multiple tree's worth of data at a time. I received enthusiastic responses to my draft versions since most doctors have never looked at the data like this, and no visualization (still or interactive) has been made to compare multiple trees! It should also be noted that the tree layout was checked-over by multiple cardiologists to guarantee that the inherent anatomical information was effectively being encoded and displayed.

A user can choose between the conventional "rainbow" color scheme (the default one since it is used in all the literature, thus I did not want to confuse the users too much!), and my own alternate color scheme. This alternate color scheme I tried to design to better reflect the data: it is diverging data so I made a diverging color scheme (high vs. low shear stress - in particular the doctors only care about ESS below ~2), I made sure to have a strong luminosity component, and I chose a pop-out color (i.e. red) as the "high risk" ESS color indicator. Thus I made a gray-scale based color map where it gets red above a certain critical threshold to make the diseased areas more eye-catching.

In the alternate "individual" mode, only one artery is displayed at a time. This allows the user to take care at studying particular arteries at high resolution. I made sure to keep these images at the same relative scale so that a user can easily tell a large from a small artery. I also made sure to have the program remember the state of the tree and individual modes so that one can switch between them without losing their viewing state. I chose to include a small non-interactive version of the whole tree in the lower right corner to help keep the displayed artery in context as well as using it as the primary way to let the user navigate branches. I also decided, for both modes, to display the shear stress and circumference values near the mouse-over for easy readability as well as on the color/size scales to help put the values in context.

Contact Info