3. Visualization

Pulmonary embolism (PE) can be detected by checking the shape of pulmonary arteries in three-dimensional CTA images, since PE has darker gray colors in CTA images than normal arteries and it makes the irregular shape of pulmonary arteries. Some locations of pulmonary artery containing PE look like squashed blood vessel. Therefore, artery-vein separation and correct boundary detection are important steps to detect PE from three-dimensional CTA images. Intensity-based boundary detection may not be good idea for PE detection, because of the partial volume effect, in which several materials are mixed within a voxel. In this project, we distinguish pulmonary arteries and veins and extract the blood vessel boundaries using 1st and 2nd derivatives as well as intensity values.

Artery-Vein Separation

1st step: Voxel classification and labeling
2nd step: Graph representation and pulmonary trunk detection
Graph representation: Image space has been changed to graph representation. Each box represents blood vessel width and each line shows the connections between boxes. In this step, we detect the pulmonary trunk that has two more main branches. Green represents the pulmonary trunk and blue and red boxes shows the main two branches.
3rd step: Pulmonary artery detection and visualization
Graph representation of the whole trees. The heart and blood vessel boundaries are extracted using 2nd directional derivatives.
4th step: Detecting abnormalities of pulmonary arteries (ongoing).