The greedy fireball: A three-dimensional skeleton algorithm with an application to a human heart fascicle

This dissertation presents a new three-dimensional skeleton algorithm and its application to a human heart fascicle (the basic unit of heart muscle organization). The problem addressed is one of biological shape. The techniques applied are from the computer science area of image processing.; The raw data consists of microscope images of serial sections of a human heart wall. These are used to reconstruct the fascicle in three dimensions. From this reconstructed fascicle, the greedy fireball algorithm calculates a three-dimensional skeleton. This skeleton is used to provide insight into the size and three-dimensional structure of the fascicle.; The image processing concepts are based on the medial axis transform of Blum and the two-dimensional grassfire model of Leymarie and Levine. This work extends the two-dimensional grassfire model to three dimensions. It makes use of the three-dimensional Euclidean distance transform developed by Mulliken. It extends the greedy algorithm concept of Williams and Shah to three dimensions.; The three-dimensional reconstruction and display were done with SurfDriver software developed by Scott Lozanoff of the University of Hawaii and David Moody of the University of Alberta (Moody and Lozanoff, 1999). This project tests the limits of current computing tools available for the Windows environment.; This project successfully demonstrated the three-dimensional skeleton generation on a single fascicle. The technique appears to be promising for use on other subjects. This project sheds some light on the attempt to apply computer science techniques to real biological structures.