Dynamic shape detection and analysis of deformable structures in biomedical imaging

The goal of this work is to discuss a three-step-approach to detect, analyze and synthesize a shape given an image, or a sequence of images. Chapter 1 discusses what is a shape. The concept of shape is fuzzy: before delving with more complex topics we need at least to agree on what we call “shape”. Chapter 2 briefly present the biological case studies we used along the work. Chapter 3 deals with the single shape detection problem, the easiest problem one could face: given a single image with a single shape of interest, how do we design an algorithm to detect it? Chapter 4 extends the single shape detection approach to reticular shapes, a kind of shapes common in biological images. Chapter 5 extends the single shape detection approach (and its reticular analogous) to a sequence of images, exploiting the temporal coherence. Chapter 6 analyzes the shape, developing new metrics and measures, while chapter 7 closes the work dealing with the synthesis step. A special section is chapter 8, which covers the Toolbox we developed to detect shapes. The Toolbox is meant to provide functions reusable on a plethora of problems. Conclusions and future works are discussed in chapter 9.