Desktop version

Home arrow Health

From Anatomy to Computational Anatomy Hidekata Hontani and Yasushi Hirano

Introduction

Today, the dissection of human bodies rarely brings new knowledge of human anatomy but is required for observing the anatomical structures of each patient. One can observe the structures without dissecting the patient body by observing detailed medical images. Imaging plays a very important role in medicine because it enables the observation of the form and structure of the organs specific to each living patient. For accurate assessment, one needs to identify the boundaries of the organs and doctors, e.g., radiologists, identify the organ boundaries in images based on the knowledge of the anatomy of human body, and imaginarily reconstruct the 3D boundaries of the organs in their heads. The imaginary reconstruction of the organs, though, is not useful for obtaining geometric information of the organs and for computer-aided diagnosis (CAD) systems or other clinical applications. One needs to explicitly describe the 3D boundaries of the organs, but it is prohibitively time consuming to describe the boundaries of the organs by manually labeling organ regions in given images.

Computational anatomy (CA) is the study of computational methodologies for medical image analysis, and one of the main purposes of the analysis is to accurately and automatically segment all of the organs included in the images: The goal is to label every voxel in the images with the name of the organ to which the corresponding voxel belongs. For every modality with known spatial resolution, one can generate a set of labels of the organs; one should identify their regions in the images. Let the label be denoted by l, (i = 1,2, ••• , M), and let the set be denoted by L = {f | i = 1,2, ••• , Mg where M denotes the total number of the labels. Given a medical image and the set of the labels, L, it is required to label voxel as the name of the corresponding organ, li. Labeling the organs in this way and identifying the boundaries of the organs require prior knowledge of human anatomy and of image patterns that are needed for accurate segmentation. Computational models of the organs supply the prior knowledge for the segmentation. In the reminder of this section, some important topics fundamental in the medical image segmentation are described.

 
Source
< Prev   CONTENTS   Source   Next >

Related topics