Search

Organ segmentation is important in diagnostic medicine to make current decision{support tools more effective and effcient. Performing it automatically can save time and labor. In this paper, a method to perform automatic identification of seed points for the segmentation of organs in three{dimensional (3D) non{annotated, full-body magnetic resonance (MR) and computed tomography (CT) volumes is presented. It uses 3D MR and CT acquisitions along with corresponding organ annotations from the Visual Concept Extraction Challenge in Radiology (VISCERAL) banchmark. A training MR or CT volume is first registered affinely with a carefully{chosen reference volume. The registration transform obtained is then used to warp the annotations accompanying that training volume. The process is repeated for several other training volumes. For each organ of interest, an overlap volume is created by merging the warped training annotations corresponding to it. Next, a 3D probability map for organ location on the reference volume is derived from each overlap volume. The centroid of each probability map is determined and it represents a suitable seed point for segmentation of each organ. Afterwards, the reference volume can be affinely mapped onto any non-annotated volume and the mapping applied to the pre{computed volume containing the centroid and the probability distribution for an organ of interest. Segmentation on the non{annotated volume may then be started using existing region{growing segmentation algorithms with the warped centroid as the seed point and the warped probability distribution as an aid to the stopping criterion. The approach yields very promising results.

Selected readings in vision and graphics, 067, ISBN:978-3-86628-387-9

Selected readings in vision and graphics, 066, ISBN:978-3-86628-374-9

Selected Readings in Vision and Graphics, 63, ISBN:978-3-86628-343-5

Selected readings in vision and graphics, 62, ISBN:978-3-86628-330-5

Selected readings in vision and graphics, 064, ISBN:978-3-86628-352-7

Selected readings in vision and graphics, 54, ISBN:978-3-86628-281-0

ISBN:978-3-86624-476-4

Selected readings in vision and graphics, 043, ISBN:3-86628-114-5

Selected readings in vision and graphics, 042, ISBN:3-86628-112-9

ISBN:3-7281-3065-6, ISBN:1-58603-613-0, ISBN:978-3-7281-3065-5, ISBN:978-1-58603-613-3

Selected readings in vision and graphics, 021, ISBN:3-89649-840-1

We firstly review some approaches to recover the 3D shape of human figures and the related movements. Then we present an improved and reliable version of an easy approach to recover 3D model of humans using just one frame or a monocular video sequence. A simplification of the camera model based on the collinearity condition is required, due to the absence of stereo view. This leads to a perspective or orthographic projection and requires additional information or constraints to solve the problem. The human figure is reconstructed in a skeleton form and visualized with the portable VRML language. Moreover a surface fitting is performed using a predefined 3D model obtained from laser scanner measurements.

Selected readings in vision and graphics, 010, ISBN:3-89649-641-7

Selected readings in vision and graphics, 011, ISBN:3-89649-681-6

CS technical report, 309