Knowledge-based segmentation of calvarial tumors with automatic parameter screening

This thesis addresses the problem of automatic segmentation of calvarial tumors from Computed Tomography (CT) images and open issues related to validation in the medical image segmentation. The motivation for this work is based on the development of the CRANIO system for computer and robot assisted craniotomy under development at the Chair of Medical Engineering, Helmholtz-Institute for Biomedical Engineering, RWTH Aachen University. Calvarial tumors comprise different tissue types, occupy a wide range of image intensities, have frail borders, and may often exhibit isolated islands of the bone within the soft tissue. Therefore, modeling and segmentation of the calvarial tumors is a challenging task. In the first part of the thesis, commonly used statistical and geometrical validation metrics are presented, followed by feature analysis and definition of requirements for an application oriented figure-of-merit. Finally, a novel statistical metric is proposed and evaluated. Intensity and shape appearance of calvarial tumors is analyzed to obtain information that might be used in a knowledge-guided segmentation algorithm. Different intensity distribution models are presented and statistically compared. The models are integrated in a knowledge-driven segmentation algorithm, based on the level set variational framework. A novel level set speed function, combining image and a priori knowledge terms is proposed. The impact of modeling approaches on the outcome is investigated, followed by a clinical validation study. The major obstacle for using automated segmentation algorithms in medical practice is their incapability to capture the biological and image quality variability. To overcome this problem, segmentation algorithms are guided with an inherent parameter set. In this thesis, a framework for an automatic statistical parameter screening is proposed and validated. Finally, some practical considerations for the integration of the proposed methods in a medical system and guidelines for the further work are given. This work offers three major contributions: a medically-oriented segmentation validation metric, a knowledge-guided extension of the level set segmentation framework, and an automatic parameter screening algorithm.

Grantor dc:publisher

Publikationsserver der RWTH Aachen University

Year dc:date

2007

Author dc:creator

• Popovic, Aleksandra

Contributors dc:contributor

• Aach, Til

• info:eu-repo/classification/ddc/620
• Medizintechnik
• Bildverarbeitung
• Dreidimensionale Bildverarbeitung
• Neurochirurgie
• Ingenieurwissenschaften
• Computer unterstützte Chirurgie
• Roboter unterstützte Chirurgie
• Level Sets
• Wissensbasierte Bildverarbeitung
• medical image processing
• neurosurgery
• computer and robot assisted surgery
• knowledge-based image processing

Statement dc:rights

• info:eu-repo/semantics/openAccess

Language dc:language

eng

Identifier

https://publications.rwth-aachen.de/search?p=id:%22RWTH-CONV-112614%22

OAI record oai:identifier

https://publications.rwth-aachen.de/oai2d?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Apublications.rwth-aachen.de%3A50050