Statistical Intensity Prior Models with Applications in Multimodal Image Registration [Elektronische Ressource] / Christoph Gütter. Betreuer: Joachim Hornegger

Christoph Guetter

Sujets

• Informatik
• Mathematics
• Gesundheit

Informations

• Publié par: friedrich-alexander-universitat_erlangen-nurnberg
• Publié le: 01 janvier 2011
• Nombre de lectures: 20
• Langue: English
• Poids de l'ouvrage: 18 Mo

Extrait

Statistical Intensity Prior Models with
Applications in Multimodal Image
Registration
Der Technischen Fakultät der
Universität Erlangen–Nürnberg
zur Erlangung des Grades
DOKTOR–INGENIEUR
vorgelegt von
Christoph Gütter
Erlangen — 2011Deutscher Titel:
Statistische Modelle für Intensitäts-basiertes Vorwissen mit
Anwendungen in der multimodalen Bildregistrierung
Als Dissertation genehmigt von der
Technischen Fakultät der
Universität Erlangen-Nürnberg
Tag der Einreichung: 19.07.2010
Tag der Promotion: 28.01.2011
Dekan: Prof. Dr.-Ing. habil. R. German
Berichterstatter: Prof. J. Hornegger
Prof. Dr. D. Cremers
Prof. Dr.med. T. KuwertAcknowledgment
I would like to thank my research advisor, Prof. Joachim Hornegger, for his friendship,
encouragement, and direction during the course of my research. From the very first
moment, he has been an excellent source of ideas and stimulating discussions. I am
very thankful to him for the opportunity to earn my degree in a truly mixed, industrial
and academic, environment. I am very indebted to my research advisor at Siemens
Corporate Research (SCR), Dr. Chenyang Xu, for his invaluable guidance in both
academic and industrial topics that provided me with a manifold research education.
His trust, constant motivation, and insights carried me through the ups and downs of
research and decision making. I truly appreciate his caring supervision and protection
as a supervisor and friend and all the inspiring brainstorming sessions that we had
over the years. I am very grateful to Prof. Daniel Cremers for not only being a
committee member but also for being an instrumental part in this thesis. His research
expertise and enthusiasm greatly inspired me, and I truly appreciate working together.
Our very visual and stimulating discussions that made the most complex problem
seem so approachable are unforgettable. I am also very thankful to Prof. Torsten
Kuwert for providing his medical expertise to this thesis and for being a committee
member.
I am very grateful to Dr. Frank Sauer, leader of the Global Technology Field
Medical Imaging, and Dr. Jens Guehring, Program Manager Interventional Imaging,
for their continued support at SCR in pursuing this research. Among my close
industrial and clinical collaborators, I would like to mention Dr. Hans Vija, Dr. Timor
Kadir, Dr. Jerome Declerck, and Dr. Timm Dickfeld. Our joint projects further
enhanced my knowledge about numerous product-related and clinical topics and this
thesis would be incomplete without the tremendous amount of clinical image data
provided by them. I was privileged to work with a number of great colleagues on
exciting projects supporting my research, including Dr. Christophe Chefd’Hotel, Prof.
Kazunori Okada, Dr. Dieter A. Hahn, Gabriele Wolz, Zhe Fan, Christoph Vetter,
Kalpit Gajera, and Ponraj Chinnadurai.
Ioweatremendousamountofgratitudetomyfriendsandunofficialthesisreviewers:
Ponraj, Klaus J. Kirchberg, and Dr. Timo Kohlberger. Their input and constructive
comments helped shape this thesis to its final form. All of my friends and colleagues
whose support and encouragement have made my experience at the SCR research lab
a memorable one far away from home and family.
Finally I would like to thank my parents Angelika and Stefan Gütter and my sister
Juliane Gütter back home for the eternal support that I have received from them.
Without the sacrifices they have made and the hardships that they went through, I
would have never reached where I am at the moment. I would especially like to thank
my wife Marcela Gütter who with all her patience, support, and love has given me
the strength to achieve this goal. This dissertation is dedicated to them.
Christoph GütterAbstract
Deriving algorithms that automatically align images being acquired from different
sources(multimodalimageregistration)isafundamentalproblemthatisofimportance
to several active research areas in image analysis, computer vision, and medical
imaging. In particular, the accurate estimation of deformations in multimodal image
data perpetually engages researchers while playing an essential role in several clinical
applications that are designed to improve available healthcare. Since the field of
medical image analysis has been rapidly growing for the past two decades, the
abundance of clinical information that is available to medical experts inspires more
automatic processing of medical images.
Registering multimodal image data is a difficult task due to the tremendous
variability of possible image content and diverse object deformations. Motion patterns
in medical imaging mostly originate from cardiac, breathing, or patient motion (i.e.
highly complex motion patterns), and the involved image data may be noisy, furnished
with image reconstruction artifacts, or rendered with occluded image information
resulting from imaged pathologies. A key problem with methods reported in the
literature is that they purely rely on the quality of the available images and have,
therefore, difficulties in reliably finding an accurate alignment when the underlying
multimodal image information is noisy or corrupted.
In this research, we leverage prior knowledge about the intensity distributions
of accurate image alignments for robust and accurate registration of medical image
data. The following contributions to the field of multimodal image registration are
made. First, we developed a prior model called integrated statistical intensity prior
model that incorporates both current image information and prior knowledge. It
shows an increased capture range and robustness on degenerate clinical image data
compared to traditional methods. Second, we developed a generalization of the first
model that allows for modeling all available prior information and greater accuracy
in aligning clinical multimodal image data. The models are formulated in a unifying
Bayesian framework that is embedded in the statistical foundations of information
theoretic similarity measures. Third, we applied the proposed models to two clinical
applications and validated their performance on a database of approximately 100
patient data sets. The validation is performed using a systematic framework and
we further developed a criteria for assessing the quality of non-rigid or deformable
registrations.
The experiments on synthetic and real, clinical images demonstrate the superior
performance, i.e. in terms of robustness and accuracy, of statistical intensity prior
models to traditional registration methods. This suggests that fully automatic mul-
timodal registration (i.e. rigid and non-rigid) is achievable for clinical applications.
Statistical intensity prior models deliver great accuracy from a “relatively small”
amount of prior knowledge when compared to traditional machine learning approaches
that is appealing in both theory and in practice.Übersicht
Die Herleitung von Algorithmen, die automatisch Bilder aus unterschiedlichen Aufnah-
mequellen registrieren können, man spricht auch von Multimodaler Bildregistrierung,
ist ein fundamentales Problem von grosser Bedeutung für Forschungsgebiete in der
Bildanalyse, der Computervision und der medizinischen Bildgebung. Insbesondere,
die genaue Berechung von Deformationen in multimodalen Bilddaten beschäftigt
fortwährend Wissenschaftler und spielt zur gleichen Zeit eine immens wichtige Rolle
in verschiedenen klinischen Anwendungen, die zu einer höheren Qualität des Gesund-
heitswesens beitragen sollen. Da das Gebiet der medizinischen Bildanalyse in den
letzen zwei Jahrzehnten rapide gewachsen ist, verlangt die Fülle der klinischen Informa-
tionen, die den Experten in der Medizin zur Verfügung stehen, nach mehr Automation
der Algorithmen in der medizinischen Bildverarbeitung.
Die Registrierung von multimodalen Bilddaten in der Medizin ist eine komplizierte
Aufgabenstellung aufgrund der hohen Variabilität der möglichen Bildinhalte und der
Mannigfaltigkeit der vorkommenden Objektverformungen. Die Bewegungsmuster in
der medizinischen Bildgebung begründen sich meist in Herz-, Atem-, oder Patien-
tenbewegungen (d.h. sehr komplexe Bewegungsmuster), und die zugrundeliegenden
Bilddaten können verrauscht sein, mit Bildrekonstruktions-Artefakten versehen sein,
oder verdeckte Bildinformation, das aus einigen Krankheitsbildern resultiert, in der
Modalität aufweisen. Ein Kernproblem der Methoden, die in der Fachliteratur vorhan-
den und bekannt sind, ist, dass sie auf die Qualität der vorhandenen Bilder angewiesen
sind. Dadurch kann es schwierig werden, eine zuverlässige und akkurate Registrierung
von verrauschten oder fehlerbehafteten, multimodalen Bilddaten zu erreichen.
In dieser Forschungsarbeit nutzen wir Vorwissen über die Intensitätsverteilun-
gen von exakten vorhergehenden Registrierungen aus, um eine robuste und genaue
Registrierung von multimodalen medizinischen Bilddaten zu erreichen. Die folgen-
den Beiträge zum Gebiet der multimodalen Bildregistrierungen werden von der vor-
liegenden Arbeit gemacht. Erstens, wir haben ein integriertes statistisches Model
für intensitäts-basiertes Vorwissen entwickelt, das Bildinformationen der zu registri-
erenden Bilder und das V miteinander verbindet. Das Mod