Computer aided diagnosis methods for coronary CT angiography [Elektronische Ressource] = Computergestützte Diagnoseverfahren für CT-Koronarangiographiedaten / vorgelegt von Matthias Teßmann

Matthias Tessmann

Sujets

• Informatik

Informations

• Publié par: friedrich-alexander-universitat_erlangen-nurnberg
• Publié le: 01 janvier 2011
• Nombre de lectures: 14
• Langue: English
• Poids de l'ouvrage: 8 Mo

Extrait

Computer Aided Diagnosis Methods
for Coronary CT Angiography
Matthias Teßmann
Dissertation - 2011Computer Aided Diagnosis Metho ds
fo r Co rona ry CT Angiography
Computergestützte Diagnoseverfahren
für CT K o rona rangiographiedaten
Der Technischen Fakultät der
Universität Erlangen–Nürnberg
zur Erlangung des Grades
Doktor–Ingenieur
vorgelegt von
Matthias Teßmann
Erlangen — 2011Als Dissertation genehmigt von
der Technischen Fakultät
der Universität Erlangen–Nürnberg
Tag der Einreichung: 21.01.2011
Tag der Promotion: 22.03.2011
Dekan: Prof. Dr.-Ing. Reinhard German
Berichterstatter: Prof. Dr. Günther Greiner
Prof. Dr. Willi A. KalenderRevision 1.00
©2011, Copyright Matthias Teßmann
All Rights Reserved
Alle Rechte vorbehalteni
Abstract
Cardiovascular diseases are the number one cause of death in the world.
As a consequence, cardiovascular diseases are a major health and economic
problem. Any actions taken to support the clinical process during diagnosis,
treatment and aftercare procedures are therefore strongly desirable. Today
medical imaging techniques play a key role for this purpose. Especially car-
diac imaging has high demands on the imaging modality with respect to
spatial and temporal resolution. The image quality that can be acquired by
computed tomography is almost at pace with traditional catheter based an-
giography. However, analysis of the data is a manual and time consuming
process. Hence, a quick evaluation of the images is required in order to pro-
vide optimal patient care. Especially the identification of small structures
like plaques contained in the coronary arteries of the heart is difficult. In
this thesis, methods were examined that approach this problem. The foun-
dation of the presented algorithms is a robust segmentation of the coronary
arteries within the data. Based on this segmentation, methods that operate
on its results have been developed that allow the classification of patholo-
gies along the vessels. A learning-based approach has been implemented
and used to identify diseased regions along the arteries. The resulting al-
gorithm is capable of quickly detecting the location of soft- and calcified
plaques in the data. Besides the detection of the location and the type of
plaques, their quantification is important with respect to risk assessment. A
fully automatic, threshold based segmentation and scoring method for calci-
fied plaque is presented that delivers similar results than those obtained by
manual segmentation from a radiologist. Finally, a snake-based segmenta-
tion algorithm for soft-plaques in CT angiography data has been examined.
This approach generates a boundary hull along the whole vessel and extracts
a radius distribution curve from that data. Thereby, it is possible to detect
and quantify the narrowing of vessel lumen in the presence of a soft-plaque.
Overall, the algorithms presented in this thesis and the software products
that were developed in conjunction with it could contribute significantly to
the provision and improvement of computer aided diagnostic methods for
the analysis of coronary artery disease in CT data.iii
Contents
Abstract i
Contents iii
List of Figures vii
List of T ables xi
A ckno wledgements xv
I Intro duction 1
1 Motivation 3
1.1 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 The Human Hea rt 7
2.1 Anatomy and Function . . . . . . . . . . . . . . . . . . . . . 7
2.1.1 Heart Cycle . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.2 Vascular System . . . . . . . . . . . . . . . . . . . . . 11
2.2 Cardiac Diseases . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Ca rdiac Diagnostics and Imaging 17
3.1 Electrocardiography . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Echocardiography . . . . . . . . . . . . . . . . . . . . . . . . 18iv
3.3 Cardiac Catheterization . . . . . . . . . . . . . . . . . . . . . 20
3.4 Computed Tomography . . . . . . . . . . . . . . . . . . . . . 22
3.4.1 Basic Principle of CT Imaging . . . . . . . . . . . . . 22
3.4.2 Hounsfield Units . . . . . . . . . . . . . . . . . . . . . 26
3.4.3 Cardiac CT . . . . . . . . . . . . . . . . . . . . . . . . 26
I I Computer Aided Diagnosis Metho ds fo r Ca rdiac CT Data 29
4 Ca rdiac CT Prep ro cessing 31
4.1 Heart Isolation . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2 Coronary Tree Segmentation . . . . . . . . . . . . . . . . . . 33
4.3 Automatic Tree Labeling . . . . . . . . . . . . . . . . . . . . 36
4.3.1 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5 Lea rning-Based Stenosis Detection 45
5.1 Inductive Learning and Pattern Classification . . . . . . . . . 45
5.1.1 Decision Stumps . . . . . . . . . . . . . . . . . . . . . 46
5.1.2 Boosting . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Feature Extraction Strategies for Stenosis Representation . . 51
5.2.1 Vascular Sampling . . . . . . . . . . . . . . . . . . . . 52
5.2.2 Multi-Resolution Extension . . . . . . . . . . . . . . 55
5.2.3 Simple Feature Values . . . . . . . . . . . . . . . . . . 57
5.2.4 Haar-like Feature Values . . . . . . . . . . . . . . . . 59
5.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.4.1 Basic Approach . . . . . . . . . . . . . . . . . . . . . 66
5.4.2 Multi-Resolution Extension . . . . . . . . . . . . . . 68v
6 Co rona ry Calcium Detection and Quantification 73
6.1 Clinical Procedures and Previous Work . . . . . . . . . . . . 73
6.2 Automatic Calcium Scoring . . . . . . . . . . . . . . . . . . . 77
6.2.1 HU-Threshold Determination . . . . . . . . . . . . . 78
6.2.2 Detecting Calcifications . . . . . . . . . . . . . . . . . 82
6.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7 Co rona ry Soft-Plaque Detection and Quantification 95
7.1 Active Contour Models . . . . . . . . . . . . . . . . . . . . . 96
7.1.1 Solving the Energy Equation . . . . . . . . . . . . . . 97
7.1.2 Gradient Vector Flow as External Image Energy . . . 99
7.2 Detection of Soft-Plaques . . . . . . . . . . . . . . . . . . . . 101
7.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
I I I Summa ry 115
8 Discussion 117
8.1 Learning-Based Stenosis Detection . . . . . . . . . . . . . . 117
8.2 Calcium-Plaque Detection and Quantification . . . . . . . . 119
8.3 Soft-Plaque Detection and Quantification . . . . . . . . . . . 121
9 Conclusion 125
Bibliography 129
Kurzfassung 141