DEVELOPMENT OF AN IMAGING SYSTEM FOR THE CHARACTERIZATION OF THE THORACIC AORTA.



Similar documents
4D Magnetic Resonance Analysis. MR 4D Flow. Visualization and Quantification of Aortic Blood Flow

Imaging of Thoracic Endovascular Stent-Grafts

Part-Based Recognition

Clinical Training for Visage 7 Cardiac. Visage 7

State of Stress at Point

Image Segmentation and Registration

Classification of Fingerprints. Sarat C. Dass Department of Statistics & Probability

Essential Mathematics for Computer Graphics fast

We can display an object on a monitor screen in three different computer-model forms: Wireframe model Surface Model Solid model

Computer Graphics. Geometric Modeling. Page 1. Copyright Gotsman, Elber, Barequet, Karni, Sheffer Computer Science - Technion. An Example.

Twelve. Figure 12.1: 3D Curved MPR Viewer Window

2. MATERIALS AND METHODS

APPLYING COMPUTER VISION TECHNIQUES TO TOPOGRAPHIC OBJECTS

Voxar 3D TM. A suite of advanced visualization and analysis software tools

AgoraLink Agora for Life Science Technologies Linköpings Universitet Kurs i Fysiologisk mätteknik Biofluidflöden

Graphing in SAS Software

Manifold Learning Examples PCA, LLE and ISOMAP

Finding skewness and deskewing scanned document

B2.53-R3: COMPUTER GRAPHICS. NOTE: 1. There are TWO PARTS in this Module/Paper. PART ONE contains FOUR questions and PART TWO contains FIVE questions.

Image Analysis for Volumetric Industrial Inspection and Interaction

Poker Vision: Playing Cards and Chips Identification based on Image Processing

Geometric Modelling & Curves

Face detection is a process of localizing and extracting the face region from the

Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 269 Class Project Report

Ultrasound in Vascular Surgery. Torbjørn Dahl

Tutorial on Exploratory Data Analysis

Page 1. Introduction The blood vessels of the body form a closed delivery system that begins and ends at the heart.

Administrative. Patient name Date compare with previous Position markers R-L, upright, supine Technical quality

Morphological analysis on structural MRI for the early diagnosis of neurodegenerative diseases. Marco Aiello On behalf of MAGIC-5 collaboration

VIIRS-CrIS mapping. NWP SAF AAPP VIIRS-CrIS Mapping

Geometry Enduring Understandings Students will understand 1. that all circles are similar.

AN INVESTIGATION INTO THE USEFULNESS OF THE ISOCS MATHEMATICAL EFFICIENCY CALIBRATION FOR LARGE RECTANGULAR 3 x5 x16 NAI DETECTORS

16 Circles and Cylinders

Development of Simulation Tools Software

Template-based Eye and Mouth Detection for 3D Video Conferencing

Automatic Reconstruction of Parametric Building Models from Indoor Point Clouds. CAD/Graphics 2015

MODERN VOXEL BASED DATA AND GEOMETRY ANALYSIS SOFTWARE TOOLS FOR INDUSTRIAL CT

GeoGebra. 10 lessons. Gerrit Stols

Solving Simultaneous Equations and Matrices

Nonlinear Iterative Partial Least Squares Method

A SECURE CLIENT APPLICATION USING RETINAL IMAGE MATCHING

Diagrams and Graphs of Statistical Data

3D Modelling and Volume Rotation of 2D Medical Images

3. Interpolation. Closing the Gaps of Discretization... Beyond Polynomials

ENG4BF3 Medical Image Processing. Image Visualization

Blood Vessel Classification into Arteries and Veins in Retinal Images

RUN-LENGTH ENCODING FOR VOLUMETRIC TEXTURE

Classroom Tips and Techniques: The Student Precalculus Package - Commands and Tutors. Content of the Precalculus Subpackage

Introducing MIPAV. In this chapter...

Pro/ENGINEER Wildfire 4.0 Basic Design

MA 323 Geometric Modelling Course Notes: Day 02 Model Construction Problem

Medical Image Processing on the GPU. Past, Present and Future. Anders Eklund, PhD Virginia Tech Carilion Research Institute

Freehand Sketching. Sections

Visualisatie BMT. Introduction, visualization, visualization pipeline. Arjan Kok Huub van de Wetering

Cone Beam Reconstruction Jiang Hsieh, Ph.D.

RACE I Rapid Assessment by Cardiac Echo. Intensive Care Training Program Radboud University Medical Centre NIjmegen

A Simple Feature Extraction Technique of a Pattern By Hopfield Network

Technical Drawing Specifications Resource A guide to support VCE Visual Communication Design study design

CATIA Wireframe & Surfaces TABLE OF CONTENTS

Copyright 2011 Casa Software Ltd. Centre of Mass

Robust NURBS Surface Fitting from Unorganized 3D Point Clouds for Infrastructure As-Built Modeling

Going Big in Data Dimensionality:

R Graphics Cookbook. Chang O'REILLY. Winston. Tokyo. Beijing Cambridge. Farnham Koln Sebastopol

3D City Modelling from LIDAR Data

Grazing incidence wavefront sensing and verification of X-ray optics performance

South Carolina College- and Career-Ready (SCCCR) Pre-Calculus

VALLIAMMAI ENGNIEERING COLLEGE SRM Nagar, Kattankulathur

Multivariate Normal Distribution

Java Modules for Time Series Analysis

Overview of the Cardiovascular System

Visualizing Data: Scalable Interactivity

096 Professional Readiness Examination (Mathematics)

Avizo Inspect New software for industrial inspection and materials R&D

AN EXPERT SYSTEM TO ANALYZE HOMOGENEITY IN FUEL ELEMENT PLATES FOR RESEARCH REACTORS

2.2 Creaseness operator

Content. Chapter 4 Functions Basic concepts on real functions 62. Credits 11

5: Magnitude 6: Convert to Polar 7: Convert to Rectangular

Lecture 9: Introduction to Pattern Analysis

OBJECT TRACKING USING LOG-POLAR TRANSFORMATION

3D Model of the City Using LiDAR and Visualization of Flood in Three-Dimension

Human Anatomy and Physiology II Laboratory

PHY121 #8 Midterm I

Multi-slice Helical CT Scanning of the Chest

V-PITS : VIDEO BASED PHONOMICROSURGERY INSTRUMENT TRACKING SYSTEM. Ketan Surender

ax 2 by 2 cxy dx ey f 0 The Distance Formula The distance d between two points (x 1, y 1 ) and (x 2, y 2 ) is given by d (x 2 x 1 )

The process components and related data characteristics addressed in this document are:

12.5 Equations of Lines and Planes

Object Recognition and Template Matching

Thinking ahead. Focused on life. REALIZED: GROUNDBREAKING RESOLUTION OF 80 µm VOXEL

Data Exploration and Preprocessing. Data Mining and Text Mining (UIC Politecnico di Milano)

Transcription:

DEVELOPMENT OF AN IMAGING SYSTEM FOR THE CHARACTERIZATION OF THE THORACIC AORTA. Juan Antonio Martínez Mera Centro Singular de Investigación en Tecnoloxías da Información Universidade de Santiago de Compostela citius.usc.es

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

The aorta is the main artery of the body, is the larger vessel throughout the body and is characterized by a thick elastic wall. 1/65

2/65

3/65

4/65

5/65

The thoracic aorta can be affected by different diseases: Dissection Atherosclerosis Aortitis Aneurysm... 6/65

a) b) An aneurysm is local and permanent dilation of a portion of the vessel. This pathologic process affects the wall. 7/65

The aortic aneurysm is the most common pathology. Its incidence is de 6/100.000. Risk of rupture 74 %. Its mortality ratio is 90 %. It is a degenerative disease. 8/65

When there is an aneurysm? The diameter of the aorta should not exceed a maximum diameter of 5.5 cm in the ascending aorta. For the descending aorta (5.5-6.0 cm). The diameters have to be calculated in the normal planes to the wall of the vessel. 9/65

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

To diagnose a patient in the hospital: The patients are scanned with a CT. We obtain several images from inside the patient. 10/65

To check if a patient has an aneurysm a commercial software (General Electric)is used. This comercial software calculates the normal diameters. 11/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 12/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 13/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 14/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 15/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 16/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 17/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 18/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 19/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 20/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 21/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 22/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 23/65

But this comercial software has several drawbacks: The specialist must indicate several points within the vessel. The radiologist uses about 20 minutes to analyze a case. The result may depend on where the radiologist put the initial points 24/65

After, the specialist approves several windows of the comercial program and gets this graph. Finally the specialist have to check if this chart is correct, and remove from the graph the branches of the aorta. 25/65

26/65

The results may be differents in the same clinical case. a b 27/65

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

Aim Develop a system to segment and analyze the thoracic aorta. Without intervention of the specialist. System that provides reproducible results that can not get a specialist. 28/65

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

Method The method is based on the following steps: 29/65

Method We apply the Hough transform to determine the correct position of the thoracic aorta. 30/65

Method The Hough transform was applied at the most caudal slice of the descending aorta. 31/65

Method In this area, a slice s approximate shape is a circle. The Hough transform enables pre-set shapes to be found in an image. Hough transform involves convolving a circular mask with the image given. 32/65

Method Volumetric region-growing technique is used to segment the section Z 1 33/65

This volumetric region growing consists of the use of the 17 pixel neighbors corresponding to the previous and current slices for the calculation of the mean and variance. X Y Z 34/65

The transition from section Z 1 to section Z 2 is characterised by a increase in the size of the segmented region. This region can be approximated by an ellipse in which one of the foci will be used to segment the section Z 2. 35/65

Algorithm based on level set is used to segment the section Z 2 36/65

We use level set algorithm, because the aorta is connected to the heart, which is filled with blood. For this reason, the boundary between the heart and the aorta is fuzzy. 37/65

We applied a pre-processing stage to enhance the border of the image, and next applied a level set algorithm. (a) (b) (c) (d) 38/65

The basic idea underlying level set can be stated as follows: Given a 2D closed curve, its representation by a single 2D function is not possible. A solution consists of using a 3D function intersected by a plane at a given height. 39/65

Thoracic Aorta Comercial method (CHUS) Aim. Method. Results. Conclusions. Method So, we obtain this result of the segmentation on a clinical case with hybrid method of segmentation. 40/65

Method 41/65

Once we have the segmentation of the aorta, we need to calculate the normal plane of the vessel. For this reason, we need to calculate the vessel centerline. 42/65

Method The centerline is extracted using two different techniques: In sections descending aorta and ascending aorta, the centerline is calculated using the center of mass of the segmented region on each slice. CMX = i=n i=0 GL i x i GL i CMY = i=n i=0 GL i y i GL i (1) 43/65

In aortic arch iterative erosion of volumetric image data is applied. This erosion process involves four basic templates. Black symbol are object to be eroded and white symbol is the background. The result to erode corresponds to the centerline of the aortic arch. 44/65

Method 45/65

Method The points calculated in the previous step are interpolated to smooth the centerline. (B-Splines) 46/65

Method 47/65

Method The points of the centerline are used to determine the normal vectors corresponding to the cross-sections at each point on the line. 48/65

Method 49/65

Once normal planes are calculated. Variance-covariance matrix of these planes is calculated too. 50/65

Method The eigenvalues of the matrix are the vectors that indicate the orientation of the principal axes. Thus, to obtain the value of maximum and minimum diameters, we compute the size of segmentation in these directions. 51/65

Method To determine real dimensions of a vessel diameter, we applied a scale factor. The distance between the slices of the acquisitin system used. The different dimensions of the voxel. PCA is calculated in two dimensions and the vectors must be transformed to 3D. 52/65

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

Results The results have been divided into: Segmentation results Diameter results. 53/65

Thoracic Aorta Comercial method (CHUS) Aim. Method. Results. Conclusions. Segmentation Results Manual segmentation was considered the gold standard to compare with automatic method. Manual segmentation was performed by two expert vascular radiologists in clinical cases. Database was made of 32 clinical cases y 14.128 images. 54/65

Thoracic Aorta Comercial method (CHUS) Aim. Method. Results. Conclusions. Segmentation Results Manual segmentation was performed in a graphics tablet equipped with software and a light pen for manually tracing borders. 55/65

Segmentation Results Only 5 % cases in both plots exceeded the confidence interval (± 1.96σ). There were no statistically significant differences. 56/65

The DSC factor was calculated to check the overlapping areas : DSC = 2 A B A B V(A) V(B) DSC error = 2 V(B) (2) (3) DSC=0 indicated that the two areas or volumes did not overlap, DSC=1.0 indicated complete overlap. Expert Asc. Arch Desc. Mean A 96.15 % 94.75 % 94.76 % 95.22 % B 94.54 % 92.88 % 91.82 % 93.08 % 57/65

Pearson coefficient and Intraclass coefficient were calculted to check the segmentation. Pearson coefficient was greater than 0,9 for each section of the aorta by radiologist A and 0,8 for radiologist B. Intraclass coefficient was between 0,8 and 0,85 in all regions of the aorta for each segmentation. 58/65

Diameters Results To check the results of diameters we designed this phantom to test the rotational invariance. 59/65

Results Diameters Phantom images were obtained by CT at different spatial projections (X,Y,Z). No statistically significant differences were found between the three subsets. 60/65

Thoracic Aorta Comercial method (CHUS) Aim. Method. Results. Conclusions. Results Diameters The diameter of the aorta was calculated with the two methods (the proposed and the commercial) in different clinical cases. Bland Altman plot Difference(COMERCIAL AUTOMÁTICO) 5 0 5 10 30 35 40 45 50 55 60 Mean of All No statistically significant differences were found. 61/65

Maximum and minimum diameters were calculated in a normal section. Pictures a)c) were calculated by PCA, and the diameters of the pictures b)c) were calculated with multidirectional analysis. 62/65

PCA method avoids the branches of thoracic aorta. 63/65

Thoracic Aorta Comercial method (CHUS) Aim. Method. Results. Conclusions. Clinical case/ the volumetric reconstruction of segmentation/ graph of diameter. 64/65

1 Thoracic Aorta 2 Comercial method (CHUS) 3 Aim. 4 Method. 5 Results. 6 Conclusions.

Conclusions Alternative to manual segmentation. Highly reproducible, while manual segmentation is not reproducible. The method developed automates the calculation of the diameter of the aorta. Shown the potential of PCA for determination of diameters, and to prevent lateral branches. 65/65

final Se acabo