• No results found

Synthetic MRI for visualization of quantitative MRI

N/A
N/A
Protected

Academic year: 2021

Share "Synthetic MRI for visualization of quantitative MRI"

Copied!
90
0
0

Loading.... (view fulltext now)

Full text

(1)

Institutionen för medicin och hälsa

Department of Medical and Health Sciences

Master Thesis

Synthetic MRI for visualization of quantitative MRI

Examensarbete utfört i medicinsk teknik vid Tekniska högskolan i Linköping

av

Erika Peterson

LITH-IMH/RV-A--10/001--SE

Linköping 2008

Department of Medical and Health Sciences Linköpings tekniska högskola Linköpings universitet Linköpings universitet SE-581 83 Linköping, Sweden 581 83 Linköping

(2)
(3)

Synthetic MRI for visualization of quantitative MRI

Examensarbete utfört i medicinsk teknik

vid Tekniska högskolan i Linköping

av

Erika Peterson

LITH-IMH/RV-A--10/001--SE

Supervisor: Marcel Warntjes

CMIV, Linköpings universitet

Examiner: Peter Lundberg

IMH, Linköpings universitet

(4)
(5)

Avdelning, Institution

Division, Department

Division of Medicine and Health

Department of Medical and Health Sciences Linköpings universitet

SE-581 83 Linköping, Sweden

Datum Date 2008-09-04 Språk Language  Svenska/Swedish  Engelska/English   Rapporttyp Report category  Licentiatavhandling  Examensarbete  C-uppsats  D-uppsats  Övrig rapport  

URL för elektronisk version

http://www.imh.liu.se http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-ZZZZ ISBNISRN LITH-IMH/RV-A--10/001--SE

Serietitel och serienummer

Title of series, numbering

ISSN

Titel

Title

Syntetisk MRT som visualisering av kvantitativ MRT Synthetic MRI for visualization of quantitative MRI

Författare

Author

Erika Peterson

Sammanfattning

Abstract

Magnetic resonance imaging (MRI) is an imaging technique that is used in hospitals worldwide. The images are acquired through the use of an MRI scanner and the clinical information is provided through the image contrast, which is based on the magnetic properties in biological tissue. By altering the scanner settings, images with different contrast properties can be obtained. Conventional MRI is a qualitative imaging technique and no absolute measurements are performed. At Center for Medical Imaging and Visualization (CMIV) researchers are developing a new MRI technique named synthetic MRI (SyMRI). SyMRI is based on quantitative measurements of data and absolute values of the magnetic properties of the biological tissue can be obtained.

The purpose of this master thesis has been to take the development of SyMRI a step further by developing and implementing a visualization studio for SyMRI imaging of the human brain. The software, SyMRI Brain Studio, is intended to be used in clinical routine. Input from radiologists was used to evaluate the imaging technique and the software. Additionally, the requirements of the radiologists were converted into technical specifications for the imaging technique and SyMRI Brain Studio. Additionally, validation of the potential in terms of replacing conventional MRI with SyMRI Brain Studio was performed. The work resulted in visualization software that provides a solid formation for the future development of SyMRI Brain Studio into a clinical tool that can be used for validation and research purposes. A list of suggestions for the future developments is also presented. Future clinical evaluation, technical improvements and research are required in order to estimate the potential of SyMRI and to introduce the technique as a generally used clinical tool.

Nyckelord

Keywords magnetic resonance imaging, absolute quantification, synthetic magnetic resonance imaging, visualization

http://urn.kb.se/resolve?urn=urn:nbn:se: liu:diva-102651

(6)
(7)

Abstract

Magnetic resonance imaging (MRI) is an imaging technique that is used in hospi-tals worldwide. The images are acquired through the use of an MRI scanner and the clinical information is provided through the image contrast, which is based on the magnetic properties in biological tissue. By altering the scanner settings, images with different contrast properties can be obtained. Conventional MRI is a qualitative imaging technique and no absolute measurements are performed. At Center for Medical Imaging and Visualization (CMIV) researchers are developing a new MRI technique named synthetic MRI (SyMRI). SyMRI is based on quanti-tative measurements of data and absolute values of the magnetic properties of the biological tissue can be obtained.

The purpose of this master thesis has been to take the development of SyMRI a step further by developing and implementing a visualization studio for SyMRI imaging of the human brain. The software, SyMRI Brain Studio, is intended to be used in clinical routine. Input from radiologists was used to evaluate the imaging technique and the software. Additionally, the requirements of the radiologists were converted into technical specifications for the imaging technique and SyMRI Brain Studio. Additionally, validation of the potential in terms of replacing conventional MRI with SyMRI Brain Studio was performed.

The work resulted in visualization software that provides a solid formation for the future development of SyMRI Brain Studio into a clinical tool that can be used for validation and research purposes. A list of suggestions for the future de-velopments is also presented. Future clinical evaluation, technical improvements and research are required in order to estimate the potential of SyMRI and to introduce the technique as a generally used clinical tool.

(8)
(9)

Acknowledgments

I wish to acknowledge and thank all those people who helped and inspired me throughout the work! A special thanks to:

Marcel Warntjes, my supervisor, for his never-ending optimism, knowledge,

teaching and support.

Janne West, for being an invaluable source of information when it comes to

just about anything concerning computers.

Peter Lundberg, for serving as my examiner and giving me useful tips

con-cerning the written report.

All the people at CMIV, for an encouraging and inspiring research

environ-ment with heaps of passion and nice coffee breaks.

The Radiologists who devoted some of their time on giving valuable input to

this master thesis.

Mum, Dad and Simon, for support and encouraging discussions about what

to do with my life. And thanks to you, Mum, for wise suggestions regarding the scientific work.

Marcus, for not caring too much about what I do, just liking me as I am.

(10)
(11)

Contents

1 Introduction 3

1.1 Problem Description . . . 3

1.1.1 Previous Research . . . 4

1.2 Thesis Objectives . . . 4

1.2.1 SyMRI Brain Studio . . . 4

1.2.2 Radiologist Interaction . . . 5

1.2.3 Validation of the Technique . . . 5

1.2.4 Research Questions . . . 5

1.2.5 Scope . . . 5

1.3 Previous Knowledge . . . 6

1.4 Method . . . 6

1.5 Target Audience . . . 6

1.6 Outline of the Report . . . 6

2 The MRI Scanner 9 2.1 Magnetic Resonance Imaging . . . 9

2.1.1 The net magnetisation, ˆM0 . . . 10

2.1.2 The RF-pulse & its B1-field . . . 11

2.1.3 Spatial dependence - The Gradient Coils . . . 12

2.1.4 Spin Relaxation . . . 13

2.1.5 From Signal Detection to Image Processing . . . 16

3 Conventional Contrast-Weighted Imaging 19 3.1 Contrast Weighted Imaging . . . 19

3.1.1 PD-weight . . . 20

3.1.2 T1-weight . . . 20

3.1.3 T2-weight . . . 21

3.1.4 FLAIR . . . 21

4 Quantitative MRI 23 4.1 The Concept of qMRI . . . 23

4.2 MR Parameters to Quantify . . . 23

4.2.1 Tissue Characterisation . . . 24

4.3 Data Acquisition . . . 25

4.3.1 QRAPMASTER . . . 25 ix

(12)

x Contents

4.4 Fitting the Data to the Mathematical Models . . . 25

4.4.1 Quantification Maps . . . 26

4.4.2 SyMRI BrainStudio Fitting Tool . . . 27

4.5 Synthetic MRI . . . 27

4.5.1 Synthetic Contrast Weighted MRI . . . 27

4.5.2 Tissue Segmentation . . . 28

4.5.3 Normalization . . . 28

5 Methods 29 5.1 SyMRI Brain Studio . . . 29

5.1.1 System Environment . . . 29

5.2 Radiologist Interaction . . . 29

5.2.1 Radiologist Interaction I . . . 30

5.2.2 Radiologist Interaction II . . . 30

5.2.3 Radiologist Interaction III . . . 31

5.2.4 Radiologist Interaction IV . . . 31

5.2.5 Radiologist Interaction V . . . 31

5.3 Validation of the Technique . . . 31

6 SyMRI Brain Studio 33 6.1 Architecture . . . 33

6.2 Main Features . . . 34

6.2.1 Quantification Maps . . . 34

6.2.2 Display of SyMRI Contrast-Weighted Images . . . 36

6.2.3 Graphical User Interface . . . 39

6.2.4 Autoscale . . . 40

6.2.5 Colormap . . . 40

6.2.6 Colorbar . . . 40

6.2.7 The Font . . . 41

6.3 SyMRI Brain Studio Releases . . . 41

6.3.1 v.1.0.0 . . . 41 6.3.2 v.3.1.0 . . . 41 6.3.3 v.3.2.0 . . . 42 6.3.4 v.4.1.0 . . . 42 7 Radiologist 43 7.1 Radiologist Interaction I . . . 43 7.1.1 Drawbacks . . . 43 7.1.2 Advantages SyMRI . . . 46

7.1.3 Future Potential of SyMRI . . . 47

7.2 Radiologist Interaction II . . . 48

7.2.1 The Clinical Routine - an MRI Examination of the Brain. . 49

7.2.2 Interaction with SECTRA IDS 5 . . . 49

7.2.3 Autoscale . . . 49

7.2.4 The SyMRI Software . . . 50

(13)

Contents xi

7.4 Radiologist Interaction IV . . . 50

7.4.1 Follow up - Drawbacks . . . 50

7.5 Radiologist Interaction V . . . 51

8 Validation of the Technique 53 8.1 Required Characteristics . . . 53

8.2 Potential of SyMRI Brain Studio . . . 54

8.2.1 Scan Time . . . 54

8.2.2 The All-in-One Approach . . . 54

8.2.3 Quantitative Measurements . . . 54

8.3 SyMRI Today . . . 55

9 Discussion 57 9.1 SyMRI Brain Studio . . . 57

9.1.1 Design Decisions . . . 58

9.2 Radiologist Interaction . . . 61

9.2.1 Selection of Radiologists . . . 61

9.2.2 Data Sets . . . 61

9.2.3 Lack of Anatomic Detail in SyMRI images . . . 62

9.2.4 MRI Knowledge - MRI Experience . . . 62

9.3 Validation . . . 62

10 Conclusions and Future Research 65 10.1 Research Questions . . . 65

10.2 Future Research . . . 66

10.3 Development Suggestions . . . 67

10.3.1 SyMRI SECTRA PACS IDS5 Plug in & SyMRI Brain Studio 67 Bibliography 69 A Magnetic Resonance 71 A.0.2 Nuclear Spin . . . 71

A.0.3 Magnetic Properties & Energy Levels of Nuclear Spin . . . 72

A.0.4 A Macroscopic Net Magnetization . . . 74

(14)
(15)

List of Figures

2.1 MRI Scanner . . . 9

2.2 Spin-Echo Pulse Sequence . . . 10

2.3 The Net Magnetization, ˆM0 . . . 11

2.4 90◦ pulse . . . 11

2.5 Slice Selecting Gradient . . . 12

2.6 Phase Encoding Gradient . . . 12

2.7 Frequency Encoding Gradient . . . 13

2.8 Spin Relaxation . . . 13 2.9 FID . . . 14 2.10 T1-relaxation curves . . . 15 2.11 T2-relaxation curves . . . 15 2.12 K-space . . . 17 3.1 PDw image . . . 20 3.2 T1w image . . . 20 3.3 T2w image . . . 21 3.4 FLAIR image . . . 22 4.1 qMRI vs MRI . . . 24 4.2 T1, T2 and PD . . . 24

4.3 QRAPMASTER Scanner Sequence . . . 26

4.4 Fit of Data . . . 26

4.5 T1-map . . . 27

6.1 The Visualization Pipeline . . . 34

6.2 Quantification Maps: T1, T2, PD . . . 34

6.3 Quantification Maps: B1, Mean Erros . . . 35

6.4 Region of Interest . . . 35

6.5 R1R2-plot . . . 36

6.6 SyMRI Brain Studio: Default Four Viewport Display . . . 37

6.7 Navigation Window . . . 37 6.8 Fat Suppression . . . 38 6.9 T1Enhanced Image . . . 38 6.10 Popup Menu . . . 39 6.11 Colorbar . . . 41 7.1 Pixel Resolution . . . 46 7.2 Artifacts SyMRI . . . 47

A.1 Nuclear Spin . . . 72

A.2 Nuclear Spin Orientations for H+. . . . 72

A.3 Hydrogen Nuclei in a Magnetic Field . . . 73

(16)

2 Contents

List of Tables

6.1 MR parameter values: Brain Tissue . . . 36 6.2 Default Scanner Parameters . . . 37

(17)

Chapter 1

Introduction

Magnetic Resonance Imaging (MRI) is an important and widely used medical imag-ing technique used in the routine clinical workflow worldwide, particularly for soft tissue visualizations in neurological applications. The technique was first put into clinical use in the 1980s [1] and it has developed rapidly ever since. In 2007, ap-proximately 40 million MRI examinations were performed [2]. An advantage of MRI compared to imaging modalities such as x-ray and computed tomography is the absence of ionizing radiation. Instead, MRI uses magnetic fields and radio frequency pulses to yield diagnostic images. At Center for Medical Imaging Sci-ence and Visualization (CMIV) researchers are developing a new MRI technique named synthetic MRI (SyMRI). Based on the quantification of four magnetic res-onance (MR) parameters SyMRI provides a new approach to MRI, where quanti-tative rather than qualiquanti-tative data is measured. The technique is believed to have a promising future and the long-term goal is to develop and establish SyMRI as a new MRI technique providing faster examinations with improved clinical information [3].

1.1

Problem Description

SyMRI is based on quantitative measurements of data opposed to conventional contrast-weighted MRI where no absolute values are measured. The quantita-tive approach provides additional information through the absolute values given. Moreover, in theory a single quantitative measurement makes it possible to post-synthesize an infinite number of contrast-weighted images that with the conven-tional technique each need a separate scan to be acquired. The approach of quanti-tative MRI (qMRI) has been discussed since the end of the 1990s [4], but has until now been constrained from being generally used in clinical applications. The main reasons for this have been the absence of measurement methods and data process-ing techniques that provide adequate information within a clinical acceptable time frame [3].

(18)

4 Introduction

1.1.1

Previous Research

At CMIV, a scanning technique allowing fast quantification of the four MR pa-rameters longitudinal relaxation (T1), transverse relaxation (T2), proton density

(PD) and the amplitude of the local B1-field has been developed. The scanner

sequence, Quantification of Relaxation times And Proton density by Multi-echo Acquisition of a Saturation recovery using TSE Read-out (QRAPMASTER), al-lows the volume of a head to be examined in about five minutes [3], which is a clinically acceptable time. Hence, QRAPMASTER provides the raw data needed for SyMRI. An additional development in the direction to integrate SyMRI into the clinical workflow is the development of the SyMRI SECTRA IDS 5 Plugin. The plug-in provides a framework for the development of fitting tools and visualization studios for SyMRI as a plug-in to the SECTRA Picture Archiving and Commu-nication System (PACS) workstation IDS 5, used by hospitals worldwide. The plug-in has an implemented cardiac visualization studio, Cardiac Studio, which is under clinical evaluation [5].

1.2

Thesis Objectives

In spite of the promising theories there are persisting challenges that need to be overcome in order to introduce SyMRI as a widespread clinical tool. First of all radiologists have no experience in interpretation based on quantified data and they are familiar with and completely rely on the conventional set of contrast-weighted images. This master thesis aims to take the development of SyMRI further by implementing a visualization studio for SyMRI of the brain and work on the inter-action between the imaging technique and the radiologists. The project is divided into three somewhat integrated parts, the development and implementation of a visualization studio for SyMRI of the brain (SyMRI Brain Studio), the analysis and optimization of the interaction between the radiologists and the technique (Radiologist Interaction), and a validation on how many of today’s conventional MRI examinations that can be replaced with the new technique in the future (Validation of the Technique).

1.2.1

SyMRI Brain Studio

The work with SyMRI Brain Studio includes the continuous development and implementation of the visualization tool SyMRI Brain Studio in SyMRI SEC-TRA IDS 5 Plugin (Visual C++ 6, Microsoft 1995). SyMRI Brain Studio should allow radiologists to compare SyMRI with conventional contrast-weighted MRI but also have features taking advantage of the extended potential of SyMRI. The work and the code should be implemented and documented in a way that enables smooth continuous development of SyMRI Brain Studio into software for clinical use. Brain Studio should include:

• Post-exam synthesis of conventional contrast-weighted images (T1w, T2w,

(19)

1.2 Thesis Objectives 5

• Display of quantification maps.

• A Graphical User Interface (GUI) allowing the user to experience the en-hanced features and possibilities of qMRI.

• Aim to become a visualization studio for routine clinical examinations using SyMRI.

1.2.2

Radiologist Interaction

By working on the interaction between the radiologist and SyMRI, the aim is to increase the understanding on how SyMRI visualizations can be developed in order to optimize their clinical use. This part of the work should:

• Serve as a background and source of information for the continuous devel-opment of SyMRI Brain Studio during the time frame of this master thesis. • Convert demands and requirements from radiologists into technical

require-ments on SyMRI Brain Studio.

• Present suggestions on further developments of SyMRI SECTRA IDS 5 Plug-in and SyMRI BraPlug-in Studio.

1.2.3

Validation of the Technique

The validation phase should result in an estimate on how many of today’s con-ventional MRI examinations that can be replaced by the new technique in the future.

1.2.4

Research Questions

• How can the quantitative data set be introduced and presented in a visual-ization tool in an efficient way?

• What technical improvements are required to satisfy the demands of the radiologist?

• How should the GUI of SyMRI Brain Studio be developed in order to opti-mize the interaction between SyMRI and the radiologist?

• How many of today’s conventional MRI examinations can be replaced with the new technique in the future?

1.2.5

Scope

Factors such as the time frame (30 ECTS credits, 20 full-time weeks) and previous knowledge within the area of research set the limitations of this master thesis. The project does not aim to present a fully developed clinical visualization tool for SyMRI of the brain, but to provide a basis for such a future development.

(20)

6 Introduction

1.3

Previous Knowledge

The author possesses four and a half year of undergraduate studies within the area of engineering biology, including basic knowledge in image processing, computer programming, scientific visualizations and medical imaging. The author has no previous knowledge in the clinical use of MRI, the quantification of MR data or GUI development.

1.4

Method

The thesis work has been performed in an iterative manner by combining literature studies, software development and input from radiologists and researchers. A weekly MRI course as well as a practical introduction to the MRI scanner was followed in order to get an enhanced understanding within the area of MRI. A more detailed method specification can be found in chapter 5.

1.5

Target Audience

The report is written in an attempt to reach an as wide audience as possible although the focus is individuals with technical background and an interest in MRI. No previous knowledge in MRI is required but basic knowledge in imag-ing, anatomy and physics will help the understanding. The abbreviation list in Appendix B of the report is present in order to straighten out possible question marks.

1.6

Outline of the Report

• Introduction

Chapter 1 provides the reader with an introduction to this master thesis. The chapter is divided into a number of sections including problem description, thesis objectives, previous knowledge of the author, method used, target audience and outline of the report.

• Background

Chapters 2, 3 and 4 will together accommodate the reader with the technical background needed to understand the work of this master thesis. Chapter 2 describes how the properties of the phenomenon forming the basis of MRI, the magnetic resonance, is used to create medical images using an MRI scanner. Chapter 3 explains how MRI is used to yield conventional contrast-weighted MRI images with today’s conventional MRI techniques. Chapter 4 explains the concept of qMRI, which forms the basis of SyMRI through the quantitative data acquisition and the fitting of the data to mathemat-ical models. Furthermore, the chapter describes the approach of SyMRI visualizations.

(21)

1.6 Outline of the Report 7

• Methods

Chapter 1 gives a short overview of the general method used, while chapter 4 provides a more detailed description of the methods used in the different parts of the work.

• Result

Chapter 6, 7 and 8 explain the different steps and outcomes of the work. Chapter 6 describes the features and development of SyMRI Brain Studio. Chapter 7 give details about the radiologist interactions and their outcome, while chapter 8 contains the validation of the technique and its potential to replace conventional contrast-weighted MRI in the clinical routine.

• Discussion

Chapter 9 contains a discussion regarding the work and outcome of the different parts of this master thesis.

• Conclusions and Future Development

In chapter 10 the research questions are answered and future work within the area is proposed together with future development suggestions for the SyMRI Sectra IDS 5 plugin and SyMRI Brain Studio.

(22)
(23)

Chapter 2

The MRI Scanner

The chapter introduces magnetic resonance imaging (MRI). In the end of the chap-ter, the reader should have a basic understanding on how MR enables the genera-tion of medical contrast weighted images through the applicagenera-tion of magnetic fields and radio frequency pulses (RF-pulses) to biological tissue.

2.1

Magnetic Resonance Imaging

Figure 2.1. MRI scanner.

A schematic picture of a MRI scanner is shown in Fig 2.1. During the clinical examination, the patient is lying on the patient table inside the bore of the mag-net. The magnet supplies a large homogeneous and static magnetic B0-field, which

causes a magnetization ˆM0 of the patient volume. Through a scanner pulse

se-quence MRI images are given. The pulse sese-quence contains hardware instructions causing the components of the MRI scanner to stimulate magnetic interaction in a pre-defined manner in the patient volume (Fig 2.2). The first line in the pulse sequence displayed in Fig 2.2 describes the pulse (Section 2.1.2). The RF-pulse tips ˆM0 into a plane where it can be processed and measured. The three

following rows describe the behaviour of the three magnetic field gradients used to 9

(24)

10 The MRI Scanner

distinguish different spatial locations within the patient volume (Section 2.1.3). The last section of the pulse sequence contains the data acquisition, where the analogue digital converter (ADC) converts the continuous analogue signal to digi-tal sample points. The pulse sequence is then repeated for each required data line by altering the phase encoding gradient at each repetition. The period between two repetitions is called repetition time (TR). The design of pulse sequences is an entire research field alone aiming to get the optimal image coverage in shortest time possible.

Figure 2.2. Spin-echo pulse sequence illustrating the hardware instructions of the MRI

scanner.

2.1.1

The net magnetisation, ˆ

M

0

The net magnetization of the patient volume, Mˆ0 (Fig 2.3), is predominantly

formed through the magnetic properties of hydrogen nuclei in biological tissue. In presence of the external magnetic field ˆB0, the hydrogen nuclei, which consist

of a single proton, will precess around ˆB0 with a certain precessional frequency,

the Larmor frequency (fL). Nuclei precessing with the same fL are said to be in

magnetic resonance and are referred to as an isochromat. Inside ˆB0, a thermal

equilibrium of the isochromat is reached, giving an excess number of nuclei that precess in the positive direction of ˆB0. The excess of spins cause the net

mag-netisation ˆM0 to appear. Mˆ0 precesses around ˆB0, but is often illustrated as a

net vector in a rotating frame (Fig 2.3). A deeper introduction to the quantum physics underlying MRI is available in Appendix A.

(25)

2.1 Magnetic Resonance Imaging 11

Figure 2.3. In presence of a magnetic field B0a resulting net magnetization will appear due to an excess number of spins in the direction of ˆB0.

2.1.2

The RF-pulse & its B

1

-field

The RF-pulse induced by the RF-coil will apply an oscillating magnetic field, ˆB1,

in the xy-plane orthogonal to ˆB0. Net magnetization caused by isochromats in

res-onance with the rotational frequency of ˆB1will then experience a static magnetic

field causing the magnetization to additionally precess around ˆB1. In the rotating

frame a tipping of ˆM0 can be seen proceeding between two states, ˆM0 and − ˆM0.

By choosing the bandwidth of the RF-pulse, it can be determined which isochro-mats in an inhomogeneous magnetic field that will experience a static B1-field.

Figure 2.4. 90◦pulse in the rotating frame of reference.

The duration of the RF-pulse determines the tip-angle α according to α =γB1t .

An RF-pulse that will tip the net magnetization into the xy-plane is referred to as a 90◦pulse (Fig 2.4). The RF-pulse will cause the spins to be in phase coherence pointing in the same direction in the xy-plane [1]. However, immediately after the RF-pulse different relaxation mechanisms will cause the magnetization to relax back to its thermal equilibrium M0. The relaxation mechanisms are crucial for

(26)

12 The MRI Scanner

2.1.3

Spatial dependence - The Gradient Coils

The three orthogonal gradient coils inside the MRI scanner are applying magnetic field gradients across the patient volume in order to make the magnetization spatial dependent according to:

ˆ

Bi= ˆB0+ ˆGT × ˆri (2.1)

ˆ

GT is the summation of gradients at a location ˆri [6]. The application of

gradients will create isochromats throughout the patient volume with distinguished resonance frequencies:

fi= γ(B0+ ˆGT × ˆri) (2.2)

Slice Selecting Gradient

Figure 2.5. Slice Selecting Gradient.

The slice selecting gradient (Gss) is introduced at the time of the RF-pulse in

order to make the RF-pulse slice selective (Fig 2.5). The thickness of the slice is proportional to the RF-pulse bandwidth and inversely related Gss, SLICEwidth=

RFbandwidth

γGss [1]. The direction of the slice is determined by the direction of the gradient. After Gssa rephasing gradient is needed in order to realign the spins due

to the dephasing caused by the excitation. In multi-slice imaging where a number of slices are used to visualize the patient volume Gss is normally kept constant

while alternating the frequency of the RF-pulse.

Phase Encoding Gradient

Figure 2.6. Phase Encoding Gradient.

The phase encoding gradient (Gpe) gives spatial phase variation and is applied

(27)

2.1 Magnetic Resonance Imaging 13

sample dephase until Gpe is turned off after which they return to their original

frequency keeping their phase angle. The phase differences cause phase encod-ing in one of the in-plane directions of the slice. The phase differences remain until another gradient is applied or the MR-signal decays due to T2-relaxation

(Section 2.1.4)[1].

Frequency Encoding Gradient

Figure 2.7. Frequency Encoding Gradient.

The frequency encoding gradient or read out gradient, Gro, will add or

sub-tract from the magnetization along the second dimension within the image slice (Fig 2.7). The gradient is applied during the data acquisition making the signal consisting of a number of different frequencies corresponding to different locations within the second in-plane axis of the slice.

2.1.4

Spin Relaxation

Figure 2.8. Spin relaxation back to thermal equilibrium.

Spin relaxation is the process following an RF-pulse when the isochromats re-lease and redistribute absorbed energy as they go back to their thermal equilibrium state ˆM0 (Fig. 2.8). Two different relaxations can be measured, the longitudinal

relaxation (T1-relaxation) and the transverse relaxation (T2-relaxation). The

re-laxations are dependent on two distinguished relaxation mechanisms, those who transfer energy away from the spins to the lattice and those who redistribute en-ergy within the spin system itself [4]. The relaxation is dependent on a number of

(28)

14 The MRI Scanner

tissue specific properties such as intra- and intermolecular interactions. Together with the proton density, the spin relaxation will form the basis of image contrast in MRI.

Bloch Equations

The behaviour of the magnetization during the excitation through RF-pulses and the following relaxation has been modelled in a set of differential equations by Bloch [6]. Bloch’s equations are solely based on classical mechanics and are con-sidering the net magnetization. The Bloch equations and their solutions are shown in the equations below.

d ¯M dt = γ ¯M × ¯B = γ   (MyBz− MzBy)i (MzBx− MxBz)j (MxBy− MyBx)k   (2.3) Mz(t) = [Mz(o) − Mo]et/T1+ M0

Mx(t) = [Mx(0)cos(ωot) + My(0)sin(ωot)]et/T2

My(t) = [Mx(0)sin(ωot) − My(0)cos(ωot)]et/T2

Mxy =

q

M2

x+ My2

After a 90◦ RF-pulse, the solutions to the Bloch equations models Mxy as

a decaying signal oscillating with the Larmor frequency, while Mz exponentially

grows back to M0. The precessing magnetic field Mxywill induce a voltage in the

RF-coil and the signal, a free induction decay (FID) can be modelled according to

S(t) = S0e−t/T2eiωLt (Fig. 2.9). In MRI, either the FID or one or several echoes

created from the FID are measured.

(29)

2.1 Magnetic Resonance Imaging 15

Figure 2.10. The T1-relaxation curve of two tissues with different T1.

T1-relaxation

The exponential recovery of Mz back to its thermal equilibrium state M0 after a

RF-pulse is called longitudinal relaxation, spin-lattice relaxation or T1-relaxation.

The T1-value (T1) serves as an absolute measurement of the T1-relaxation and

is the time when Mz has recovered to 63% of |M0 − Mz0| (Fig 2.10). Apart

from molecular dependencies, T1 is dependent on the scanner field strength and

temperature [1]. The inverted T1 is often referred to as T1-relaxation rate (R1).

T1-relaxation is induced as the isochromats release their energy to the

surround-ing lattice and is strongly correlated with molecular motion. Molecular motion in surrounding molecules will cause local magnetic fields tumbling with different frequencies. Tumbling with fL perpendicular to B0will due to the resonance

con-dition induce energy transfer from the spin system to its surroundings.

T2-relaxation

Figure 2.11. The T2-relaxation curve of two tissues with different T2.

The transverse relaxation, spin-spin relaxation or T2-relaxation is present in the

xy-plane after the RF-pulse. The T2-value (T2) serves as an absolute measurement

of the T2-relaxation and is the time when Mxy has decreased to 37% of the value

immediately after a RF-pulse (Fig 2.11). T2 is dependent on the scanner field

strength as well as temperature. The inverted T2is referred to as the T2-relaxation

rate (R2) [1]. T2-relaxation is affected by relaxation mechanisms redistributing

energy within the spin system as well as the energy transfer from the spins to the lattice. Hence, T2is always shorter than or equal to T1. The T2-relaxation caused

(30)

16 The MRI Scanner

lose the phase coherence gained through the RF-pulse. The relaxation is induced by fluctations in the magnetic field experienced by the isochromats, making the precess with slightly different frequencies.

2.1.5

From Signal Detection to Image Processing

The pulse sequence makes it possible to distinguish the signal from different spatial locations within the patient volume. The signal measured and sampled are often one or several echoes from the appearing FID. The time between the exciting RF-pulse and the created echo are called echo time (TE). Echoes are created in two major ways using spin echo (SE) or gradient echo (GE) techniques. Both techniques create an echo of the signal in the transverse xy-plane that the receiving coil detects by measuring the induced voltage (Eq. 2.4).

emf = −d

dt[

I

Sample

MtotBrec] (2.4)

emf = electro motive force, Brec= receiving coil sensitivity

A schematic image of a SE sequence is shown in Fig. 2.2. The SE is formed using a refocusing 180◦ pulse that flips the magnetization around the y-axis at time T E2 after the initial RF-pulse. The pulse will refocus the transverse spins that have dephased due to T2-relaxation. The spins will now come back into

phase coherence creating an echo at time T E according to SSE = S0e−T E/T2.

In GE sequences, a negative gradient lobe is used to form the echo. Turbo spin echo (TSE) is an approach where several refocusing pulses are applied during each repetition in order to form several echoes from a single excitation pulse.

Analog to Digital Converter

The analogue-to-digital converter (ADC) digitalizes the emf and stores it in nu-meric form in a computer. The raw data space where the data storage is done is called k-space.

K-space

K-space is a two or three dimensional data space used in MRI, with one frequency encoding and one phase encoding direction. K-space is often looked upon as a trajectory path for the phase encoding and frequency encoding gradients. The rows in k-space are collected throughout the repetitions of the pulse sequence as the gradients are changed. The middle of k-space collects low frequencies containing SNR and contrast and the outer rows will collect high frequencies containing edges, boundaries and image resolution [1]. The number of sample points in the frequency encoding direction of k-space will determine the number of sample points collected during each echo. The number of sample points in the phase encoding direction will determine how many times the sequence has to be repeated to fill all the lines in k-space. The number of data points sampled and the image resolution

(31)

2.1 Magnetic Resonance Imaging 17

determines the field of view (FoV). The reconstruction from k-space to a spatial image is made through a 2D Fourier transform (Fig 2.12).

Figure 2.12. An MRI image and corresponding k-space.

Sensitivity encoding (SENSE) is a technique giving shorter acquisition times by not acquiring data to all lines in k-space, instead several receiving coils are used. This will cause aliasing that appears when signals are superimposed on each other due to a too large FoV. Through knowledge of the sensitivity of the receiving coil elements and a coil sensitivity that varies in space, a defined equation system can be generated (Eq. 2.5). The system can be used to unwrap the pixels and a complete image can be restored.

mi= s1ip1 + s2ip2 (2.5)

miis the measured signal of a coil, s is the sensitivity of the receiving coil and p is the signal at each pixel.Another technique to achieve faster image acquisition

is echo planar imaging (EPI). With EPI a number of lines are collected in k-space during each repetition using gradients.

(32)
(33)

Chapter 3

Conventional

Contrast-Weighted Imaging

The chapter will give the reader an insight on how the conventional set of contrast-weighted MRI images is achieved and used is in today’s clinic.

3.1

Contrast Weighted Imaging

In conventional contrast-weighted MRI, tissue can be distinguished through dif-ferences in image pixel intensity. The image contrast is caused by difdif-ferences in signal amplitude at different locations of the patient volume. The signal ampli-tude is dependent on proton density (PD), T1-relaxation and T2-relaxation and a

number of scanner parameters. The scanner parameters TE and TR are used to create images with fixed contrast behaviour. TE and TR can be varied to produce images whose contrast is mainly dependent on T1-relaxation, T2-relaxation, PD or

a combination of these. The images are said to have a certain T1-weight (T1w),

T2-weight (T2w) and PD-weight (PDw). TE determines the amount of T2w. With

a relatively long TE the signal amplitude will be dependent on the T2-relaxation.

Tissue compartments with short T2 will dephase faster than tissue compartments

with long T2 and hence have a more attenuated signal at the formed echo. With

short TE, very limited T2w is present. TR is the time between the applications of

the excitation RF-pulses and determines the amount of T1w in the image. With

a long TR, all tissue compartments will have time to relax back to their thermal equilibrium M0before the application of an additional RF-pulse. Contrarily, with

a short TR only tissue compartments with short T1 will have time to recover

be-fore a new RF-pulse is applied. Hence, the signal from tissue compartments with a long T1will be attenuated.

An additional not yet mentioned scanner parameter is the inversion time (TI). TI is present in the case of an inversion recovery sequence, with an inversion prepulse at time TI before the exciting RF-pulse. The inversion pre-pulse can be used to cancel out signals from a certain tissue. The inversion pre-pulse is an 180◦-pulse

(34)

20 Conventional Contrast-Weighted Imaging

that initially inverts the magnetization M0. The application of the RF-pulse at

time TI will cancel out signals from tissue passing the zero-signal line at that time.

3.1.1

PD-weight

(a) (b)

Figure 3.1. A PDw image is given using long TR and short TE.

P Dw is achieved using long TR and short TE, which keep the T1w and T2w

to a minimum (Fig 3.1.1). P Dw-images are not normally a part of general MR protocols of the brain. PD is strongly correlated with water content, which is rather similar in brain tissue, therefore P Dw-images do not yield very good image contrast for brain applications.

3.1.2

T

1

-weight

(a) T1w Image (b) NavigationT1w

(35)

3.1 Contrast Weighted Imaging 21

T1w-images are often referred to as anatomical scans since the T1-relaxation in

different brain tissues are varying yielding good contrast between different tissues in the brain (Fig 3.1.2). In T1w-images the boundaries between tissues can be

seen clearly. Fluids have almost no signal intensity due to the long T1-values

while water based tissues appear mid-grey and fat based tissue very bright due to the short T1. The signal from PD and T1-relaxation counteracts each other in

T1w-images, since grey matter has higher PD but longer T1 than white matter.

3.1.3

T

2

-weight

(a) T2w Image (b) Navigation T2

Figure 3.3. A T2-weighted image is given using long TR and long TE.

Fluids get the highest pixel intensity in T2w-images due to long T2-relaxation

while water- and fat-based tissues appear mid-grey. In brain imaging T2w images

are often referred to as ’pathology scans’ since many pathological processes in the brain result in an increased water content, making them easy to spot in T2w

images as the image intensity is increased.

3.1.4

FLAIR

FLAIR images (Fig 3.4) are achieved by applying an inversion pre-pulse to a T2

w-image in order to cancel out the signal from cerebrospinal fluid (CSF). FLAIR images show brain interfaces very good and make it possible to distinguish cere-brospinal fluid (CSF) from other tissues or diseases that appear bright in T2

w-images [8].

The Image - Pixel by Pixel

An image stack used in MRI consists of a number of images representing the slices collected throughout the patient volume. The FoV and the number of sample points measured in the frequency encoding direction will together with the number of repetitions determine the image resolution. Each pixel in the image represents the signal from a voxel of the patient volume. Partial volume is a phenomenon

(36)

22 Conventional Contrast-Weighted Imaging

Figure 3.4. In the FLAIR image the signal from CSF is cancelled out.

that arises when a voxel contains a number of tissue compartments with different relaxation rates.

(37)

Chapter 4

Quantitative MRI

The chapter explains the concept of qMRI throughout the process of data acquisi-tion, fitting of the data and post processing through the rendering of SyMRI images and the display of quantification maps.

4.1

The Concept of qMRI

While conventional contrast-weighted MRI is a qualitative technique where the intensity of a pixel in the final image have no absolute value, qMRI is a quantitative approach taking MRI into the area of measurement science [4]. In qMRI the quantitative and absolute values of the different MR parameters are measured for each voxel within the slices in the patient volume. The measured values can later be used for different kind of data analyses and image rendering. To illustrate the difference between the two techniques, conventional MRI can be described as a snap-shot imaging technique where a scanner sequence with predefined settings sample signals at time-points were good contrast between clinical important tissues are given. In the quantitative approach, the scanner sequence is used to collect data points in order to estimate the behaviour of the complete relaxation within the tissue and measure T1, T2 and PD (Fig 4.1). QMRI opens the possibility for

reproducible and comparable measurements allowing reliable multi-centre studies and studies of disease progress and response to treatment in patients. There are researchers claiming that qMRI opens the possibilities for a paradigm shift in the medical MR science. Until now clinical qMRI has been constrained from being generally used, due to excessive scan times and extensive post-data processing [3].

4.2

MR Parameters to Quantify

In qMRI the three MR parameters of clinical importance are T1, T2 and PD

(Fig. 4.2). T1and T2provide information about the spin relaxation of the biological

tissue while PD corresponds to proton density of the tissue. From a technical point of view, other MR parameters might be necessary to measure in order to

(38)

24 Quantitative MRI

Figure 4.1. QMRI enables tissue to be

accurately characterized through the ab-solute values of the MR parameters. In the figure the quantitative values of the four pixels indicates white matter or white matter looking tissue. The information is given regardless of the greyscale mapping of the image and neighbouring pixel inten-sities. Using conventional contrast MRI, the intensities of the pixels will not supply any quantitative information on the char-acteristics of the tissue. The clinical inter-pretation needs to be based on the image contrast and given knowledge about the specific scanner sequence used.

get accurate and reproducible measurements. One such parameter is the B1-field

strength to correct for B1-field inhomogeneity resulting in inaccurate flip-angles.

Figure 4.2. The graph shows T1, T2and PD and how they are related to the relaxation behaviour of tissue.

4.2.1

Tissue Characterisation

By creating a 3D Cartesian grid with PD, R1 and R2 on the axes, tissue specific clusters can be identified in order to characterize tissue. At the moment, several research projects are aiming to investigate and map the MR parameter values in healthy and diseased tissues in order to use the absolute values as diagnostic tools, both for characterisation of disease and normal tissue. For example, T1 has been

shown to be affected in a number of neurological diseases such as multiple sclerosis (MS), intracranial tumours, stroke and dementia [4].

(39)

4.3 Data Acquisition 25

4.3

Data Acquisition

In qMRI, the scanner sequence performs the signal acquisition at a number of time-points during the magnetic relaxation following the RF-pulses. The collected data points are used to fit the data to the relaxation behaviour predicted by the Bloch equations. The scanner sequence has to provide enough data to achieve acquired dynamic range of the parameters to be measured as well as sufficient SNR and resolution of the image. Due to the long scan times needed to achieve good SNR and reliable measurements, many scanners sequences will quantify only one of the MR-parameters. As an example, the present gold standard for T1 quantification

is the inversion recovery sequence. By repeating the scan and measure the signal with a number of different TIs, the complete T1-relaxation can be estimated. Multi

echo acquisition methods are needed in order to measure T2 and PD [4]. The

sequence used for data acquisition within the synthetic imaging project at CMIV is QRAPMASTER [3].

4.3.1

QRAPMASTER

QRAPMASTER performs simultaneous quantification of T1, T2, PD and B1-field

strength throughout the patient volume. The sequence was developed in order to scan a head within a clinical acceptable period of five minutes. The T1

ments are given through a spoiled saturation pulse, which also allows measure-ments of the B1-field strength. By dividing the scanner sequence into a saturation

block and an acquisition block performing saturation and acquisition on different slices QRAPMASTER provides fast enough measurements. The number of scans and the delay times (TD) in between them determine the dynamic range of T1. T2

is measured using a fast multi echo gradient spin echo sequence (GRaSE) which uses EPI. The number of echoes and the spacing in between them determine the dynamic range of T2. Additionally, the sequence includes a REST slab that is

placed a distance away from the image volume. The REST slab saturates the signal from blood to prevent motion artefacts. This will cause all flowing blood to appear black in the images. A schematic picture of the sequence can be seen in Fig 4.3. PD is given in relation to M0, which is calculated from T1, T2 and

B1-field strength.

4.4

Fitting the Data to the Mathematical Models

After the data acquisition, fitting of the data to the relaxation models are needed to calculate T1, T2 and PD. The most basic way to fit T1 and T2 is to assume

a monoexponentional relaxation curve in each measured voxel. By doing so the relaxation within each voxel can be fitted by a least square fit to a monoexponential curve (Fig. 4.4), minimum two data points are needed. When using the golden standard for T1measurement as described in Section 4.3 the data can be fitted to

the following equation S(T I) = S0(1 − 2e−T I/T 1) where S0 is the signal achieved

(40)

26 Quantitative MRI

Figure 4.3. Schematic schedule of the QRAPMASTER scanner sequence (Ref [3], Page

321)

Figure 4.4. The data points are fitted to the mathematical models describing the relaxation.

In reality multiexponentional relaxation is common due to partial volume ef-fects within the voxels. Simplification to a monoexponential fit is reasonable when the relaxation exchange between the compartments within the voxel is larger than the relaxation rate within each compartment [4]. In brain tissue, T1is often

mono-exponential while T2is bi- or multiexponential. If multiexponential relaxations are

fitted to monoexponential curves the data will be scanner parameter dependent.

4.4.1

Quantification Maps

Quantification maps are used to visualize the measured MR parameter values in each voxel. In Fig 4.4.1, a T1-map is displayed were the color of the pixels

correspond to T1 in milliseconds. The image analysis of the quantified values can

be made in a number of ways including for example a region of interest (ROI), histogram analysis or voxel based group mapping [3].

(41)

4.5 Synthetic MRI 27

Figure 4.5. T1-map (Image retrieved from Contrast Predictor)

4.4.2

SyMRI BrainStudio Fitting Tool

The SyMRI BrainStudio Fitting Tool is used to create T1, T2 and PD maps from

the data acquired with QRAPMASTER. The SyMRI fitting tool uses a least square fit to a monoexponential curve for T1 and T2. B1-field strength is given from the

saturation pulse and is together with T1 and T2 used to calculate an appropriate

M0. PD is then given from T1, T2, M0and a number of scaling factors in order to

correct for a number of scanner parameters.

4.5

Synthetic MRI

SyMRI uses T1, T2and PD maps to post-synthesize contrast-weighted MRI images.

4.5.1

Synthetic Contrast Weighted MRI

The image contrast that would have been given through a conventional scanner sequence can be post-synthetisized using Eq. 4.1 [3]. The equation predicts the signal intensity from a voxel with a certain T1, T2 and PD given the scanner

settings TE and TR. This can be done since the quantified parameters give enough information to predict the complete relaxational behaviour of the tissue. To post-synthesize a contrast-weighted image that would have been given from an inversion recovery sequence, Eq. 4.2 is used in order to take account for the influence of the inversion prepulse [3]. S ∝ P D 1 − e −TR/T1 1 − e−TR/T1cosαe −TE/T2 (4.1) S ∝ P D1 − 2e −TI/T1+ e−TR/T1 1 − e−TR/T1cosα e −TE/T2 (4.2)

(42)

28 Quantitative MRI

4.5.2

Tissue Segmentation

Synthetic images can also be calculated on a purely quantitative basis using tissue characterisation. In this way images can be displayed in a way that is not possible in conventional contrast-weighted MRI. The 3D Cartesian grid with PD, R1 and R2 on the axis will provide a foundation for tissue segmentation and the creation of contrast-weighted images with certain tissues suppressed.

4.5.3

Normalization

In all conventional contrast-weighted images there is a certain amount of PDw since the proton density supply the origin of the signal detected by the MRI scanner. Since the PD contribution from each voxel is known using qMRI the images can be normalized excluding the contrast dependence on PD giving images with pure

(43)

Chapter 5

Methods

The chapter describes the methods used in this master thesis throughout the im-plementation and development of SyMRI Brain Studio, the interaction with radi-ologists and the validation of the technique.

5.1

SyMRI Brain Studio

SyMRI Brain Studio was developed and implemented in an iterative manner in parallel with the continuous development of the SyMRI Plugin framework (Syn-thetic MR Technologies AB, 2007) and the improvement of the SyMRI Brain fitting tool and the scanner sequence settings. A number of versions of SyMRI Brain Studio were developed throughout the work. SyMRI Brain Studio v.1.0.0 was the first version developed, which was used for a first validation of an on-site radiologist (Sec 5.2.1). Together with user input from other radiologists and researchers SyMRI Brain Studio v.3.1.0 was developed and sent to a number of off-site hospitals for validation. SyMRI Brain Studio v.3.1.0 was additionally in-stalled at CMIV for research and validation purposes and later upgraded with the successive versions v.3.2.0., v.3.3.0 and v.4.1.0.

5.1.1

System Environment

SyMRI Brain Studio is developed as a visualization studio for SyMRI SECTRA IDS 5 Plugin in Microsoft Visual C++ 6.0. The SyMRI SECTRA Plugin provides a framework for SyMRI fitting tools and visualization modes using raw data stored as DICOM [9] files in SECTRA IDS 5.

5.2

Radiologist Interaction

Radiologists with an interest in SyMRI were asked to participate in the work. Since the technique introduces a new concept of MRI, the most important issue was to find radiologists prepared to put their time on evaluating a not yet clinical

(44)

30 Methods

introduced imaging technique. The meetings with the radiologists were arranged for two reasons. First, there was an interest in evaluating the technique from a radiologist’s point of view, and to convert the demands from the radiologist into technical demands on SyMRI. Secondly, there was an interest in getting familiar with and understand the radiologist’s working environment. The interaction with the radiologists was done under the principle that it is easier to learn about the users’ need and problems by learning the way they are working, rather than to ask questions about what they would expect from a new quantitative MRI technique.

5.2.1

Radiologist Interaction I

Title: Comparison: Conventional T1w, T2w and FLAIR vs SyMRI

Software: SyMRI Brain Studio v.3.0.0, Contrast Predictor v.1.1 beta, SECTRA

IDS 5

Resolution: Conventional MRI 0.8mm, SyMRI: 1 mm Slice thickness: Conventional MRI 3mm, SyMRI: 5mm

The aim with the meeting was to get an idea on how an experienced radiologist would sense the new technique and find it in comparison to conventional contrast-weighted MRI. A week prior to the meeting, the radiologist had been given a short introduction to SyMRI Brain Studio and Contrast Predictor in order to be able to use and interact with the software independently. During the meeting the radi-ologist looked at SyMRI images alone as well as compared SyMRI to conventional contrast-weighted MRI. Information was also gathered regarding preferences on the user interface as well as windowing options and the possibilities of auto con-trast. Additionally, an observation was made on how the radiologist interacts with IDS 5. During the day, a number of the SyMRI unique concepts were intro-duced in order to observe the radiologist’s reaction on features such as contrast optimization, post-examination tissue nulling and PD normalisation that are only possible using SyMRI. The observer (i.e. the author) was taking basic notes in a continuous manner in order to collect as many thoughts and observations as possible without interfering with the workflow. The notes were put together into a document [10] which were read through and commented by the radiologist before finished. The document was written in Swedish to prevent language difficulties to interfere with the findings. The documentation was written in such a way that it could be used to evaluate future versions of SyMRI Brain Studio. Problem with the data transfer resulted in only one subject with available data for conventional contrast weighted MRI and SyMRI through SyMRI Brain Studio. Additionally six data set for SyMRI visualizations in Contrast Predictor were used from which four also were available in PACS. In total seven subjects were investigated, out of them three were diagnosed with multiple sclerosis (MS) and three with brain tumours.

5.2.2

Radiologist Interaction II

Title: How are IDS 5 used by the radiologist today? Software: SECTRA IDS 5

(45)

5.3 Validation of the Technique 31

The meeting was made as an observation and discussion in order to observe how the radiologist uses conventional contrast-weighted MRI in the clinical routine. The radiologist was asked to explain the routine clinical work using a standard protocol for a brain examination. Additionally an observation was made on how the radiologist interacted with the SECTRA IDS 5. Notes were taken during the session and the findings where put together into a document [11].

5.2.3

Radiologist Interaction III

Title: Introduction to SyMRI

Software: SyMRI Brain Studio v.4.1.0, SECTRA IDS 5

The radiologist was briefly introduced to the concept of SyMRI and a discussion about the SyMRI images and features was hold. Three different SyMRI data sets were looked at.

5.2.4

Radiologist Interaction IV

Title: Comparison: Conventional T1w, T2w and FLAIR vs SyMRI - follow up Software: SyMRI Brain Studio v.3.1.0, Contrast Predictor v.1.1 beta, SECTRA

IDS 5

Resolution: Conventional MRI 0.8mm, SyMRI: 1 mm Slice thickness: Conventional MRI 3mm, SyMRI: 5mm

A follow up was made on Radiologist Interaction I in order to see what progress that had been achieved concerning the drawbacks found during the first meet-ing (Section 7.1). The data set that were used in Radiologist Interaction I was re-looked upon and additionally two data sets available for comparison between conventional contrast-weighted MRI and SyMRI Brain Studio.

5.2.5

Radiologist Interaction V

Title: Response Leiden University Medical Centre (LUMC) Resolution: Conventional MRI 0.8mm, SyMRI: 1 mm Slice thickness: Conventional MRI 3mm, SyMRI: 5mm Software: SyMRI Brain Studio v.3.1.0

SyMRI Brain Studio was sent to LUMC in Leiden, Holland for validation on a 3T MRI scanner and to analyse the potential for SyMRI imaging of infants and small children. Their response was used as an additional input for this master thesis.

5.3

Validation of the Technique

The validation was made as a discussion based on the information and knowledge gained from the work with the master thesis.

(46)
(47)

Chapter 6

SyMRI Brain Studio

SyMRI Brain Studio is a visualization studio developed to visualize quantitative MR data of the brain using quantification maps and SyMRI. This chapter explains the features of Brain Studio, its design and the successive development of the dif-ferent versions of the software.

6.1

Architecture

Brain Studio is implemented in C++ as a class inheriting the SyMRI Plugin’s virtual CSyMRIImage class. Necessary modifications in the source code of the SyMRI Plugin framework were made as described in the extensibility chapter in the SyMRI SECTRA IDS 5 Plugin Design document [5]. The architecture of SyMRI Brain Studio is divided into three sections, each with local functions implemented to customize the visualization mode and create the final user interface.

• Rendering

Overridden functions: DrawImage()

The rendering section calculates and visualizes the SyMRI images and quan-tification maps based on the user defined settings. A synthetic T2w image is

displayed as default. • Heads Up Display (HUD)

Overridden functions: DrawHUD(), SetUpHudData()

The HUD gives the end-user information about patient data, the quantitative data inside the region of interest (ROI) as well as the SyMRI settings of displayed contrast images.

• User Event Handler

Overridden functions: HandleMenuSelection(), ModifyMenu(), OnKeyDown(), OnLMouseDown(), OnLMouseUp(), OnMouseMove().

User events are handled through menu selections, short commands and mouse bottom clicks.

(48)

34 SyMRI Brain Studio

Figure 6.1. The visualization pipeline contains several steps ranging from the initial

data acquisition to the rendering of images. SyMRI Brain Studio renders and visualizes the data through the quantification maps provided by the SyMRI fitting tool.

An overview of the SyMRI visualization pipeline can be seen in Fig 6.1. The raw data is retrieved and handled as DICOM images by the SyMRI plugin framework. The quantified data is retrieved by the fitting tool and used by the visualization studio were the rendering of quantification maps and post-synthesized images are made. The SyMRI framework enables a debug mode that has been used for error search.

6.2

Main Features

6.2.1

Quantification Maps

The quantification maps display the quantified MR data pixel by pixel throughout each slice of the brain. In the four viewport default mode a T1-map, a T2-map

and a PD-map is rendered and displayed together with the latest active contrast weighted SyMRI image (Fig 6.2). The quantification maps are visualized using a rainbow colormap and are autoscaled based on the absolute T1, T2 and PD

values present in human brain (Table 6.1). Two additional quantification maps are accessible through the popup menu, the B1-map displaying variations in the

B1-field and the mean-error-map supplying information about estimated errors in

the fitting algorithm (Fig 6.3).

Figure 6.2. Default display of quantifica-tion maps using four viewports.

Upper left: T2w Upper right: T1-map Lower left: T2-map Lower right: PD-map

(49)

6.2 Main Features 35

Figure 6.3.

Additional quan-tification maps. Left: B1 map

Right: Mean Error Map

Region of Interest

The region of interest (ROI) is used to display information about the quantitative MR parameters of the pixels (Fig 6.4). The size of the ROI can be determined by the user, but are limited to a rectangle with a minimum of six viewport pixels. The measurement values of the pixels inside the ROI are displayed in the upper right corner of the HUD, displaying the mean and standard deviation of T1, T2,

PD, the pixel intensity value inside the ROI and the ROI position in viewport coordinates. The relaxation rates within the ROI are plotted in the R1R2-plot.

Figure 6.4. ROI and the supplied info displayed in the HUD

R1R2-plot

In the R1R2-plot, the relaxation rates are forming the two axes in a 2D Cartesian

grid (Fig 6.5). Clusters characterising white matter, central white matter, internal capsule, putamen, grey matter and CSF are defined based on the mean values in Table 6.1 and are indicated in the plot as ellipses. The area of each cluster is based on empirical studies and on-going projects at CMIV working on the classification of the clusters given by the QRAPMASTER sequence. The bounding lines in between the clusters illustrate relaxation values possible caused by partial volume effects. The relaxation rates inside the ROI are plotted in the graph and each data point is stretched over an area corresponding to 0.05 ± 0.01s−1 in the R1

-direction and 0.4 ± 0.08s−1 in the R2-direction. The intensity value given each

pixel is linearly related the maximum distance in the x- or y-direction from the actual data point. Finally the pixel intensities are scaled in the interval [-mean/2

(50)

36 SyMRI Brain Studio

mean/2] and are displayed with a threshold discriminating intensity values initially below zero. This is done in order to get a weighted plot.

Figure 6.5. The relaxation rates are plotted and specific clusters indicate different types

of brain tissue

Table 6.1. MR parameter values for brain tissue.

Tissue T1 T2

White Matter 570 75 Central White Matter 630 85 Internal Capsule 665 68

Putamen 800 73

Grey Matter 1100 96

CSF 3800 1800

Fat 320 90

6.2.2

Display of SyMRI Contrast-Weighted Images

The contrast images are rendered based on the measured MR parameters and the current settings for TE, TR (Eq. 4.1), and TI (Eq. 4.2) in the case of a simulated inversion pre-pulse. The default settings with values corresponding to generally used scanner parameters (Table. 6.2) are accessed through the popup menu and keyboard accelerators. The parameters can be modified by the end user through menu options and the left mouse bottom to render any contrast-weighted image possible. The additional image settings: colormap, interpolation, modulus/real and autoscale are modified through the popup menu and keyboard accelerators. Fig 6.6 show the four viewport default display with a T1w image, a T2w image, a

(51)

6.2 Main Features 37

Figure 6.6.

Upper left: default T1w Upper right: default T2w

Lower left: default PDw

Lower right: default FLAIR

Table 6.2. Default Scanner Parameters

Image TR TE TI Mode T1w 350 10 - -T2w 4500 100 - -PDw 6000 10 - -FLAIR 6000 120 200 Modulus Navigation Window

The navigation window shows the current contrast-weight of the image in the viewport (Fig 6.7) and is displayed in the HUD in order to navigate the user to the current image-weight while freely selecting the scanner parameters TR and TE.

Figure 6.7. The navigation window allows for user guidance when adjusting the current

(52)

38 SyMRI Brain Studio

Fat Suppression

Menu options and keyboard accelerators for fat suppression are available (Fig. 6.8). When calculating the image a mask is applied that excludes the intensity contribu-tions from fat. The fat suppression is unique to SyMRI and can not be achieved on the same basis in conventional imaging as the selection is based on the measured MR parameters.

(a) T2w Image (b) T2w Image with Fat

Sup-pression

Figure 6.8. Fat Suppression is used to exclude the signal contribution from fat.

T1 Enhanced Image

(a) T1w Image (b) Enhanced T1w Image

Figure 6.9. The contrast between white and grey matter is enhanced through the PD-normalization.

By excluding the PD-contribution in Eq 4.1 & Eq. 4.2 a PD-normalized image can be calculated neglecting the signal contribution from PD. This will create enhanced contrast between grey matter and white matter in T1w images since the

(53)

6.2 Main Features 39

6.2.3

Graphical User Interface

The GUI is divided into a number of features, the popup menu, the HUD and the keyboard accelerators.

PopUp Menu

The pop-up menu is divided into four sections (Fig 6.10). The image settings section is partly pre-defined by the SyMRI framework and handles general image settings: pan, zoom, interpolation of the image, linking between slice selection in IDS5 and the plugin, autoscale, colormap, modulus or real image, linked viewports, settings of the R1R2-plot and the number of viewports displayed. The second

section allows easy to use pre-defined default values for SyMRI contrast-weighted images, either one by one or as a four viewport default option supplying the user with the conventional set of contrast images for typical brain applications in addition with a P Dw image (Fig. 6.6). The third menu section visualizes the quantification maps one by one or as a four-viewport default option (Fig. 6.2). The fourth section provides options for modification of the simulated scanner parameter settings TE, TR and TI as well as the simulation of an inversion pre-pulse and qMRI features in terms of fat suppression, PD-normalization and T1enhancement.

Additionally there is a stack tile sub-menu and a close bottom provided by the SyMRI framework.

Figure 6.10. Popup Menu

Heads Up Display

The HUD supplies the user with information regarding the examination, the pa-tient, the image settings and the quantified data inside the ROI. Information on the

References

Related documents

Byggstarten i maj 2020 av Lalandia och 440 nya fritidshus i Søndervig är således resultatet av 14 års ansträngningar från en lång rad lokala och nationella aktörer och ett

Utöver vår revision av årsredovisningen har vi även utfört en revision av styrelsens och verkställande direktörens förvaltning för Synthetic MR AB (publ) för år 2017 samt

The Board of Directors and Managing Director of SyntheticMR AB (publ), organization number 556723-8877, hereby issue the annual report for the financial year 2017.. SyntheticMR AB

The difference between the conventional and synthetic method is that using synthetic MRI only require one sequence to generate images of the different weightings.. This is due to

1600, 2017 Department of Radiological Sciences. Linköping University SE-581 83

A vast amount of congenital syndromes are ac- companied by ear involvement 83, 84. 3D ultrasound is useful to delineate the pinna, it helps to rule out local malformations

Box and Tiao (1975) classified the pattern of intervention effects as “step” and “pulse”, where the former implies a constant intervention effect over time. Figure 1 depicts

A novel estimation method for physiological parameters in dynamic contrast- enhanced MRI: application of a distributed parameter model using Fourier-