Resonance Imaging using Stimulated Echo Acquisition Mode
Pulse Sequence Development
Utsläckning av blodsignal genom applicering av stimulerade ekon vid faskontrastavbildning med magnetisk resonanstomografi
Pulssekvensutveckling
Reenalyn Borromeo
School of Engineering Sciences in Chemistry, Biotechnology and Health KTH Royal Institute of Technology
Place for Project
Stockholm, Sweden
Karolinska University Hospital
Examiner
Matilda Larsson
Department of Biomedical Engineering and Health Systems KTH Royal Institute of Technology
Reviewer
Matilda Larsson
Department of Biomedical Engineering and Health Systems KTH Royal Institute of Technology
Supervisors
Alexander Fyrdahl
Department of Clinical Physiology
Karolinska Institute and Karolinska University Hospital Andreas Sigfridsson
Department of Clinical Physiology
Karolinska Institute and Karolinska University Hospital
Doppler echocardiography is the conventional method for measurement of myocardial motion. However, the same clinical parameters can be measured with phase contrast magnetic resonance imaging (MRI). Suppression of the blood signal with black-blood methods can reduce flow-related artifacts that may affect quantitative measurements with phase contrast MRI. Conventional blood suppression techniques are not time efficient and a potential approach to achieve the black-blood effect in phase contrast imaging is through application of stimulated echo acquisition mode (STEAM).
This thesis describes the development of a pulse sequence where a STEAM-based preparation was combined with conventional phase contrast imaging to achieve a black- blood effect in the produced images. The results of the performed imaging experiments showed that black-blood contrast was achieved with the proposed pulse sequence, and the blood suppression was also maintained during the cardiac cycle. Myocardial tissue velocity measurement with the suggested approach showed good agreement with conventional phase contrast imaging. It was concluded that black-blood phase contrast imaging can be achieved through the application of a STEAM-based preparation.
Keywords
MRI, phase contrast, black blood, STEAM, myocardial motion
Dopplerekokardiografi är den konventionella metoden för mätning av myokardiell rörelse. Liknande mätningar kan dock utföras genom faskontrastavbildning som är en metod inom magnetisk resonanstomografi. Signal från blod kan orsaka flödesrelaterade bildartefakter vid faskontrastavbildning som kan påverka hastighetsmätningar i myokardiet. Utsläckning av blodsignalen kan mitigera artefakternas påverkan på kvantitativa mätningar som utförs med faskontrast. Konventionella metoder för utsläckning av blodsignalen är inte tidseffektiva och en potentiell metod för att åstadkomma blodutsläckning vid faskontrastavbildning är applicering av en preparation baserad på stimulated echo acquisition mode (STEAM).
I detta arbete utvecklades en pulssekvens där en STEAM-baserad preparation kombinerades med en traditionell faskontrastsekvens. De resulterande bilderna visade att blodutsläckning hade åstadkommits och att denna effekt kunde bibehållas under hjärtcykelns gång. Hastighetsmätningar utfördes även i myokardiet och var jämförbara med traditionell faskontrast. Slutsatsen var att utsläckning av blodsignalen vid faskontrastavbildning är möjligt genom applicering av en STEAM- baserad preparation.
Nyckelord
MRI, magnetisk resonanstomografi, faskontrast, blodutsläckning, STEAM, myokardiell rörelse
I would like to thank my supervisors Alexander Fyrdahl and Andreas Sigfridsson for the opportunity to write my thesis at Karolinska CMR and for providing me with excellent guidance. I would also like to thank Daniel Loewenstein for his help during the imaging process. A special thanks to Alex, who always did his best to help me in my confusion and could spend hours supervising me through video calls. Your dedication for your work is inspiring and I am certain that you will go far.
Thank you to my KTH group supervisor Rodrigo Moreno for answering my endless number of questions and to everyone in my seminar group who gave me helpful feedback. I would also like to thank my reviewer and examiner Matilda Larsson for understanding my situation and giving me the opportunity to present this thesis at a later time.
I am very grateful for the support of my family and friends throughout my studies.
My parents and siblings have been a great support system and I am very lucky to have had them through all the highs and lows. I would also like to thank my boyfriend Sami whose encouragement kept me going despite the challenges. Thank you for sitting with me throughout this entire process and for keeping me sane! You always manage to make everything feel better.
This thesis is dedicated to my beloved Lola, who is my motivation and inspiration for doing better in life. I will always do my best to make you proud.
Acronyms
31 Introduction
41.1 Purpose & Aim . . . 5
1.2 Delimitations . . . 5
2 Method
6 2.1 Pulse Sequence Development . . . 62.1.1 Velocity Encoding Gradients . . . 6
2.1.2 Black-Blood Preparation Module . . . 8
2.1.3 Data Acquisition & Pulse Sequence Parameters . . . 10
2.2 Imaging Experiments . . . 12
2.2.1 Simulator experiments . . . 12
2.2.2 Volunteer Imaging . . . 12
2.2.3 Image Reconstruction . . . 14
2.3 Image Data analysis . . . 15
2.3.1 Image Intensity Measurements in Blood . . . 15
2.3.2 Contrast Measurements . . . 16
2.3.3 Velocity Measurements . . . 17
3 Results
18 3.1 Volunteer Imaging . . . 183.1.1 Magnitude Images . . . 18
3.1.2 Phase Images . . . 21
3.2 Region of Interest Measurements . . . 24
3.2.1 Blood Image Intensity . . . 25
3.2.2 Contrast Values . . . 27
3.2.3 Velocity Measurements . . . 29
4 Discussion
315 Conclusions
35 5.1 Future Work . . . 35A State of the Art
36 A.1 Magnetic Resonance Imaging . . . 36A.1.1 MR Physics . . . 36
A.1.2 The MR signal . . . 37
A.1.3 Gradient Moments . . . 38
A.2 Phase Contrast Imaging . . . 39
A.2.1 Physical Principles of Phase Contrast . . . 39
A.2.2 Image Acquisition . . . 41
A.2.3 Phase Contrast Applications . . . 42
A.3 Black-Blood Imaging . . . 43
A.3.1 Blood Suppression Methods . . . 43
A.3.2 Black-Blood Applications . . . 46
A.4 Stimulated Echo Acquisition Mode . . . 47
A.4.1 Stimulated Echoes . . . 47
A.4.2 Displacement Encoding with Stimulated Echoes . . . 48
A.4.3 Signal-to-Noise Ratio in STEAM . . . 49
A.5 Related Work . . . 51
B Complementary Material
52 B.1 Additional Images . . . 52B.2 Image Reconstruction Code . . . 55
References
56ADC analog-to-digital converter
CMR cardiovascular magnetic resonance
CSPAMM complementary spatial modulation of magnetization DENSE displacement encoding with stimulated echoes
ECG electrocardiography
FLASH fast low angle shot
MRI magnetic resonance imaging
RF radiofrequency
ROI region of interest
RV right ventricle
SAR specific absorption rate
SNR signal-to-noise ratio
STEAM stimulated echo acquisition mode
TE echo time
TI inversion time
TM mixing time
TR repetition time
VENC velocity encoding
Introduction
Cardiovascular disease is one of the leading causes of death in the world [1], making it important to accurately be able to quantify clinical parameters, such as ventricular volumes and blood flow [2]. Delineation of vascular structures and the heart valves is also relevant for assessments of pathological heart conditions [3–5]. Magnetic resonance imaging (MRI) can be used to quantify clinically relevant parameters for diagnosis and treatment in cardiovascular disease [6]. Phase contrast imaging is an established cardiovascular magnetic resonance (CMR) imaging method for quantification of blood flow parameters [2] where two dimensional phase contrast MRI is the most widely used technique for measurement of blood velocities [7].
Myocardial motion is also important in clinical evaluations, and is conventionally done using Doppler echocardiography [7], but is possible to assess with phase contrast MRI [2]. Though these measurements can be disturbed by flow-related artifacts that degrade the image quality and prevent accurate delineation of the myocardium [2, 8].
Suppression of the blood signal with black-blood imaging can mitigate the effects of flow related artifacts [9, 10]. It allows improved delineation of myocardial tissue for visualization of the heart chambers or vascular structures [10]. Established approaches to achieve this effect include double inversion recovery [11], which only achieves blood saturation at a set time point in the cardiac cycle, or spatial presaturation [8] which prolongs the imaging time, hampers the temporal resolution and has a diminishing effect over the cardiac cycle as unsaturated blood flows back into the imaging slice.
Some work has also suggested a velocity selective excitation, which could potentially excite spins under a suitable velocity threshold [12]. Another option is using stimulated echo acquisition mode (STEAM) which is normally used for strain imaging but has an inherent black-blood effect [13].
A potential approach to achieve quantification of myocardial tissue velocities with MRI is using the conventional phase-contrast method combined with the inherent black- blood effects of STEAM. This way, the black-blood effect can be maintained through the cardiac cycle, and the temporal penalty of using spatial presaturation is avoided. This could eliminate flow-related artifacts, improve delineation of the myocardium and allow quantitative measurements of myocardial motion with high temporal resolution without blood flow interfering with the measurement. The mentioned benefits can contribute to improvement of detection, diagnosis and treatment of pathological conditions, such as diastolic dysfunction [14].
1.1 Purpose & Aim
The purpose of this thesis report is to describe the development and evaluation of a proposed STEAM-based black-blood phase contrast imaging technique, and to present the resulting conclusions about what has been achieved in the process. The aims of the thesis project are to combine the black-blood effect of STEAM with phase contrast imaging, maintain the black blood effect throughout the cardiac cycle and measure myocardial tissue velocities.
1.2 Delimitations
The main focus was technical development with a clinical evaluation pending. The imaging experiments were also performed with a single volunteer.
Method
The following sections describe the development of the proposed pulse sequence and the imaging experiments performed to determine whether the black-blood effect was achieved without affecting the velocity measurements. The resulting images were visually evaluated and quantitatively analyzed through image intensity and velocity measurements.
2.1 Pulse Sequence Development
The initial approach was to build and test the pulse sequence in a Bloch simulator before exporting it to an MR scanner. However, it was concluded that the black-blood effect could not be adequately simulated with available Bloch simulation software. Instead, it was decided to implement the pulse sequence and run it on a clinical scanner. The pulse sequence was therefore developed in a proprietary vendor-provided coding environment.
The following sections describe the implementation process.
2.1.1 Velocity Encoding Gradients
A pulse sequence diagram describes the wave forms of the radiofrequency (RF) pulses and magnetic field gradients that are applied to generate an image in MRI, an example is presented in figure 2.1.1. Gradients are applied for spatial localization of the signal along three directions: slice selection, phase encoding and readout. Data sampling occurs when the analog-to-digital converter (ADC) is active, which is represented by the waveform on the last line of the pulse sequence diagram in figure 2.1.1.
The starting point of the pulse sequence development was a fast low angle shot (FLASH) sequence. It is a gradient echo sequence that utilizes RF pulses with a low flip angle α and short repetition time (TR), which places the signal in what is similar to a steady state. The FLASH pulse sequence is simple, as seen in figure 2.1.1, which made it a good foundation to build on.
Figure 2.1.1: An example of a FLASH pulse sequence diagram inspired by Frahm et al. [15].
Imaging performed with this pulse sequence is characterized by the use of a small α together with a short TR.
The phase of the complex MR signal is sensitive to motion, which is utilized in phase contrast MRI. The method allows for measurement of the velocity component perpendicular to the image slice [2]. Phase contrast imaging requires a flow compensated and flow sensitive acquisition, since image subtraction is performed to obtain a velocity map [2].
A bipolar velocity encoding gradient was implemented into the FLASH pulse sequence where the velocity encoding (VENC) value could be set to enable flow compensation or sensitivity, see figure 2.1.2. The bipolar gradient was added in the slice direction, which would allow for measurement of through-plane motion. The first lobe of the bipolar gradient was combined with the slice rephaser to reduce scan time. The longest possible duration of the velocity encoding gradient was set to the duration of the gradient corresponding to the lowest VENC set by the operator, as lower VENC values require larger bipolar gradients.
Figure 2.1.2: A bipolar velocity encoding gradient (marked in yellow) was added to the FLASH pulse sequence to enable phase contrast imaging. The addition of a bipolar gradient in the slice direction would allow for measurement of velocities along this direction.
2.1.2 Black-Blood Preparation Module
STEAM sequences enable storing of the magnetization in the longitudinal direction until imaging is initiated and also have an inherent black-blood effect [16]. A STEAM- based method known as displacement encoding with stimulated echoes (DENSE) is used to measure strain in the myocardium [17] and involves the application of a position encoding gradient before the signal is stored, as well as a position decoding gradient applied during imaging.
Only using STEAM to saturate the blood signal will result in a recovery of the blood signal over the cardiac cycle. To obtain continuous blood saturation in a STEAM pulse sequence, complementary spatial modulation of magnetization (CSPAMM) can be used. Several published papers have described the use of STEAM-based black- blood preparation prior to image acquisition [10]. The preparation in this report will be referred to as the STEAM preparation. A pulse sequence diagram of the proposed sequence is presented in figure 2.1.3.
Figure 2.1.3: A schematic overview of the proposed pulse sequence. Two non-selective 90° RF pulses were added at the beginning of each heartbeat. The encoding and decoding gradients are marked in blue. The gradient moment of the encoding/decoding pair is known as the encoding strength (ke). The center of the encoding gradient coincides with the point in between the two 90° pulses. The time between the two non-selective pulses, as well as the time between the slice-selective pulse and the stimulated echo, have to be equal to TE/2. The time between the second non-selective pulse and the first excitation pulse is called the mixing time (TM). Crusher gradients (marked in red) were added in all directions at the end of the black-blood preparation as well as image acquisition to eliminate any remaining magnetization.
The STEAM preparation was implemented by the addition of two non-selective 90°
RF pulses where the initial phase of the second RF pulse was alternated each time the modulation was applied at the beginning of each heartbeat. An encoding and decoding gradient of equal magnitudes but opposing sign were added to modulate the signal along the direction of motion [18]. The encoding strength was the same for both gradients and denoted ke. The signal from blood flowing in this direction would be spoiled during the mixing time (TM) between the second and third RF pulse. The proposed sequence measures through-plane myocardial tissue velocities, and the modulation gradients were therefore applied along the slice direction to attain a black-blood effect in the resulting images.
Finally, the option of adding and removing the STEAM preparation was also implemented into the code for the purpose of reverting the sequence to a conventional bright-blood sequence. The parameters available to modify during the subsequent imaging experiments were the echo time (TE), TM, VENC and ke. The TR between two excitation pulses was always chosen to be as short as possible to obtain the highest possible temporal resolution.
2.1.3 Data Acquisition & Pulse Sequence Parameters
To capture the dynamics of the cardiac cycle, data acquisition was performed under breath hold with segmented k-space sampling combined with electrocardiography (ECG) retro-gating, see figure 2.1.4. Each segment contains a set of k-space lines, and the retro-gating was implemented by continuously sampling one segment over the entire cardiac cycle. Within each segment, each line was sampled twice, once for each VENC value.
Figure 2.1.4: A schematic overview of the data acquisition process with the proposed pulse sequence. Data acquisition was initiated by the detection of an R-wave in the ECG signal.
The STEAM preparation was applied once per heartbeat, represented by the grey bar after the R-wave. The sign of the second RF pulse in CSPAMM was alternated at each heartbeat. The preparation was followed by imaging until detection of a new trigger and each heartbeat was divided into a number of time frames, or cardiac phases, represented by the blue bars. In each cardiac phase, a segment of k-space was sampled, and every line in the segment was encoded with two different VENC values. Each unique line in k-space was read a total of four times.
The STEAM preparation was applied once per heartbeat and image acquisition was repeated until a new trigger was registered in the ECG signal. The sign of the second RF pulse in CSPAMM was also alternated when a new trigger was detected. Thus, for each line in k-space, a total of 4 acquisitions were required: a flow compensated and flow encoded acquisition for each CSPAMM preparation. The imaging parameters for the proposed pulse sequence are described in table 2.1.1.
Table 2.1.1: A summary of the implemented pulse sequence parameters.
Parameter Unit Min Max Increment
Segmentation Factor - 1 64 1
Velocity Encoding (VENC) cm/s 0 - 0.1
Encoding Strength (ke) (mTm)· µs 0 - 0.1
Mixing Time (TM) ms 0.01 10 0.01
The number of unique k-space lines within a segment is referred to as the segmentation factor and is a parameter set by the operator. The segmentation factor determines the temporal resolution as well as the breath-hold time. A VENC value of 0 cm/s corresponds to a flow compensated acquisition. The bipolar gradient amplitudes were calculated by functions in the coding environment based on the time parameters set by the user. The input to the gradient amplitude calculating functions were gradient moment values. The strength of the CSPAMM encoding ke is usually expressed with the unit cycles/pixel. In this case, however, it was expressed as a gradient moment for cohesiveness.
2.2 Imaging Experiments
Simulator and imaging experiments were performed to see if the pulse sequence would run on a scanner and if images could be produced with the proposed pulse sequence based on phantom and volunteer imaging.
2.2.1 Simulator experiments
Prior to running the sequence on a clinical scanner, the sequence was simulated using the proprietary simulation software provided by the vendor. This was to investigate that the sequence behaved as expected, and to ensure that no safety-related limits were exceeded. The simulator was also used to ensure that the expected VENC and ke
values were achieved. To simulate a trigger signal, a rudimentary trigger function was therefore implemented in the code, where a certain heart rhythm was assumed.
2.2.2 Volunteer Imaging
Phantom imaging was performed prior to imaging a volunteer to check if the sampled data could be used to produce an image, but also to assure that it would be safe to scan a volunteer by measuring the applied RF-power and estimating the specific absorption rate (SAR) through the MR scanner. The phantom was a stationary spherical ball containing a nickel sulfate (NiSO4) solution. An external ECG-signal was simulated by the scanner to emulate the heartbeat of a patient.
After successful simulations and phantom imaging, a healthy male volunteer was scanned. Informed consent was obtained. Development and testing of novel pulse sequences has been approved by the Swedish Ethical Review Authority (DNR:
2011/1077-31/3). Images were acquired with the subject lying supine in an 1.5T MR scanner (Aera, Siemens Healthcare, Erlangen, Germany) with an anterior body array coil and a posterior spine array coil. ECG-triggering was achieved using electrodes placed on the subject. Imaging was performed during breath hold that lasted approximately 20 seconds, which corresponds to a heart rate around 60 bpm.
Since through-plane motion was imaged, short axis slices were acquired. The image slice was placed below the mitral valve, as shown in figure 2.2.1.
Figure 2.2.1: Schematic images of the slice placement during volunteer imaging. a) A four chamber view of the heart showing the left (LA) and right atrium (RA), as well as the left (LV) and right ventricle (RV). The image slice, marked in blue, was placed under the mitral valve.
b) Short-axis view of the previously described image slice.
The matrix size was 64×64 pixels, with a field-of-view of 300×300 mm and a slice thickness of 5 mm, resulting in non-isotropic voxels of size 4.7 × 4.7 × 5 mm. The operator set parameter values used during imaging of the subject are presented in table 2.2.1. The TR/TE were 9.22 ms/6.75 ms and TM was set to 5.0 ms.
Table 2.2.1: The parameter values used during volunteer imaging. A conventional phase contrast acquisition, meaning a CSPAMM encoding value of 0, was also obtained to serve as a reference for comparison with the acquisitions where the STEAM preparation was applied.
Segmentation Factor (Seg)
CSPAMM
Encoding (ke) Flip Angle (FA) VENC 1 VENC 2
6 0 (mTm )· µs 10° 0 cm/s 30 cm/s
6 1000 (mT
m )· µs 10° 0 cm/s 30 cm/s
6 1500 (mTm )· µs 10° 0 cm/s 30 cm/s
6 2000 (mTm )· µs 10° 0 cm/s 30 cm/s
6 1000 (mTm )· µs 15° 0 cm/s 30 cm/s
6 1500 (mTm )· µs 15° 0 cm/s 30 cm/s
6 2000 (mTm )· µs 15° 0 cm/s 30 cm/s
High CSPAMM encoding values were used in order to fully suppress the signal from blood. A relatively small α was used to mitigate signal loss at the end of the cardiac cycle and a VENC of 30 cm/s was used to encompass the expected velocity of the myocardium.
2.2.3 Image Reconstruction
Volunteer imaging resulted in 7 raw data sets that were was exported from the MR scanner for image reconstruction. The images were reconstructed in MATLAB (R2020a, MathWorks, Natick, MA, USA). The reconstruction process for a single data set started with a Fourier transform to acquire images from k-space. During data sampling, the cardiac cycle was divided into a number of time frames which were referred to as cardiac phases. Each image data set consisted of a magnitude and phase image of all cardiac phases. Two subsequent image subtractions were performed after the Fourier transform.
A complex subtraction was first done to eliminate the non-modulated component that arises during STEAM imaging, which resulted in two images with black-blood contrast.
The next step was subtraction of the stationary tissue phase, yielding a phase-contrast image where all phase contributions were due to moving spins. To maintain the phase information after coil combination, coil sensitivity maps were estimated [19], and the coil combination was performed as:
I =X
j
Ij· Cj∗ (2.1)
where I was the coil-combined image, Ij the image from coil j and Cj the estimated coil sensitivity map for coil j. Figure 2.2.2 shows the image reconstruction process followed in order to attain the resulting magnitude and phase image for each cardiac phase. The image reconstruction code can be found under section B.2 in Appendix B.
Figure 2.2.2: The image reconstruction workflow for one cardiac phase in an image data set. A Fourier transform was performed to obtain images corresponding to each parameter combination, followed by two image subtractions. A complex subtraction and a phase subtraction were performed to acquire the final phase contrast image. Both the magnitude and phase image of each cardiac phase underwent the same process.
2.3 Image Data analysis
The resulting images were analyzed to evaluate the black-blood effect and region of interest (ROI) measurements were performed to measure the image intensity in the blood pool, calculate the image contrast between the myocardium and blood, and measure myocardial velocity. A significance test was also performed for each type of ROI measurement to investigate if there was a mean difference between the reference and the acquisitions where the STEAM preparation was applied.
2.3.1 Image Intensity Measurements in Blood
The image intensity of blood was measured in an ROI placed in the blood pool of the right ventricle (RV) over the entire cardiac cycle for each encoding strength, including no CSPAMM encoding. The image intensity measurements were performed in the magnitude images. The ROI was manually drawn in the blood pool of the right ventricle (RV) and a binary mask containing 4 voxels was extracted from the magnitude image of one cardiac phase and used in all measurements, see figure 2.3.1. The image intensity mean of each cardiac phase was calculated and the values were plotted to determine the effect of the encoding strength on the blood suppression.
Figure 2.3.1: A schematic image of the ROI used for image intensity measurement in the blood pool of the right ventricle (RV). The binary mask consisted of 4 voxels and was extracted from the magnitude image of one cardiac phase. The left ventricle (LV) is also seen in the image.
A Wilcoxon signed-rank test was also performed for the calculated mean image intensity values. Each acquisition with the proposed sequence was compared with no CSPAMM encoding, i.e. conventional phase contrast acquisition, to investigate if there was a significant difference in image intensity.
2.3.2 Contrast Measurements
Relative intensity differences between the myocardium and the blood pool in the magnitude images were evaluated by placing an ROI in the septal wall and in the right ventricular blood pool, see figure 2.3.2. The image contrast between the myocardium and blood was calculated by taking the absolute difference of the mean intensity in both tissues. The difference was then normalized against the image intensity mean of all cardiac phases. The calculated contrast values for each image acquisition were then plotted against the cardiac phase. A Wilcoxon signed-rank test was also performed where each acquisition with the proposed sequence was compared with no CSPAMM encoding to investigate if there was a significant difference in contrast.
Figure 2.3.2: Schematic image of the position of each ROI used for the contrast measurements. Two binary masks consisting of 4 voxels each were extracted from the image and used for the measurements. One ROI was placed in the septal wall of the myocardium and another in the blood pool of the right ventricle (RV). The left ventricle (LV) is also seen in the image.
2.3.3 Velocity Measurements
An ROI that consisted of 4 voxels was placed in the septal wall in the phase images for velocity measurement. The mean value in each cardiac phase was calculated for all CSPAMM encoding strengths used during volunteer imaging. The measured values in all cardiac phases were then plotted for each image acquisition. A Wilcoxon signed- rank test was also performed where the mean velocity values of each acquisition with the proposed sequence was compared with conventional phase contrast to investigate if there was a significant difference between the data sets.
Figure 2.3.3: Schematic image of the ROI used for velocity measurement in the septal wall of the myocardium. The measurements were performed using a binary mask extracted from the phase image of one cardiac phase and it consisted of 4 voxels. The left (LV) and right ventricle (RV) are seen in the image.
Results
The sections in this chapter present the images produced from the imaging experiments performed with a single volunteer using the proposed pulse sequence. Plots based on the values from the ROI measurements are also presented along with the results from the significance tests.
3.1 Volunteer Imaging
The images in this section are the results from the image acquisitions where α was set to 10°, since the same value was used in the reference acquisition. Additional images from the acquisitions where α was set to 15° can be found under section B.1 in Appendix B.
3.1.1 Magnitude Images
The magnitude images from the reference acquisition are presented in figure 3.1.1 and the gray scale corresponds to the signal intensity. In each cardiac phase shown in figure 3.1.1, the blood signal is present and can be observed in both ventricles of the heart, as well as in the surrounding blood vessels. Flow artifacts are also present and appear as a dark area under the blood pool.
Figure 3.1.1: The magnitude images of all cardiac phases from the conventional phase contrast acquisition that served as a reference. The number at the bottom left corner of each image states the cardiac phase shown. In each image, the presence of the blood signal is seen in both ventricles of the heart. Flow artifacts appear in the form of a dark area under the blood pool of the heart.
The magnitude images from the acquisition with ke set to 1000 and α = 10° are presented in figure 3.1.2. It is apparent that the signal strength decreases during the cardiac cycle. The fading signal leads to a signal-to-noise ratio (SNR) that decreases with time.
Figure 3.1.2: The magnitude images corresponding to ke= 1000 and α = 10°. The number at the bottom left corner of each image states the cardiac phase.
The magnitude images from the acquisition with ke= 1500 and α = 10° are presented in figure 3.1.3. Similar to the images in figure 3.1.2 where ke = 1000, the signal fades during the cardiac cycle which results in a low SNR, especially in the later cardiac
phases.
Figure 3.1.3: The magnitude images corresponding to ke= 1500 and α = 10°. The number at the bottom left corner of each image states the cardiac phase.
The magnitude images from the acquisition with ke= 2000 and α = 10° are presented in figure 3.1.4. Similar to the previously described acquisitions where ke = 1000 and ke = 1500, the signal fades throughout the cardiac cycle.
Figure 3.1.4: The magnitude images corresponding to ke= 2000 and α = 10°. The number at the bottom left corner of each image states the cardiac phase.
Visual comparison with the reference suggests that the SNR is lower in the magnitude images where the STEAM preparation was applied, as seen in figure 3.1.2, 3.1.3 and 3.1.4. The low SNR comes with the inherently low signal strength in STEAM imaging.
An important observation in the magnitude images of figure 3.1.2, 3.1.3 and 3.1.4,
is that the blood signal was suppressed in both ventricles of the heart in the images produced with the proposed sequence. The black-blood effect was also maintained over the course of the cardiac cycle, despite the fading signal. The early cardiac phases in the black-blood acquisitions show that the dark area under the blood pool is not as prominent compared to the reference images in figure 3.1.1. Figure 3.1.5 highlights the difference in blood signal when the STEAM preparation was applied. The suppression of the signal in the surrounding blood vessels can also be observed in figure 3.1.5.
Figure 3.1.5: The magnitude images of cardiac phase 2 from the conventional phase contrast acquisition and an acquisition where a STEAM preparation was applied with ke = 2000 and α = 10°. a) The phase contrast acquisition where the blood signal is present in the ventricles of the heart and the areas marked in orange show blood signal in blood vessels. b) Application of the STEAM preparation suppressed the blood signal in both ventricles as well as the blood vessels marked in orange.
3.1.2 Phase Images
The phase images from the reference acquisition are presented in figure 3.1.6. The gray scale in the phase images represents velocities within the VENC range of ±30 cm/s used in the imaging experiments described in section 2.2. In the phase images of the reference acquisition in figure 3.1.6, velocities can be measured in the ventricles. Flow artifacts are also present and can be seen as increased noise in the area under the blood pool.
Figure 3.1.6: The phase images of all cardiac phases from the conventional phase contrast acquisition. The number at the bottom left corner of each image states the cardiac phase shown.
Velocities within the VENC range of±30 cm/s are seen in the ventricles of the heart.
The phase images from the acquisition with ke set to 1000 and α = 10° are presented in figure 3.1.7. The noise in certain areas of the phase images corresponds to an absence of signal, while the smooth gray areas represent stationary tissue. The noise in the ventricles of the heart and surrounding blood vessels indicates the absence of blood signal.
Figure 3.1.7: The phase images corresponding to the encoding strength ke set to 1000 and α set to 10°. The number at the bottom left corner of each image states the cardiac phase shown.
The phase images from the acquisition with ke = 1500 and α = 10° are presented in figure 3.1.8. Similar to the images in 3.1.7, the blood signal is absent in the ventricles of the heart.
Figure 3.1.8: The phase images corresponding to the encoding strength ke set to 1500 and α set to 10°. The number at the bottom left corner of each image states the cardiac phase shown.
The phase images from the acquisition with ke = 2000 and α = 10° are presented in figure 3.1.9. Similar to the previously described sets of phase images seen in figure 3.1.7 and 3.1.8, the blood signal is also suppressed in the ventricles for this encoding strength.
Figure 3.1.9: The phase images corresponding to the encoding strength ke set to 2000 and α set to 10°. The number at the bottom left corner of each image states the cardiac phase shown.
The phase images presented in figure 3.1.7, 3.1.8 and 3.1.9 share a similar pattern where the noise increases during the cardiac cycle, which is due to the fading signal strength that was described in section 3.1.1. The fading intensity in the magnitude images corresponds to an increase of noise in the corresponding phase images.
Compared to the reference in figure 3.1.6, the phase images from all the STEAM preparation acquisitions show suppression of the blood signal in the ventricles and the surrounding blood vessels. The images suggest that the blood signal was absent in areas where it was previously present, and an example is presented in figure 3.1.10.
Figure 3.1.10: The phase images from cardiac phase 2 from the reference acquisition and an acquisition where a STEAM preparation was applied with the encoding strength 2000 and flip angle 10°. a) Different blood velocities are present in the reference images and can be measured in the ventricles. The areas marked in orange show velocities in the surrounding blood vessels.
b) Application of the STEAM preparation suppressed the blood signal in both ventricles and the blood vessels marked in orange, making noise appear in these areas due to the absence of signal.
The black-blood effect in the magnitude images was maintained throughout the cardiac cycle when the STEAM preparation was applied. The blood suppression is also visible in the phase images since noise appears in the ventricles of the heart in all cardiac cycles presented in figure 3.1.7, 3.1.8 and 3.1.9.
3.2 Region of Interest Measurements
The ROI in the septal wall of the myocardium and RV blood pool used for image intensity measurement are shown in figure 3.2.1 together with the ROI used for velocity measurements that was also located in the septal wall of the myocardium. The plots of the blood image intensity, contrast and velocity included the values from cardiac phase 1 to 8, since the reference acquisition had 8 data points. The values measured for the acquisitions where α = 15° were also included in the plots.
Figure 3.2.1: The ROIs used for image intensity and velocity measurement. a) The myocardial mask in the septal wall of the myocardium, marked in white. b) The mask in the blood pool of the RV, marked in white. c)The ROI that was used for velocity measurements (white) in the septal wall of the myocardium. Each ROI consisted of 4 voxels.
3.2.1 Blood Image Intensity
The plot of the image intensity in blood is presented in figure 3.2.2. The values in the reference curve appear to be higher compared to those of the black-blood acquisitions.
Figure 3.2.2: Plot of the image intensity values for blood in cardiac phase 1 to 8, measured in an ROI placed in the RV blood pool. The horizontal axis shows the cardiac phase and the longitudinal axis shows the image intensity value. The reference is represented by the dashed black line (PC), while the black-blood acquisitions are represented by the colors seen in the plot legend on the top right corner.
The image intensity of blood in figure 3.2.2 appears to be lower in the first cardiac phase and decreases in the later cardiac phases. It can also be noted that the initial intensity value seems to be lower when the encoding strength increases, suggesting that a higher encoding strength suppresses the signal more. The plot in figure 3.2.2 also shows that the blood signal is kept low through all the cardiac phases. The results from the performed Wilcoxon signed-rank tests with the image intensity values in the RV are summarized in table 3.2.1.
Table 3.2.1: The resulting values from the Wilcoxon signed-rank test based on the measured image intensity values in the blood for phase contrast and each CSPAMM encoding strength in cardiac phase 1 to 8. Data set 2 follows the format: CSPAMM encoding strength_flip angle.
The null hypothesis was that there was no difference in signal intensity. With a significance level of 5%, h = 1 indicated that the null hypothesis was rejected and h = 0 indicated that the null hypothesis was not rejected.
Data Set 1 Data Set 2 p-value h
PC 1000_10 0.01 1
PC 1500_10 0.01 1
PC 2000_10 0.01 1
PC 1000_15 0.01 1
PC 1500_15 0.01 1
PC 2000_15 0.01 1
The graph in figure 3.2.2 shows that there is a difference in the image intensity of blood between the reference acquisition and black-blood acquisitions, and the results from the significance test confirms this as well since h= 1 when each black-blood acquisition is compared with the conventional phase contrast acquisition.
3.2.2 Contrast Values
The plot of the contrast values is presented in figure 3.2.3. A high contrast value between the myocardium and blood allows for more accurate delineation of the myocardium.
Figure 3.2.3: Plot of the contrast values calculated based on the mean image intensity values measured in cardiac phase 1 to 8 for all image acquisitions. The horizontal axis shows the cardiac phase and the longitudinal axis shows the calculated contrast value. The reference is represented by the dashed black line (PC), while the black-blood acquisitions are represented by the colors seen in the plot legend on the top right corner.
The contrast values seem to increase with the encoding strength, which could be related to the observation described in section 3.2.1 where the blood signal strength decreases as keincreases. An increasing difference in relative image intensity between myocardium and blood leads to a higher contrast value. The results from the performed Wilcoxon signed-rank tests with the contrast values are summarized in table 3.2.2.
Table 3.2.2: The resulting values from the Wilcoxon signed-rank test based on the contrast values of the reference and each CSPAMM encoding strength in cardiac phase 1 to 8. Data set 2 follows the format: CSPAMM encoding strength_flip angle. The null hypothesis was that there was no difference in contrast. With a significance level of 5%, h = 1 indicated that the null hypothesis was rejected and h = 0 indicated that the null hypothesis was not rejected.
Data Set 1 Data Set 2 p-value h
PC 1000_10 0.15 0
PC 1500_10 0.01 1
PC 2000_10 0.04 1
PC 1000_15 0.05 0
PC 1500_15 0.01 1
PC 2000_15 0.01 1
According to table 3.2.2 the difference between the reference and the acquisitions with ke = 1000 and α = 10°, as well as α = 15°, was not significant. The two acquisitions are represented by the blue and green curves in figure 3.2.3 with contrast values close to those of the reference. A mean difference could however be confirmed for the other black-blood acquisitions with higher CSPAMM encoding strengths.
3.2.3 Velocity Measurements
The plot of the mean velocity in each cardiac phase in all acquisitions is presented in figure 3.2.4. The application of the STEAM preparation should not interfere with the myocardial tissue velocity measurement, which entails that the values measured with the black-blood acquisitions should be as similar to the reference values as possible.
Figure 3.2.4: Plot of the measured mean velocity within the ROI for cardiac phase 1 to 8 of all acquisitions. The horizontal axis shows the cardiac phase and the longitudinal axis shows the image intensity value. The reference is represented by the dashed black line (PC), while the black-blood acquisitions are represented by the colors seen in the plot legend on the top right corner.
The velocity curves in figure 3.2.4 are centered around the reference, suggesting that the velocity measurements with the proposed sequence are not completely distorted. The results from the performed Wilcoxon signed-rank tests are summarized in table 3.2.3.
The significance test results confirm that there is no mean difference in velocity between the reference and all acquisitions except for the one with ke = 1000 and α = 15°.
Table 3.2.3: The resulting values from the Wilcoxon signed-rank test based on the measured mean velocity values with phase contrast and each CSPAMM encoding strength in cardiac phase 1 to 8. Data set 2 follows the format: CSPAMM encoding strength_flip angle. The null hypothesis was that there was no difference in mean velocity. With a significance level of 5%, h
= 1 indicated that the null hypothesis was rejected and h = 0 indicated that the null hypothesis was not rejected.
Data Set 1 Data Set 2 p-value h
PC 1000_10 0.02 1
PC 1500_10 0.01 1
PC 2000_10 0.04 1
PC 1000_15 0.15 0
PC 1500_15 0.02 1
PC 2000_15 0.01 1
Discussion
Blood suppression is evident in the magnitude images produced with the proposed pulse sequence, as seen in figure 3.1.2, 3.1.3 and 3.1.4. The signal difference in blood when phase contrast was compared to the acquisitions with STEAM preparation is also apparent in the plot of the image intensity in figure 3.2.2. Furthermore, the results from the Wilcoxon signed-rank test in table 3.2.1 confirm the signal difference as well.
All CSPAMM encoding strengths utilized in the imaging experiments were sufficiently high to ensure that the black-blood effect was present in all acquisitions where the STEAM preparation was applied. Careful selection of the encoding strength is not crucial as it is only necessary to set the encoding strength high enough to achieve a black-blood contrast. This allows focus to be placed on the optimization of other imaging parameters instead.
According to figure 3.2.3, the contrast seems to increase with the encoding strength ke which may be due to the larger gradient amplitudes causing more effective suppression of the blood signal by inducing intravoxel phase dispersion. The decrease in blood signal with increasing encoding strength can also be observed in figure 3.2.2 where the initial image intensity value decreases when ke increases. Considering the resulting values from the significance test in table 3.2.2, in the cases where the p-value ≥ 0.05, no exact conclusions could be drawn, which was the case for ke = 1000in combination with α = 10° as well as α = 15°. However, a mean difference between conventional phase contrast and the black-blood acquisitions could be stated for the higher encoding strengths. While increasing ke may result in more effective suppression of the blood signal, setting the value too high may also lead to myocardial signal loss.
A disadvantage with STEAM is the signal loss that occurs during the cardiac cycle. The fading signal is a result of the magnetization being consumed throughout the imaging process. A possible solution is to use a ramped flip angle scheme, in which the flip angle is increased to compensate for signal loss. However, this may only serve to make the images more visually appealing, with little effect on the measurements of quantitative values [20]. It is also possible to do more SPAMM preparations, known as N-SPAMM [21], however as the blood saturation shown in this work appears to be satisfactory, additional SPAMM would only serve to prolong the breath hold.
It can be observed in the magnitude images of figure 3.1.2, 3.1.3 and 3.1.4 that the blood saturation is maintained during the cardiac cycle. Figure 3.2.2 also shows that the blood is suppressed in all cardiac phases. This is an advantage of the proposed pulse sequence compared to more conventional black-blood methods such as double inversion recovery and spatial presaturation where the black-blood effect is only attained at certain time points or only for a short time span. The STEAM preparation does not have to be repeated as frequently compared to the conventional methods in order attain a black-blood effect that lasts during cardiac cycle, which can reduce the total scan time.
It is important that myocardial velocity measurements are not obscured by the application of the STEAM preparation. The mean velocity values plotted in figure 3.2.4 are centered around the line representing the conventional phase contrast acquisition, which suggests similar accuracy in the measurements. The results from the statistical calculations presented in table 3.2.3 also suggest similar accuracy. Good agreement between phase contrast and the black-blood acquisitions can be confirmed for all CSPAMM encoding strengths and flip angle combinations except for the acquisition with ke= 1000paired with flip angle 15°. Only 8 data points from each acquisition were used, which contributed to a low statistical power. A larger number of data points could allow for more proper evaluation of the accuracy in the velocity measurements.
Adding the CSPAMM preparation does not obscure the velocity measurements, but loss in SNR is a problem, especially in the later cardiac phases as seen in figure 3.1.7, 3.1.8 and 3.1.9. Some SNR could be regained using a prolonged acquisition, however, this would increase the breath hold duration. The problem could potentially be solved using a free-breathing approach, though it is unclear what effect that would have on the quantitative values. Doppler echocardiography, which is considered the gold standard for myocardial tissue velocity measurements is routinely performed under free breathing. Another possible solution could be to perform the measurement under free breathing combined with respiratory triggering.
The high temporal resolution in cine phase contrast imaging makes double inversion recovery, or inversion recovery methods in general, not ideal for black-blood imaging of the cardiac cycle mainly due to the black-blood contrast only being achieved at a certain time point. De Rochefort et al. [12] have previously described velocity selective gradients that could potentially be applied in black-blood imaging through saturation of the signal from spins with a certain velocity. However, it is speculated that the time penalty of performing this method is far greater than the time penalty of performing STEAM imaging. In the sequence proposed by De Rochefort et al. the excitation portion alone had a duration of 36 ms [12], which is considerably longer than the TR of 9.22 ms utilized in this work.
Cartesian black-blood phase contrast imaging has previously been performed with spatial presaturation [22]. Spatial presaturation requires a saturation module to be applied every TR [9, 22] or repeated very close in time. The STEAM preparation in this work was approximately 8.38 ms, while previous papers have described the use of saturation modules that were 11 ms [9] or 12 ms [23] long. The STEAM preparation is only applied once per heartbeat, while each TR in spatial presaturation is accompanied by a fix time cost due to the saturation module. A thorough comparison of the time efficiency of different black-blood methods is considered to be outside the scope of this project. However, a simple hypothetical experiment could be performed to give an idea of the time gain that comes with the proposed method. Considering a
Cartesian readout, the number of k-space lines acquired during a heartbeat could be multiplied with the duration of a spatial presaturation module, and subtraction of the duration of a single STEAM preparation would then correspond to the time gain. The frequent application of the saturation module will prolong the scan time and hamper the temporal resolution in phase contrast imaging.
A limitation of the study was that the imaging experiments were performed with a single healthy volunteer. The results from the imaging experiments have shown that black-blood imaging is possible with the pulse sequence developed in this thesis and a larger investigation on multiple subjects is warranted. The blood flow can vary in patients with pathological heart conditions, and a study with both healthy subjects and patients could be of interest to observe how different types of flow may affect the blood suppression. More subjects could also provide useful insights for parameter optimization, but this is left for future work.
To the author’s knowledge, no previous work has been published about a STEAM-based preparation combined with phase contrast imaging to achieve a black-blood effect as in the pulse sequence developed in this paper. The results produced in this thesis project showed that the STEAM preparation is compatible with phase contrast imaging. The preparation in the proposed sequence could potentially be implemented into existing sequences, such as the method proposed by Fyrdahl et al. [14] where myocardial tissue velocity measurement was performed with an accuracy comparable to Doppler echocardiography. The STEAM approach could be used as an alternative to spatial presaturation to allow for imaging with higher temporal resolution and quantitative measurements without the disturbance of blood flow.
Conclusions
In this work, a pulse sequence was developed where a STEAM-based preparation was combined with a conventional phase contrast sequence to achieve black-blood contrast.
The black-blood effect was attained in the produced images and the blood suppression could be maintained during the cardiac cycle, which may allow for black-blood imaging with the high temporal resolution of phase contrast MRI. Velocity measurements with the proposed sequence were not obscured by the application of the STEAM preparation when compared to conventional phase contrast. Furthermore, it was shown that careful selection of the encoding strength is not crucial for achieving satisfactory saturation of the blood signal, thus allowing a wide latitude for optimization of other imaging parameters. This work showed that black-blood phase contrast imaging is possible by using a stimulated echo approach.
5.1 Future Work
The developed pulse sequence could be compared with other blood suppression methods, such as spatial presaturation or velocity selective excitation. A comparison with Doppler echocardiography could also be performed in a larger cohort of healthy volunteers or even patients with heart failure to determine the clinical relevance of performing black-blood phase contrast. Whether the extra time spent on blood saturation, which increases the imaging time by a factor of 2, actually translates into a clinical benefit could also be investigated. An advantage with the proposed sequence is that both strain and velocity data can be extracted from the same acquisition, as it is a fully functioning DENSE sequence, albeit only in the slice direction and with a limited mixing time. The implications of performing longitudinal strain measurements using this data was outside the scope of this paper and is thus left for future work.
State of the Art
This state of the art provides an overview of magnetic resonance imaging (MRI) and concepts within the field that are relevant for this thesis project. The central topics are phase contrast imaging, black-blood methods and stimulated echo acquisition mode (STEAM). The purpose is to present what has been done within each area and provide a context for the experiments performed in the thesis project.
A.1 Magnetic Resonance Imaging
MRI is a non-invasive imaging modality that allows acquisition of detailed tomographic images without the risks of ionizing radiation. MRI is not limited to static structures, motion can also be imaged and techniques in cardiovascular magnetic resonance (CMR) imaging have made it possible to visualize and quantify myocardial motion and blood flow.
A.1.1 MR Physics
Elementary particles have an inherent characteristic called spin, that generates a magnetic moment and the spin of the hydrogen nucleus is used in MRI. When proton spins are placed in a strong external magnetic field (B0), such as the fields used in MRI, a distribution of spins is created. More spins will be aligned parallel to B0, which in turn results in a net magnetization in the same direction. If the magnetization is tilted away from the direction of B0, the vector will precess around the direction of the field. The precession occurs with a frequency called the Larmor frequency (ωL), which is given by:
ωL= γ· B0 (A.1)
Where γ is the gyromagnetic ratio for hydrogen. In MRI, a radiofrequency (RF) pulse is applied to excite the spins and change the orientation of the net magnetization vector, which can then be divided into a longitudinal and transversal component. Once the RF pulse has ended, the magnetization will return to equilibrium. The recovery of the longitudinal component is called T1 relaxation, and the decay of the transverse magnetization is known as T2 relaxation.
Spatial localization during imaging is done with magnetic field gradients that introduce a linear variation of the magnetic field, as well as the Larmor frequency, in the imaging volume. The change in Larmor frequency due to a gradient (ωG) along the x-axis can be written as:
ωG = γ G(t)x(t) (A.2)
where G is the amplitude of the applied gradient. Integrated over time t, the frequency change caused by a gradient will correspond to a change in phase (ϕG) [24]:
ϕG(t) = Z
γ G(t)x(t) dt (A.3)
A.1.2 The MR signal
In MRI, the transverse magnetization Mxy forms the signal. The MR signal S is complex valued and depends on many different factors, such as spin position ⃗r, spin density ρ, T1 and T2 relaxation time constants. The MR signal can be described by the equation:
S(⃗r, t) = f (M 0(⃗r, 0), T1, T2, ρ)
| {z }
Magnitude factor
· |{z}eiϕ
Phase factor
(A.4)
where M 0 is the initial value of the net magnetization. The complex MR signal does also have a phase ϕ and there are various sources that can contribute to this phase.
Information can be encoded into the signal phase with the help of magnetic field gradients, it is therefore desired to minimize the contributions from other sources. Phase from eddy currents can be minimized [25] and other sources such as Maxwell terms [26]
can be corrected for, but an off-resonance phase ϕof f will in practice always be present.
The off-resonance phase can come from B0 field inhomogeneity and susceptibility effects [27]. The phase of the MR signal can be written as:
ϕ = ϕof f + γ Z τ
0
G(t)⃗ · ⃗r(t) dt
| {z }
ϕG⃗
(A.5)
where ϕG⃗ is the gradient phase contribution [28] and τ is the duration of the gradient.
The evolution of the spin position ⃗r(t) can be expressed as a Taylor series:
⃗
r(t) = ⃗r0+ ⃗v0t + 1
2⃗a0t2+ ... (A.6)
which can be substituted into equation A.2 [29]. The third and higher order terms are often small and can be disregarded, resulting in the following expression for the phase of the complex MR signal:
ϕ = ϕof f + γ Z τ
0
G(t)⃗ · [⃗r0+ ⃗v0t] dt (A.7)
A.1.3 Gradient Moments
A gradient moment is used to describe the effect of a gradient on the MR signal. The zeroth gradient moment is related to position, the first is related to velocity, the second is related to acceleration and so forth. The zeroth gradient moment M0 is the gradient amplitude integrated over time:
M0 = Z τ
0
G(t) dt⃗ (A.8)
while the first gradient moment M1 is given by:
M1 = Z τ
0
G(t)⃗ · t dt (A.9)
The gradient moments are usually evaluated at the time of the echo, meaning
τ
=TE.Substituting the two previous integrals in A.8 and A.9 into equation A.7, the phase of
the MR signal along a certain direction can be expressed as:
ϕ = ϕof f+ γ Z τ
0
G⃗ · ⃗r0 dt + γ Z τ
0
G⃗ · ⃗v0t dt = ϕof f+ γM0· ⃗r0+ γM1· ⃗v0 (A.10)
Gradient moments are often considered when designing pulse sequences and can be designed for applications in flow imaging, which will be described in the following section.
A.2 Phase Contrast Imaging
MRI is inherently sensitive to motion [6] and the phase of the MR signal can be encoded with velocity information [28]. Phase contrast imaging allows for quantification of blood flow and myocardial tissue velocity in CMR [2, 6].
A.2.1 Physical Principles of Phase Contrast
Two-dimensional phase contrast MRI allows measurement of the velocity component perpendicular to the image plane [2], also known as through-plane velocity encoding [30]. Bipolar gradients are used to create phase shifts for moving spins that are proportional to their velocities [2]. Only the spins that have moved along the direction of the applied gradient will accrue a phase, while the phase of static tissue will remain the same [28, 29].
Figure A.2.1: Phase shift induced by a bipolar magnetic field gradient. Spins that move (red) along the direction of the applied bipolar gradient will accrue a phase, which will in turn be proportional to the velocity of the spins. Stationary spins (green) will not accrue a phase.
The phase contribution from gradients can be designed with the M0 and M1 values so that pulse sequences can compensate for flow [31] or become flow sensitive [29].
Different types of motion during data acquisition can cause artifacts in the resulting images, and blood flow while imaging the heart can cause ghosting artifacts along the phase encode direction or signal loss due to intravoxel dephasing of spins [2, 31].
Figure A.2.2: Flow artifacts in a short-axis image of the heart with the image plane placed below the mitral valve, showing both ventricles. a) In the magnitude image, a lower signal intensity is apparent within the area marked in red. b) In the same area of the corresponding phase image, also marked in red, more noise is apparent.
Flow-related artifacts can obscure quantitative measurements with phase contrast MRI and flow compensation during data sampling can reduce the artifacts in the resulting images [2]. A pulse sequence with an M0 and M1 that are both zero at the time of the echo is said to be flow compensated, since there will be no additional phase contribution from moving spins [2].
ϕ = ϕof f+γM0· ⃗r0+γM1· ⃗v0 (A.11) The phase of the complex MR signal mentioned in A.1.2 can be encoded with information about flow [28]. It is possible to create flow sensitive pulse sequences where the moving spins introduce a phase shift that is proportional to the velocity.
The gradients have to be designed so that M0 is zero, but M1 is non-zero.
ϕ = ϕof f+γM0· ⃗r0+ γM1· ⃗v0 (A.12)
A.2.2 Image Acquisition
Image acquisition will result in a magnitude image and a velocity map where the signal intensity corresponds to a specific velocity. The velocities are proportional to the phase accrual that was caused by the spins movement through the bipolar gradients [2]. Two acquisitions are needed and it is typically a flow compensated and a flow sensitive acquisition [2]. Phase difference reconstruction is then performed where the phase images are subtracted [2], resulting in zero phase shift for static tissue and non-zero phase shift for moving spins [28, 29]. The phase contributions from other sources ϕof f is the same for both data sets and will cancel out in the subtraction [2, 30]. The phase difference ∆ϕ can be written as: [28].
∆ϕ = γ· ∆M · v (A.13)
where ∆M is the difference in first gradient moment and v is the velocity of the moving spins. The velocities are scaled so that the maximum and minimum velocity correspond to a phase shift of ±
π
[2]. The velocity encoding (VENC) is the maximum velocity in the image and is a parameter set by the operator [2]. In flow measurements with phase contrast MRI, wraparound artifacts arise when the measured velocity surpasses the maximum or minimum velocity that is encoded in the pulse sequence [2].Figure A.2.3: Relating the phase shift to the velocity of moving spins and image intensity in phase contrast. A phase shift of±
