Towards quantitative cardiac imaging
Citation for published version (APA):
Gutteling, J. W. A. (2009). Towards quantitative cardiac imaging. (School of Medical Physics and Engineering
Eindhoven; Vol. 2009004). Technische Universiteit Eindhoven.
Document status and date:
Published: 01/01/2009
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
Where innovation starts
Technische Universiteit
Eindhoven
University ofTechnology
School of Medical Physics
and Engineering Eindhoven
TU
Appendices
Towards
SMPE/e nr 2009/004
Certificate date
:
Spr
i
ng 2010
Quantitative Cardiac lmaging
Ir. Job W.A. Gutteling
25-08-2009
Appendices part of rep011: Towards Quantitative Cardiac Imaging
SMPE/e nr: 2009/004
Date: spring 2010
CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN
Gutteling, Job
Towards Quantitative Cardiac Imaging
/
by Job Gutteling - Eindhoven : Technische
Universiteit Eindhoven, 2009.
-
(School of Medica! Physics and Engineering Eindhoven
: project reports; 2009/004)
ISBN 978-90-386-1954-5
NUR 954
Keywords: Cardiology / 3D ultrasound
/
Magnetic Resonance Imaging
/
Software
Appendices - Contents
LITERATURE RESEARCH AND STUDY
APPENDIX A-APPROPRIATE IMAGING FOR VARIOUS CARDIAC PROBLEMS APPENDIX B - LITERATURE STUDY -VOLUMETRY
APPENDIX C- LITERATURE STUDY- HARDWARE AND SOFTWARE
METHODOLOGY
APPENDIX D - 3D ECHOCARDIOGRAPHY APPENDIX E - CARDIAC MRI
APPENDIX F - ABOUT ECG GATING AND TRIGGERING APPENDIX G - CARDIAC MRI ACQUISITION PARAMETERS APPENDIX K- THE RESOLUTION OF VOXEL BASED DATASETS APPENDIX L - CONTOUR DETECTION IN MEDICAL DATA
ACQUISITION AND ANALYSIS PROTOCOLS
APPENDIX M - ANALYSIS PROTOCOL CAAS MRV APPENDIX N - TOMTEC MR ACQUISITION
RAWDATA
APPENDIX R- RAW DATA HEALTHYVOLUNTEERS APPENDIX S- RAW DATA PATIENTSTUDY
REFERENCES AND PROCEEDINGS
APPENDIX Y - SUBMITIED POSTERS, ABSTRACTS AND PAPERS FROM THIS STUDY APPENDIX Z - REFERENCES
Appendix A
Appropriate lmaging for various cardiac problems
lntroduction Current literature calls cardiac MRI the golden standard for left ventricular function analysis. The question arises what this means, as cardiac MRI in general is still not widely used in hospitals, so the golden standard is not available for the majority of
examinations.
Further proof of this statement can be appreciated with the table below, that lists the appropriate imaging study for several cardiac clinical problems, adapted from the standard work Radiology, second edition (2007) [reference].
Clinical Problem Most cardiac problems
Congestive heart failure (new or worse)
Congestive heart failure (chronic)
Hypertension (suspected essential)
Hypertension
(suspected renal artery stenosis) Left ventricular ejection fraction Chest pain or shortness of breath (suspected pulmonary embolism) Shortness of breath
(suspected cardiac origin)
Acute chest pa in
(suspected myocardial infarction [6 hours]) Chronic chest pa in
(suspected cardiac origin) Coronary ischemia
Congenital heart disease Endocarditis
Valvular disease Pericardial effusion Constrictive pericarditis
lmaging Study lnitial posteroanterior and lateral chest radiography
Chest radiography and ejection fraction; wall motion evaluation by nuclear medicine or echocardiography
Chest radiography
Nuclear medicine renogram or magnetic resonance angiogram
Gated nuclear medicine blood pool study or echocardiography Chest radiography and nuclear medicine ventilation/perfusion scan
Chest radiography and echocardiography or nuclear medicine myocardial perfusion study
Electrocardiography, chest radiography, and coronary angiography
Electrocardiography, chest radiography, nuclear medicine myocardial perfusion or coronary angiogram
Electrocardiography; if negative then stress electrocardiogram, nuclear medicine, myocardial perfusion study, or stress echocardiogram; if
positive, then coronary angiogram
Chest radiography, echocardiography, or cardiac catheterization
Echocardiography Echocardiography Echocardiography
Aortic trauma
Thoracic aortic dissection
Abdominal aortic aneurysm
Deep venous thrombosis Carotid bruit
Angiography or computed tomography with contrast Computed tomography with contrast or transesophageal ultrasonography
Computed tomography with contrast if symptomatic; ultrasonography for screening or follow-up
Duplex ultrasonography
Duplex ultrasonography; if high-grade stenosis, then contrast angiography
Almost none of these examinations can only be assessed by MRI, although for several cardiac studies, MRI examinations are available. In bold, left ventricular ejection fraction is listed as having either MUGA or 20 Echo as imaging standard. These are of course current standards, as opposed to the possible standard of the future that MRI may well be.
The most important reasons not to use MRI are spatial resolution and duration of the examination, as 20 echo, CT and conventional radiography are
fa
ster and have higher spatial resolutions, although contrast especially with 20 echo is bad compared to MRI.Conclusion MRI may be the imaging tool of the future for cardiac imaging, but as of today there are only a few cardiac studies that are routinely performed with MRI.
Appendix B
Literature study-volumetry
lntroduction lt could be beneficia! to patients if CMR and 30 echo examinations provided intercomparable results.
In literature, several studies were performed to compare the two methods. This is an analysis of those studies. Goal is to find if a direct comparison is already possible, or what needs to be done to make this comparison happen.
This appendix is an adaption from a report by Ellemiek Wintjes, who performed this literature study as apart of her master thesis [reference Z-15]
In current literature, searches were performed on studies that compared 30 echocardiography and CMR. Ten relevant articles were found. Other findings included comparison between 20 en 30 echocardiography, or 20 echo versus either CMR, CT, MUGA or SPECT, sometimes even multiple modalities or multiple analysis methods on one modality. Some studies dealt with either young or special patient populations and were therefore not included. Also, 3T MRI scanners were excluded, so that resolution and contrast differences are not an issue.
Results
Table B-1 shows results from the ten studies. All of them performed at least one comparison between 3DE and CMR, and sometimes two groups of 30E measurements were included, mostly when two different software packages were used for analysis, or two different acquisition techniques for measuring.
One thing that is immediately obvious from all studies is that they mainly report group averages. Reported parameters are the number of patients and the fraction of them that are men, and LV volumes and EF with their respective means and standard deviation (SD). When the latter we re given as a standard error of the estimate (SEE), a correction was performed using the number of patients, so all va lues are stated as SO. Table B-1- Literature results from ten studies performed comparing 3DE and CMR.
EDV (ml) ESV (ml) EF (%)
n mean sd mean sd mean sd
1 Jenkins 2007 CMR 30 168 54 86 50 50 13 3DE 30 153 31 78 26 49 7 3DE 30 142 33 73 42 46 9 2 Jenkins 2006 CMR 110 180 55 93 50 50 13 3DE 110 136 35 72 28 48 10 3DE 110 165 28 83 22 51 8 3 Jenkins 2004 CMR 50 172 53 91 53 50 14 3DE 50 168 29 88 18 50 7 4 Caiani 2005 CMR 20 164 64 94 55 47 16 3DE 20 150 65 89 48 42 17 3DE 20 141 57 79 48 47 16 5 Caiani 2005b CMR 46 168 70 99 69 46 19 3DE 46 162 68 96 64 45 17 6 Krenning CMR 39 218 70 125 69 45 5 3DE 39 198 60 116 58 43 13 3DE 39 200 67 117 65 44 15 7 Soliman 2007 CMR 53 175 51 74 51 61 17 3DE 53 165 50 69 48 61 18 3DE 53 150 48 63 44 61 18 8 Qi CMR 58 139 59 79 57 47 16 3DE 58 117 53 64 46 48 12 9 Lee CMR 25 190 97 93 87 56 15 3DE 25 191 93 97 87 60 15 10 N ikitin 2005 CMR 64 195 72 117 68 44 16 3DE 64 202 74 121 66 43 15
Unfortunately, group results do not necessarily reflect the results an individual patient would have had. Mainly, these results show the large variation in cardiac function that a patient population will exhibit. Same studies report correlation coefficients to indicate whether results are consistent, but it would have been preferable to have some idea as to what error margins can occur for one patient when he/she undergoes 3DE and CMR examinations.
300 -::; .§. 2 50 (tl E ~ 200
g
150 100 50 0 ~ 250!
QI E 200 :::s ö>
150 100 50 0 80 c 70 0 ·.;::; u ~ 60 L.I. c ..g 50 ~ w-40 30 20 10 0 n:::110 n;;30•
- ·-···· IL•
111 •tMH •.mE à /\v~·1,)~ç Ml-: A'Ji>fAJ,;:f'311F 0 +MR •JDE . &.Av.~·.:gc~~. · .Ave>zge 3GEn"30 n=110
~
f
.
0 2 n=30 n=110•
•
•
•
- ---· 0 2 n=•ó i>:Sû -1!
4 5 6 7 8 9 10 n=39 n=64 n=25 n=46 n=20 n=50 n=53f
n=53~
f
3 4 5 6 7 8 9 10 n=53l
•
MR
•
3~
n=2S n=50 n=20 n=46 n=53 n=39 n=64•
•
~
.
•
1•
3 4 5 6 ï 8 9 10Fig. B-1-(A) End-diastolic Volumes for all ten literature studies, (B} end-systolic volumes
Seven out of the ten studies conclude that 3DE LV volumes are lower than their CMR counterpart, but EF va lues are si mi lar because of the inherent relationship. The other three studies find no significant differences whatsoever. This means that 70% of the researchers were aware that differences existed between the methods, but they did not quantify them. lnstead, they published group ave rages in which the small
differences are negligible compared to the huge error margins. Additionally, it is not enough to know that EF is roughly equal, as the LV volumes are also of clinical importance: they may indicate hypertrophy.
Since it is very interesting to knowhow different the va lues may be in the diagnosis of one patient when
both methods are used, a new study has to be performed. This is the main goal of this report.
Conclusion Several studies al ready investigated comparison between 3DE and CMR, but they reported group ave rages and only qua litative results of the volume differences that they did observe.
This means a new study that investigates these differences for an individual patient may be very useful to know whether both methods can be intercomparable.
Appendix C
Literature study - hardware and software
lntroduction In literature, several studies were performed to compare CMR and 3DE examinations. This is an analysis of those studies, regarding the hardware and software that was used to perform those studies.
All the studies use real-time 30 transthoracic echocardiography, so datasets could either be analysed online on the 30E machine or offline using a software package. Table 1 gives an overview of the 30 ultrasound machines and the analysis software used by the different studies. Most studies use a Philips Sonos 7500 and some the more recent Philips iE33. For the analysis of 3DE images Tomtec 40 analysis is the most popular. Custom software includes CIM that is used only on the university in Queensland, Australia and other nondescript software. Table C-1 summarizes these findings.
Table C-1 -3DE hardware and software
Year 3DE machine
Jenkins 2007 Philips Sonos 7500 2006 Philips iE33 2004 Philips Sonos 7500 Caiani 2005 a Philips Sonos 7500 2005 b Philips Sonos 7500
Krenning 2007 Philips iE33 and Philips Sonos 7500 Soli man 2007 Philips iE33 and Philips Sonos 7500
Qi 2007 Philips iE33
Lee 2001 Volumetrie Machine (Ouke University) Nikitin 2005 Philips Sonos 7500
Analysis software
Tom Tee Qlab
v
v
v
v
v
v
v
v
Customv
v
v
v
The results for CMR are less uniform than for 3DE. The scanner type is either one of three major manufacturers, Philips, Siemens or GE, as can be seen in table C-2.
The software with which the analysis needs to be done, however, seems to have no standard, as multiple packages are used. Only one third party package is used, Medis Mass, all other software is either custom-made or from the sa me company that custom-made the MR scanner used to generate the images. This means that there is still a need for standardization with respect to the software used in CMR analysis.
In regard to the time that is required to perform the scan and the analysis, 30E is on average a lot quicker with a scanning time of 3 minutes and analysis time requiring 10-15 minutes (including time where the computer is calculating and no input is needed), while CMR takes at least 15 minutes to make a scan (not
including some 20 minutes worth of positioning and planning) and an equal amount of time to do the
Table C-2 - CMR hardware and software
Year MRmachine Anal~sis software
Custom GE Me dis Siemens
Jenkins 2007 Siemens 1.5 T
v
2006 Siemens 1.5 Tv
2004 Siemens 1.5 Tv
Caiani 2005 a GE 1.5 Tv
2005 b GE 1.5 Tv
Krenning 2007 GE 1.5 Tv
Soli man 2007 GE 1.5 Tv
Qi 2007 2007 Siemens 1 .5 Tv
v
Lee 2001 Philips ACS 1.5 Tand GE 1.5 T
v
Nikitin 2005 GE 1.5 T
v
One thing that was not mentioned in literature, but is interesting nonetheless, is the question of who is doing
the analysis. Of course, all analyses in literature were performed by 'skilled, experienced observers', but
echocardiography is the turf of cardiologists and sonographers, while MRI examinations are performed by technicians and analyzed by radiologists. Practically, this would mean that different people will review their
'own' dataset, and not both datasets. lt would be useful to let them experience the other method as well, or
even let a single person do all the analyses. This is added as a point of attention for our own study.
Conclusions 3DE hardware and software is mostly limited to Philips machines and Tomtec or
Philips software for post-processing.
CMR hardware is always from one of three major vendors, while software can be anything. The re is still need fora third party product that is able to hand Ie all cardiac MR examinations.
CMR acquisition time is longer than for 3DE (no surprise), while analysis time is also
longer (not necessary?).
In the future, wil! there be only one analyst to perform the analysis both methods, or will they remain separate domains?
Appendix D
30 Echocardiography
lntroduction lt is important to knowhow echocardiography developed over the years. This appendix deals with the basic physics and practical advantages and disadvantages of 30
echocardiography.
In this report, Real-time 30 echocardiography is one of the methods used to image the heart of patients. This tool is based on the propagation of soundwaves in the MHz range (inaudible to the human ear) through the human body. Upon reaching a transition between two tissue types, the sound waves will be partially reflected back to the body's surface. These reflected waves can then be detected by the sa me device that emitted the sound waves, called a transducer, and are converted into images. Echo studies are carried out using specialized ultrasound machines. In the Máxima Medical Center, the Philips iE33 was used, see figure 0-1. Since its inception, cardiac ultrasound has been a very popular, easy to use method to image cardiac anatomy and function and blood flow through the heart.
Reconstructed 30 echocardiography indicates that a 30 volume is generated from a set of 20 images. This was the first generation of 30 imaging, as computer power and
Fig. D-1 - The Philips iE33, ultrasound machine, specifically designed for echocardiography
transducer technology was not able to process the large amounts of data in real-time. In real-time 30E, aso-called phased-array matrix transducer is used. In this transducer, multiple recordings are automatically performed to cover the full left ventricle, without moving the transducer.
Ultrasound of different frequencies (in adults usually 2 to 5 MHz) is transmitted by piezoelectricity (see figure 0-2) from the transducer (or probe) which is placed on the patients anterior chest. This method is cal led transthoracic echo (TIE), as opposed to methods that measure from within the esophagus (transesophageal echo) or even blood vessels (intravenous echo). The patient usually lies in the left lateral decubitus position, this means the patient is lying on his left side with his left arm under his head.
This is advantageous to open a large enough acoustic window through the ribs, cal led the apical window, as the entrance point will be approximately the apex of the heart, see figure 0-3. This is the only way that the wedge-shaped 30 volume captures the whole left
ventricle.
The wedge-shaped dataset is often divided into 4 different parts, each part rotated over a certain amount of degrees and acquired in one heartbeat. This is done to obtain the optima! contrast over the whole dataset (volume of interest), as it is still impossible to acquire all data in sufficient quality within one heartbeat. There is another acquisition option that does enable one-heartbeat full datasets, but then the resolution and contrast make it unusable for further qua ntitative analysis.
Tension C ompresslon
+
}
Fig. D-2 - The piezoelectric effect, where pressure and electric potential are made exchangeable. A pressure on the material generates an electric signa!, while an electric signa! makes the material either compress or extend (positive versus negative)
Fig. D-3 - The most common acoustic windows: the apical window is normally used for 30 echocardiography
The four parts are triggered to every other heart beat of the patient to al low recalibration of the transducer and storage of the data. One entire volume is made in a single end-expiratory breath-hold lasting about 10 seconds.
The iE33 ultrasound machine has three different settings for the density in the 30 mode. This density is related to the image resolution, or line density. The line density sets the volume of the image displayed and the pyramidal shaped volume. The higher the density, the smaller the volume. Other settings for the 30 mode include Full Volume Optimization (FV Opt) Live 30 and Non-triggered Full Volume. FV Opt con trol enables a change in resolution of the image to see a larger volume. Table 0-1 gives an overview of the effects of the FV Opt settings. Frame Rate increases the number of acquisition beats so that each subvolume is approximately halved, enabling an enhancement in frame rate. Acq Beats minimizes the number of acquisition beats, decreasing the time it takes to acquire a Full Volume.
Table D-1 -Philips iE33 Full Volume Optimization settings (source: iE 33 user manual) Density VolumeSize Acquisition Beats Frame Rate
Low Largest volume Large volume and Fast with large volume and medium image quality medium image quality Medium Large volume and good Medium volume and Fast with medium volume
image quality good image quality and good image quality High Medium volume and Small volume and best Fast with small volume and
Strangely enough, there seems to be little advantage of measuring a longer time to increase resolution, as it is compromised by movement artefacts, and reconstruction averaging effects. The faster measurements therefore show quick and fairly reproducible results. See figure 0-4 for typical image quality.
Fig. D-4 - Typical 3DE image quality from apical window with a X3-1 transducer. Left image shows end-diastolic frame, right image the end-systolic frame. This is a 4 chamber view, so only one 'slice' of the total volume.
Conclusion 30 echocardiography is a promising method that allows an easy and fast estimation of heart function and anatomy. Limitations are lower image quality compared to regular 20 echo and low contrast, which will probably impede the contour detection.
Appendix E
Cardiac MRI
lntroduction Cardiac MRI is an umbrella term for several methods that investigate cardiac function or
anatomy using magnetic resonance imaging. In this appendix, it is explained which
specific protocol is used in this project, how imaging is performed and what the
advantages and disadvantages of the method are.
Cardiac MR properties
Magnetic Resonance lmaging uses a powerful magnetic field to align the nuclear magnetization of (usually) hydrogen atoms in
water in the body. Radiofrequency fields are then used to
systematically alter the alignment of this magnetization, causing
the hydrogen nuclei to produce a rotating magnetic field
detectable by the scanner. Additional magnetic fields can be used
to manipulate this field and the resulting signals can reconstruct
an image of the body.
The main advantage of MRI is that it provides much greater Fig. E-1- Philips lntera 1.5 T MRI scanner
contrast between the different soft tissues of the body than echo and even CT do, making it especially useful
in several fields, including cardiovascular studies. Another advantage is that MRI is very versatile: different
magnetic field switching sequences can produce very different contrast options, which can be very useful to investigate different pathologies.
MRI also has a few drawbacks: it is expensive, slow and complicated to use (and sometimes to interpret). Also,
some patients may experience claustrophobia in the confines of the scanner tube and pacemakers and comparable devices are counterindications because of the magnetic field.
Cardiac MRI Acquisition
For the creation of a cardiac MRI dataset several steps are required.
1. Patient preparation
a. Placing the patient on the MR table in the scanner bore.
b. Placement of MRI-compatible ECG leads (normal leads would interfere with the magnetic
field and also be interfered by it. This could lead to heating by magnetic induction and burn the patient). The ECG leads are required for the gating of the acquisition as will be
explained in appendix F.
c. Placement of the receiving coil (can be a special (dedicated) cardiac coil or a general body
coil. The first provides a better signal to noise ratio (SNR), because the coil elements are grouped around the heart reg ion, so more of the magnetization signa! can be received)
2. Setting the scanning parameters. This includes the shimming plane, the imaging planes, the slice
thickness, the number of cardiac phases and several other parameters.
3. The actual scanning and storage of the acquired data.
In genera!, these steps are needed for all cardiac imaging studies, however there are several different pathologies that can be detected using CMR, each using its own planning and parameter settings.
Examinations can focus on viability of the heart muscle, heart anatomy, heart valve function, coronary
arteries, aortic valve function, blood flow, perfusion. The cardiac imaging method that is focused on in this
report is general left ventricular function, for which a series of images of multiple angles of the left ventricle is
Planning of the imaging planes for Cardiac Function (Cine) lmaging
When performing an LV function analysis MRI study, one of the first settings is the selection of the reg ion of interest (ROi) and main imaging planes.
The ROi is of course the whole volume (cube) that encloses the heart and is a combination of slices through
the three main planes of the heart. These planes depict the left ventricle in three orthogonal planes: the
horizontal long axis (LA 4-chamber view), vertical long axis (LA 2-chamber view), and short axis (SA) planes.
Regularly an additional fourth view is used: a 3-chamber view or left ventricular outflow tract (LVOT). Since
this view is required by the TomTec software, we..w.iU include the 3-chamber view in the planning process.
Fig. E-2 -Typical imaging planes in cardiac function MR. From left to right: 2-chamber view, 3-chamber view, 4 chamber view, short axis view and the corresponding views in an anatomical model.
The process of obtaining the desired planes is complex. This is caused by the fact that, in contrast to most
other MRI applications, the imaging planes used in cardiac MRI are defined with respect to the orientation of
the heart. These imaging planes are double oblique (this means they have arbitrary angles in all planes)
relative to the conventional axes of imaging (axial, sagittal and coronal}, and they differ from subject to
subject depending on the particular orientation of the left ventricle, which can vary with respect to the body.
The actual selection of the imaging planes is done with the help of the
interactive scan
mode available onPhilips scanners that are customized to do cardiac imaging studies. This mode Iets the user select the
appropriate imaging plane by interactively showing the current image plane as a scout image and the effect
on it when adjusting angles or position. In essence, this is already a scanning sequence that is so fast it can be
visualized realtime, therefore is does not have clinical image quality.
Step 1 Finding a transaxial image through the left ventricle and septum. The approach starts with
routine scout images, which include standard corona! images (vertical plane that divides body in
anterior, or front, and posterior, or back) of the chest. The coronal slices help to assess the position
and orientation of the heart, since this can differ from person to person. From this coronal view, a
transaxial image is positioned through the heart. This image should depict bath left and right
ventricles and the interventricular septum (muscular heart wall between the ventricles).
Step 2 Defining a long axis (2-chamber scout) trom the transaxial image. Based on the axial image
identified in step 1, an oblique coronal slice is positioned through the left ventricle that is parallel
to the interventricular septum and passes through the left ventricular apex. This will produce a
Step 3 Obtaining a short axis view from the 2-chamber scout and the transaxial image. In subjects whose hearts are vertically oriented, the 2 chamber scout view may serve as a good vertical long axis view. In genera!, for true vertical long axis, additional adjustment is needed (see step 5). A plane aligned perpendicular to the long axis of the heart on bath the 2-chamber scout view and the original transverse image gives a short axis view. The short axis view is used for subsequent (re)positioning of the long axis.
Step 4 Defining a horizontal long axis view from a short axis and 2-chamber scout. Using the short axis and the 2-chamber scout, the horizontal long axis (4-chamber view) can be positioned by bisecting the left ventricle in the horizontal plane. As a guide to the horizontal plane, the line
should bisect bath the left and right ventricles and be parallel to the diaphragm.
Step 5 Obtaining a vertical long axis view from the horizontal long axis and short axis. Using the horizontal long axis view (4-chamber) and the short axis view, a true vertical long axis view can be defined by bisecting the left ventricle in the vertical plane. The resulting image shows the left ventricle, left atrium and mitral valves.
Step 6 The 3-chamber view can be acquired by positioning a long axis view orthogonal to the short axis view, similar to the 4-chamber view but tilted obliquely through the left ventricular outflow tract. The resulting views will re present the three (or four) views needed to be able to calculate cardiac function in the software packages. Theoretically, only the SA stack of images could be enough, but it was found that LA
slices improved the accuracy and reproducibility of the measurements [reference Z-21 ]. See figure E-2 for
typical views and figure E-3 for the conventions on axis delineation.
Shor1 axis /
Horizonlal
Shor1 axis /
Ver1ical
long axis 3-Chamber view
Fig. E-3 - Conventional imaging planes of the heart, corresponding to the geometry shown in figure E-2.
Choosing the scanning sequence
Optimizing the scanning sequence is a science in itself. This is one of the most challenging aspects of MRI research worldwide. Based on tissue properties, the existence of movement, desired image quality, contrast
-to-noise ratio, the absence of all kinds of artefacts and several other parameters, the choice of which sequence to use fora specific purpose is paramount to its success.
Cardiac lmaging is a special case of MRI, since the heart is a moving organ, unlike most other internal organs, and large amount of blood are flowing through it, disturbing the magnetic field gradients that are applied to select slices. This means it is very difficult to scan with sequences normally used for anatomy.
Synchronization of acquisitions with motion
After the positioning of imaging planes the actual acquisition can start. The data is typically collected across
multiple heart beats, as most imaging sequences are not fast enough to acquire everything in a single
heartbeat. For some sequences, such as spin echo anatomie imaging or contrast-enhanced infarct imaging,
data are typically collected at the end of the cardiac cycle, when there is minimal motion of the heart du ring
diastole. For functional imaging, such as cine gradient echo imaging or phase contrast flow q uantification,
data is collected throughout the whole cardiac cycle and then partitioned into separate frames. Each frame
corresponds toa short segment of the cardiac cycle and reflects a snapshot of the heart du ring the cardiac
cycle. When viewed togetherrrra cinematic loop these produce a beating heart video clip. In this way cine
MRI enables the viewer to assess cardiac motion.
Cine Gradient Echo lmaging
With Cine Gradient echo (GRE), the sa me slice of tissue is imaged at multiple time points in the cardiac cycle.
The multiple frames are viewed in a cinematic loop, which gives an appreciation for flow and function of the
heart and vessels. Compared to 'normal' spin echo methods, cine GRE pulse sequences are more complex,
because they require that several parameters be adapted to the patient's heart rate and breath holding capabilities.
Cine GRE sequencesare designed to evaluate flow and function of the myocardium and blood vessels. They
are referred to as brig ht-blood sequences, because the signa 1 intensity of the blood is brig ht compared to the
myocardium. How exactly the image contrast is generated depends on the specific pulse sequence.
Standard, so-called spoiled GRE sequences use saturation of stationary tissues, while inflow of blood leads to its high signa! intensity. Steady state free precession (SSFP) sequences produce contrast between blood and
myocardium based on tissue relaxation time differences, that are independent offlow. This means the SSFP
sequences benefit from lower acquisition time or higher spatial resolution, once potential sources of artifacts are eliminated. This makes this sequence very valuable for cardiac imaging.
To achieve optimal MR images the scanned object has to be as motionless as possible. The heart is far from
motionless and, as it is located in the chest cavity, it is also influenced by breathing. This results in a modified
scanning procedure which synchronizes the acquisition with cardiac and breathing motion. Synchronization
with breathing is achieved by scanning du ring end expiratory breath holds. Before a scan starts the subject is
instructed to inhale, exhale and to suspend respiration. Because the time respiration can be suspended is
normally limited to 15 to 20 seconds, the duration of a scan is limited.
Details on this procedure, which is called either triggering or gating, are covered in appendix F.
Conclusion Cardiac MRI is a useful method that can employ different sequences to image the heart
in various ways, depending on the suspected pathology of the patient.
Si nee heart orientation is unique for every patient however, it is never easy to set up a
Appendix F
About ECG gating and triggering
lntroduction Because the MRI k-space data are collected across several heart beats and all cardiac
phases need to be imaged, the acquisition has to be synchronized with cardiac motion. Only if this synchronization is optima!, the images produced accurately reflect the state of the heart during its different stages of contraction and relaxation and have minimal motion artifacts. Synchronization with cardiac motion is achieved with
electrocardiographic (ECG) gating or triggering.
Gating or triggering?
The terms gating and triggering can be confusing, because they are aften used interchangeably. Generally, gating refers to any means relating MR data acquisition to the phase of the cardiac cycle during which the data were acquired. Gating can be either prospective or retrospective (see figure F-1 ).
Triggering is one form of prospective gating, in which the MR sequence is initiated with the R wave. When the data acquisition for the given R-R interval is completed, the scanner waits for the next R wave (see figure F-2). lmaging that beg ins immediately after the R wave starts just before the on set of ventricular systole. For some sequences diastolic images may be desired. To obtain diastolic images with R-wave triggering, a trigger delay can be introduced. This delay of at least 150-250 msec is introduced between the detection of the R wave and the start of imaging.
Many sequences lso be obtained by retrospective gating. Retrospective gating means that the data are acquired continuously, along with a recording of the ECG tracing. After the acquisition, the imaging data are sorted based on the time of the echoes relative to the R-wave. Retrospectively
Prospective gating
1 - - - -Acquisilion - - - + Trigger -j window Retrospective gating
Acquisilion
-Fig. F-1 - schematic drawing of prospective gating sequence compared to retrospective gating. ECG triggering is one form of prospective gating, where acquisition is started at the R peak and last fora set amount of time.
R-R interval
r--t---+---1
Trigger Acquisition Trigger
de lay window window
gated sequences provide information about imaging through the entire cardiac cycle, including the full duration of diastole, provided that the patient's heart rhythm is sufficiently regular on a beat-to-beat basis. Compared to prospectively gated sequences, the image reconstruction of
retrospectively gated sequences is more complex and computationally intensive.
Fig. F-2 - Definition of trigger delay, acquisition window and R-R interval
With retrospective gating, the tempo ral spacing of the frames can be defined by the user, regardless of the true or effective TR of the seq uence. An electrocardiog ram (ECG) tracing de pi cts the electrical activity of the heart. A P-wave, QRS complex and T wave are aften identifiable. The P-wave represents atrial depolarization and the onset of atrial contraction. The QRS complex reflects the electrical activity associated with ventricular depolarization preceding systole. The onset of left ventricular systolic contraction occurs about 50 msec after
the R wave, and contraction lasts for about 150-250 msec. The T wave represents repolarization of the ventricle. Until the next QRS, the ventricle remains in diastole.
There are two main reasons why the synchronization may fail in the scanner: patient arrhythmias and failure
of the system to detect the R wave for triggering. For ECG-triggered sequences, data acquisition assumes a
regular heart rate. Following the R wave, the system begins collecting data, portions of which are assigned to
different k-space do mains corresponding to different time points in the cardiac cycle. lf the heart rate is regular, then all the data collected shortly after the R wave will reflect the left ventricle in systole, while the
data toward the end of the R-R interval will image the ventricle in diastole. lf the heart rate is irregular and a
second heartbeat co mes much earlier as expected, it is possible that the data collected toward the end of the acquisition window which should correspond to diastole, will instead reflect systole. This corrupts the data as the k-space of the images which should depict diastole now contains a mix of systolic and diastolic data. The
acquisition time in subjects with irregular heart rates is also langer than expected, because not all heartbeats
can be used to collect data. Although an occasional irregular beat is tolerable, frequent irregularities cause poor quality and misleading images.
The most problematic source of artifactual triggering is due to the magnetohydrodynamic effect of moving blood within the magnetic field. Electrical charges moving through a magnetic field induce a voltage. Blood contains many charged particles like Na+ and Cl-, among others. When these ions move through blood
vessels in the setting of a magnetic field, a voltage can be detected, particularly du ring systole or the ST
portion of the ECG tracing. Distortion of the ST portion and peaking or elevation of the Twaves results in
faulty triggering wherever the Twave is higher than the R wave. Triggering off the Twave means that much of systole is missed.
Vectorcardiographic (VCG) triggered or gated approaches reduce artifactual triggering from the
magnetohydrodynamic effect. With VCG, the electrical activity of the heart is depicted bath temporally and
spatially, using measured signal from all three leads. Because the orientation of the electrical axis of the heart is different from the artifacts associated with the magnetohydrodynamic effect, the vectorcardiogram is more accurate at detecting cardiac activity. Occasionally, adequate ECG tracings can not be obtained, perhaps because of a subject's body habitus or other interference with signal measurement, such as a large
pericardial effusion. Peripheral pulse gating is a via bie alternative when central gating is not possible. Like
plethysmography, peripheral pulse gating detects the pulse wave of blood as it transits through the finger.
Typically peripheral pulse gating monitors are clipped to the fingertips or toes. Only MR-compatible
peripheral pulse monitors should be used.
Conclusion Several methods of gating, either prospective (triggering) or retrospective, can be used to acquire useful datasets with cardiac MRI. lt is important to know their limitations and
AppendixG
Acquisition parameters
lntroduction
MRI acquisition
This appendix gives an overview of the CMR and 3DE acquisition parameters that were used in this study
Acquisition parameters for the standard protocol
The standard protocol for the MRI acquisition in this study is the following: Long axis: 2, 3 and 4 chamber view one slice per view
Short axis stack (Sax): 12 slices, no slice gap
MRI Scanner settings
Field of view: 40 by 40 cm (45 x 45 cm for 4 chamber view only) Matrix size: 256 by 256
Coil: SENSE body coil
Sequence: SSFP sequence for Philips scanners: B-TFE sequence TR and TE for SAx: TR=3.47 ± 0.07 ms and TE=l.74 ± 0.03 ms TR and TE for LA: 3.31
±
0.06 ms and 1.66±
0.03 msThis results in cine loops containing 50 frames for the LA axis slices and 12 times 50 frames for the SAx stack. A problem with MRI is the determination of the EDV and ESV. Because the moment of closure of the mitral valves is difficult to determine in MRI (partial volume effects and interpolation of several cardiac cycles), the EDV is usually taken as the volume of the LV cavity in the frame just before closure of the mitral valve, or just after the QRS peak on an ECG, or when the LV cavity is largest. For the ESV the volume of the LV with the smallest LV cavity size is taken. In this study EDV is defined as the volume in the first frame of the cine MRI loop.
Global adaptations of the protocol for Tomtec MR
In order to make the MR data acquisitions compatible with the Tomtec MR software, some changes
were made to the standard acquisition protocol as was discussed in Appendix E:
every dataset needs to have a stack of short axis (SAx) images and three equiangular long axis views (a 2 chamber, 3 chamber and 4 chamber view), which means that they are rotated over 60 degrees around the LA. This LA runs through the apex and the center of the mitral valve.
Another requirement is that the number of frames in the LA images and the SA images is the same. For the planning of the equiangular long axis views a template was used, because radial scanning is not available on the scanner used. This template is placed on the computer screen.
30 echocardiography acquisition
Patient positioning and transducer placement are crucial to acquire the best image. This completely relies on the experience of the sonographer and is therefore very difficult to put into protocol. This means that this is also the main source of error, especially in reproducibility.
Besides the image settings that were discussed in Appendix D, the ultrasound machine determines most parameters depending on the patient. ECG monitoring is used to trigger the acquisition, while patient breath hold is used
!O
minimize movement artifacts during acquisition. Depending on the patient, the number of breath holds can be increased or decreased by imaging larger sub volumes of data (and thus fewer in total) per breath hold. Standard setting is medium density and four sub volumes.The resulting number of frames is different for each dataset, but mostly between 13 and 17. This covers about 70% of the heart cycle, as most of the diastole is often not acquired/shown.
Conclusion CMR ad 3DE acquisition protocols are discussed. CMR is far more difficult to operate, but is also better standardized. The ease of 3DE is also its drawback; every acquisition leaves room for error.
Appendix K
The resolution of voxel based datasets
lntroduction This appendix will explain and show the intrinsic spatial and tempora! resolution of {30) datasets, which may help understand the accuracy of an imaging modality.
Image resolution describes the detail an image holds. Higher resolution means more image detail. Basically,
resolution quantifies how close lines can be to each other and still be visibly resolved. In the case of three
dimensional datasets, each image voxel represents a certain volume in space. In most techniques, all voxels in the dataset will have the sa me dimensions. These dimensions can aften be defined or adjusted by the user, based on the detail required for the method at hand. Especially, MRI and CT are flexible when it comes to resolution, although they obviously have an intrinsic maxi mal spatial resolution that, for MRI, is determined by the gradient field strength. Additionally, the available time for acquisition in, for instance, a heart cycle and the signal-to-noise ratio(SNR) that is needed will influence the resolution that can be successfully managed without compromising image quality.
Spatial resolution in cardiac MRI datasets
For example, all cardiac MR datasets in this study were acquired using a 256x256 matrix in a field of view (FOV) of 40x40 cm and a slice thickness of 1 cm. Voxel si ze therefore was 1.56xl .56x10 mm, which means that a complete dataset of the heart consists of 10-12 short axis 'slices' of 256x256 voxels with dimensions 1.56x1.56x10 mm.
This means that a hypothetical anatomical feature that measures 1 x1 xl mm, will only cover 4% of the volume of one voxel, and therefore they could be easily lost if they have poor {tissue) discrimination (contrast). Fortunately, the main areas of interest in the heart are much larger, and are therefore likely to show up in the data. Even with adequate contrast however, the SNR might be too low to notice the contrast. Additionally, the feature may not exactly be covered in a single voxel, as the voxel boundaries are just like a grid. The signal may therefore even be divided among multiple voxels and be even more spread out and indiscernible as a result. This phenomenon is called a partial volume effect.
Fig. K-1 - Typical short axis view of the heart, erop of original image with 256x256 matrix and
voxel size 1.56xl.56x10 mm. The third
dimension is slice thickness, that cannot be seen
in this picture. No interpolation was used on this
picture.
Cardiac MRI sequences in this study need contrast between heart wall {muscle tissue) and blood (the
endocardial border). The 10-12 short axis slices more or less show a circular ventricular cavity surrounded by
the heart wall; muscle tissue. Figure K-1 shows a typical result. Because the purpose of these image is to enable calculation of endocardial (cavity) volumes during systole and diastole, the better the resolution of the interface between blood and muscle, the better the accuracy of the measurement. More on this is covered in appendix L.
From figure 1, one can see that an exact delineation of the interface/border is not possible. There is a gradient from low signa! in the heart wall to higher signal in the blood, so the real position will remain unclear. To add even more ambiguity, the contrast in every voxel is based on the signal from a slice of tissue 1 Omm thick, which means that a lot of averaging is occurring from possibly several different tissues in the volume. Especially with a heart wall, curved as it is, this influences accuracy (see figure K-2).
Another aspect that has to be taken into account is the repeated
motion of the heart, that emphasizes the parti al volume effects
even more. As fast as some MR sequences are, the heart will probably move during the time the acquisition takes.
Tempora! resolution in cardiac MRI
The issue of heart contraction during acquisition, which is of
course inevitable for in-vivo measurements, leads to the
importance of tempora! resolution. In analogy to spatial resolution, tempo ral resolution is the discrimination of each acquisition in time. A higher tempora! resolution, depicted in frames per second, means the acquisition time has to be short. Of course, the limiting factor will be the minimal time that is needed to acquire enough data fora single image, or in the case of MRI, a line of data. Every image consists of multiple lines of data. In cardiac MR, there is a precarious balance between spatial resolution (=the amount of lines needed), tempora! resolution
(=the number of image frames that depict one heartbeat cycle)
and the ability of a patient to hold his breath.
For example, to image one slice of the heart with 256x256 voxels
and 50 frames per second, assuming the heart rate is 60bpm,
gives every frame a maxi mal acquisition time of 20 ms in which
256 lines of data need to be sampled. This is impossible to achieve
in a single heartbeat, so the data is acquired over multiple heartbeats and then merged toa single image. Of course, the patients has to remain still during this time, and also has to hold
Fig. K-2 - Typical example of how in-plane resolution is much better than long axis resolution, based only on short axis stacks. This clearly demonstrates that LA views can help contrast and thus contour detection.
his breath, so the required time cannot be too long, or motion artifacts will ruin the image. When the spatial resolution is increased in an attempt to create better edge delineation, the number of lines to be acquired becomes even larger, and so does the time the patient has to hold his breath. Since the patients that undergo a cardiac MRI examination are aften not in the best of health, holding their breath for langer than 12 seconds may be difficult. So optimization of the tempora! and spatial resolution leads to faster imaging and coincidentally better quality as unnecessary motion artifacts are avoided.
Spatial resolution in echocardiography
In echocardiography, spatial resolution is similarly defined as in MR, although the datasets themselves
can not be directly compared. Echo transducers have a fixed number of piezocrystals, or elements, that define
their maximum number ofvoxels and resolution in one direction (or two, in case of 30 matrix transducers
that don't have a line of elements but a grid). However, since the sound waves are sent to and received from the tissue, the depth resolution is based on the sample frequency of the transducer, which is again
dependent on pulse duration. The faster this sampling is performed, the smaller the details that can be resolved. Complicating factor is the fact that several transducers (convex transducers) emit divergent beams
of sound waves, which results in the axial spatial resolution to decrease with increasing distance from the
transducer surface, as one beam has to cover a larger area.
Temporal resolution in echocardiography
Tempora! resolution in echocardiography is limited by sample frequency, which in turn is limited by processing power of the ultrasound machine. In recent years, better computer processing has enabled time-resolved three dimensional imaging, aided by new transducers that could acquire these datasets without
having to reposition the transducer, yielding more reproducible results. Tempora! resolution is also influenced by patient heart ra te, since by current standards 20 frames per second is the limit. Patients with higher heart rates will therefore have fewer frames per heart cycle, so motion effects may become more pronounced.
Conclusion Spatial and tempora! resolution play an important role in the reproducible acquisition of datasets with both CMR and 30 echo. A balance has to be found so better resolution does not interfere with available examination time or risks unnecessary motion artifacts.
Appendix L
Contour detection in medical data
lntroduction As the limitations of resolution were discussed, this can now be applied to the process offinding or drawing contours for segmentation in medical data
All software that is used in this report, bath the analysis of 3DE and CMR datasets, performs contour
detection of endocardial borders based on segmentation of tissues, mainly by contrast and depending on
the quality of the software, probably some other advanced computational properties. This means that each software package has a segmentation algorithm that uses one or more mathematica! operations such as the
tracing of edges, tracing local gradients and curvature in contrast or the seeding of (global) contrast areas.
Unfortunately, all the software packages have copyrighted, classified, algorithms that can therefore not be directly compared to each other. However, based on the options featured in each software package, some information can be obtained from the complexity of each of the algorithms.
3DE software: Philips Qlab 6
Philips Qlab (see figure L-1) requires the user to rotate the dataset to an predefined, standardized orientation and define anatomical landmarks that will later be used by the algorithm. Once the landmarks are selected, the contour and ventricular
mesh model are drawn automatically, largely confined within
the landmarks the user specified. This leads to the hypothesis
that the algorithm used a predefined contour shape that can be slightly altered based on the input of the user. However, it can not be completely ignored or overruled by the user, which is illustrated by the lack of adjustment options for the contour
once it has been drawn by the software. Of course, the
anatomical landmarks can be reassigned, and a new contour will be drawn, but this leads toa trial and error approach that
in some cases causes noticeable deviations from what the user Fig. L-1- Interface of Philips Qlab
would judge to be the desired contour based on qualitative assessment.
In conclusion, this algorithm works most of the time, but is
really rigid and no adjustments are possible.
3DE software: TomTec LV Analysis 2.6
The TomTec software (see figure L-2) requires the user to manually adjust/rotate the dataset so that the apex of the
heart and the an nu lus of the mitral valve are vertically
aligned. Also, the correct long axis views must be chosen to
prevent foreshortening, although should this occur it can be
easily adjusted. When all views are aligned, three pairs of contours have be drawn manually, although, after the first pair, the software will suggest what the contour of the other two might be. When the contours have been drawn, the
software will then calculate a provisional full ventricular mesh
model. Now, using various views and different editing options, the user may adjust the provisional results and when satisfied, finalize the calculation. Another option of the software is to adjust the volume that the algorithm views as being either cavity or wall, displayed as a percentage. lncreasing the percentage will expand the whole contour towards the wall, thereby increasing the volume. Although these shifts are in the sub-voxel range, the cumulative increase over all voxels will lead to significantly smaller or larger volumes if the setting is adjusted by five or more percentage points.
In conclusion, the algorithm will produce reproducible results, especially when all contours are manually drawn and possibly adjusted, instead of using the suggested contours. The possibilities to change the contour suggest the algorithm is more sophisticatectttran the Qlab algorithm.
CMR software: Tom Tee LV MR Analysis 1.0 The MR analysis software from TomTec (see figure L-3) works similarly to the echo analysis. Although the algorithm has the same
philosophy behind it (drawing contours in long axis images), it will probably operate
completely differently, since the CMR dataset is nota continuous block of data, at least not in the long axis slice direction. Also, the contrast in CMR images is much better than in 3DE datasets, with a higher SNR as well. The stack of short axis images have to be registered to any long axis slice that was acquired, the minimum
being three, so as to be able to merge the long Fig. L-3 - Interface of Tomtec LV MR Analysis
axis slice with the short axis slices. After
drawing of the contours, the algorithm again produces a provisional mesh, that can be altered in various ways, and then a final contour appears. Based on the shape of the contour mesh, the mitral valve is correctly excluded from the volume, but it is rather hard to see which part of the left ventricular outflow tract (LVOT) is included in the volume. Also, there is not an option to exclude papillary muscles from the contour volume. The algorithm seems sophisticated and precise enough to reproduce results within a very small error margin. As is discussed elsewhere, the results may not be completely accurate, as LVOT and papillary muscles volume inclusion is unclear.
CMR software: PieMedical CAAS MRV 3.2
The fourth program used in this report is
PieMedical CAAS MRV. While the software is very versatile, it is much more cumbersome for left ventricular analysis than TomTec LV MR. CAAS MRV has several additional functions compared to the other software, including drawing
epicardial contours to be able to calculate heart mass, the option to exclude papillary muscles (automatic detection possible) and the option to detect a LV contour fully automatically. Support for other CMR techniques such as flow
quantification, viability and perfusion is also included.
Since the option to automatically calculate LV volume is present in the software, it was tested how well this performed. A healthy volunteer was scanned 9 times in close interval (2 weeks). Heart rate was al most equal during each scan. The data was imported into the software and the automatic segmentation was performed., along with three other possible analysis methods. Results can be seen in table L-1 below. Because both automatic method did not provide satisfactory results, all analysis was performed with the semi-automatic method, because the manual method costs too much time (15 minutes or more), especially compared to Tomtec (3-4 minutes).
Table L-1 - Results for four analysis methods in CAAS MRV. Both automatic methods have a much larger variability than the semi-auto and manual method. Moreso, the automatic+LA method has a significantly other mean than the other measurements, which is a strong indication that this option should never be used.
dataset 1 2 3 4
s
6 2a (60fr) 3a (5122) 6a (60 fr) Mean Error ConclusionCAAS MRV Analysis method (Ejection Fraction percentages)
Automatic Automatic+ LA Semi-auto Manual
70,2 45,8 55,4 57,9 49,7 54,4 62,5 62,6 65,6 50,4 57,5 60,4 54,8 40,3 57,5 60,3 55,6 48,3 56,5 60,3 46,4 56,3 59,2 59,8 40,9 59,3 59,6 36,3 45,8 56,9 60,2 37,0 43,9 56,0 57,9 53,6 46,2 57,5 59,8 12,3 4,4 2,2 1,4
Looking at the software, some differences in algorithms, reliability and flexibility can be seen, but all packages provide a usable means to perform cardiac function analysis.
Appendix M
Analysis Protocol CAAS MRV
lntroduction This is a user manual in Dutch that has been written for MR technicians so they could
learn to use CAAS MRV by themselves, as it is planned to use this software in all future examinations
LV Volume/massa analyse
Alle te volgen inleidende stappen zijn: 1. log in op pc rode suite
2. stop de grote USB dongle in 1 van de USB poorten (kleinere is voor QFLOW)
3. start CAAS MRV 3.2
4. laad een dataset, door de MAP te selecteren van een patiënt (dus niet 1 bestand, maar
een map, het programma analyseert zelf de inhoud van map door DICOM)
5. data van proeven Job/Ward +patiënten van studie staan al op d:\MRI data.
6. Voor nieuwe data: het snelste is om een map aan te maken op D:\MRI data en de
cd-inhoud te kopiëren (laden gaat dan stukken sneller)
$New analysis -" '.1.
_IOB..: } Le11 Ventrlcte
Daadwerkelijke analyse
na het kiezen van de map met patiëntdata, klik op SCAN kies uit de geladen data:
a. een stack SA cine beelden b. een LA 4CH view
c. een LA 2CH view
Na het selecteren van elke serie moet deze toegewezen worden aan SA, 2CH of 4CH afhanke+ij-k
van
welke serie het betreft (als er meer 2CH of 4CH opnamen zijn, kies de beste)Klik als je klaar bent op Import
2CH/CINE ss 3CH/CINE ss
CH1CltlE"
LV volumes
Om een LV volume en ejectiefractie uit te kunnen rekenen, heeft het programma een aantal contouren nodig ('rondjes') in de korte as beelden. De lange as beelden dienen als een soort extra controle, waarmee de positie van het LV wordt bevestigd (ivm bewegingen mitraalklep etc).
Er moeten een aantal handelingen worden verricht, die in totaal zo'n 15 minuten kost per patient (dit neemt af tot 7-8 minuten als je geoefend hebt).
1. EDV en ESV aangeven
2. LA correctie ('U'tjes' tekenen)
3. SA contouren ('rondjes' tekenen)
4. Propageren van de getekende contouren voor overige hartfasen
:
I
EFEfH{EIH±f83
ED en ES aangeven
Je ziet in de interface hierboven dat links een overzicht is van alle SA cinebeelden, horizontaal de fasen (meestal 50) en verticaal het aantal slices door het hart (meestal 12).
1. Selecteer met de rechtse muisknop de ED fase en kies 'mark as ED' (altijd lste kolom,
omdat we daarop triggeren)
2. de hele ED fase wordt blauw gearceerd
3. Selecteer een ES fase-kolom en kies 'Mark as ES' (zelf bepalen, door bijv in slice 6 alle
fasen te controleren)
LA correctie {'U'tjes' tekenen)
1. Teken in zowel ED en ES de endo- en epicardiale contour in zowel het 2ch als 4ch view.
a. Let hierbij goed op de positie van het mitraalklepvlak.
b. Je tekent dus steeds 2 U'tjes per view, en 4 beelden, dus acht U'tjes
c. Als je klaar bent, zie je onder in beeld 'lnterpolation +LA volume based correction
correctly applied'
2. Je tekent de endocontour door
a. de endocontour knop te kiezen (zie pijl in onderstaand plaatje, lijnen zijn rood) b. beide kanten van de mitrale annulus aan te klikken
c. dan een punt bij de apex
d. en vervolgens zoveel punten langs het endocard als nodig is
3. Je tekent de epicontour met dezelfde stappen als hierboven, maar dan de epicard-tool
SA contouren ('rondjes' tekenen)
Er zijn twee manieren die succesvol het volume en EF kunnen berekenen.
1. handmatig zowel ED als ES intekenen (grofweg 12x2 rondjes plus evt papillairen) 2. handmatig ED intekenen en propageren over alle fasen (toont ook contractieprofiel in
resultaten) (grofweg 12 rondjes plus papillairen en twaalf keer klikken op 'propagate')
Doe de volgende stappen
1. Teken in de ED phase in de slices van de korte as stack handmatig de endocardiale contouren met de endocontour knop (en evt papillair spieren met de knop daarnaast) 2. Doe hetzelfde voor de ES fase als je optie 2 hebt gekozen (vergeet de papillair spieren
niet)
Propageren van de getekende contouren voor overige hartfasen
Als je optie 2 hebt gekozen, kan je deze stap overslaan. Zo niet, dan klik je voor elke ED contour-slice die je hebt getekend, in het overzicht met de rechtermuisknop en klikt dan op 'propagate slice endocard >right' (zie screenshot hieronder)
HET RESULTAAT
Als het goed is zie je nu al in het beeld EF, SV en CO staan, maar als je op 'Results' klikt (zie pijl in screenshot hierboven), dan zie je ook de EDV en ESV staan.
Eventueel kan je de berekende getallen exporteren in een bestand dat in Excel ingelezen kan worden: klik dan op File> export data. Opslaan als: "kiesjenaam.csv" (let op in welke map je dit bewaart!)
Dit is alleen een handleiding voor LV volume en EF. CAAS kan ook gebruikt worden voor viabiliteit, perfusie, QFlow en massa, maar dit is nog niet beschreven.
Conclusion Using this manual, all people with basic understanding of CMR and cardiac anatomy can
Appendix N
TomTec MR Acquisition
lntroduction This appendic contains an acquisition guide for Tomtec 40 LV-Analysis MR, as
settings deviate from the standard procedure in MMC, as well as the competing software, CAAS MRV in this case
In order to use the TomTec "40 LV-Analysis MR"-software the following criteria for data acquisition and data format must be fulfilled (see below)
1) Data format
• All series must have the same number of phases! (Retrospective gated acquisition
recommended) • Transfer syntax Uncompressed Explicit Little endian • Data Export
In order to accelerate the data import to Research-Arena the data should be exported with a "OICOM.dir" - file.
2) Data acquisition - Geometry The following series are required:
1. Stack of short-axis slices
2. 6-8 rotational (equiangular) long axis- slices (minimum: 3 LAX slices) CAVE:
1. Data acquisition geometry (rotational long axis) affects the analysis quality of the software 2. The rotational long axis must intersect the apex (in all slices) as accurate as possible!
3) Data acquisition - Respiration Artifacts
CAVE: Respiration artifacts must be minimized.
lt is recommended to perform respiratory gated acquisitions or to acquire the data very properly in end-expiratory. This is mainly important during acquisition of the long axes.
4) Image Quality
• 'Sharp' and high-contrast images are necessary to perform accurate endocardial border
tracking
• The field of view (FOV) should be 100%
5) Appendix
How to acquire the long axis slices: As shown below, the long axis slices should be acquired by rotating around the long axis of the left ventricle (The three long axes intersect in one straight line). The long axes should be acquired equiangular.
Left ven ricular
Appendix R
Raw data healthy volunteers
lntroduction
This appendix lists all raw data from the study with two healthy volunteers
.
Data is
grouped per subject, per observer and per software package
.
Subject 1 MR data (only MR data was acquired for this subject)
Analist 1 Tomtec
