imaging protocols for cardiac ct - mdct.net

The incorporation of multiple detectors into spiralcomputed tomography (CT) scanners has expand-ed the clinical role of CT in cardiac imaging, in-cluding coronary CT angiography (CTA). Ad-vances in both the speed at which the X-ray sourcerotates and the number of detectors have im-proved the ability of CT to resolve smaller anatom-ic detail and have enabled imaging of the nativecoronary arterial tree. At present, and for at leastthe near future, CT is the most robust modality tononinvasively image the coronary arteries. CTAcontributes largely to cardiovascular diagnoses,but one of the most important and one of the mostpromising contributions is its high negative pre-dictive value for coronary artery disease (CAD).That is, using the protocol detailed in this chapter,CAD can be reliably excluded in minutes withoutarterial catheterization. Moreover, in a single CTacquisition, native coronary imaging can be ex-tended to include the beating myocardium, valve

motion, ventricular outflow tracks, and coronarybypass grafts. In this chapter, in addition to detail-ing a basic cardiac imaging protocol, examples ofexaminations are illustrated.

Introduction

Protocols for electrocardiogram (ECG)-gated car-diac CT have evolved with rapid improvement intechnology. The technique has progressed fromearly cardiac CT [4-slice multidetector CT (MDCT)with 1-s gantry rotation] to current standards(ECG-gated 64-slice MDCT with gantry rotationtimes as low as 330 ms). Technology has developedat a rapid rate, fueled primarily by the promise of arobust, noninvasive method of performing diag-nostic coronary angiography (Fig. 1). AdditionalMDCT imaging includes coronary bypass graftsand evaluation of cardiac valves.This chapter fo-

III.1Imaging Protocols for Cardiac CT

Frank J. Rybicki and Tarang Sheth

Fig. 1a, b. Multiplanar reformatted CTA images demonstrate the major branches of the left coronary arterial system. This patient pre-sented with atypical chest pain, and coronary artery disease was excluded noninvasively

a b

98 MDCT: A Practical Approach

cuses on the coronary CTA protocol and also de-scribed how the basic protocol can be modified orextended for problem solving.

Temporal Resolution

Successful cardiac imaging by any modality relieson the ability of the technology to produce motionfree images or to scan faster than the heart beats.Thus, cardiac CT is founded on (1) imaging fasterthan the heart beats, or (2) slowing cardiac mo-tion. Temporal resolution is the metric that meas-ures imaging speed. For a CT scanner with a singlephoton source, the temporal resolution is one halfof the CT gantry rotation time. This is because im-age reconstruction requires CT data acquired fromone half (180°) of a complete gantry rotation. Atthe time of publication, all manufacturers havegantry rotation times less than 500 ms, with a min-imum of 330 ms.With a 330-ms gantry rotation, anECG-gated cardiac image can be reconstructed(using single-segment reconstruction describedbelow) with CT data acquired over 165 ms of thecardiac cycle. Thus, the reconstructed images dis-play the average of the cardiac motion over the165 ms during which the data was acquired. This ishow ECG gating enables coronary CTA. Withoutgating, cardiac images are nondiagnostic becausethe reconstruction “averages” the motion over theentire RR interval – 1,000 ms for a patient with aheart rate of 60 beats per minute.

Temporal resolution can be improved in single-source scanners by adopting a so-called “multiseg-ment” image reconstruction. The principle under-lying multisegment reconstruction is that the ac-quisition over several heart beats is summed to ob-tain the one half gantry (i.e., 180°) CT data. For ex-ample, in a two-segment reconstruction, two heartbeats are used to generate a single axial slice, andthus the temporal resolution is halved. Similarly, iffour heat beats are used (four segment reconstruc-tion), only 45° of data are used from each heartbeat. This would yield a four-fold reduction in thetemporal resolution. Since multiple heart beats areused to fill the 180° of gantry rotation necessaryfor the reconstruction, stable periodicity of theheart is essential. Moreover, multisegment recon-struction requires a lower CT pitch, resulting ingreater data oversampling and a higher radiationdose. Radiation considerations and a simple for-mula to estimate effective patient dose are given inan upcoming section.

A recent approach to improving temporal reso-lution involves the use of two independent sourcesand two independent (64-slice) detector systems(Siemens Definition; Siemens Medical Solutions,Erlangen, Germany). The second X-ray source ispositioned 90° from the first X-ray source, and the

second detection system is positioned 90° from thefirst detection system. With respect to temporalresolution, the practical consequence of this CTconfiguration is that 180° of gantry rotation can beachieved in half the time (e.g., 82.5 ms as opposedto 165 ms). This improvement in the temporal res-olution is expected to eliminate the need for multi-segment reconstruction. In fact, in patients with ahigher heart rate or a heart rate that is difficult tocontrol with beta blockade (described below), theCT pitch can be increased without compromisingimage quality.

Beta Blockade for Heart-Rate Control

As suggested from the discussion on temporal res-olution, beta blockade is an important componentof most cardiac CT examinations. A useful rule ofthumb for the target heart rate is “the first numberis a 5” – i.e., an ideal heart rate between 50 and 59beats per minute.While this goal is not achieved inevery patient, it provides a useful reference frame.IV metoprolol is routinely administered at our in-stitution; with cardiac monitoring, 5-mg incre-ments are given every 5 min up to a total dose of25 mg. Doses greater than 15 mg are rarely needed.Beta blockade can be safely performed by a radiol-ogist or a cardiologist. An alternative approach in-volves the use of oral beta blockade. Although thisapproach has the disadvantage of a longer serumhalf life, most patients arrive for the study with aheart rate already in the target range. This can sim-plify patient preparation on site and has the po-tential to increase patient throughput. The tradeoffis the extra step of premedicating the patient andissues surrounding patient compliance.

In theory, using the multisegment reconstruc-tion approach described above, beta blockade canoften be avoided because using multiple heartbeats in the reconstruction enables the scanner tohave an effective temporal resolution in the rangeof 40–50 ms. However, when multisegment recon-struction is used, image quality becomes highlydependent on cardiac beat-to-beat variability. Inour experience, multisegment reconstructionworks well in patients with high heart rates whoare being studied for clinical indications where thehighest image quality may not be required, for ex-ample, coronary bypass graft location and patency.For coronary CT angiography, beta blockade is stillrecommended.

ECG Gating

ECG gating refers to the simultaneous acquisitionof both the patient’s electrocardiogram (ECG)

III.1 • Imaging Protocols for Cardiac CT 99

tracing and CT data (Fig. 2a, b). By acquiring bothpieces of information, CT images can be recon-structed using only a short temporal segment pe-riodically located in the same location of the RRinterval over multiple cardiac cycles. The durationof the temporal segment is equal to the temporalresolution of the scanner. Each temporal segmentof the RR interval is named by its “phase” in thecardiac cycle; the most commonly used nomencla-ture is to name the percentage of a specific phasewith respect to its position in the RR interval. Forexample, if a manufacturer enables reconstructionof 20 (equally spaced) phases, they would typicallybe named 0%, 5%, 10% . . . 95%, beginning withone R wave and ending with the following R wave.The period in which the heart has the least motionis usually (but not always) in mid diastole, near aphase between 55% and 75%. Thus, under the as-sumption that the position of the heart remainsconsistent over the RR intervals during which CTdata is acquired, cardiac motion is minimized byproducing images from the same phase over mul-tiple cardiac cycles. This explains why ECG gatingtypically fails to freeze cardiac motion in patientswith an irregular rhythm, such as atrial fibrilla-tion. Consequently, atrial fibrillation patientsrarely have diagnostic cardiac CT examinations,

and it is our policy to not perform coronary CTAin this population.

If only static (as opposed to cine) images aredesired, image reconstruction can usually be per-formed over a small number of phases for whichmotion is minimized. (This is in contrast to cineimaging where images are reconstructed in allparts of the cardiac cycle and then played, in cinemode, to demonstrate cardiac motion.) The imagereconstruction phases used for interpretationmust account for differences in movement of theleft and right coronary arterial systems. Becausecoronary arterial motion is not synchronous, thephase of the cardiac cycle that proves best for di-agnosis of the left main and left anterior descend-ing artery is often different than the phase thatproves most diagnostic for the right coronary ar-tery. Moreover, it is often necessary to view morethan one phase to best assess the full extent of anindividual artery and its branches (e.g., the left an-terior descending and the diagonal branches).

The most complete cardiac CT examinationsinclude cine imaging. In cine cardiac CT, imagesare reconstructed in periodic phases throughoutthe cardiac cycle to yield information regarding amoving structure. For example, cardiac CT offersan outstanding assessment of the aortic valve and

Fig. 2a-c. a Electrocardiogram (ECG) gating as demonstrated on a Somatom Sensation 64 cardiac computed tomography (CT) scanner(Siemens Medical Solutions, Erlangen, Germany). Continuous ECG tracing is displayed on the console. In this case, minimum, maximum,and average heart rate is 60 beats per minute (top left). Thus, the width of the RR interval is 1,000 ms. The gray vertical bars indicate thatportion of the cardiac cycle used in the reconstruction. As discussed in the text, the width of the gray bar is the temporal resolution of thescan. For this single-segment reconstruction, the width is half the gantry rotation time, or 165 ms. The term “–400 ms” refers to the factthat the center of the gray bars is located 400 ms before the second of the two R waves in the RR interval. b Enlarged view of a single RRinterval. To provide a simple demonstration of how the RR interval is divided, six segments are illustrated. In clinical imaging, CT scannersdivide the RR interval into a number of segments between 10 and 20. The gray block at the very bottom emphasizes that each recon-structed image uses only a small portion of the cardiac cycle. The gray block is positioned in diastole; its center is approximately 65%between the R waves, i.e., 650 ms elapse between the first R wave and the center of the block. This reconstruction is the most common-ly used to visualize the left coronary arterial system. If this reconstruction does not provide the most optimal images, additional recon-structions, either earlier or later phases, are performed. Since the right and left coronary arterial system are asynchronous, it is sometimesthe case that evaluation of the right system is best performed using images closer to systole. c Electrocardiogram (ECG)-based tube cur-rent modulation or ECG pulsing. The ECG tracing is identical to the one illustrated in a. Yellow bars under the tracing correspond to thegray bars and correlate where, with current modulation, the optimal tube current (e.g., effective mAs = 650) will be used. Red lines showtimes that correspond to portions of the cardiac cycle where the X-ray CT tube current is minimized. ECG pulsing can reduced patient effec-tive radiation dose by 30–50%

a

cb


aortic root (Fig. 3). In addition to the fact that cineimaging can be used to assess valve motion, CT isby far the best imaging modality to identify andquantify calcification, and thus both structure andfunction can be well characterized in a singlebreath-hold CT acquisition. Cine CT can also beused to assess ventricular-wall motion. In compar-ison with cardiac magnetic resonance (MR), thegold standard for global and regional-wall motionabnormalities, CT has less contrast to noise, andimages typically have greater artifact owing topoorer temporal resolution. However, it is impor-tant to emphasize that cine CT is not a separateimage acquisition. The entire CT data set (coro-nary, valve, myocardium, pericardium) is acquiredin a single breath hold; cine CT is simply part ofthe image postprocessing. It is also important tonote that the most common contraindications forcardiac CT (e.g., impaired renal function as meas-ured by glomerular filtration rate or alternativelyby serum creatinine) differ from those for MR(pacemaker), and thus CT can often be used forpatients who cannot have MR.

Finally, it is important to note that future CTequipment with up to 256 slices is expected to per-form whole-heart coverage with a single half-gantry (180°) rotation. This approach holds thepromise of a subsecond cardiac scan. In additionto the fact that patient radiation would be de-creased, this would provide the ability to perform

multiple scans over the same injection of iodinatedcontrast material and thus create the opportunityfor a host of additional studies (e.g., myocardialperfusion) that are, at present, largely in the do-main of cardiac MR and nuclear cardiology.

Patient Irradiation

In some cardiac CT applications, for example, thelocation and patency of bypass grafts (Figs. 4 and5), all diagnostic information can typically be ob-tained from reconstruction of only a single phaseof the cardiac cycle in mid diastole. However, asemphasized above, in cardiac CT, image data is ac-quired throughout the cardiac cycle. Thus, forstudies such as bypass graft analyses, the CT data(and the radiation used to acquire that data) in theremaining “unused” phases is wasted.

Because cardiac CT requires ECG gating with aCT pitch less than 1, the patient radiation in car-diac CT is higher than that for CT of any otherbody part. While dose should be a considerationfor all patients undergoing CT, it is essential thatdiscussions regarding CT dose are based on soundprinciples. The risk most commonly cited as acause for concern is the development of a fatal ra-diation-induced neoplasm. While sparse, all hu-man data and antidotal reports to date support alatency period of no less than 20 years for a radia-

Fig. 3a, b. Reformatted ECG-gated cardiac CT images in a patient status post aortic valve replacement. Note that the patient has a pace-maker (noted by the right heart wires in 3b), and thus magnetic resonance imaging (MRI) was contraindicated. a Image through themechanical valve while it is open demonstrates multiple surrounding collections of contrast, characteristic of pseudoaneurysm. b orthog-onal view again demonstrates abnormal contrast to the right of the valve. This patient required emergent surgery with successful place-ment of a new valve. Images courtesy of Scott Koss, MD

a b


Fig. 4a-c. ECG-gated CT images from a patient status post leftinternal mammary artery to left anterior descending artery coro-nary bypass grafting. This patient was scheduled for a repeatbypass graft. CTA with reformatting is now performed routinely todetect cases such as this one where the graft becomes adherent tothe posterior table of the sternum. a Axial image demonstratesproximity of the internal mammary to the sternum. b Sagittal andobliquely reformatted images are essential in the evaluation ofthese patients. In this case, the graft is demonstrated to be patentand too close to the sternum for a repeat thoracotomy through thesternal incision; an alternate surgical approach was required forthis patient. c Selected image from a three-dimensional (3-D) vol-ume rendering again demonstrates the course of the graft. Volumerendering is often more appealing to our referring clinicians andcan help in the communication of important findings

a

c

b

Fig. 5. 3-D volume-rendered image from a patient with normalinternal mammary arteries who had undergone saphenous veincoronary bypass grafting. Note that the left-sided vein graft isbifurcated. The vein graft to the right coronary territory is single. Ina patient with only saphenous vein grafts, it is essential to imagethe entire course of the internal mammary arteries since, whennormal, they will be used for redo coronary artery bypass


tion-induced neoplasm. For this reason, for thepurpose of radiation dose, it is important to sepa-rate patients into two groups: those with a life ex-pectancy of roughly 20 years or less, and thosewith a longer life expectancy. In the former group,the only dose consideration of any consequence isthe radiation that could cause a skin burn (the on-ly short-term complication of any consequence).X-ray skin burns are extremely uncommon, par-ticularly in CT (even for ECG-gated studies), andwould be the consequence of multiple exams re-peated at short-term intervals. Thus, for this subsetof patients, radiation dose should be a lesser con-sideration in determining a modality for coronaryimaging.

For patients with a life expectancy muchgreater than 20 years, ECG-based tube currentmodulation (also called ECG pulsing) representsone strategy to lower overall patient radiation bymodulating the tube current over the course of thecardiac cycle (Fig. 2c) so that the desired diagnos-tic tube current is delivered in diastole while thecurrent is reduced for the remainder of the cardiaccycle.

Current modulation is featured on newer CTscanners and is important in many cases (e.g., pe-diatric patients). However, the decision to incorpo-rate current modulation should be made carefullysince the potential drawbacks are significant. First,once current modulation is used, images subse-quently reconstructed during phases with low tube

current will be noisy (Fig. 6). That is, reducing thetube current results in the production of fewer X-rays, and subsequently fewer X-rays pass thoughthe patient and reach the detection system. Sec-ond, current modulation eliminates the potentialto reconstruct high-quality cine imaging sinceevery phase of the cardiac cycle will not have the“full” tube current. Thus, if cine imaging is desired,current modulation cannot be utilized. Anotherpotential drawback concerns the identification ofincidental findings (e.g., bicuspid aortic valve) onstatic imaging acquired with current modulation.In these cases, it is impossible to perform postpro-cessing of a high-signal cine loop for a more com-plete evaluation.

While cine cardiac CT can provide a useful ad-junct to high spatial resolution anatomic data, CThas poor temporal resolution and ventricular im-age contrast when compared with steady-state freeprecession (SSFP) cardiac MR (CMR). For this rea-son, SSFP cine CMR remains the gold standard toassess cardiac function and to evaluate cardiacmasses. However, CT is far more accessible, it iseasier to perform, and a cardiac pacemaker is not acontraindication. Moreover, there are many car-diac masses that can be well or better seen on CT(Fig. 7). Hence, CT has become not only an adjunctto cardiac MR but in some cases the diagnostic testof choice (Fig. 8).

Since patient dose in CT is so frequently dis-cussed and has great potential to be misquoted, the

Fig. 6. ECG-based tube current modulation, or ECG pulsing. The top demonstrates how the operator selects current modulation from theconsole of the Somatom Sensation 64 cardiac computed tomography (CT) scanner (Siemens Medical Solutions, Erlangen, Germany). Theimage on the left is a two-chamber view (note the normal mitral valve that is well demonstrated with ECG gating) that is reconstructedat 65% of the RR interval. The right-hand, two-chamber view was reconstructed from 10% of the RR interval, where the X-ray CT tube cur-rent was dramatically reduced. Subsequently, this image suffers from high noise, the consequence of fewer photons received at the detec-tor


fundamentals of CT dose, including the dose fromcardiac CT, are described here. There are three dif-ferent parameters used to describe, quantify, andcalculate the dose:1. CT dose index (CTDIvol) [1]. CTDIvol units are

milligray (mGy) [3].2. Dose-length product (DLP) [2]. Units of DLP

are mGy � centimeters.3. Effective dose [3]. Units of effective dose are

milliSievert (mSv).The numerical value for CTDI is determined by

measuring dose in a cylindrical phantom. Al-though the phantom should somehow reflect theattenuation of a human body, the CTDI is not usedto make a statement regarding an individual pa-tient’s dose. Rather, it is used to compare differentscan protocols, optimize scan protocols, and com-

pare protocols used on different CT scanners. Incontrast to parameters such as tube current, CTDIvalues reflect delivered dose since parameters,such as scanner geometry and filtration, are con-sidered.

CTDIvol describes dose for a single rotation.DLP characterizes CT exposure over a completefield of view (FOV); DLP is defined as the productof the CTDIvol and the craniocaudal extent (Z-ax-is length) of the scan. Even though DLP reflectsmost closely radiation dose for a specific CT ex-amination, it is important to keep in mind thatDLP is a function of patient size (i.e., how much Z-axis coverage is required to complete the CT scan).Therefore, CTDIvol should be used to optimize ex-am protocols.

While CTDIvol and DLP enable evaluation of

Fig. 7a, b. Left atrial myxoma. a Four-chamber echocardiogram demonstrates the round lesion adjacent to the intra-atrial septum. bECG-gated CT shows the mass with higher spatial resolution, well depicting the attachment point of the mass with the intra-atrial sep-tum. Note that the left heart is well opacified with contrast material and the right heart is filled with saline. This is the goal in the timingof the dual injection protocol (contrast followed by saline)

a b

Fig. 8. Axial ECG-gated four-chamber image in a patient with apacemaker. The patient has a history of renal cell carcinoma, andechocardiography demonstrated an ill-defined echogenic mass.This image from the single 12-s CT acquisition excludes a metasta-tic deposit as the source of the finding on ultrasound. Fat splaysthe left and right atria, diagnostic of lipomatous hypertrophy ofthe intra-atrial septum. CT is the most rapid and accurate imagingmodality to demonstrate both fat and calcium


CT scanners and comparison of protocols acrossmanufacturers, these values only characterize thescanner. It is the effective dose, a weighted sumover the organ doses [4], that quantifies patientdose. Since the dose to an individual organ cannotbe measured directly, it is difficult to determine theeffective dose of a CT scan. However, methods havebeen described to estimate effective dose frommeasurable values [5–8]. A simple estimation isthat the effective dose is the product of DLP and aconversion factor, EDLP,that is specific to a body re-gion. For example, for the chest, the EDLP =0.017 mSv×mGy-1×cm-1[2].

As an example of a how this is used in clinicalpractice, consider the following hypothetical coro-nary CTA. The user console will give the CTDIvol.A typical value for a high-dose scan without cur-rent modulation would be on the order of 60 mGy.Assume that the craniocaudal extent of the scan is 15 cm, a typical value for a normal-sized heart.For this study, the DLP = 60 mGy×15 cm =900 mGy×cm. Given that the heart is in the chest,the appropriate conversion factor is EDLP =0.017 mSv×mGy-1×cm-1, and the effective dose forthis patient is 60 mGy×15 cm x 0.017 mSv×mGy-1

×cm-1 ~ 15 mSv. Note that, as discussed above, dosevalues in gated exams can be reduced with “ECGpulsing”. The degree of reduction is a function ofthe patient’s heart rate but is typically 30–50%.

Image Acquisition Time

Improved temporal resolution decreases the timeof the CT examination. This is important not onlyfor decreasing the effect of cardiac motion but alsofor completing the examination in a breath hold.Scan time becomes a factor for cardiac and as-cending aorta imaging because of the requiredECG gating. The image data must be “oversam-pled” since, for the reconstruction of each interval,only a small portion of the cardiac cycle is used.Data oversampling for cardiovascular applicationsdifferentiates it from all other MDCT scans thatcan capitalize on undersampling and interpolationin image reconstruction to dramatically decreasescan time.

CT pitch (a unitless parameter) is most accu-rately characterized as the distance the patientmoves through the scanner in a single gantry rota-tion divided by the width of the X-ray beam used.Because such a small part of the RR interval isused to reconstruct an entire image, significantoverlap along the craniocaudal extent of the pa-tient is required, translating into a pitch between

0.2 and 0.35, or an oversampling rate between 5:1and roughly 3:1. In addition to cardiac imaging,ECG gating is routinely required for CTA of the as-cending aorta to eliminate artifacts from aorticmotion that can be confused with pathology.

The practical consequence of oversampling isthat scan time (craniocaudal imaging over approx-imately 15 cm) is far greater than nongated scan-ning of the same Z-axis region of any other bodypart. This is one great benefit of scannersequipped with a larger number of detectors, whichallow coverage of a larger craniocaudal territoryper rotation. In the extreme case, craniocaudalcoverage can be large enough that the entire heartis covered in a single rotation. The number of de-tectors and focal spots determines the number ofslices obtained per gantry rotation. That is, 64-slicecoronary CTA (Figs. 9 and 10) can be achievedwith 32 detectors and a dual focal spot or 64 detec-tors and a single focal spot. The important factorsare temporal resolution (determined by gantry ro-tation time), slice thickness, and quality of the X-ray CT tube.

Increasing either number of slices, thickness ofdetectors, or both increases “Z-axis coverage” perrotation and thus decreases scan time. For exam-ple, for coverage of the heart, a 4-slice scanner mayrequire a 35-s breath hold while a 64-slice acquisi-tion (performed with the same gantry rotationtime and detector width) on the same patient mayrequire only 15 s.

While thicker detectors decrease scan time byproviding more Z-axis coverage per rotation, in-creasing detector width for cardiac applications isundesirable since it degrades the spatial resolutionof the examination. In general, spatial resolutionrefers to the ability to differentiate two structures.In practical terms, spatial resolution refers to thethinnest axial slices that can be reconstructed fromconfiguration of the detectors. The CT industry,driven by the promise (and the competition) ofselling scanners to noninvasively image coronaryarteries, has dramatically improved spatial resolu-tion by producing detection systems that can re-construct submillimeter images. Thinner slices(higher spatial resolution) correspond to longerscan times; however, all imaging applications donot require the highest spatial resolution. For ex-ample, myocardial and ascending aortic imagingrarely requires submillimeter slices. In particular,when a patient is expected to have difficulty withbreath holding, imaging should be performed withthicker slices to maximize the diagnostic informa-tion available.


Scanning Parameters

As with ECG gating, image data oversampling, andspatial resolution, modern cardiac CT pushes thelimits of technology with respect to the X-raysource. The two main scanning parameters thatdetermine the number of photons used are the ef-fective milliampere second (mAs) and kilovolts(kV). Effective mAs is defined as mAs divided byCT pitch and is proportional to X-ray CT tube cur-rent and scan time. Effective mAs and kV are setby the operator at the time of image acquisition.Typical values of effective mAs and kV are

550–700 and 120, respectively. However, in order toavoid reconstruction of images with significantnoise, larger patients require more photons andthus higher settings. The most common way tomaintain diagnostic images is to increase the ef-fective mAs. Cardiac imaging of very obese pa-tients is often limited. All X-ray CT tubes have a“limit” to the number of photons that can be pro-duced, and for a particular application, all meth-ods to decrease image noise (using the tube limit,increasing image thickness, scanning a smaller re-gion) should be considered.

As described above, Z-axis spatial resolution is

Fig. 9a-c. Diagnostic quality coronary computed tomography angiography. a Oblique multiplanar reformatted image of the proximalright coronary artery. One of the major advantages of coronary CTA in comparison with digital subtraction angiography is the ability toobtain an orthogonal view through any lesion. This is the most important step in image interpretation. b The upper right-hand imageframed in red corresponds to the more proximal right coronary artery. There is no coronary artery disease at this level. c The lower right-hand image framed in yellow corresponds to the more distal right coronary artery lesion. This orthogonal view is obtained at the centerof the noncalcified (soft) plaque and demonstrates a greater than 70% stenosis

a

b

c

Fig. 10. Curved multiplanar reformatted image of the right coro-nary artery. Curved multiplanar images track (either automated,semiautomated, or manually) the center of the coronary arterythough a long segment of its course and then display this longsegment on a single image. While curved multiplanar reformattedimages have no added information with respect to multiple short-segment standard reformatted images, they have the advantagethat a large amount of data is displayed on a single image.However, it is essential to note that curved multiplanar reformat-ted images rely heavily on a precise placement of the center line.In our experience, interpretation of curved multiplanar reformat-ted images created from an imprecise center line is the most com-mon source of error in image interpretation. This image demon-strates a complex (combination of calcified and noncalcified)plaque in the right coronary artery. At the calcified central compo-nent (white arrow), the stenosis was 50%


determined by image slice thickness. For all mod-ern scanners, the resolution is submillimeter. Al-though details of image interpretation are beyondthe scope of this article, it is important to point outone major advantage of CT in comparison withother imaging modalities, such as catheter angiog-raphy, is the ability to perform multiplanar recon-structed images. Quality of reconstructed images isinversely related to image slice thickness, and it isbeneficial to reconstruct images on so-called“isotropic data”; that is, CT data sets where spatialresolution is equal in the X, Y, and Z-directions. Asan example of the impact of spatial resolution, con-sider native coronary CTA acquired with perfectECG gating and no respiratory motion. In this set-ting, a 3-mm coronary artery reconstructed with0.4 mm isotropic voxels spans seven or eight high-quality pixels (3 mm/0.4 mm) in any direction. Thisexplains why properly performed CTA can differ-entiate between <50% and >50% stenosis but can-not grade a stenosis more precisely.

Scan Range and Image Field of View

Scan range in the Z-axis should include the anato-my of interest and allow for variations induced byboth breath holding and the possibility thatpathology can extend in both the cranial and cau-dal directions. In coronary imaging, occasionally,the left main and proximal left anterior descend-ing arteries course superiorly over 1–2 cm, afterwhich these vessels follow their usual path. Evenin these situations, when imaging the native coro-naries, the superior border of the scan should be

set at the top of the carina, and the inferior bordershould scan through the entire inferior wall of theheart. Ideally, the planned field of view (FOV)should include several slices of the liver to accountfor cardiac displacement during breath holding.Since CT data is acquired in the craniocaudal di-rection, obtaining a small amount of CT data infe-rior to the heart does not affect image quality. Ex-tended craniocaudal coverage is required for spe-cific applications (Fig. 11). The most common ap-plications are evaluation of coronary bypass graftsand of chest pain. For bypass grafts, imaging ex-tends cranially to include the origins of the inter-nal mammary arteries from the subclavian arter-ies. It is important to image the full extent of bothsides, as course, caliber, and patency is importantin the assessment of patients who have an internalmammary graft as well as those in whom the in-ternal mammary artery is being considered for by-pass.

The operator also specifies the FOV in the XYplane for coronary CT reconstruction. Choosingan FOV in the XY plane that is smaller than 24 cmis not recommended, as these images can be noisy.Typical values range between 24 cm and 30 cm,and it is almost always the case that coronary ar-tery reconstruction will include the most impor-tant anatomy in the mediastinum.

In every case, complete CT reconstruction witha full FOV should be performed, followed by “skin-to-skin” interpretation in lung, mediastinum, andbone windows. Patients have significant “inciden-tal” findings, including cases of acute pulmonaryembolism and lung masses invading the chest wall,that can be the source of chest pain.

Fig. 11. Topogram of a car-diovascular patient. For imag-ing of the native coronaryarteries (blue range), imagingshould extend from the top ofthe carina though the inferiorof the heart. Ideally, imagingshould include a few slices ofthe liver to ensure that theinferior of the heart is covered.For bypass graft imaging (yel-low range), the superior aspectof the range is extended toinclude both subclavian arter-ies and the origin of the inter-nal mammary arteries. Inchest-pain imaging, the entirechest, as well as the full extentof the aorta, is imaged


Contrast Material

While intravenous contrast can be administeredwith either a single or a dual injection system, dualinjection (iodinated contrast followed by saline) isrecommended for coronary CTA. Saline is used toavoid dense opacification of the right heart andpotential artifacts that can limit interpretation ofthe right coronary artery. In addition, the injectionof saline after iodinated contrast pushes the iodi-nated contrast to its anatomic destination andhelps to minimize dilution of the contrast as itpasses through the central veins.

It is now standard to inject contrast material atrates of at least 5 cc/s, and most centers have an in-jection rate between 6 cc and 7.5 cc/s. This rate ofdelivery requires that the IV be placed in the ante-cubital vein and be at least 20 gauge (usually 18guage is required). A right-arm injection is pre-ferred since contrast material injected into the leftarm often fills the left brachiocephalic vein at thetime of image acquisition, and dense opacificationof the brachiocephalic vein can be the source of ar-tifact in the anterior mediastinum. Contrast mate-rial volume is determined by contrast injectionrate and scan time required to cover the cranio-caudal extent of the heart. As an example, consideran 18-s scan of the native coronaries plus bypassgrafts. With an injection rate of 5 cc/s, an adequatevolume of contrast media would be 90 cc (18 s ×5 cc/s). Since the administration of contrast mate-rial after completion of image acquisition is of nobenefit, it is important to perform this calculationso that excessive contrast does not fill the rightheart and subsequently induce image artifact.

With respect to timing the contrast injection,there are two general methods that can be used:bolus tracking and a test bolus. Both are illustratedin the setting of imaging the native coronary arter-ies. As previously mentioned, the superior border

of the region to be imaged is set at the top of thecarina. The axial slice at this position is usually2–4 cm above the origin of the left main coronaryartery. In patients with normal cardiac output, novenous obstruction, and whose arms are posi-tioned over the head and above the right heart, thetypical transit time from the right antecubital veinto the ascending aorta at the level of the carina isbetween 17 s and 23 s. In bolus tracking, the con-trast injection begins with the scanner prepared torepetitively image a region of interest (ROI) in theascending aorta at the axial slice defined by the po-sition of the carina. Roughly 10 s after the contrastinjection begins, images of the same slice are ac-quired while enhancement of the ascending aortais monitored; that is, the bolus is “tracked” in theascending aorta just above the coronary ostia.Once enhancement reaches a preset threshold(typically 200 HU above baseline attenuation inthe ascending aorta), craniocaudal diagnostic im-ages are acquired, beginning at the axial locationwhere bolus tracking was performed.

A test bolus uses a separate injection to timethe diagnostic injection. For an injection rate of6 cc/ s, a typical test bolus would be 12 cc of con-trast followed by 30 cc of saline. As with bolustracking, the ROI is chosen in the ascending aortaat the level of the carina. However, 10 s after the be-ginning of the test injection, scans separated by 1 sare used to plot enhancement versus time to in-clude the time of peak enhancement (Fig. 12).Once the optimum delay is determined from theplot, the diagnostic images are obtained with a sec-ond contrast injection. When using a test bolus, itis important that the test injection mirrors the di-agnostic injection. In particular, injection ratesshould be the same. Also, most centers routinelygive nitroglycerin (0.4 mg sublingually) for coro-nary vasodilatation to all patients undergoing na-tive coronary CTA. While the effect of nitroglyc-

Fig. 12. Typical appearance of timing bolusplot for coronary CTA. The region of interest isapproximately 3 cm above the origin of the leftmain coronary artery. Contrast (12 cc) followedby saline (30 cc) is administered at a rate of6 cc/s, and images are acquired every second.There are no data points for the first 10 s of thetiming bolus since imaging is not performedwhile the contrast passes from the venous sys-tem to the pulmonary arterial system. In thisexample, the contrast in the ascending aortapeaks at 18 s (arrow). The subsequent coronaryCTA will use a 21-s delay. The rationale is thatthe additional 3 s will ensure that the coronaryarteries have time to fill with contrast beforethe CT data is collected

Ho

un

dsf

ield

un

its

Time (s)

0 5 10 15 20 25 30 35

400

300

200

100

0


erin on cardiac output (and thus iodinated con-trast transit time) is typically negligible, the nitro-glycerin should be administered before the test in-jection.

While using a test bolus has the disadvantagethat a separate injection is required, there are dis-tinct advantages in coronary imaging. First, unlikebolus tracking, when a test bolus is used, actualtime to peak enhancement can be obtained. Thiswill better ensure that when the diagnostic contrastinjection is performed, peak enhancement will beachieved. Second, the test injection tests the qualityof the intravenous access. Finally, since the test bo-lus can be performed with the breathing instruc-tions that will be used in the diagnostic injection, itallows the patient to “practice” the exam, and it en-ables the operator to visualize variation in heartrate during a breath-hold IV contrast injection.

Image Reconstruction

Single-segment (one heartbeat) or multisegment(greater than one heartbeat) retrospective imagereconstruction can be performed, the later strate-gy yielding an improvement in temporal resolu-tion at the expense of greater data oversamplingand more patient irradiation. In addition to choos-ing a single- versus multisegment algorithm, theoperator can choose a reconstruction kernel for aparticular application. Reconstruction with addi-tional kernels is most often done in the evaluationof patients with one or more coronary stents

(Fig. 13). At present, stenosis within a stent cannotbe quantified reliably. However, sharper imagingkernels (i.e., closer to a bone algorithm than a softtissue algorithm) can be used to “sharpen” edgesand determine that a coronary stent is not occlud-ed. This can provide useful information in patientswho present with chest pain post-stent placement.

Another technique that can be used to improveimage quality is ECG editing. This refers to the abil-ity to manually modify and/or eliminate a recon-struction phase in one or in a few RR intervals(Fig. 14). ECG editing is most commonly used whena patient has a premature ventricular contraction(PVC) during image acquisition. Since the recon-structed phase of the cardiac cycle is triggeredfrom the high amplitude of the R wave, reconstruc-tion software can mistake a PVC for an R wave. Re-constructed slices that correspond to RR intervalswith this error will suffer from severe motion arti-fact since these slices will be reconstructed over adifferent part of the cardiac cycle than the remain-der of the scan. Since coronary CTA data is over-sampled (CT pitch <1), reconstruction can be per-formed after removal of a PVC, often yielding adramatic improvement in image quality. This canbe a critical step for patients who have a PVC, par-ticularly for those who do not have CAD. Since thehigh negative predictive value of coronary CTA de-pends on acquiring diagnostic images through thefull extent of the major coronary arteries, elimina-tion of a short segment of severe motion artifactcan enable the interpreting physician to determinethat a study is normal.

Fig. 13. For patients with coronary artery stents, both image artifacts and the spatial resolution of coronary CTA limit the interpretation.a Curved multiplanar reformatted image in a patient with a stent in the proximal left anterior descending artery. This image is reconstruct-ed with a standard coronary imaging kernel. b Image reconstructed with a kernel closer to a bone algorithm shows sharper edges and lessartifact from the high attenuation stent. The in-stent lumen is better visualized

a b


Summary

Imaging protocols for cardiac CT have been revo-lutionized by recent advances in technology. In ad-dition to the widespread availability of ECG gat-ing, CT equipment with submillimeter resolution,64 slices per rotation, and gantry rotation timesless than one-half second are available from allmajor vendors. Cardiac imaging protocols aremore complicated than CT scanning of other bodyparts. Strict adherence to the protocol is requiredto maintain image quality. However, for the clini-

cian familiar with CT, superior diagnostic imagescan be obtained routinely.

Acknowledgments

The authors gratefully acknowledge useful discus-sions with Bernhard Schmidt Ph.D., particularlywith respect to the section of CT radiation dose.

Before ECG Editing

After ECG Editing

Fig. 14. ECG editing in a patient with suboptimal gating. The top right image demonstrates artifact that is explained by the locations(rectangles in the top left image) in the ECG where reconstruction was performed. Analysis of ECG tracing reveals that for some RR inter-vals, noise in the ECG is of great enough amplitude the the reconstruction algorithm mistook noise for an R wave. Thus, the different axiallevels of the reconstruction reflect different phases of the cardiac cycle, rendering the study nondiagnostic. The bottom left image repre-sents the ECG after editing; the location of the reconstruction has been manually placed to correspond with a relatively quiescent periodin diastole. The subsequent ECG edited reconstruction (bottom right image) showed normal coronary arteries, using the high negativepredictive value of coronary CT angiography (CTA) to eliminate coronary artery disease as a source of this patient’s chest pain. Image post-processing courtesy of Melissa Ende, Siemens Medical Solutions

a

b


References

1. International Electrotechnical Commission. Med-ical electrical equipment. Part 2–44: Particular re-quirements for the safety of X-ray equipment forcomputed tomography. IEC publication 60601-2-44(2002). International Electrotechnical Commission(IEC) Central Office: Geneva. Accessed 17 Jan 2006

2. European Guidelines on Quality Computed Tomog-raphy. EUR16262 (2002). http://www.drs.dk/guide-lines/CT/quality/index.htm. Accessed 17 Jan 2006

3. Morin R, Gerber T, McCollough C (2003) Radiationdose in computed tomography of the heart. Circula-tion 107:917–922

4. International Council on Radiation Protection(1991) 1990 recommendations of the InternationalCommission on Radiological Protection. Publica-

tion 60. Annals of the ICRP. Pergamon, Oxford5. Schmidt B, Kalender WA (2002) A fast voxel-based

Monte Carlo method for scanner- and patient-spe-cific dose calculations in computed tomography.Physica Medica XVIII(2):43–53

6. Zankl M, Panzer W, Drexler G (1991) The calcula-tion of dose from external photon exposure usingreference human phantoms and Monte Carlo meth-ods. Part IV: Organ doses from tomographic exam-inations. GSF report 30/91. GSF, Neuherberg

7. Kalender WA, Schmidt B, Zankl M, Schmidt M(1999) A PC program for estimating organ dose andeffective dose values in computed tomography. EurRadiol 9:555–562

8. Stamm G, Nagel HD (2002) CT-Expo – ein neuar-tiges Programm zur Dosisevaluierung in der CT.Fortschr Röntgenstr 174:1570–1576

imaging protocols for cardiac ct - mdct.net

Documents