Physics of Cardiac Imaging with Multiple-Row Detector CT PDF (AAPM/RSNA Physics Tutorial)

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. AAPM/RSNA PHYSICS TUTORIAL...

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. AAPM/RSNA PHYSICS TUTORIAL 1495 AAPM/RSNA Physics Tuto- rial for Residents Physics of Cardiac Imaging with Multiple-Row Detector CT1 Mahadevappa Mahesh, MS, PhD Dianna D. Cody, PhD Cardiac imaging with multiple-row detector computed tomography (CT) has become possible due to rapid advances in CT technologies. Images with high temporal and spatial resolution can be obtained with multiple-row detector CT scanners; however, the radiation dose asso- ciated with cardiac imaging is high. Understanding the physics of car- diac imaging with multiple-row detector CT scanners allows optimiza- tion of cardiac CT protocols in terms of image quality and radiation dose. Knowledge of the trade-offs between various scan parameters that affect image quality—such as temporal resolution, spatial resolu- tion, and pitch—is the key to optimized cardiac CT protocols, which can minimize the radiation risks associated with these studies. Factors affecting temporal resolution include gantry rotation time, acquisition mode, and reconstruction method; factors affecting spatial resolution include detector size and reconstruction interval. Cardiac CT has the potential to become a reliable tool for noninvasive diagnosis and pre- vention of cardiac and coronary artery disease. © RSNA, 2007 Abbreviations: ECG ⫽ electrocardiographic, MPR ⫽ multiplanar reformation, 3D ⫽ three-dimensional RadioGraphics 2007; 27:1495–1509 Published online 10.1148/rg.275075045 Content Codes: 1From the Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, 601 N Caroline St, Baltimore, MD 21287-0856 (M.M.); and the Department of Imaging Physics, University of Texas M. D. Anderson Cancer Center, Houston, Tex (D.D.C.). From the AAPM/RSNA Physics Tutorial at the 2005 RSNA Annual Meeting. Received March 12, 2007; revision requested April 4 and received May 21; accepted June 8. M.M. receives research support from Siemens; D.D.C. is a speaker for the Medical Technology Management Insti- tute, Milwaukee, Wis. Address correspondence to M.M. (e-mail: [email protected]). © RSNA, 2007 1496 September-October 2007 RG f Volume 27 Number 5 Introduction lution (⬍0.75 mm), improved temporal resolu- Coronary heart disease (CHD) represents the tion (80 –200 msec) (11), and electrocardio- major cause of morbidity and mortality in West- graphically (ECG) gated or triggered mode of ern populations (1). In 2003 alone, more than acquisition, the current generation of CT scan- 450,000 deaths were associated with CHD in the ners (16 – 64-row detectors) makes cardiac imag- United States; that translates to nearly one of ev- ing possible (12–17) and has the potential to ac- ery five deaths (1). The economic burden on curately characterize the coronary tree. health care due to CHD is also enormous. Ad- The purpose of this article is to describe the vances in multiple-row detector computed to- physics of cardiac imaging with multiple-row de- mography (CT) technology have made it feasible tector CT. The factors affecting temporal and for imaging the heart and to evaluate CHD non- spatial resolution are discussed along with scan invasively. Calcium scoring, CT angiography, acquisition and reconstruction methods, recon- and assessment of ventricular function can be struction algorithms, reconstruction interval, performed with multiple-row detector CT (2). pitch, radiation dose, and geometric efficiency. Coronary artery calcium scoring (3,4) allows pa- tients at intermediate risk for cardiovascular Key Issues in Cardiac Imag- events to be risk stratified. Coronary arterial anat- ing with Multiple-Row Detector CT omy and both noncalcified and calcified plaques The primary challenge required to image a rap- are visible at coronary CT angiography (5,6). idly beating heart is that the imaging modality Vessel wall disease and luminal diameter are de- should provide high temporal resolution. It is picted, and secondary myocardial changes may necessary to freeze the heart motion in order to also be seen (5–7). image coronary arteries located close to heart Even though the prime diagnostic tool to muscles, since these muscles show rapid move- evaluate and treat CHD is coronary angiography ment during the cardiac cycle. Since the most (an x-ray fluoroscopy– guided procedure) and the quiescent part of the heart cycle is the diastolic fluoroscopically guided images are considered the phase, imaging is best if performed during this standard of reference for comparing coronary ar- phase. Hence, it is required to monitor the heart tery image quality, the procedure is invasive and cycle during data acquisition. The subject’s elec- requires longer examination times than CT an- trocardiogram (ECG) is recorded during scan- giography, including patient preparation and re- ning because the image acquisition and recon- couping times. Images obtained with 16-row and struction are synchronized with the heart motion. 16⫹-row detector CT scanners are increasingly Also, the imaging modality should provide high becoming comparable to those of the standard of spatial resolution to resolve very fine structures reference (8). such as proximal coronary segments (right coro- The prospect of imaging the heart and coro- nary ascending and left anterior descending arter- nary arteries with CT has been anticipated since ies) that run in all directions around the heart. the development of CT more than 3 decades ago. These requirements impose greater demands on The lack of speed and poor spatial and temporal multiple-row detector CT technology. One of resolution of previous generations of CT scanners the primary goals of the rapid development of prevented meaningful evaluation of the coronary CT technology has been to achieve these de- arteries and cardiac function. Most early assess- mands in order to make cardiac CT imaging a ments of the coronary arteries with CT were per- clinical reality. formed with electron beam CT, developed in the early 1980s (9). Electron beam CT has been used Understanding the mostly for noninvasive evaluation of coronary Physics of Cardiac Imaging artery calcium (10), but other applications in- To better demonstrate and understand the neces- cluding assessment of coronary artery stenosis sity for high temporal resolution in cardiac imag- have been reported in limited cases. However, ing, Figure 1 shows how the length (in time) of electron beam CT is expensive and is not widely the diastolic phase changes with heart rates. The available. least amount of cardiac motion is observed during Recent advances in CT technologies, especially the diastolic phase; however, the diastolic phase multiple-row detector CT, have dramatically narrows with increasing heart rate. With rapid changed the approach to noninvasive imaging of heart rates, the diastolic phase narrows to such an cardiac disease. With submillimeter spatial reso- extent that the temporal resolution needed to im- age such subjects is less than 100 msec. The de- RG f Volume 27 Number 5 Mahesh and Cody 1497 ment has been to obtain similar spatial resolution in all directions, also expressed as isotropic spatial resolution (18). In addition, a sufficient contrast-to-noise ratio is required to resolve small and low-contrast structures such as plaques. In CT, low-contrast resolution is typically excellent. However, it can degrade with the increasing number of CT detec- tors in the z direction due to increased scattered radiation that can reach detectors in the z direc- tion. It is important to achieve adequate low-con- trast resolution with minimum radiation expo- sure. The need to keep radiation dose as low as reasonably possible is essential for any imaging modality that uses ionizing radiation. Overall, cardiac imaging is a very demanding Figure 1. Diagram shows the range of diastolic application for multiple-row detector CT. Tem- regions for varying heart rates. The desired tem- poral, spatial, and contrast resolution must all be poral resolution for cardiac CT is approximately optimized with an emphasis on minimizing radia- 250 msec for average heart rates of less than 70 beats per minute; for higher heart rates, the de- tion exposure during cardiac CT imaging. sired temporal resolution is approximately 100 msec. Temporal Resolution There are a number of factors that influence the sired temporal resolution for motion-free cardiac temporal resolution achieved with multiple-row imaging is 250 msec for heart rates up to 70 beats detector CT scanners. Among them, the key fac- per minute and up to 150 msec for heart rates tors are the gantry rotation time, acquisition greater than 100 beats per minute. Ideally, mo- mode, type of image reconstruction, and pitch. tion-free imaging for all phases requires temporal resolution to be around 50 msec. The standard of Gantry Rotation Time reference for comparing the temporal resolution Gantry rotation time is defined as the amount of obtained with multiple-row detector CT is fluo- time required to complete one full rotation (360°) roscopy, wherein the heart motion is frozen dur- of the x-ray tube and detector around the subject. ing dynamic imaging to a few milliseconds (1–10 The advances in technology have considerably msec). Therefore, the demand for high temporal decreased the gantry rotation time to as low as resolution implies decreased scan time required 330 –370 msec. The optimal temporal resolution to obtain data needed for image reconstruction during cardiac imaging is limited by the gantry and is usually expressed in milliseconds. rotation time. The faster the gantry rotation, the The demand for high spatial resolution that greater the temporal resolution achieved. How- enables the visualization of various coronary seg- ever, with increasing gantry rotation speed, there ments (such as the right coronary artery, left ante- is also an increase in the stresses on the gantry rior descending artery, and circumflex artery) that structure, since rapid movement of heavy me- run in all directions around the heart with de- chanical components inside the CT gantry results creasing diameter is high. These coronary seg- in higher G forces, making it harder to achieve a ments range from a few millimeters in diameter further reduction in gantry rotation time. In fact, (at the apex of the aorta) and decrease to a few even a small incremental gain in the gantry rota- submillimeters in diameter as they traverse away tion time requires great effort in the engineering from the aorta in all directions. The need to im- design. age such small coronary segments requires small In the past, the minimum rotation time was as voxels, and this is key to cardiac imaging with high as 2 seconds; in the past few years, gantry multiple-row detector CT. Spatial resolution is rotation time has decreased steadily to less than generally expressed in line pairs per centimeter or 400 msec. As discussed in the previous section line pairs per millimeter. Like temporal resolu- and in Figure 1, since the currently available gan- tion, the standard of reference for comparing spa- try rotation time is not in the desired range for tial resolution is the resolution obtained during obtaining reasonable temporal resolution, various fluoroscopy. However, one of the major goals of multiple-row detector CT technology develop-  1498 September-October 2007 RG f Volume 27 Number 5 Figure 2. During the prospective ECG-triggered scan mode, the pa- tient’s ECG is continuously moni- tored but the x-rays are turned on at predetermined R-R intervals to ac- quire sufficient scan data for image reconstruction. The table is then moved to the next location for fur- ther data acquisition. These types of scans are always sequential and not helical and result in a lower patient dose because the x-rays are on for a limited period. Calcium scoring scans are typically performed in this scan mode. methods have been developed to compensate, resolution that can be achieved in the partial scan such as different types of scan acquisitions or im- mode of acquisition is slightly greater than half of age reconstructions to further improve temporal the gantry rotation time. Once the desired data resolution. are acquired, the table is translated to the next bed position and, after a suitable and steady heart Acquisition Mode rate is achieved, the scanner acquires more pro- For imaging the rapidly moving heart, projection jections. This cycle repeats until the entire scan data must be acquired as fast as possible in order length is covered, typically 12–15 cm (depending to freeze the heart motion. This is achieved in on the size of the heart). multiple-row detector CT either by prospective With multiple-row detector CT, the increasing ECG triggering or by retrospective ECG gating number of detectors in the z direction allows a (19). larger volume of the heart to be covered per gan- try rotation. For example, using a multiple-row Prospective ECG Triggering.—This is similar detector CT scanner capable of obtaining 16 axial to the conventional CT “step and shoot” method. sections (16 rows of detectors with 16 data acqui- The patient’s cardiac functions are monitored sition system channels in the z direction) and with through ECG signals continuously during the each detector width of 0.625 mm, one can scan a scan. The CT technologist sets up the subject 10-mm (16 ⫻ 0.625-mm) length per gantry rota- with ECG monitors and starts the scan. Instruc- tion. Similarly, with a 64-section multiple-row tions are built into the protocol to start the x-rays detector CT scanner (64 rows of detectors with at a desired distance from the R-R peak, for ex- 64 data acquisition system channels) and each ample at 60% or 70% of the R-R interval. The detector 0.625 mm wide, one can scan about 40 scanner, in congruence with the patient’s ECG mm per gantry rotation. Typically, the cardiac pulse, starts the scan at the preset point in the region ranges from 120 to 150 mm, which can be R-R internal period (Fig 2). The projection data covered in three to four gantry rotations with a are acquired for only part of the complete gantry 64-row detector CT scanner. This has a major rotation (ie, a partial scan). advantage in terms of the decreased time required The minimum amount of projection data re- for breath holding to minimize motion artifacts quired to construct a complete CT image is 180° (critical when scanning sick patients). Teaching Point plus the fan angle of the CT detectors in the axial One of the advantages of the prospective trig- plane. Hence, the scan acquisition time depends gering approach is reduced radiation exposure, Teaching on the gantry rotation time. The best temporal because the projection data are acquired for short Point periods and not throughout the heart cycle. Tem- poral resolution with this type of acquisition can range from 200 to 250 msec. Prospective trigger- RG f Volume 27 Number 5 Mahesh and Cody 1499 Figure 3. During the retrospective ECG-gated scan mode, the patient’s ECG is continuously monitored and the patient table moves through the gantry. The x-rays are on continu- ously, and the scan data are col- lected throughout the heart cycle. Retrospectively, projection data from select points within the R-R interval are selected for image recon- struction. Radiation dose is higher in this type of scan mode compared with that in the prospective trigger- ing mode. Pos ⫽ position. ing is the mode of data acquisition used for cal- resolution. Temporal resolution with this type of cium scoring studies, since calcium scoring analy- acquisition can range from 80 to 250 msec. sis is typically performed in axial scan mode. The The disadvantage of the retrospective gating scan technique such as tube current (milliam- mode of acquisition is the increased radiation Teaching peres) for a calcium scoring protocol can be quite dose, because the data are acquired throughout Point low, yielding low radiation dose, since calcium the heart cycle, even though partial data are actu- has a high CT number and is easily visible even ally used in the final image reconstruction. Also, with a noisier background. Also, each data set is since this scan is performed helically, the tissue obtained during the most optimal ECG signal to overlap specified by the pitch factor is quite low, reduce motion artifacts. indicating excessive tissue overlap during scan- ning, which also increases radiation dose to the Retrospective ECG Gating.—Retrospective patients. The need for low pitch values or exces- gating is the main choice of data acquisition in sive overlap is determined by the need to have cardiac coronary artery imaging with multiple- minimal data gaps in the scan projection data re- row detector CT. In this mode, the subject’s quired for image reconstruction. The need for ECG signals are monitored continuously and the low pitch values is discussed in detail in the sec- CT scan is acquired continuously (simulta- tion on pitch. neously) in helical mode (Fig 3). Both the scan projection data and the ECG signals are re- Reconstruction Method corded. The information about the subject’s heart Cardiac data acquired with either prospective cycle is then used during image reconstruction, ECG triggering or retrospective ECG gating are which is performed retrospectively, hence the used in reconstructing images. High temporal name retrospective gating. The image reconstruc- resolution images are obtained by reconstructing tion is performed either with data corresponding the data either with partial scan reconstruction or to partial scan data or with segmented reconstruc- with multiple-segment reconstruction. tion. In segmented reconstruction, data from differ- Partial Scan Reconstruction.—Among the ent parts of the heart cycle are chosen, so that the methods of image reconstruction in cardiac CT, sum of the segments equates to the minimal par- the most practical solution is the partial scan tial scan data required for image reconstruction. This results in further improvements in temporal 1500 September-October 2007 RG f Volume 27 Number 5 Figure 4. Differences between partial scan reconstruction versus multiple-segment reconstruction. Top: During partial scan reconstruc- tion, sufficient data from a pre- scribed time range within the R-R interval of one cardiac cycle are se- lected for reconstruction. Bottom: In multiple-segment reconstruction, sufficient data segments of the same phase from multiple cardiac cycles are selected for image reconstruc- tion. Higher temporal resolution (TR) can be achieved with this type of reconstruction. reconstruction. Partial scan reconstruction can source CT, and some are even considering devel- be used for both prospective triggering and retro- oping multiple x-ray source CT scanners. spective gating acquisitions. The minimum amount of data required to reconstruct a CT im- Multiple-Segment Reconstruction.—The pri- age is at least 180° plus the fan angle of data in mary limitation to achieving high temporal reso- any axial plane. This determines the scan time to lution with the partial scan approach is the gantry acquire projection data needed for partial scan rotation time. To achieve even higher temporal reconstruction and also limits the temporal reso- resolution, multiple-segment reconstruction was lution that can be achieved from an acquisition. developed (12). The principle behind multiple- The CT detectors in the axial plane of acquisition segment reconstruction is that the scan projection Teaching extend in an arc that covers at least a 30°– 60° fan data required to perform a partial scan recon- Point angle. Therefore, during partial scan reconstruc- struction are selected from various sequential tion, the scan data needed for reconstruction are heart cycles instead of from a single heart cycle obtained by rotating the x-ray tube by 180° plus (Fig 4). This is possible only with a retrospective the fan angle of the CT detector assembly (Fig 4). gating technique and a regular heart rhythm. The If the gantry rotation time is 500 msec, the CT projection data are acquired continuously time required to obtain the minimum scan data is throughout many sequential heart cycles. slightly greater than half of the gantry rotation The multiple-segment reconstruction method time. This means that, for a gantry rotation of selects small portions of projection data from vari- 500 msec, the scan time for acquiring data for ous heart cycles, so that when all the projections partial scan reconstruction is around 260 to 280 are combined, they constitute sufficient data to msec. This value represents the limit of temporal perform partial scan reconstructions. For ex- resolution that can be achieved through partial ample, if one chooses to select half of the data set scan reconstruction. To achieve further improve- required for partial scan reconstruction from one ments in temporal resolution, the CT manufac- heart cycle and the rest from another heart cycle, turers are driving scanner gantry rotation time this results in temporal resolution that is about faster and faster. To date, the fastest commer- one-fourth of the gantry rotation time. This is cially available gantry rotation time is 330 msec. done by using projection data from two separate In such scanners, the partial scan reconstruction segments of the heartbeat cycle for image recon- temporal resolution can be as high as 170 –180 struction. Further improvement in temporal reso- msec. At the same time, the G force generated lution can be achieved by cleverly selecting pro- due to rapid gantry motion is growing exponen- jection data from three or four different heart tially and is reaching a limit for the existing tech- cycles, resulting in temporal resolution as low as nology. The demand for even higher temporal 80 msec. resolution has led to the development of dual- In general, with multiple-segment reconstruc- tion, the temporal resolution can range from a maximum of TR/2 to a minimum of TR/2M, where TR is the gantry rotation time (seconds) RG f Volume 27 Number 5 Mahesh and Cody 1501 Figure 5. Effect of temporal resolution on reconstructed images from the same patient. (a) Partial scan reconstruction with temporal resolution of approximately 250 msec. (b) Multiple-segment reconstruction (two segments) yields a temporal resolution of approxi- mately 105 msec. The stair-step artifacts are less visible and the structures in the sagittal plane have a smooth edge compared with the appearance of partial scan reconstruction. and M is the number of segments in adjacent motion can result in the degradation of image heartbeats from which projection data are used spatial resolution. This method also allows selec- for image reconstruction. Usually, M ranges from tion of different packets of data for reconstructing 1 to 4. an image for patients with irregular heart rates. Overall, the temporal resolution of cardiac CT TR depends on the gantry rotation time. A gantry TR max ⫽. rotation time of 330 –500 msec is possible with 2M 16 – 64-channel multiple-row detector CT scan- ners. With such rapid gantry rotation time, one If TR ⫽ 400 msec and M ⫽ 1, then TRmax is as can achieve a temporal resolution of 80 –250 follows: msec through multiple- and partial-segment re- construction, respectively. Temporal resolution TR improves with multiple-segment reconstruction TR max ⱖ ⱖ 200 msec. (Fig 5); however, the spatial resolution can de- 2 grade due to misregistration of motion artifacts, since projection data sets are selected from differ- If TR ⫽ 400 msec and M ⫽ 2, then TRmax is as ent heartbeats. follows: With both types of reconstruction, there is a demand for a significant amount of projection TR overlap during data acquisition, which is indi- TR max ⱖ ⱖ 100 msec. cated by the pitch. Usually, the pitch ranges from 4 0.2 to 0.4 for cardiac CT protocols. This is quite different from routine body CT protocols, which If TR ⫽ 400 msec and M ⫽ 3, then TRmax is as are typically performed with pitch values of 0.75– follows: 1.50. TR TR max ⱖ ⱖ 67 msec. Spatial Resolution 6 There are a number of factors that influence the spatial resolution achieved with multiple-row de- The advantage of multiple-segment recon- tector CT scanners. Among them are the detector struction is the possibility to achieve high tempo- size in the longitudinal direction, reconstruction ral resolution. The disadvantage is that because algorithms, and patient motion. projection data sets are obtained from different heartbeat cycles, a misregistration due to rapid 1502 September-October 2007 RG f Volume 27 Number 5 Effect of Detector Size The effect of detector size in the z direction or out-of-plane spatial resolution is very significant and has become one of the driving forces in the advancement of multiple-row detector CT tech- nology. Also, larger volume coverage in combina- tion with a larger number of thin images requires more detectors in the z direction, which is the hallmark of technological advance in multiple- row detector CT. On the other hand, scan plane or axial spatial resolution has been very high from the beginning and is dependent on the scan field of view (SFOV) and image reconstruction matrix. Axial pixel size is the ratio of SFOV to image ma- Figure 6. Detector array designs for multiple- trix; for example, for a conventional 512 ⫻ 512 row detector CT scanners that can yield 64 sec- matrix, the transverse pixel size for a 25-cm tions per gantry rotation. SFOV is 0.49 mm. On the other hand, the longi- tudinal or z-axis resolution mainly depends on the image thickness. The z-axis spatial resolution tient exposure. The reason for decreased recon- (image thickness) ranges from 1 to 10 mm in con- struction interval (or increased overlap) is to im- ventional (nonhelical) and in helical single-row prove z-axis resolution, especially for three-di- detector CT. With multiple-row detector CT, the mensional (3D) and multiplanar reformation z-axis detector size is further reduced to submilli- (MPR) images. If the reader is making a diagnosis meter size. based on only axial images, reconstruction inter- Initially, with the introduction of multiple-row val is not an issue. But frequently, physicians are detector CT technology, the thinnest detector also reading MPR and 3D images; this is espe- size was 0.5 mm and there were only two such cially true for cardiac CT. detectors. However, within a few years, the tech- For example, in a single examination the same nology improved to provide 16 of these thin de- cardiac data set (acquired at 0.5-mm detector tectors, ranging from 0.625 to 0.5 mm. With 64- configuration) was reconstructed with three dif- section multiple-row detector CT scanners, the ferent values of reconstruction interval (Fig 8). detector array designs are as shown in Figure 6; Overlapping axial images results in a relatively 64 thin detectors (0.625 mm) are currently avail- large number of images but can also result in im- able, resulting in z-axis coverage of up to 40 mm proved lesion visibility in MPR and 3D images per gantry rotation (11). The increased spatial without increasing the patient dose. For routine resolution with multiple-row detector CT scan- MPR and 3D applications, a 30% image overlap ners is demonstrated in Figure 7. Cardiac CT is generally sufficient (1-mm section thickness images can be comparable in delineating details with 0.7-mm reconstruction interval). For cardiac of the coronary vessels to the cardiac images ob- images, at least 50% overlap is desirable (0.5-mm tained with fluoroscopy (8). section thickness with 0.25-mm reconstruction interval). Reconstruction Interval It should be recognized that too much overlap The reconstruction interval defines the degree of results in a large number of images, increases re- overlap between reconstructed axial images. It is construction time, can result in longer interpreta- independent of x-ray beam collimation or image tion periods, and can put undue strain on image thickness and has no effect on scan time or pa- handling overhead costs (image transfer, image display, image archiving, etc) with no significant gain in image quality. Overall, spatial resolution in the axial or x-y plane has always been quite high and is on the RG f Volume 27 Number 5 Mahesh and Cody 1503 Figure 7. Images from cardiac CT angiography (a) and fluoroscopically guided coronary angiography (b) show a right coronary artery (long arrow) with calcification (short arrows). The spatial resolution and delineation of details of CT angiography are comparable with those of coronary angiography. (Reprinted, with permission, from reference 8.) Figure 8. Effect of reconstruction interval on image quality. All three sets of images are from the same data set reconstructed with 0.5-mm section thickness. How- ever, the reconstruction intervals are different, which affects the number of reconstructed images and 3D image quality. (a) A reconstruction interval of 0.3 mm yields 301 images and implies a 60% overlap. (b) A reconstruction interval of 5 mm yields only 19 images and results in a staggered appearance of 3D images. (c) A reconstruction interval of 0.5 mm yields 184 im- ages and results in image quality similar to that of a. Normally, a 50% overlap is sufficient for optimum im- age quality for MPR and 3D images. order of 10 –20 line pairs per centimeter. The z- per centimeter. The efforts toward obtaining iso- axis spatial resolution is influenced by the detec- tropic resolution are leading further developments tor size, reconstruction thickness, and other fac- in multiple-row detector CT technology. tors such as pitch and is around 7–15 line pairs 1504 September-October 2007 RG f Volume 27 Number 5 Figure 9. Pitch is defined as the ratio of feed per gantry rotation to the total x-ray table width. This definition is applicable to bothbeam row detector CT and multiple-row detectorsingle- (21). I ⫽ table travel (millimeters) per CT N ⫽ number of active data acquisitionrotation, T ⫽ single data acquisition channel width channels, meters), W ⫽ beam width (millimeters). (milli- Figure 10. Graphs demonstrate the necessity for scanning at low pitch values during helical Pitch cardiac CT data acquisition. If the table feed be- comes greater than the beam width, it results in The concept of pitch was introduced with the ad- a data gap, which is detrimental for image vent of spiral CT and is defined as the ratio of reconstruction. table increment per gantry rotation to the total x-ray beam width (20,21) (Fig 9). Pitch values cardiac CT protocols require injecting beta-block- less than 1 imply overlapping of the x-ray beam ers (23) to lower the subject’s heartbeat within the and higher patient dose; pitch values greater than desirable range of less than 70 beats per minute. 1 imply a gapped x-ray beam and reduced patient When the subject’s heart rates are rapid and dose (18). Cardiac imaging demands low pitch difficult to control, the diastolic ranges are values because higher pitch values result in data smaller, so images are reconstructed using mul- gaps (Fig 10), which are detrimental to image tiple-segment reconstruction in order to improve reconstruction. Also, low pitch values help mini- temporal resolution. With multiple-segment re- mize motion artifacts, and certain reconstruction construction, the number of segments used in the algorithms work best at certain pitch values, reconstruction further restricts the pitch factors. which are lower than 0.5 in cardiac imaging. Typical multiple-row detector CT pitch factors used for cardiac imaging range from 0.2 to 0.4. The pitch required for a particular scanner is Pⱕ 冉 NM 冊 N⫹M⫺1 T R TRR , affected by several parameters, as shown in the fol- where N ⫽ number of active data acquisition lowing equation. For single-segment reconstruc- channels, M ⫽ number of segments or subse- tion (partial scan acquisition), the pitch factor is quent heart cycles sampled, TR ⫽ gantry rotation influenced heavily by the subject’s heart rate (22). time (milliseconds), and TRR ⫽ time for a single 冉 冊 heartbeat (milliseconds). For a heart rate of 60 N⫺1 TR Pⱕ , beats per minute with a TR of 400 msec, N ⫽ 16, N TRR ⫹ TQ and M ⫽ 2, the required pitch is ⬃0.21; similarly, for M ⫽ 3, the required pitch is ⬃0.15. where N ⫽ number of active data acquisition The pitch factor plays a significant role in im- channels, TR ⫽ gantry rotation time (millisec- proving both the temporal and spatial resolution Teaching onds), TRR ⫽ time for a single heartbeat (millisec- but at the same time has a dramatic effect on the Point onds), and TQ ⫽ partial scan rotation time (milli- overall radiation dose delivered during a cardiac seconds). For heart rates of 45–100 beats per CT examination. Because radiation dose is in- minute (TRR of 1333– 600 msec), TR of 500 msec, versely proportional to the pitch, the low pitch and TQ of 250 –360 msec, the required pitch fac- values characteristic of cardiac CT protocols sub- tor ranges from 0.375 to 0.875. At higher pitch, stantially increase radiation dose to patients un- there are substantial data gaps. As a result, most dergoing cardiac imaging with multiple-row de- tector CT. RG f Volume 27 Number 5 Mahesh and Cody 1505 Typical Effective Doses for Various Cardiac Imaging and Routine CT Procedures Imaging Procedures Modality Effective Dose (mSv)* Cardiac procedures Calcium scoring Electron beam CT 1.0–1.3 Multiple-row detector CT 1.5–6.2† Cardiac CT angiography Electron beam CT 1.5–2.0 Multiple-row detector CT 6†–25 Cardiac SPECT with 99mTc or 201T1‡ Nuclear medicine 6.0–15.0 Coronary angiography (diagnostic) Fluoroscopy 2.1†–6.0 Chest radiography Radiography 0.1–0.2 Routine CT procedures Head CT Multiple-row detector CT 1–2 Chest CT Multiple-row detector CT 5–7 Abdominal and pelvic CT Multiple-row detector CT 8–11 *10 mSv ⫽ 1 rem. †Indicated data are from reference 25. ‡SPECT ⫽ single photon emission CT, 99mTc ⫽ technetium 99m, 201T1 ⫽ thallium 201. In cardiac CT imaging, the need for high spa- (Table). Across the board, radiation doses are tial and temporal resolution in turn requires the higher with multiple-row detector CT compared pitch values to be as low as 0.2– 0.4. This results with the doses delivered with electron beam CT in a radiation beam overlap of nearly 80%– 60%, and fluoroscopically guided diagnostic coronary respectively, and an increase in radiation dose of angiography and similar procedures (25). up to a factor of five times compared to a pitch of One approach to reduce the high dose associ- 1. Hence, proper choice and optimization of pitch ated with retrospective gating is called ECG dose factor is critical in cardiac imaging. In fact, the modulation (27) and is directed at reducing the demand for reducing radiation dose and faster tube current during specific parts of the cardiac scans is driving the technology to introduce either cycle, particularly during systole, where image an even higher number of thin detectors in the z quality is already degraded by motion artifacts direction (256 rows) or flat panel technology so and these portions of the cardiac cycle are not that the entire cardiac area can be covered in a used in the image reconstruction. When dose single gantry rotation without the necessity for modulation is implemented, a 10%– 40% dose tissue overlap (low pitch values). reduction (27) can be achieved; however, the sav- ings must be evaluated for each specific CT pro- Radiation Risk tocol. It is important that any steps taken to re- One of the disadvantages of cardiac imaging with duce radiation exposure should not jeopardize the multiple-row detector CT is its use of ionizing image quality, because poor image quality may radiation. The radiation dose delivered is highly result in repeat scans, which would result in addi- dependent on the protocol used in cardiac CT tional radiation doses to patients. (24). Among the most widely known protocols such as calcium scoring studies, the effective dose Geometric Efficiency is relatively small, 1–3 mSv (25). However, for Dose efficiency (also called geometric efficiency) retrospective gating, used for coronary vessel ste- is of particular concern with the earlier four-chan- nosis assessment and CT angiography, effective nel multiple-row detector CT scanners, for which doses of 8 –22 mSv and higher have been re- the x-ray photon beam has to be quite uniform as ported. By comparison, the radiation dose of an it strikes the detector array. This requirement uncomplicated diagnostic coronary angiography means that the natural shadowing of the beam study performed under fluoroscopic guidance (penumbra) attributable to the finite-sized focal ranges from 3 to 6 mSv (11,25) and for typical spot is intentionally positioned to strike the neigh- body CT protocols ranges from 2 to 10 mSv (26) boring nonactive detector elements. Thus, some amount of radiation transmitted through the pa- tient does not contribute to image generation. The width of the penumbra is fairly constant with each scanner, generally in the range of 1–3 mm. 1506 September-October 2007 RG f Volume 27 Number 5 Figure 11. Left anterior oblique (a) and anterior (b) MPR images show cardiac pulsation artifacts due to a rapid heartbeat. (Reprinted, with permission, from reference 29.) Figure 12. Banding artifacts due to an increased heart rate from 51 to 69 beats per minute. Coronal (a) and sagittal (b) reformatted images of the heart obtained from CT data show banding artifacts (arrowheads). (Reprinted, with permission, from reference 29.) The proportion of radiation wasted relative to most common artifacts are due to cardiac pulsa- the overall width of the x-ray beam varies with the tion (29). Figure 11 shows disconnect in the lat- protocol used. If very thin images are required eral reconstructed image due to pulsation. These and the overall x-ray beam width is small—5 mm, types of artifacts are minimized by multiple-seg- for example—then the proportion of wasted ment reconstruction or by scanning at even x-rays could be 20%– 60% (resulting in a dose higher temporal resolution on the order of 50 efficiency of 40%– 80%). If thin images are not msec. The second types of artifacts are the band- required and a wider x-ray beam can be used—20 ing artifacts due to increased heart rate during the mm, for example—then the proportion of wasted scan. In the example shown in Figure 12, the x-rays would be 5%–15% (resulting in a dose effi- heart rate varied from 51 to 69 beats per minute ciency of 85%–95%). Dose efficiency may be dis- during the scan and resulted in banding artifacts played on the scanner console. More recent mul- (29). The other types of cardiac artifacts com- tiple-row detector CT scanners have been engi- monly observed are due to incomplete breath neered such that this problem is very much holding. These types of artifacts are not observed diminished. on axial images but are visible on coronal or sagit- tal views (Fig 13). Artifacts When subjects with previous stents or coils in In cardiac imaging, owing to the inherent nature the coronary artery undergo CT imaging, we ob- of imaging a rapidly moving organ, there arise serve streak artifacts around these highly attenu- many unique artifacts (28,29); among them, the ating objects. Often these artifacts can dominate the artery region and obscure other structures. As shown in Figure 14, the metallic structures RG f Volume 27 Number 5 Mahesh and Cody 1507 Figure 13. Artifacts due to incomplete breath holding. (a) Axial images show no motion artifacts. (b, c) Coronal (b) and sagittal (c) reformatted images show banding artifacts. (Reprinted, with permis- sion, from reference 29.) Figure 14. Streak artifacts visible in the presence of a stent. Thin-slab maximum in- tensity projection image (a), MPR image (b), and thin-slab maximum intensity projection image obtained with a wide window (c) show streak artifacts (arrows in a). (Reprinted, with permission, from reference 29.) 1508 September-October 2007 RG f Volume 27 Number 5 appear on axial images with no streak artifact but possible to obtain complete data from one heart are very distinct and disturbing in coronal or sag- cycle, further diminishing the need for excessive ittal planes (29). These types of artifacts are to tissue overlaps (low pitch values) and therefore some extent handled by special artifact reduction reducing radiation dose and also reducing motion software developed by manufacturers. The artifacts. Ideally, the combination of a 256-row blooming artifacts are caused primarily by the detector assembly in the dual-source CT scanner combination of very highly attenuating objects would be phenomenal because that would not and the inherent limiting resolution of the scan- only give high temporal resolution but also high ner. spatial resolution with minimal motion artifacts. Future Directions Conclusions in Cardiac Imaging Cardiac imaging is a highly demanding applica- The demand for higher temporal and spatial reso- tion of multiple-row detector CT and is possible lution has already led to the development of a only due to recent technological advances. Un- dual-source CT scanner (30) and a 256-row de- derstanding the trade-offs between various scan tector CT scanner (31). In dual-source CT, there parameters that affect image quality is key in opti- are two x-ray tubes positioned 90° apart provid- mizing protocols that can reduce patient dose. ing 64 axial sections for a complete gantry rota- Benefits from an optimized cardiac CT protocol tion, which yields further improvement in tempo- can minimize the radiation risks associated with ral resolution. As mentioned earlier, the mini- these cardiac scans. Cardiac CT has the potential mum amount of data needed to reconstruct an to become a reliable tool for noninvasive diagno- image is 180° plus the fan angle; therefore, with sis and prevention of cardiac and coronary artery two tubes positioned at 90°, it is sufficient to ac- disease. quire data for one-fourth of a gantry rotation and then coordinate the data from two sets of detec- References tors to reconstruct the image. This can yield tem- 1. American Heart Association. International cardio- poral resolution as low as one-fourth the gantry vascular disease statistics. Dallas, Tex: American Heart Association, 2003. rotation speed. With scanner gantry rotation 2. Mahnken AH, Wildberger JE, Koos R, Gunther speeds at below 330 msec, the temporal resolu- RW. Multislice spiral computed tomography of tion can be as low as 80 msec. With this scanner, the heart: technique, current applications, and the pitch factor may be increased for higher heart perspective. Cardiovasc Intervent Radiol 2005;28: rates with a potential to reduce radiation dose. 388 –399. 3. Budoff MJ, Achenbach S, Duerinckx A. Clinical Similarly, the demand for higher spatial resolu- utility of computed tomography and magnetic tion has led to the development of a 256-row de- resonance techniques for noninvasive coronary tector CT scanner, which can cover the entire angiography. J Am Coll Cardiol 2003;42:1867– heart in one gantry rotation (12.8-cm beam width 1878. at isocenter). In the 256-row detector CT scan- 4. Ulzheimer S, Kalender WA. Assessment of cal- cium scoring performance in cardiac computed ner, the 256 detectors in the longitudinal direc- tomography. Eur Radiol 2003;13:484 – 497. tion can cover an area of 12.8 mm per gantry ro- 5. Dewey M, Borges AC, Kivelitz D, et al. Coronary tation and therefore can eliminate the need for artery disease: new insights and their implications overlapping pitch. In this type of scanner, it is for radiology. Eur Radiol 2004;14:1048 –1054. 6. Pannu HK, Jacobs JE, Lai S, Fishman EK. Coro- nary CT angiography with 64-MDCT: assessment of vessel visibility. AJR Am J Roentgenol 2006; 187:119 –126. RG f Volume 27 Number 5 Mahesh and Cody 1509 7. Schoenhagen P, Halliburton SS, Stillman AE, et 19. Desjardins B, Kazerooni EA. ECG-gated cardiac al. Noninvasive imaging of coronary arteries: cur- CT. AJR Am J Roentgenol 2004;182:993–1010. rent and future role of multi– detector row CT. 20. Mahesh M, Scatarige JC, Cooper J, Fishman EK. Radiology 2004;232:7–17. Dose and pitch relationship for a particular multi- 8. Hoffmann MH, Shi H, Schmid FT, Gelman H, slice CT scanner. AJR Am J Roentgenol 2001; Brambs HJ, Aschoff AJ. Noninvasive coronary im- 177:1273–1275. aging with MDCT in comparison to invasive con- 21. International Electrotechnical Commission. Medi- ventional coronary angiography: a fast-developing cal electrical equipment, part 2-44: particular re- technology. AJR Am J Roentgenol 2004;182:601– quirements for the safety of x-ray equipment for 608. computed tomography. IEC Publication No. 9. McCollough CH, Morin RL. The technical design 60601-2-44. Geneva, Switzerland: International and performance of ultrafast computed tomogra- Electrotechnical Commission, 2002. phy. Radiol Clin North Am 1994;32:521–536. 22. Ohnesorge B, Becker C, Flohr T, Reiser MF. 10. Detrano RC. Coronary artery scanning using elec- Multi-slice CT in cardiac imaging. Heidelberg, tron beam computed tomography. Am J Card Im- Germany: Springer, 2002. aging 1996;10:97–100. 23. Pannu HK, Alvarez W Jr, Fishman EK. Beta- 11. Mahesh M. Cardiac imaging: technical advances blockers for cardiac CT: a primer for the radiolo- in MDCT compared with conventional x-ray an- gist. AJR Am J Roentgenol 2006;186(6 suppl 2): giography. In: Boulton E, ed. US cardiology 2006: S341–S345. the authoritative review of the clinical and scien- 24. Gerber TC, Kuzo RS, Morin RL. Techniques and tific issues relating to cardiology with perspectives parameters for estimating radiation exposure and on the future. London, England: Touch Briefings dose in cardiac computed tomography. Int J Car- (http://www.touchcardiology.com), 2006; 115– diovasc Imaging 2005;21:165–176. 119. 25. Hunold P, Vogt FM, Schmermund A, et al. Ra- 12. Flohr TG, Schaller S, Stierstorfer K, Bruder H, diation exposure during cardiac CT: effective Ohnesorge BM, Schoepf UJ. Multi– detector row doses at multi– detector row CT and electron- CT systems and image-reconstruction techniques. beam CT. Radiology 2003;226:145–152. Radiology 2005;235:756 –773. 26. Morin RL, Gerber TC, McCollough CH. Radia- 13. Gerber B, Rosen BD, Mahesh M, Araujo LI, St tion dose in computed tomography of the heart. John Sutton M, Lima JAC. Physical principles of Circulation 2003;107:917–922. cardiovascular imaging. In: St John Sutton M, 27. Jakobs TF, Becker CR, Ohnesorge B, et al. Multi- Rutherford J, eds. Clinical cardiovascular imaging: slice helical CT of the heart with retrospective a companion to Braunwald’s heart disease. Phila- ECG gating: reduction of radiation exposure by delphia, Pa: Elsevier-Saunders, 2004; 1–77. ECG-controlled tube current modulation. Eur 14. Klingenbeck-Regn K, Flohr T, Ohnesorge B, Radiol 2002;12:1081–1086. Regn J, Schaller S. Strategies for cardiac CT imag- 28. Choi HS, Choi BW, Choe KO, et al. Pitfalls, arti- ing. Int J Cardiovasc Imaging 2002;18:143–151. facts, and remedies in multi– detector row CT cor- 15. Klingenbeck-Regn K, Schaller S, Flohr T, Ohne- onary angiography. RadioGraphics 2004;24:787– sorge B, Kopp AF, Baum U. Subsecond multi- 800. slice computed tomography: basics and applica- 29. Nakanishi T, Kayashima Y, Inoue R, Sumii K, tions. Eur J Radiol 1999;31:110 –124. Gomyo Y. Pitfalls in 16 – detector row CT of the 16. Nikolaou K, Flohr T, Knez A, et al. Advances in coronary arteries. RadioGraphics 2005;25:425– cardiac CT imaging: 64-slice scanner. Int J Car- 438; discussion 438 – 440. diovasc Imaging 2004;20:535–540. 30. Flohr TG, McCollough CH, Bruder H, et al. First 17. Pannu HK, Flohr TG, Corl FM, Fishman EK. performance evaluation of a dual-source CT Current concepts in multi– detector row CT (DSCT) system. Eur Radiol 2006;16:256 –268. evaluation of the coronary arteries: principles, 31. Mori S, Endo M, Obata T, et al. Clinical poten- techniques, and anatomy. RadioGraphics 2003; tials of the prototype 256-detector row CT-scan- 23(spec issue):S111–S125. ner. Acad Radiol 2005;12:148 –154. 18. Mahesh M. Search for isotropic resolution in CT from conventional through multiple-row detector. RadioGraphics 2002;22:949 –962. RG Volume 27 Volume 5 September-October 2007 Mahesh and Cody Physics of Cardiac Imaging with Multiple-Row Detector CT Mahadevappa Mahesh, MS, PhD, and Dianna D. Cody, PhD RadioGraphics 2007; 27:1495–1509 Published online 10.1148/rg.275075045 Content Codes: Page 1498 The minimum amount of projection data required to construct a complete CT image is 180° plus the fan angle of the CT detectors in the axial plane. Page 1498 One of the advantages of the prospective triggering approach is reduced radiation exposure, because the projection data are acquired for short periods and not throughout the heart cycle. Page 1499 The disadvantage of the retrospective gating mode of acquisition is the increased radiation dose, because the data are acquired throughout the heart cycle, even though partial data are actually used in the final image reconstruction. Page 1500 The principle behind multiple-segment reconstruction is that the scan projection data required to perform a partial scan reconstruction are selected from various sequential heart cycles instead of from a single heart cycle (Fig 4). Page 1504 The pitch factor plays a significant role in improving both the temporal and spatial resolution but at the same time has a dramatic effect on the overall radiation dose delivered during a cardiac CT examination. Because radiation dose is inversely proportional to the pitch, the low pitch values characteristic of cardiac CT protocols substantially increase radiation dose to patients undergoing cardiac imaging with multiple-row detector CT.