Automated Identification of Minimal Myocardial Motion for Improved Image Quality on MR Angiography at 3 T

Transcription

1 Automated Identification of Minimal Myocardial Motion on MRA Cardiac Imaging Technical Innovation Ali Ustun 1,2 Milind Desai 1,3 Khaled Z. Abd-Elmoniem 2 Michael Schar 1,4 Matthias Stuber 1,2,5 Ustun A, Desai M, Abd-Elmoniem KZ, Schar M, Stuber M Keywords: cardiac imaging, heart, MR angiography, MRI, MR technique DOI: /AJR Received March 6, 2006; accepted after revision August 1, This work was supported by a biomedical engineering grant from the Whitaker Foundation (RG ), a grant from the Donald W. Reynolds Foundation, and a grant from the National Institutes of Health (HL61912). M. Stuber is compensated as a consultant by Philips Medical Systems, the manufacturer of equipment described in this presentation. The terms of this arrangement have been approved by the Johns Hopkins University in accordance with its conflict-of-interest policies. 1 Department of Radiology, Johns Hopkins University Medical School, JHOC 4223, 601 N Caroline St., Baltimore, MD Address correspondence to M. Stuber. 2 Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD. 3 Department of Cardiovascular Medicine, Cleveland Clinic Foundation, Cleveland, OH. 4 Philips Medical Systems, Cleveland, OH. 5 Department of Medicine, Johns Hopkins University Medical School, Baltimore, MD. WEB This is a Web exclusive article. AJR 2007; 188:W283 W X/07/1883 W283 American Roentgen Ray Society Automated Identification of Minimal Myocardial Motion for Improved Image Quality on MR Angiography at 3 T OBJECTIVE. Imaging during a period of minimal myocardial motion is of paramount importance for coronary MR angiography (MRA). The objective of our study was to evaluate the utility of FREEZE, a custom-built automated tool for the identification of the period of minimal myocardial motion, in both a moving phantom at 1.5 T and 10 healthy adults (nine men, one woman; mean age, 24.9 years; age range, years) at 3 T. CONCLUSION. Quantitative analysis of the moving phantom showed that dimension measurements approached those obtained in the static phantom when using FREEZE. In vitro, vessel sharpness, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were significantly improved when coronary MRA was performed during the software-prescribed period of minimal myocardial motion (p < 0.05). Consistent with these objective findings, image quality assessments by consensus review also improved significantly when using the automated prescription of the period of minimal myocardial motion. The use of FREEZE improves image quality of coronary MRA. Simultaneously, operator dependence can be minimized while the ease of use is improved. ne of the major problems associated with coronary MR angiogra- O phy (MRA) is cardiac motion resulting from both respiration and natural cardiac contraction and relaxation. Breath-hold and navigator techniques have been used to minimize respiratory motion artifacts [1]. However, the quality and diagnostic value of coronary MRA remain variable [2]. The motion originating from cardiac contraction and relaxation contributes to this variability. The average displacement of the left coronary arteries during a time period of 120 milliseconds may be as much as 6 mm [3, 4]. Throughout the entire cardiac cycle, a total displacement of 9 and 18 mm for the left and right coronary arterial systems, respectively, has been measured using biplanar angiography [5]. On MRI, signal sampling during a period of rapid coronary motion leads to blurring that adversely affects the diagnostic quality of the images. The duration of the acquisition window can be abbreviated to minimize this effect. However, this change is associated with prolonged scanning times. Evidently, myocardial motion is not constant throughout the entire cardiac cycle. Therefore, careful selection of both the position and the duration [6] of the MRI acquisition window plays an important role in minimizing the effects of blurring on coronary MRA [7]. Furthermore, these rest periods must be determined on a per-subject basis [8]. Although visual assessment of this quiescent period on MR cine images is likely to be superior to empiric formulas [9], visual inspection is still subjective. A powerful calibration scan based on navigator echoes was introduced by Wang et al. [9]. However, this approach necessitates meticulous geometric planning of the navigator position and has currently not found widespread use. As an alternative, automated image-based cross-correlation analysis for multi heart phase images has recently been reported for the identification of the period of minimal myocardial motion [8]. Hereby, a frame-toframe correlation method on a selected volume encompassing the heart was used. In these studies, the trigger delays found by visual assessment and those identified by an automated algorithm were compared. However, the effect of these strategies on coronary MR image quality was not ascertained to our knowledge. In the present work, we tested the hypothesis that coronary MRA performed at a computer-identified time delay after the R wave of the ECG leads to better objective and subjective image quality when compared with that using a visually identified period of minimal myocardial motion. For the auto- AJR:188, March 2007 W283

2 Fig. 1 Screen shot shows FREEZE software tool. Rectangle on axial image shows user-defined region of interest (ROI) in which motion is analyzed. FREEZE operates on user-friendly interface in which user can zoom in or out and pan and level image; user can also select different time frames and pick most suitable frame for ROI selection. After ROI selection, algorithm starts automatically and result (software-prescribed trigger delay = Tds f ) is sent directly to scanner. Another option is that user can select restricted time window (on displacement vs time plot [small window]) for which period of minimal motion should be analyzed by algorithm. mated detection of that rest period, a computer algorithm, FREEZE, was developed; the name FREEZE is not an abbreviation or acronym. The adequacy of this algorithm to detect minimal motion was first tested in a moving phantom [10] at 1.5 T. Subsequently, 3-T in vivo coronary MRA images obtained using visual inspection of the period of minimal motion were compared with those obtained using FREEZE. Materials and Methods Algorithm and Implementation A previously reported [8] image-based correlation algorithm was adopted and extended to examine multi heart phase cine images on a frame-toframe basis. A manually drawn region of interest (ROI) was used as the correlation kernel, and each consecutive cine frame pair was correlated in this user-specified area (Fig. 1A). Rather than directly using these correlation values to extract motion (Nehrke K et al., presented at the 2003 annual meeting of the International Society of Magnetic Resonance in Medicine), the center of mass of each resulting correlation matrix was identified using a center-of-mass calculation. Then, the relative 2D displacement of this center of mass between successive image pairs was calculated. This results in a time series of displacement values that are proportional to the amount of motion occurring in the ROI. Subsequently, a moving average displacement is calculated using these values in a user-specified time window (i.e., the acquisition window for coronary MRA), and the time point of minimal displacement is identified and transferred to the scanner. This FREEZE algorithm (Fig. 1) was implemented using interactive data language (IDL 6.0, RSI, Inc.) on a commercial PC running Windows XP (Microsoft) that is interfaced with the scanner. The scanner was programmed to automatically use this software-prescribed time point of minimal myocardial motion for coronary MRA scanning. Phantom Experiments To test the hypothesis that FREEZE can effectively identify periods of minimal motion, phantom studies were conducted [11]. An MR-compatible sinusoidally moving phantom with ECG output and a maximum displacement of 1.4 cm (frequency, 72 cycles/min; maximum velocity, 10.2 cm/s) was developed, as described in an earlier article [10]. A curved silicone tube with an internal diameter of 1.8 mm attached to a water bottle was mounted on the moving phantom (Fig. 2). The tube was filled with mineral oil (Johnson s Baby Oil, Johnson & Johnson). A standard 2D functional cine steady-state freeprecession (SSFP) scan (temporal resolution [ t] = 14 milliseconds, n = 60 frames per R-R interval) of the moving phantom was then acquired and further processed by the FREEZE software. The duration of the time window in which FREEZE should identify minimal motion was set to approximately 70 milliseconds, as previously reported for coronary MRA [12]. In this phantom experiment, Tds is the time delay between the R wave and imaging for which minimal motion is found. Images of the moving phantom were then obtained using a conventional navigator-gated and navigator-corrected 3D high-resolution segmented k-space imaging sequence (TR/TE = 7.0/2.4, α = 35, resolution = mm, 10 k-space lines per cardiac cycle, acquisition time window [T acq ] = 70 milliseconds, field of view = mm, matrix). The trigger delay between the phantom-generated R wave and the imaging sequence was then adjusted from 100 to 800 milliseconds in increments of 100 milliseconds. For comparison, one static scan without motion of the phantom was acquired and another scan with the trigger delay prescribed by the software was acquired as well. Vessel diameter and sharpness were then analyzed using Soap-Bubble software [13]. All the phantom scans were obtained on a commercial 1.5-T system (Gyroscan Intera, Philips Medical Systems) equipped with a PowerTrak 6000 gradient system (Philips Medical Systems) (23 mt/m, 219-µsec rise time). The vessel diameter was calculated as the average value found perpendicular to a 3- cm segment of the plastic tube oriented perpendicular to the direction of motion (segments I and II in Fig. 2C). The vessel sharpness values were calculated for the same segment using the Deriche algorithm [14]. This analysis was then repeated for another 3-cm segment of the tube in parallel to the readout direction (segments III and IV in Fig. 2C). In Vivo Experiments Subjects Ten healthy human subjects (nine men, one woman; mean age, 24.9 years; age range, years) who did not have contradictions to MRI were examined. Written informed consent was obtained W284 AJR:188, March 2007

3 Automated Identification of Minimal Myocardial Motion on MRA A B C Fig. 2 Navigator-gated and navigator-corrected 3D high-resolution segmented k-space imaging of phantom (1.5 T, TR/TE = 7.0/2.4, α =35, resolution= mm, 10 k- space lines per cardiac cycle, acquisition time window [T acq ] = 70 milliseconds, field of view = mm, matrix). A, Static image of phantom. Arrow points to plastic tube, with internal diameter of 1.8 mm, that is attached to water bottle. B 0 = external magnetic field. B, Image of phantom obtained at maximum velocity (10.2 cm/s). Vessel diameter is measured approximately 2.7 mm. C, Image of phantom at trigger delay (493 milliseconds) identified using FREEZE. Vessel diameter is measured approximately 1.8 mm. Diameter in contiguous segments was quantified perpendicular and parallel to direction of motion. Measurements between I and II and between III and IV, respectively, were performed. from all participants, and the research protocol was approved by the hospital committee on clinical investigation. The study was compliant with the Health Insurance Portability and Accountability Act. MR technique All the in vivo MR scans were obtained on a commercial 3-T system (Gyroscan Achieva, Philips Medical Systems) equipped with a Dual Quasar gradient system (Philips Medical Systems) (80 mt/m, 200-µsec rise time, 16-channel parallel receiver architecture). All subjects were imaged in the supine position using a six-element cardiac phased-array coil for signal reception and vector ECG triggering [15]. To test the hypothesis that the use of FREEZE leads to improved coronary MRA image quality when compared with images obtained using visual inspection of the period of minimal myocardial motion, the following coronary MRA protocol was used. To localize the heart, a low-resolution 2D free-breathing segmented k-space gradientecho imaging scan was obtained in the transverse, coronal, and sagittal orientations (3.8/1.8, α = 20 ). The scanning duration for the scan was 12 seconds. The navigator for subsequent coronary MRA was localized at the dome of the right hemidiaphragm as identified on the transverse and coronal views of this first scout scan. The navigator was localized one third above the lung liver interface and two thirds below with a 5-mm navigator-gating window. Subsequently, low-resolution free-breathing navigator-gated and navigator-corrected 3D segmented k-space gradient-echo imaging (2.5/1.29, α = 15, resolution = mm, field of view = mm, matrix) was performed in the transverse orientation to accurately localize the left and right coronary arterial systems. Twenty-five radiofrequency excitations were performed per R-R interval (T acq = 63 milliseconds), and the total scanning duration for this scan was approximately 2 minutes during free breathing. On these images, volume targeting parallel to the right and left coronary arterial systems was obtained using a 3-point plan scan tool [16]. Next, an axial free-breathing 2D segmented k- space gradient-echo cine sequence with 4 signal averages (2.7/1.37, α =25, resolution=2 2 8 mm, field of view = mm, matrix, scan duration = 15 seconds, 50 frames/s) was used for imaging at a midventricular level. This scan was used to determine the time point (after the R wave of the ECG) of the onset of the period of minimal myocardial motion for both visual assessment (Tds v ) and FREEZE (Tds f ). To determine Tds f, an ROI of approximately pixels that included both the left and right ventricles was manually selected on an end-diastolic frame of the axial cine images. To image the coronary arteries with a high spatial resolution at Tds v and Tds f, volume-targeted navigator-gated and navigator-corrected double oblique 3D segmented k-space gradient-echo imaging was performed for the left and right coronary arterial systems (4.3/1.47, α =20, resolution= mm, field of view = mm, matrix, 16 radiofrequency excitations per R-R interval, T acq = 69 milliseconds, bandwidth = 362 Hz/pixel, scan duration = seconds depending on the navigator efficiency and heart rate, 10 slices [acquired], 20 slices [reconstructed using zero filling], fat saturation, adiabatic T2 prepulse, [TE = 50 milliseconds]) [17]. In one volunteer study, free-breathing coronary MRA using Tds f was performed with a voxel size of mm (7.5/2.3, α = 20, field of view = mm, matrix, scan duration = 906 seconds, 12 radiofrequency excitations per R-R interval, T acq = 90 milliseconds, 10 slices [acquired], 20 slices [reconstructed], fat saturation). Image Analysis Quantitative analysis of coronary MRA was performed using the Soap-Bubble tool [13] by one reviewer who was blinded to the scanning method (Tds v or Tds f ). Images obtained at Tds v and Tds f were compared for length of visible vessel, vessel sharpness, vessel diameter, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). Length measurements were performed for both the right coronary artery (RCA) and the left anterior descending (LAD) coronary artery. Vessel sharpness AJR:188, March 2007 W285

4 3.0 Diameter (mm) Static Trigger Delay Value (ms) and vessel diameter were measured in the proximal 4 cm of both the RCA and LAD. As described indepth in a related article by Etienne et al. [13], a Deriche algorithm was applied on the multiplanar reformatted coronary Soap-Bubble image. The local value in a Deriche image represents the magnitude of local change in signal intensity (derivative), which allows the user to map the vessel sharpness along the diameter of a selected vessel. SNR calculations were conducted by the same reviewer by selecting two ROIs: one in the intraaortic blood pool close to coronary ostia (ROI A ) and another one in a region anterior to the chest wall (ROI B ) where no respiration-induced motion artifacts were identified. SNR was then computed as the average signal intensity in ROI A divided by the SD of ROI B. In a similar manner, CNR was determined using three ROIs: ROI A and ROI B were identical to those used for the SNR assessment, and the third ROI was positioned on the myocardium (ROI C ). The difference between the signal intensities in ROI A and ROI C was then divided by the SD in ROI B to calculate CNR. The qualitative analysis of coronary MRA was first performed individually by two reviewers blinded to how images were obtained (i.e., with Tds v or Tds f ). This was followed by a consensus review by the same reviewers using a previously reported 1- to 4-point scale to assess image quality: 1, coronary artery not visible or visible with markedly blurred borders or edges; 2, coronary artery visible with moderately blurred borders or edges; 3, coronary artery visible with mildly blurred borders or edges; and 4, coronary artery visible with sharply defined borders or edges [18]. Tds A Fig. 3 Graphs show diameter (A) and sharpness (B) values of plastic tube obtained at different instances of sinusoidal motion of phantom. A, No significant diameter change was measured in segment that is oriented parallel to motion direction (dashed line). However, diameter changes considerably in segment oriented perpendicular to motion direction (solid line). Arrow identifies diameter value in image obtained using trigger delay found by FREEZE software tool. B, There is no significant change in sharpness in segment oriented parallel to motion direction (dashed line). However, sharpness changes substantially in segment that is oriented perpendicular to motion direction (solid line). Arrow identifies sharpness value in image obtained using trigger delay found by FREEZE. Sharpness (%) Statistical Analysis Quantitative and qualitative results from evaluation of images obtained using visual assessment and FREEZE were compared using the paired Student s t test. A p value of < 0.05 was considered statistically significant. The correlation between grades assigned by two blinded readers was tested using the Pearson s correlation coefficient method. According to that method, for a two-tailed test with a significance level of 0.01 to be statistically important, the r xy value that is, the Pearson s method result must be above This was taken into consideration for the interpretation of reviewer consensus. For statistical comparison of image quality grading, Wilcoxon s rank sum test was used. Results Phantom Experiments Phantom tube diameter measurements in the readout direction of the images acquired at discrete time points in the moving phantom ranged from 1.8 to 2.6 mm (44% overestimation) as shown in Figure 3A. On the static image that was acquired at baseline, 1.8 mm was measured (Fig. 3A) as expected. In the moving phantom, the tube diameter measured on the image that was obtained using Tds was 1.8 mm as well. Orthogonal to the direction of motion, a constant diameter of 2.2 mm (22% overestimation) was found for all time points, the static condition, and Tds. For tube interface definition (i.e., sharpness), the highest values were found for the nonmoving (i.e., static) phantom (72%) and for the software-prescribed trigger delay in the moving phantom (73%). The lowest Static Trigger Delay Value (ms) Tds sharpness value (40%) was found during the highest velocity in temporal coincidence with the maximum overestimation of the tube diameter (Figs. 2B and 3B). Perpendicular to the readout direction, 60% sharpness was found independent of the phantom velocity (Fig. 3B). In Vivo Experiments Using FREEZE, the identification of the period of minimal myocardial motion, including data transfer from and to the scanner was less than 1 minute. The mean vessel length obtained from images acquired with Tds v was 65.2 mm. The mean vessel length measured on the images that were acquired using FREEZE was 83.2 mm (Table 1). The mean vessel sharpness for the proximal coronary arteries was 43.8% using Tds v, whereas it amounted to 46.3% with FREEZE (Table 1). In the same segments, the vessel diameter on the images obtained using Tds v was 2.9 mm compared with 3.0 mm using FREEZE (p < 0.04) (Table 1). The average SNR measured on images acquired at Tds v was 29.5 arbitrary units (AU) in comparison with 32.0 AU obtained from the images acquired using FREEZE. Simultaneously, the average CNR was 21.0 and 23.0 AU (Table 1), respectively. Using Pearson s correlation coefficient, there was certain reviewer consensus for the qualitative grading of the images obtained using visual assessment and FREEZE (r xy = 0.76). The median quality grade of the images obtained with visual assessment and those obtained with FREEZE was 3 on the B W286 AJR:188, March 2007

5 Automated Identification of Minimal Myocardial Motion on MRA TABLE 1: Quantitative In Vivo Analysis of FREEZE Software Tool and Visual Assessment Method Used for Analysis Characteristic Being Evaluated Visual Assessment FREEZE Software Tool p a Vessel length (mm) < 0.01 Mean SD Vessel sharpness for the proximal coronary arteries (%) < 0.02 Mean SD Vessel diameter (mm) < 0.04 Mean SD Signal-to-noise ratio (AU) < 0.02 Mean SD Contrast-to-noise ratio (AU) < 0.04 Mean SD Note AU = artitrary units. Student s t test. scale described earlier, which ranges from 1 to 4. The results of the Wilcoxon s rank sum test suggest that there is a significant difference in image quality grades between images obtained with Tds f and those obtained with Tds v (p < 0.02), with the Tds f images being superior. In Figure 4, three sets of images each set composed of one image obtained using visual identification and the other, FREEZE are compared for image quality, sharpness, and vessel length. In Figure 4B, the left coronary system obtained using Tds f has both higher visual vessel definition and contrast in comparison with the image in Figure 4A that was obtained with Tds v. In another subject shown in Figure 4D, a longer contiguous segment of the RCA is observed with Tds v. Finally, Figures 4E and 4F show one of the cases in which a strong disagreement between Tds and Tds v was found. Figure 5 displays all the Tds v and Tds f values from this study (r = 0.57). The average Tds v and Tds f values were 645 ± 72 and 592 ± 136, respectively (p =0.2). In two cases, an end-diastolic time window was identified visually, whereas FREEZE suggested an end-systolic time point for coronary MRA data acquisition (Fig. 5). Objective and subjective image quality grades improved consistently in these two cases when the FREEZE-prescribed trigger delay was used (Figs. 4E and 4F). Discussion A 3-T scanner offers the potential for enhanced spatial resolution. However, to take full advantage of the higher magnetic field strength, residual myocardial motion must be constrained and other potential challenges must be considered. At 3 T, increased susceptibility artifacts (B 0 inhomogeneity), B 1 inhomogeneity, distorted ECG recordings due to the enhanced magnetohydrodynamic effect, a substantial increase in power deposition, and prolonged T1 values are challenges that need to be overcome for successful coronary MRA. To minimize the adverse effects of B 0 inhomogeneity on image quality, we used local volume shimming in conjunction with segmented k-space gradient-echo imaging rather than SSFP imaging. To remove artifacts originating from B 1 inhomogeneity, we used an adiabatic T2 prepulse for contrast generation between the coronary blood pool and the myocardium [17]. To avoid signal loss due to a prolonged T1, we lowered the radiofrequency excitation angle to 20 for coronary MRA (30 at 1.5 T). Reliable R wave triggering at a higher field strength was ensured using vector ECG technology [15]. In all of our subjects, ECG triggering was reliable and repositioning of electrodes was never necessary. However, once the 3 T specific challenges were addressed, residual myocardial motion had to be further con- strained to support better vessel conspicuity and, ultimately, imaging at a higher spatial resolution. Therefore, FREEZE was developed, and its utility was tested in the present study. In Figure 3A, a variable diameter measurement is a result of different phantom velocities during the acquisition interval. Imaging during increased phantom velocity leads to an overestimation of the diameter measurements of structures that are oriented perpendicular to the direction of motion. However, for structures that are oriented parallel to the direction of motion, no such oscillations are visible, but an overestimation of the diameter measurements occurs. This can be attributed to the anisotropic in-plane spatial resolution of mm. Similar findings can be reported for vessel sharpness, which is adversely affected if imaging is performed during high phantom velocities. Together with the variable diameter measurements as a function of phantom velocity, this emphasizes the strong need for approaches that minimize motion during the signal readout. In vivo, the use of FREEZE led to objective and subjective improvements in image quality. The SNR, CNR, vessel sharpness, and vessel length values for images obtained using FREEZE were superior to those for images obtained using visual assessment of the rest period. Consistent with these findings, visual grading also showed an improvement in image quality when FREEZE was used. A minor discrepancy was also observed between phantom and in vivo studies. Whereas the diameter measurement in the phantom was slightly reduced by the use of FREEZE, FREEZE led to a diameter increase of 0.1 mm in the in vivo study. Although this effect reached only borderline statistical significance, it may be attributed to coronary blood flow that was present in the in vivo but not in the in vitro part of the study. Interestingly, the correlation between Tds v and Tds f was not outstanding, suggesting that visual inspection may not have allowed correct identification of the period of minimal myocardial motion in all cases. We found two examples in which the visually identified trigger delay and the FREEZE-prescribed trigger delay were in strong disagreement. In these cases, end-systolic rather than late diastolic windows of minimal myocardial motion were identified by the software. In both cases, image quality was substantially improved when FREEZE was used. These findings suggests that FREEZE may be more objective than reviewers in identifying the period of minimal myocardial motion. AJR:188, March 2007 W287

6 A C E Fig. 4 Volume-targeted navigator-gated and navigator-corrected double oblique 3D segmented k-space gradient-echo images obtained in three subjects (3 T, TR/TE = 4.3/1.47, α =20, resolution= mm, field of view = mm, matrix, 16 radiofrequency excitations per R-R interval, acquisition time window [T acq ] = 69 milliseconds, bandwidth = 362 Hz/pixel, scan duration = seconds depending on navigator efficiency and heart rate, 10 slices [acquired], 20 slices [reconstructed using zero filling], fat saturation, adiabatic T2 prepulse [TE = 50 milliseconds]). A and B, Axial views of left anterior descending coronary artery in 31-year-old healthy man acquired at trigger delay of 579 milliseconds using visual assessment method (A) and at trigger delay of 675 milliseconds using FREEZE software tool (B). Conspicuity of vessel (arrows) differs between B and A. C and D, Oblique sagittal views of right coronary artery (RCA) in 29-year-old healthy man acquired at trigger delay of 614 milliseconds obtained by visual assessment method (C) and at trigger delay of 561 milliseconds obtained by FREEZE (D). Length of visible vessel (arrows) is increased in image acquired with FREEZE. E and F, Oblique sagittal views of RCA in 33-year-old healthy man acquired at trigger delay of 674 milliseconds obtained by visual assessment method (E) and at trigger delay of 193 milliseconds obtained by FREEZE (F). B D F A 2D segmented k-space gradient-echo cine sequence acquired at a midventricular level was used to determine the onset of the period of minimal myocardial motion using FREEZE. During the cardiac cycle, the displacement and velocity of the different coronary segments are not identical. However, using biplanar angiography, Johnson et al. [11] identified a period of relative quiescence during mid diastole in which proximal, mid, and distal segments of both the left and right coronary arterial systems are at rest simultaneously. This finding suggests that cine images from an axial midventricular level are adequate for determining the optimal trigger delay for coronary MRA. In the present study, we used a fixed acquisition window. However, previous studies have shown that periods of minimal myocardial motion vary among individuals and can be as long as 200 milliseconds if a patient s heart rate is lower than 60 beats per minute. This allows the duration of the acquisition window to be adjusted for abbreviated scanning times [5 7]. Although this was not part of the present study, the FREEZE software could be modified to prescribe not only the onset of the period of minimal myocardial motion, but also its relative duration. Even though an in-depth study about the selection of the ROI for the identification of the period of minimal myocardial motion was not performed, experimental repeated assessments of the period of minimal cardiac motion using FREEZE on the same volunteer resulted in no or very minor variations of the trigger delay. We expect that the use of smaller ROIs may lead to larger variations in the trigger delay. By increasing the magnetic field strength from 1.5 to 3 T, the increased SNR can theoretically be traded for enhanced spatial resolution. However, this necessitates sufficiently constrained residual myocardial motion, as already mentioned. Encouraged by the results obtained with FREEZE, we tried to reduce the voxel size in one subject to mm (Fig. 6) toward the end of our study. Although these preliminary data show this approach is feasible and promising, more experience with that protocol in both adult patients and healthy adult subjects is now needed. Limitations and Future Outlook In the present study, phantom and in vivo measurements were performed at different magnetic field strengths. The moving phantom was originally designed for use at 1.5 T. However, when first operated at 3 T, the W288 AJR:188, March 2007

7 Automated Identification of Minimal Myocardial Motion on MRA FREEZE (ms) Visual Assessment (ms) electric motor was slowed considerably because of additional torque induced by the stronger magnetic field. For these reasons, we decided to continue with the phantom studies at 1.5 T. However, at the same time it was also important to further improve coronary MRA at 3 T. To take full advantage of the higher field strength, residual myocardial motion needs to be further constrained. This goal was supported by the use of the FREEZE software tool and the encouraging results that we obtained with it in the phantom study at 1.5 T. To our knowledge, this study is the first that shows that the automated identification of the period of minimal myocardial motion results in improved image quality for free-breathing Fig. 5 Graph shows trigger delays found using FREEZE and using visual assessment. In two subjects, volume-targeted navigator-gated and navigatorcorrected double oblique 3D segmented k-space gradient-echo imaging was performed (3 T, TR/TE = 4.3/1.47, α = 20, resolution = mm, field of view = mm, matrix, 16 radiofrequency excitations per R-R interval, acquisition time window [T acq ] = 69 milliseconds, bandwidth = 362 Hz/pixel, scan duration = seconds depending on navigator efficiency and heart rate, 10 slices [acquired], 20 slices [reconstructed using zero filling], fat saturation, adiabatic T2 prepulse [TE = 50 milliseconds]). FREEZE found endsystolic trigger delay (arrows), whereas visual inspection led to diastolic acquisition interval in same subjects. Fig. 6 Navigator-gated and navigator-corrected double oblique 3D segmented k- space gradient-echo imaging sequence (3 T, TR/TE = 7.5/2.3, α =20, resolution= mm, field of view= mm, matrix, scan duration = 906 seconds, 12 radiofrequency excitations per R-R interval, acquisition time window [T acq ] = 90 milliseconds, 10 slices [acquired], 20 slices [reconstructed], fat saturation). High-resolution scan of 23-year-old healthy man acquired at trigger delay of 591 milliseconds using FREEZE software tool shows highly visible interface in region of coronary arteries (small-diameter branches [A]) as well as pericardium and lung liver interface (B). Together with ability to reveal small-diameter branching vessels, this suggests excellent suppression of both intrinsic and extrinsic myocardial motion. 3D coronary MRA. However, there are still some challenges that remain to be addressed. Heart rate variability is an important reason for the change of both temporal position and the duration of the period of minimal myocardial motion. FREEZE needs to be applied before coronary MRA to achieve the highest precision in the identification of this rest period. However, FREEZE cannot accommodate real-time changes due to heart rate variability during scanning. Therefore, real-time identification of this rest period is the next step to elaborate. Although we have not studied the effect of using this software tool to evaluate patients with coronary artery disease, the results obtained in a phantom and in healthy adult subjects suggest an improvement in image quality. Improvement in image quality is likely to support improved characterization of coronary artery disease. Conclusion FREEZE, an automated software tool that can be used to identify the period of minimal myocardial motion, improves the accuracy of dimension measurements and objective and subjective image quality in coronary MRA. Simultaneously, operator dependence can be minimized while ease-of-use is improved. Because FREEZE minimizes the adverse effects of intrinsic myocardial motion on image quality, it supports coronary MRA data acquisition with high spatial resolution and at high magnetic field strength. AJR:188, March 2007 W289

8 The FREEZE tool may easily be adapted for use on multiple vendors systems. In addition, this method may have the potential to improve coronary CT angiography as well. References 1. McConnell MV, Khasgiwala VC, Savord BJ, et al. Prospective adaptive navigator correction for breath-hold MR coronary angiography. Magn Reson Med 1997; 37: Danias PG, Hauser TH, Katsimaglis G, Botnar RM, Manning WJ. Coronary magnetic resonance angiography. Herz 2003; 28: Hofman MB, Wickline SA, Lorenz CH. Quantification of in-plane motion of the coronary arteries during the cardiac cycle: implications for acquisition window duration for MR flow quantification. J Magn Reson Imaging 1998; 8: Stainsby JA, Sussman MS, Flexman ML, Wright GA. Real-time magnetic resonance with physiologic monitoring for improved scan localization. Magn Reson Med 2005; 53: Wang Y, Vidan E, Bergman GW. Cardiac motion of coronary arteries: variability in the rest period and implications for coronary MR angiography. Radiology 1999; 213: Plein S, Ridgway JP, Jones TR, Bloomer TN, Sivananthan MU. Coronary artery disease: assessment with a comprehensive MR imaging protocol initial results. Radiology 2002; 225: Kim WY, Stuber M, Kissinger KV, Andersen NT, Manning WJ, Botnar RM. Impact of bulk cardiac motion on right coronary MR angiography and vessel wall imaging. J Magn Reson Imaging 2001; 14: Jahnke C, Paetsch I, Nehrke K, et al. A new approach for rapid assessment of the cardiac rest period for coronary MRA. J Cardiovasc Magn Reson 2005; 7: Wang Y, Watts R, Mitchell I, et al. Coronary MR angiography: selection of acquisition window of minimal cardiac motion with electrocardiography-triggered navigator cardiac motion prescanning initial results. Radiology 2001; 218: Huber ME, Stuber M, Botnar RM, Boesiger P, Manning WJ. Low-cost MR-compatible moving heart phantom. J Cardiovasc Magn Reson 2000; 2: Johnson KR, Patel SJ, Whigham A, Hakim A, Pettigrew RI, Oshinski JN. Three-dimensional, timeresolved motion of the coronary arteries. J Cardiovasc Magn Reson 2004; 6: Plein S, Jones TR, Ridgway JP, Sivananthan MU. Three-dimensional coronary MR angiography performed with subject-specific cardiac acquisition windows and motion-adapted respiratory gating. AJR 2003; 180: Etienne A, Botnar RM, Van Muiswinkel AM, Boesiger P, Manning WJ, Stuber M. Soap-Bubble visualization and quantitative analysis of 3D coronary magnetic resonance angiograms. Magn Reson Med 2002; 48: Deriche R. Fast algorithms for low-level vision. IEEE Trans Pattern Anal Machine Intel 1990; 12: Fischer SE, Wickline SA, Lorenz CH. Novel realtime R-wave detection algorithm based on the vectorcardiogram for accurate gated magnetic resonance acquisitions. Magn Reson Med 1999; 42: Stuber M, Botnar RM, Danias PG, et al. Double-oblique free-breathing high resolution three-dimensional coronary magnetic resonance angiography. J Am Coll Cardiol 1999; 34: Nezafat R, Stuber M, Ouwerkerk R, Gharib AM, Desai MY, Pettigrew RI. B1-insensitive T2 preparation for improved coronary magnetic resonance angiography at 3 T. Magn Reson Med 2006; 55: McConnell MV, Khasgiwala VC, Savord BJ, et al. Comparison of respiratory suppression methods and navigator locations for MR coronary angiography. AJR 1997; 168: W290 AJR:188, March 2007