Axial Black Blood Turbo Spin Echo Imaging of the Right Ventricle

Abstract

Black blood turbo spin echo (TSE) imaging of the right ventricle (RV) free wall is highly sensitive to cardiac motion, frequently resulting in non-diagnostic images. Temporal and spatial parameters of a black blood TSE pulse sequence were evaluated for visualization of the RV free wall. 74 patient studies were retrospectively evaluated for the effects of acquisition timing on image quality. Axial black blood TSE images were acquired on 10 healthy volunteers to assess the role of spatial misregistration on right ventricle visualization; increasing the double inversion recovery (DIR) slice thickness beyond 300% had no effect on image quality (p=0.2). 35 patient studies were prospectively evaluated with inversion times (TIs) corresponding to the mid-diastolic rest period and end-systole based on visual analysis of a four chamber cine. When TIs were chosen to be within the patients’ RV rest period, mean image quality score was significantly improved (2.3 vs. 1.86, p<0.001) and the number of clinically diagnostic images increased from 32% to 46%. Black blood TSE imaging of the RV free wall is highly sensitive to cardiac motion. Image quality can be improved by choosing TIs concordant with the rest period of the patient’s RV that may occur at mid-diastole or end-systole.

Introduction

Black blood turbo spin echo (TSE) sequences are an essential component of the cardiac MRI pulse sequence repertoire and are commonly used for tissue characterization.(1-4) TSE sequences are preferred over faster gradient echo techniques for anatomic assessment of the myocardium because of higher SNR with good T1 and T2 weighting in the absence of chemical shift artifact.(5) Black blood preparation pulses suppress the signal from flowing blood without affecting the contrast in stationary tissues. These two properties give black blood TSE imaging an important role in visualizing the morphology of the myocardium, which is altered in conditions such as in arrhythmogenic right ventricular dysplasia (ARVD).(1,6) As part of a national reading center for the United States ARVD multi-center study, we have observed non-diagnostic TSE images of the right ventricle (RV) myocardium in 20% of patients with black blood TSE images, and suboptimal quality in 60 % (n=137 patients). The TSE black blood sequence was originally optimized for left ventricle myocardium; the RV wall may be completely absent or blurred when the left ventricle is identified with a high level of sub structural detail. These significant variations in image characteristics have led to substantial errors in the interpretation of MRI right ventricular studies. This has been documented in the literature.(7)

Physiologic differences between the ventricles make black blood TSE imaging of the right ventricular myocardium more challenging than imaging the left ventricle. There is less signal from the right ventricle because its wall is much thinner than that of the left ventricle. The right ventricle contracts more along its long axis during systole compared to the left ventricle; both have similar torsional motion.(8) The rest period of the right coronary artery (123±60ms) is shorter than that of the left (162±75ms) (9,10), and the mean velocity of right coronary artery motion (61±7mm/s) is twice that of the left (32±7mm/s), paralleling RV wall motion characteristics.(11) Blood velocities are slower in the right heart compared to the left heart for much of the cardiac cycle.(12) Prior studies have demonstrated the effects of motion on right coronary imaging. Kim, et al showed that black blood imaging of the right coronary wall was only possible when acquisition occurred during patient specific diastolic rest periods.(13)

The purpose of this study was to determine appropriate MRI pulse sequence parameters for TSE black blood imaging of the right heart and prospectively evaluate their impact on diagnostic image quality in healthy volunteers and hospitalized patients.

Theory

Turbo spin echo pulse sequences have inherent blood nulling properties, even without black blood preparation pulses.(5,14) In order to contribute signal to the image, a given spin must subjected to a 90° excitation and 180° refocusing pulse. Mobile spins such as blood or moving myocardium, may move out of plane in between these two slice selective pulses and therefore will not experience the 180° refocusing pulse and generate a spin-echo. This “washout effect” leads fast moving spins to appear black in the final image. The situation is further complicated in a turbo spin echo sequence because a train of refocusing pulses is used. The k-space lines at the end of a TSE acquisition window are more sensitive to motion since more time elapses between excitation and refocusing. Flow related dephasing, intravoxel dephasing, and turbulent flow also contribute to signal loss of moving spins.(5) Despite the inherent blood suppression of TSE sequences, black blood preparation pulses are used to ensure complete and homogenous blood nulling, especially for slow moving or in-plane flow.

The relevant features of the black blood TSE pulse sequence are shown in Figure 1 and explained in detail elsewhere.(14) The first inversion recovery pulse, triggered by the R-wave, is non-slice selective, while the second inversion pulse is a slice-selective reinversion of the imaging slice of interest. Together they are known as a double inversion recovery (DIR) pulse. The DIR pulse occurs immediately after the R-wave, ensuring that the DIR pulses are delivered before systole, when the heart is ideally in the same position as it will be during readout. The thickness of the slice-selective excitation is a modifiable parameter. The reinverted slice thickness, also known as DIR slice thickness, can be increased compared to the final imaged slice volume, allowing for some spatial slice misregistration due to through-plane motion. DIR slice thickness is expressed as a percentage of the imaging slice thickness. As an additional method of mitigating misregistration errors, Keegan et al. applied a spatial offset to the reinversion pulse based on tracking of cardiac motion on a perpendicular slice. This method was tested on short-axis images of the heart, accounting for motion along the long axis.(15)

Figure 1. Black Blood Pulse Sequence.

a) Schematic representation of the volume of the right ventricle, with the approximate rest periods noted. b) Schematic ECG tracing. c) RF pulses of black blood TSE sequence. Underlined pulses are slice-selective. Tall lines = 180°, short lines = 90°, ETL = Echo Train Length, TI = Inversion Time, TE_eff = effective echo time, Acq Win = acquisition window. Note that all sequences in this study had a trigger delay of zero. d) Schematic of longitudinal magnetization (M_z) of stationary tissues within the imaged slice (myocardium, fat) and blood which moves into the imaged slice after the DIR pulse.

The choice of inversion time (TI) in a black blood TSE sequence has multiple implications for image quality. When the short-time inversion-recovery (STIR) sequence was first adapted for black blood imaging primarily for the left ventricle, Simonetti et al. advised that TI should be determined by the time required for blood to be nulled.(14) The graph of M_z in Figure 1d schematically shows the ideal choice of TI when the longitudinal magnetization of inverted blood crosses zero. Fleckenstein et al. reported the direct relationship between TR and optimum TI:

$$T{I}_{\mathrm{optimum}}=T_1{}_{\mathrm{blood}}{}^{\ast }(\ln 2-\ln (1+\exp (-TR∕T_1{}_{blood})))$$ [1]

Since the TR used in an ECG triggered sequence, 1 or 2 RR intervals, is shorter than the time required for the blood’s longitudinal magnetization to fully relax, the longitudinal magnetization available at the beginning of the next RR interval is decreased. When this reduced longitudinal magnetization is again inverted, the null point will be reached sooner.(16)

When imaging relatively stationary anatomy, such as the vasculature, sole consideration of the ideal TI will optimize image quality. However, when imaging the myocardium, consideration of cardiac motion is paramount. The ventricular volume trace in Figure 1a shows a schematic representation of myocardial motion throughout the cardiac cycle. Traditionally, positioning the black blood TSE readout within diastole was thought to be sufficient. Indeed, this is usually adequate for imaging the left ventricle. However, the normal RV longitudinal shortening at the cardiac base is up to 2cm (17), so additional consideration of wall motion is necessary. Two periods of quiescent cardiac motion are typically present: 1) at end-systole, and 2) at mid-diastole. Recent studies have shown that 26% of patients have a longer rest period at the end-systole, rather than in mid-diastole.(18) Tissue Doppler echocardiography has confirmed that the myocardium has a period of zero velocity between the end of systole and beginning of ventricular filling.(18) Timing of the rest period is complicated further in tachycardic patients as the rest period decreases rapidly with an increase in HR.(10)

Methods

Institutional Review Board approval was given for evaluation of MRI examinations of patients referred for evaluation of the right ventricle. Volunteer studies were also approved by our Institutional Review Board. All studies were conducted in a 1.5-T whole-body MR system (Magnetom Avanto; Siemens Medical Solutions, Erlangen, Germany) that was equipped with a high performance gradient system (max amplitude of 40 mT/m; max slew rate of 200 T/m/s). For signal reception, the subjects lay on top of a built-in-table spine phased-array coil (6 elements active) and had a 6 element phased array coil placed around the thorax. All TSE sequences in the retrospective analysis, volunteer studies, and prospective acquisitions utilized a scrolled linear-segmented k-space ordering scheme. All sequences used TEs positioned closer to the beginning of the acquisition window (TEs in the 40-50ms range) which resulted in a scrolling of the traditional linear segmented ordering scheme. For example, instead of the first echo occurring at ky_min and the last echo occurring at ky_max, the first echo would occur at (ky_min+1), the second to last echo would occur at ky_max, and the last echo would occur at ky_min.

Retrospective Analysis

A retrospective analysis of patient studies was performed in order to determine the relationship between imaging and patient parameters for delineation of RV structures. Between June 2005 and July 2006, 98 patients were evaluated with cardiac MRI for suspected abnormality of the right ventricle. A four chamber cine sequence (steady state free precession (SSFP), TR=3.5ms, TE=1.3ms, field of view (FOV)=360x270mm, matrix=256x192, slice thickness=8mm, flip angle=68°) was used to assess the cardiac rest period in each patient.

The rest period of the right ventricle was determined by measuring the motion of the fat within the right atrial-ventricular groove. The India-ink artifact (chemical shift artifact of the second kind(5)) between atrial-ventricular groove fat and myocardium served as a natural tag of right ventricle motion. The rest period was considered the time during which minimal motion was visually observed relative to a stationary cursor. After assessing the limits of observer measurement, this technique corresponded to approximately 1 mm or less of movement between frames. In cases where the patients’ RV had end-systolic and mid-diastolic rest periods, the longest period was recorded. The timing of the rest period relative to the R-wave trigger was noted from the trigger time field within the respective DICOM images.

The 98 axial TSE black blood images were acquired using a range of sequence parameters. In order eliminate the effects of variable TE and acquisition window length, 74 images were selected with a more limited range of parameters (TE=43-47ms, ETL=17-19, echo spacing=6.1ms, acquisition window=103.7-115.9ms, slice thickness=5mm, FOV=220-380x180-285mm, matrix=256-384x168-312). The type of ECG triggering (every R-wave or every-other R-wave) used was stored within the DICOM header. Using the RR interval recorded in the cine acquisition, and the triggering information from the image, the theoretical optimum TI time was calculated for each image using Eq. [1].

For each study, the trigger time for a mid-ventricular black blood image was recorded. The trigger time was the time after the R-wave trigger when the central line of k-space was acquired. The duration of the acquisition window was calculated by multiplying the echo train length (ETL) by the echo spacing. Using trigger time, TE, and the acquisition duration, the beginning and end of the acquisition window were calculated with respect to the R-wave trigger. The timing of the acquisition window was compared to the timing of the rest period to determine the percentage of the acquisition window that was within each patient’s longest rest period.

Volunteer Studies – TI based on nulling, variable DIR thickness

Examinations were conducted on 10 healthy volunteers to assess the effects of variations in TI and DIR slice thickness on visualization of the right ventricle. After a 4 chamber cine was acquired (SSFP, TR=3ms, TE=1.3ms, FOV=360x270mm, matrix=256x192, slice thickness=8mm, flip angle=74°), the rest period was determined by visual assessment of AV groove motion.

The theoretical optimum TI was calculated based on Eq. [1] using the subject’s mean heart rate. An axial black blood TSE image in the mid-ventricular plane was acquired using this TI. The initial parameters of the TSE sequence were TE=46ms, ETL=17, echo spacing=6.1ms, acquisition window=103.7ms, slice thickness=5mm, FOV=300x225mm, matrix=256x192, DIR thickness=300%, 180° refocusing thickness=127%, triggering every-other RR. After acquiring an image with the theoretically optimum TI, the TI was varied in 50 ms increments above and below this level. The image that best demonstrated the right ventricle wall was chosen by operator (SB) inspection, and this TI was used for subsequent acquisitions investigating the DIR slice thickness. Next, the slice thickness of the slice-selective double inversion pulse was increased to 400%, 500%, and 1000%. Refocusing pulse thickness was increased in a similar manner starting with an initial value of 127%. However, altering refocusing pulse thickness showed no effect, and further adjustments were discontinued to keep total scan duration at an acceptable length for the volunteers.

Prospective Patient Studies – TI optimized for rest period, constant DIR thickness

Between August and December 2006, 35 patients were assessed using the optimized black blood TSE protocol for the right ventricle. A four chamber cine sequence (SSFP, TR=3.5ms, TE=1.3ms, FOV=360x270mm, matrix=256x192, slice thickness=8mm, flip angle=68°) was used to assess the rest period in each patient.

A trained MRI technologist determined the rest period of each patient by observing the motion of the fat within the right atrial-ventricular groove on the four-chamber cine. The technologists also approximated the time of end-systole by observing the point at which the left ventricular myocardium was thickest. Four axial black blood TSE images in the mid-ventricular plane were obtained. Images with two different TI times were acquired: one TI was chosen near the beginning of the calculated mid-diastolic rest period, and the second TI corresponded with the end of systole. Although Eq. [1] was not directly factored into TI choice, the difference between TIs in the diastolic or end systolic rest periods and the theoretical TIs for optimum blood nulling was minimized by exploiting TI dependence on TR; end systolic images were obtained with triggering off every R wave and mid-diastolic images were triggered off every other R wave. Two black blood TSE sequences, one with each TI, were acquired without parallel imaging (TE=55ms, ETL=25, echo spacing=6.1ms, acquisition window=152.5ms, slice thickness=5mm, FOV=300x225mm, matrix=512x384, DIR thickness=300%, 180° refocusing thickness=127%). Two more images, one with each TI, were acquired with a sequence that used parallel imaging (GRAPPA, acceleration factor=2) to reduce the ETL while keeping resolution constant. (TE=43ms, ETL=17, echo spacing=6.1ms, acquisition window=103.7ms, slice thickness=5mm, FOV=300x225mm, matrix=512x384, DIR thickness=300%, 180° refocusing thickness=127%,). DIR thickness of 300% and 180° refocusing thickness of 127% were used because they adequately prevented slice misregistration in the volunteer studies.

Image Quality Analysis

The quality of myocardium visualization in each black blood TSE image was scored independently by three cardiac MRI observers (SB, RM, AM). In cases of discrepancy, the three reviewers met to reach a consensus score which was used in all further analysis. The grading system emphasized the appearance of the right ventricle wall, with a secondary emphasis on blood nulling in the RV cavity. Images were scored on a scale of 1 to 3 using the following criteria. Scores of 1 were given to ‘poor’ images when the RV myocardium was not visible and blood signal may have been present in RV cavity. These images were not clinically diagnostic. Scores of 2 had ‘fair’ quality in which the RV myocardium was visible but blurry, the edges were not delineated well, and blood signal may have been present in RV cavity. These images were not clinically diagnostic. Scores of 3 were assigned to ‘good’ quality images that showed significant contrast at the RV myocardium-blood interface and myocardium-epicardial fat interfaces. ‘Good’ images also had low blood signal in RV cavity. Scores of 3 signified clinically diagnostic images. Representative images from each score grade are shown in Figure 2.

Figure 2. Representative images from each score grade.

RV free wall myocardium indicated by arrows. a) Grade 1: RV myocardium not visible. Blood signal may be present in RV cavity. b) Grade 2: RV myocardium visible but blurry. Edges not delineated well. Blood signal may be present in RV cavity. c) Grade 3: Significant contrast between RV myocardium and blood and myocardium and epicardial fat. Low blood signal in RV cavity.

Statistical analysis

The retrospective image scores were sorted into quartiles based on the percentage of the acquisition window that fell within the rest period. A two-way nonparametric analysis of variance (ANOVA) was performed with SPSS software (SPSS Inc, Chicago, IL) to assess whether there were significant differences in image quality between timing groups. Wilcoxon rank sum tests were used to further quantify the differences.

The trigger times of the retrospective images were also compared to the ideal TI time calculated using Eq. [1]. The differences between ideal and actual TI were divided into groups in an attempt to evenly distribute the differences: actual TI before ideal TI, actual TI up to 150ms after ideal, actual TI 150-300ms after ideal and actual TI more than 300ms after ideal. ANOVA and Wilcoxon analyses were performed on the groups.

Images from healthy volunteers were grouped based on DIR slice thickness. Image scores were compared using a paired Wilcoxon signed-rank test. Prospectively acquired images on patients were grouped based on TI choice and use of parallel imaging. Image scores were compared using a paired Wilcoxon signed-rank test. Summary statistics are presented as means ± standard deviation.

Results

Retrospective analysis

The 74 studies examined retrospectively represented a range of physiologic and scanner dependent parameters. 80% (n=58) of patients had their longest rest period in diastole. The mean beginning, end, and duration of the diastolic rest periods were 649± 88 ms, 796 ± 131 ms, and 146 ± 90 ms. 20% (n = 15) of patients had their longest rest period at end-systole. The mean beginning, end, and duration of the systolic rest periods were 299 ± 40 ms, 377 ± 43 ms, and 78±33 ms. Mean acquisition window was 105±19 ms. Trigger time varied from 245 to 938 with a mean of 597±171 ms. After calculating the percentage of the acquisition window within the rest period for each image, the images were divided into quartiles.

The consensus image quality score is shown in Figure 3 for each group. Two-way ANOVA showed a statistical difference between these groups (p<0.001). The highest image quality score occurred for images in the top quartile (76-100% of the acquisition within the RV rest period). Wilcoxon rank sum tests showed significant differences between the 2^nd and 4^th quartiles (1.62±0.74 vs 2.65±0.59, p=0.002).

Figure 3. Image Quality vs. Percentage of Acquisition Window in Rest Period.

All retrospective studies were sorted into quartiles based on how much of the acquisition window was within the rest period of the right ventricle, as estimated from the four chamber cine. Average image quality scores (1-3) are shown here. Significance noted from Wilcoxon rank rum test.

A secondary analysis was undertaken to examine the impact of k-space ordering. Images were divided into groups based on which regions of k-space were acquired during the rest period. ANOVA showed a statistical difference between these groups (p<0.001). The results, Figure 4, show that the best images (2.67±0.62) were acquired when the acquisition window was fully in the rest period. A significant reduction in quality was seen in acquisitions where only the periphery of k-space was included in the rest-period (2.0±0.84 vs 2.67±0.62, p=0.02).

Figure 4. Image Quality vs. Position of Acquisition Window.

Images were grouped based on which parts of the acquisition window were acquired during the rest period: bRP = before Rest Period; aRP = after Rest Period; 90 = initial 90° RF pulse; LF = Low Frequency phase-encode lines [½ ky_min to ½ ky_max] occur within rest period; HF = High Frequency phase-encode lines [ky_min to ½ ky_min, ½ ky_max to ky_max] occur within rest period; pLF/pHF = partial overlap of low/high frequency phase-encode lines occur within rest period. Average image quality scores (1-3) are shown here. Significance noted from Wilcoxon rank rum test.

Figure 5 shows the mean score for images grouped by the difference between ideal TI and actual trigger time. Images with the smallest difference between the ideal TI and the actual trigger time did not have the highest image quality scores. ANOVA showed a statistically significant difference between groups (p<0.001). Wilcoxon rank sum tests showed significant differences between the second group, TI before <= 150 ms after Ideal TI, and the last group, TI more than 300ms after ideal TI (1.71±0.69 vs. 2.5±0.80, p<0.001).

Figure 5. Image Quality vs. Relationship between the Ideal and Actual Inversion Time.

Retrospective studies were sorted into quartiles based on how the actual inversion time compared to the ideal TI calculated from the Eq. [1]. The average image score (1-3) for each quartile is plotted here. Significance noted from Wilcoxon rank sum test.

Volunteer Studies – TI based on nulling, variable DIR thickness

The scanning of healthy volunteers (HR 47 – 74 bpm) showed that blood nulling could be achieved at a wide range of TIs away from the optimum TI calculated by Eq. [1]. In an attempt to improve RV visualization, TIs were advanced further into diastole, average TI = 640 ± 60 ms, beyond the ideal calculated optimum TI, average 499 ± 112 ms. TIs were not purposely optimized for rest period concordance in this portion of the study. Despite adjustment for RV visualization, image quality was still less than 3 (average score 2.2) in 6 of the 10 volunteers after TI adjustment. We expected that increasing the DIR slice thickness would mitigate any slice misregistration artifacts at the expense of reduced blood nulling. However, image scores for DIR slice thicknesses of 400%, 500%, and 1000%, showed no significant change in image quality 2.3, 1.8, 2.0 (p=0.2). Higher blood signal was only observed at slice thicknesses in excess of 500%.

Prospective Patient Studies – TI optimized for rest period, constant DIR thickness

Patient results after optimization for the black blood TSE sequence are shown in Table 1. Although all 4 images were acquired for each patient, the data in Table 1 were only calculated for those patients whose longest rest period occurred during the imaged period, diastole or end systole respectively. By using parallel imaging, the percentage of the acquisition window contained wholly in the rest period was significantly increased for the diastolic and systolic images. However, parallel imaging actually decreased image quality in diastolic images and had no significant change among systolic images (diastolic, p=0.006; systolic, p=0.2). For each patient, the maximum image quality score out of the four acquisitions was recorded. An image acquired during the end systolic rest period had a higher image quality than a diastolic image in 5/35 (14%) patients. The best score for each patient was used to create Figure 7b, representing a clinical situation in which the technologist would choose the optimal TI using the patient’s longest rest period. When the best image was chosen, 46% of patients scanned prospectively had ‘good,’ clinically diagnostic images. This is a 44% improvement in the success rate of the pulse sequence since only 32% of the studies analyzed retrospectively had ‘good’ images. Likewise, the percentage of ‘poor’ images was decreased from 35% in the retrospective study to only 17% of the patients scanned prospectively. The average image score in the prospective patient study was 2.3 versus 1.97 in the retrospective analysis (P<0.001). Among acquisitions where > 75% of the acquisition window was in the rest period, the average image score further increased to 2.67.

Table 1.

Results of Prospective Patient Study

Image Timing	Diastole (n=26)	Diastole Parallel	Systole (n=9)	Systole Parallel
Mean TI	748	748	343	332
RR Trigger	2	2	1	1
ETL	25	17	25	17
Mean RP duration	160		59
Mean Score	2.19	1.97	1.67	1.78
Rest period%	47	56	40	53

Four axial black blood images were prospectively acquired for 35 patients. 26 patients had longer diastolic rest periods, 9 had longer systolic rest periods. The rest period % is the percentage of the image acquisition window that overlaps the RV rest period, averaged for all images in each group.

Figure 7. Quality of RV Visualization in Patient Studies.

Distribution of studies in each quality category from: a) the retrospective review and b) the prospective study. The data in chart b) differ from Table 1 in that the best image for each patient was used, regardless of whether it was a systolic or diastolic image; thus simulating a clinical scenario where the technician would interactively choose the optimal timing for a given patient.

Discussion

Imaging the morphology of the right ventricle myocardium is a challenging task, dependent on multiple patient specific and machine dependent factors. A prior study evaluated factors related to spatial resolution and TSE blurring in ex vivo RV tissue, (1) but non-diagnostic patient studies for the RV remain common in clinical practice. We have shown that timing of the acquisition window relative to the rest period of the right ventricle, a standard practice in coronary wall imaging(13), is a major determinant of image quality when imaging the right ventricle myocardium. The right heart is much more mobile than the left, with peak velocities up to twice as fast.(11) The RV free wall has the greatest total longitudinal displacement of any structure in either ventricle.(8) The thin, rapidly moving right ventricle free wall is a difficult target for the TSE sequence, which is inherently sensitive to motion. Long echo trains are necessary to acquire high resolution images within a single breath hold, and these will exaggerate the “washout effect” for fast moving myocardium, as mentioned earlier in the theory section. In the axial plane, RV motion is a complex combination of through and in-plane motion which may exaggerate motion sensitivity.

Although detrimental to myocardial visualization, the inherent motion sensitivity of TSE sequences has implications for choosing the inversion time. In practice, we observed that blood nulling was achieved over a wide range of TIs that were not necessarily concordant with the ideal TI time based on the formula by Fleckenstein et al. (Eq. [1]). This discrepancy is likely due to the inherent blood nulling of TSE sequences. This principle is evident when a TSE image is acquired without a black blood pre-pulse. Although there is signal from slow moving blood near the endocardial border of the blood pool, the blood signal is absent in the center of the RV cavity. This effect may have been maximized by our choice of mid-ventricle axial slices.

Slice misregistration has also been thought to contribute to the difficulty of imaging the mobile right ventricle. We initially thought that the complete absence of the right ventricle in some images (Figure 2a) was due to slice misregistration. In theory, increasing the DIR slice thickness would reduce the amount of misregistration. However, in experiments on healthy volunteers where TI was chosen primarily on the optimum for blood nulling and secondarily adjusted for right ventricle visualization, we were unable to further improve right ventricle visualization by increasing the width of the DIR reinversion slice. The detrimental effect of slice misregistration may be greater when imaging in the short-axis plane because of pronounced long-axis shortening.(19) We also noted that increasing the DIR slice thickness did not lead to a substantial increase in blood intensity. Slightly more blood artifact was seen in the LV, which was not the focus of this study. Since we used very thin slices, 5mm, the total DIR slice thickness was only 25mm at 500%. Our use of mid-ventricle axial slices may have helped maximize the inflow of fresh blood into the imaging slice, allowing the use of thicker reinversion slices with minimal blood artifact.

A prospective study was performed with individual patient optimization of inversion times based on assessments of the rest period from a four chamber cine acquisition. This intervention was successful in reducing the number of poor images acquired and increasing the number of fair and good images. Although the number of clinically diagnostic (‘good’) images was improved, just fewer than 50% of studies met this criterion. Since these data were acquired from clinical studies, many factors may contribute to this low percentage. As noted in table 1, the overlap of the acquisition window with each rest period was lower than ideal. Although the average rest period was in the prospective study was 139±108 ms, 40% of the patients (n=14) had rest periods less than 60ms. There were some inter-observer differences in rest period estimation between the time of image acquisitions and post-hoc analysis. The patient population studied was referred primarily for the evaluation of RV arrhythmia; inconsistent RR intervals during the MRI acquisition are a major cause of motion blur on the resulting images; the mean standard deviation of the RR interval during the four chamber cine was 45±20ms in volunteers and 61±92ms in the patients. Reduced breath hold capability in patient populations also adversely affect image quality compared to volunteer studies.

In the prospective patient studies reducing the overall length of the acquisition window by using parallel imaging did not affect image quality. This observation is at odds with our retrospective data, as shown in figure 3, indicating that increasing the overlap of the acquisition window in the rest period increases image quality. Limitations of the present study may explain this result. The shorter sequence with parallel imaging, ETL 17, still had an acquisition window of 103.7ms which was still longer the RV rest period is some patients. This ETL was necessary to acquire the high spatial resolutions required to visualize details of the RV morphology.(1) This study was not designed to dynamically change acquisition windows to fit the length of individual patients’ rest periods; fixed ETLs were used for all patients. Some parallel images suffered from poor quality due to the inherently lower SNR of parallel acquisitions. Image quality was slightly lower in the images taken during mid-diastole with parallel imaging. The effect of lower SNR may have been accentuated by our use of thin, 5mm, slice thicknesses. Another consideration is that not all echoes in an echo train equally contribute to image quality. Analysis of the retrospective image cohort showed that the central lines of k-space are major determinants of image quality, with less impact from the periphery. Although one might expect the initial 90° RF pulse to be important for quality, our data does not show a significant difference between images with the central region of k-space partially in the rest period with and without the 90° RF pulse occurring in the rest period.

Manual assessment of the rest period may be an insufficient means to accurately assess cardiac motion. New techniques are now available for on-line automated assessment of the rest period. These new techniques also eliminate inter-observer differences in rest period estimation, allowing uniformity between technologists at the imaging console and researchers performing post-hoc analyses. It is also possible that, in our studies, the rest period changed in between the 4 chamber cine acquisition and the black blood TSE image acquisition. Since repeating a 4 chamber cine sequence before each TSE image is impractical in a clinical protocol, a faster means of assessing the rest period may be required.

Conclusions

The data presented here indicate that timing of the rest period should be a major consideration when imaging the right ventricle using a black blood TSE sequence. The traditional method of setting the TI to be concordant with the Fleckenstein et al. optimum TI may happen to agree with the optimal timing of the pulse sequence based on rest period considerations. In other cases, this concordance will not occur and the rest period may be located later in diastole or at the end of systole. Although there are numerous challenges in imaging the RV with black blood TSE sequences, optimization of acquisition window timing on a per-patient basis is likely to aid improvement of RV visualization in a typical patient referral population.

Figure 6. Effect of TI Choice on RV Wall Visualization.

Two identically positioned axial TSE images from a single patient acquired with two different inversion times: a) TI = 545, poor quality. b) TI = 685, good quality. The patient’s RV rest period was measured to be from 642-764ms. The patient’s RR interval was 974ms, yielding a theoretically optimum TI of 616ms using every-other RR triggering.