Pediatric Radiology

, Volume 42, Issue 4, pp 431–439

Optimization of myocardial nulling in pediatric cardiac MRI

Authors

    • Department of Pediatrics, Division of Pediatric CardiologyStollery Children’s Hospital, University of Alberta
  • Ryan W. Hung
    • Pediatric Radiology, Stollery Children’s HospitalUniversity of Alberta
  • Kimberley A. Myers
    • Alberta Children’s Hospital
  • Cinzia Crawley
    • Pediatric Radiology, Stollery Children’s HospitalUniversity of Alberta
  • Michelle L. Noga
    • Pediatric Radiology, Stollery Children’s HospitalUniversity of Alberta
Original Article

DOI: 10.1007/s00247-011-2276-z

Cite this article as:
Tham, E.B., Hung, R.W., Myers, K.A. et al. Pediatr Radiol (2012) 42: 431. doi:10.1007/s00247-011-2276-z

Abstract

Background

Current protocols to determine optimal nulling time in late enhancement imaging using adult techniques may not apply to children.

Objective

To determine the optimal nulling time in anesthetised children, with the hypothesis that this occurs earlier than in adults.

Materials and methods

Sedated cardiac MRI was performed in 12 children (median age: 12 months, range: 1–60 months). After gadolinium administration, scout images at 2, 3, 4 and 10 min and phase sensitive inversion recovery (PSIR) images from 5 to 10 min were obtained. Signal-to-noise ratio (SNR) and inversion time (TI) were determined. Quality of nulling was assessed according to a grading score by three observers. Data was analysed using linear regression, Kruskal-Wallis and quadratic-weighted kappa statistics.

Results

One child with a cardiomyopathy had late enhancement. Good agreement in nulling occurred for scout images at 2 (κ = 0.69) and 3 (κ = 0.66) min and moderate agreement at 4 min (κ = 0.57). Agreement of PSIR images was moderate at 7 min (κ = 0.44) and poor-fair at other times. There were significant correlations between TI and scout time (r = 0.61, P < 0.0001), and SNR and kappa (r = 0.22, P = 0.017).

Conclusion

Scout images at 2–4 min can be used to determine the TI with little variability. Image quality for PSIR images was highest at 7 min and SNR optimal at 7–9 min. TI increases with time and should be adjusted frequently during imaging. Thus, nulling times in children differ from nulling times in adults when using standard adult techniques.

Keywords

Cardiac MRICongenital heart diseasePediatric cardiologyLate enhancement imaging

Introduction

Late gadolinium enhancement (LGE) imaging is a cardiac MRI technique used to detect myocardial fibrosis in both ischemic and nonischemic heart disease. Gadolinium (III) ion is the paramagnetic agent used that accumulates in damaged myocardial tissue and is detected as high T1 signal areas on LGE imaging [1]. In children with largely nonischemic myocardial disease (e.g., myocarditis or cardiomyopathy), it allows differentiation of the diagnoses based on the pattern of enhancement [24]. The ability to distinguish these patterns requires accurate and reliable performance of late enhancement imaging in children, which in turn depends on adequate nulling of normal myocardium [3].

During imaging, the inversion time (TI) is set to null signal from viable myocardium, which then appears black, thus emphasizing signal from necrotic tissue, which appears bright due to its gadolinium-induced shorter T1 relaxation time. The optimal TI is based on the dose of contrast administered and the elapsed time between injection and imaging. Prediction of optimal TI is made at the time of scanning, either empirically or by using scout imaging. A suboptimal T1 can reduce contrast or nullify the contrast between normal and damaged myocardium, leading to the underestimation of the extent of fibrosis [5]. Magnitude reconstructed images are very dependent on optimal TI for optimal contrast [6]. The optimal TI varies among individuals and must be set empirically for each patient. Phase sensitive reconstruction (PSIR) is a robust imaging technique that provides consistent image quality independent of TI over a range of typically used TIs [7]. This technique restores the signal polarity of the LGE image by using phase information provided by a background reference image [6]. The reference image, used to estimate background phase and surface coil field maps, is obtained at the same cardiac phase, during the same breath-hold acquisition as the LGE image [6]. The resultant image has decreased background noise and improved contrast‐to‐noise ratio. As a result, PSIR is a more forgiving technique than standard LGE magnitude reconstructions and is therefore ideal in children, where LGE nulling is often difficult. By removing the variable of TI, the use of PSIR allows the study of optimal myocardial nulling to be based on time after contrast administration, rather than a combination of the two variables.

With the widespread use of late enhancement imaging, protocols vary as to the contrast dose, inversion time and timing of acquisition [8]. Furthermore, studies on the technique and use of LGE imaging are sparse in the pediatric literature, resulting in the adoption of adult techniques of waiting 10–15 min after intravenous injection of gadolinium [9]. Studies adopting adult techniques are mainly reported in older, unsedated children [4, 10, 11] with very few reports on infants [12]. The pharmacokinetics of drugs in children are different from those in adults. Infants have larger extracellular water composition and lower fat content compared to adults resulting in a larger volume of distribution for hydrophilic drugs such as gadolinium [13]. Furthermore, children tend to have relatively faster heart rates and cardiac outputs than adults. From our experience, the adult protocol of waiting approximately 10 min after the administration of gadolinium is often too long, with adequate nulling being difficult to achieve consistently using adult protocols. Thus, there is a need to standardize and optimize LGE protocols in young children.

The aim of this study was to determine the optimal timing to perform late enhancement in anesthetised children. Our hypothesis was that the timing for optimal nulling is earlier in children than adults.

Material and methods

Patients

We performed a retrospective study of patients younger than 5 years of age undergoing sedated cardiac MRI. This consisted of 12 patients (4 boys), median age 12.5 months (range 1–60 months) with median body surface area 0.4 m2 (range 0.21–0.62 m2). Only children who underwent late enhancement imaging were included in the study. The study was approved by the hospital’s research ethics board. Studies were excluded if no contrast was administered.

Cardiac MRI protocol

Cardiac MRI examination (Fig. 1) was performed on a 1.5 T Siemens Sonata scanner (Erlangen, Germany) using a head coil for infants <4 kg, a knee coil for infants 4–10 kg, and a four-channel phase array coil for children >10 kg. All patients were intubated and ventilated under general anesthesia using variable combinations of ketamine, fentanyl and rocuronium for induction and sevoflurane during the procedure with ECG-gating. Standard cardiac protocol was applied to evaluate the anatomy and ventricular function. After hand injection of 0.2 mmol/kg of gadopentetate dimeglumine (Magnevist; Bayer, Inc., Toronto, Canada), contrast-enhanced MRA was performed in children with congenital heart diseases, where time 0 was the time that gadolinium was administered. LGE imaging was performed in all patients with free breathing technique. A segmented inversion recovery cine true FISP pulse sequence (scout) was used to determine the TI at the mid-ventricular short axis level (TR: 25–26 ms, TE: 1.1–1.2 s, BW: 965 Hz, averages 1, slice thickness 5–7 mm, FOV 119–220 × 200–320 mm, matrix 57–75 × 192). Scout sequences were performed at 2, 3 and 4 min. The scout sequence acquires 20 to 30 frames of a cine sequence after an initial inversion pulse with varying TIs using a segmented true FISP readout at multiple time intervals. Scrolling through the scout images allows the reader to select the inversion time from the image with the darkest normal myocardium and brightest blood pool (Fig. 2) [14]. The best nulled scout image at 4 min was used to determine the inversion time (TI) and 30 ms was added to the TI used for subsequent FLASH imaging to account for progressive physiological washout of gadolinium from myocardium resulting in the lengthening of the longitudinal recovery time (T1) of the tissue [15]. Late gadolinium enhancement using a TurboFLASH fast phase sensitive inversion recovery (PSIR) sequence followed from 5 to 10 min (TR 350–500 ms, TE 4.2 ms, BW 130, averages 1, slice thickness 5–7 mm, gap 10%, FOV 138–188 × 200–270 mm, matrix 130–180 × 256, no. of slices 7–12) in the short axis stack, two-chamber (vertical long axis through the LA and LV), three-chamber (long axis through the LVOT and LV) and four-chamber (horizontal long axis) planes (Fig. 3). This sequence resulted in both magnitude and phase sensitive reconstructed images. Slight variation in the matrix size and FOV relates to the application of user-defined regions of interest individualized to each patient. The TI was continually increased by 10–30 ms approximately every 1–4 min to maintain adequate nulling of normal myocardium as assessed visually by the physician performing the study. At 10 min, another scout sequence was obtained. Late enhancement imaging continued after this if all views were not yet acquired. Our protocol was designed in such a way as to address the clinical question as well as to determine the optimal timing protocol.
https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig1_HTML.gif
Fig. 1

Schematic summary of the MR imaging protocol. Series of scout images were obtained at 2, 3, and 4 min after gadolinium injection, each with varying inversion times. The last series of scout images (at 4 min) was used to select the optimal inversion time (TI) to yield the best myocardial nulling. This TI was then used to perform subsequent phase-sensitive inversion recovery (PSIR) imaging between 5–10 min. A scout image was then obtained at 10 min. Adult imaging for late gadolinium enhancement is typically performed between 10–20 min

https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig2_HTML.gif
Fig. 2

Effect of varying inversion time (TI) on scout images, with corresponding schematic. At optimal nulling TI of 232.5 ms, myocardial signal is minimized (note black or nearly black left ventricular myocardium). With lower or greater TI, some myocardial signal becomes perceptible (note grey intensity of myocardium on other images)

https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig3_HTML.gif
Fig. 3

Example of nearly optimized late gadolinium enhancement (LGE) imaging in a 3-month-old boy with dilated cardiomyopathy (patient 3) shows no significant LGE on (a) short axis, (b) four-chamber (horizontal long axis) and (c) two-chamber (vertical long axis) views. The slight apparent irregularity and increased intensity in the LV lateral wall on the short axis view is attributed to partial volume averaging and noise within the blood pool

Image analysis

Patient identifying information was removed for image analysis. Images were reviewed by three independent observers experienced in cardiac MRI and blinded to the clinical information and TI used. For obvious reasons, observers could not be blinded to the types of lesions examined. For each of the scout images, the best nulled image was chosen and the degree of nulling was given a grading score as shown in Table 1 where: 0 = none; 1 = reverse; 2 = poor, 3 = partial and 4 = good nulling. After all three observers had finished grading the images, the primary investigator recorded the TI from the best nulled scout image chosen by each observer and determined the mean TI from this. PSIR images from 5 to 10 min were also graded by the three observers and the mean TI deduced in the same fashion. We used PSIR images for analysis because it removes off-resonance artifacts and minimizes background noise, resulting in more consistent image quality. Where there were multiple PSIR images during a 1-min period, the 1st image acquired during that minute was chosen for analysis. The mean and standard deviation (SD) of the grading score was calculated. The TI used during PSIR images with a mean grade ≥3 was recorded.
Table 1

Grading score for late enhancement imaging

Grade

Degree of nulling

Myocardium appearance

Blood pool appearance

0

None

Can’t distinguish between myocardium & blood

Can’t distinguish between myocardium & blood

1

Reverse

White

Black

2

Poor

Grainy

White

3

Partial

Some black areas, some grainy areas

White

4

Good

Black

White

Signal-to-noise ratio

The corresponding magnitude images were further analysed for their signal-to-noise ratio (SNR) in accordance to a published method for signal-to-noise ratio determination [16, 17]. Because the ability to optimally null myocardium was the property of interest being studied, with lower signal intensity representing more optimal nulling, a region of interest (ROI) was drawn within the best nulled area of myocardium (signal). Background noise was calculated using the same sized ROI in two locations within air outside the body, anterior and lateral to the heart, and the results were averaged. The SNR was calculated as signal divided by the mean standard deviation of background noise [6, 16, 17]. Due to the variety of congenital heart diseases (CHD), the location of the myocardial ROI could not be standardized.

Statistical analysis

Differences in the mean grading score were analysed using the non-parametric Kruskal-Wallis test, with post hoc comparisons using the Dunn test. Interobserver variability was determined using quadratic-weighted kappa statistics according to Cohen (http://www.ncbi.nlm.nih.gov/pubmed/19673146), where poor = 0–0.20; fair = 0.21–0.4; moderate = 0.41–0.6, good = 0.61–0.80 and almost perfect agreement = 0.81–1.0. Linear regression was used to assess the association of TI with time, as well as SNR with interobserver agreement. A nonlinear regression (second order polynomial, quadratic) of SNR vs. time was performed. Statistical significance was set at P < 0.05.

Results

Patients had a variety of acquired and congenital heart diseases as shown in Table 2. The indication for cardiac MRI was anatomical evaluation in congenital heart disease (n = 7) and quantification of ventricular function in cardiomyopathy (n = 5). Late enhancement imaging was clinically indicated in the five children with cardiomyopathy to assess for evidence of fibrosis. Late myocardial enhancement was present in the RV free wall in a child with a mitochondrial cardiomyopathy (Fig. 4), while the rest were negative.
Table 2

Patient characteristics

Patient

Age at MRI (months)

BSA (m2)

Diagnosis

1

 4

0.32

Cardiomyopathy

2

60

0.62

Unbalanced AVSD

3

 3

0.35

Cardiomyopathy

4

36

0.60

Heterotaxy

5

 1

0.24

Cardiomyopathy

6

57

0.70

Aortic and pulmonary stenosis

7

 5

0.32

Unbalanced AVSD

8

31

0.46

Sinus venosus ASD

9

 1

0.21

Coarctation of the aorta

10

30

0.61

Coarctation of the aorta

11

20

0.46

Cardiomyopathy

12

 1

0.23

Cardiomyopathy

https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig4_HTML.gif
Fig. 4

Late enhancement in the RV free wall (red arrow) in a 20-month-old girl (patient 11) with a mitochondrial cardiomyopathy shows the (a) magnitude and (b) PSIR images of 2 contiguous short axis slices. The PSIR images utilize the phase information together with the magnitude data to emphasize the difference between normal and abnormal myocardium. Due to the subtraction algorithm employed, the PSIR images inherently have a lower signal-to-noise ratio than the source magnitude images

Interobserver variability

Good agreement was found in the mean grading score of the scout images at 2 min (κ = 0.69) and 3 min (κ = 0.66), and a moderate agreement at 4 min (κ = 0.57). For the PSIR images, there was a moderate agreement at 7 min (κ = 0.44), fair agreement at 8 (κ = 0.29) and 9 min (κ = 0.25), and poor agreement at other times as shown in Table 3.
Table 3

Interobserver variability showing the mean grade and Kappa scores for scout and PSIR images at each time point

Type and timing of image

Mean grade

Qw Kappa score

Agreement

Scout

2 min

3.1 ± 1.1

0.69

Good

3 min

3.1 ± 1.2

0.66

Good

4 min

3.2 ± 0.9

0.57

Moderate

10 min

3.0 ± 0.4

−0.006

Poor

PSIR

5 min

3.4 ± 0.2

0.001

Poor

6 min

2.8 ± 0.6

0.11

Poor

7 min

3.2 ± 0.7

0.44

Moderate

8 min

2.9 ± 0.7

0.29

Fair

9 min

3.1 ± 0.7

0.25

Fair

Mean grading score

There were no statistically significant differences in the mean grading score of the scout images at 2, 3 and 4 min (P = 0.78). Typical scout images obtained at these times are shown in Fig. 5. For the PSIR images, the highest graded image was seen at 7 min (grade = 3.2 ± 0.72), and was not significantly higher than at other times. Poor image grading correlated with poorer agreement, thus as grades of nulling became less than 3, the differentiation between partial and poor nulling became more difficult, which simply highlights the difficulties of selecting ideal nulling time in children.
https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig5_HTML.gif
Fig. 5

Representative scout images in a 3-month-old boy with dilated cardiomyopathy (patient 3) at 2, 3 and 4 min demonstrate that images were visually similar consistent with results showing no differences in Kappa

Correlation of inversion time and scout time

There was a significant, positive linear correlation between mean TI and scout time (r = 0.61, P < 0.0001) (Fig. 6). When analyzing the scout images, the mean increase in TI per min was 18 ± 12 ms from 2 to 3 min, 15 ± 12 ms from 3 to 4 min and 41 ± 30 ms from 4 to 10 min. For PSIR images with a grading score ≥3, the mean TI change was 5 ± 8 ms per min.
https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig6_HTML.gif
Fig. 6

Linear regression of best agreed TI (mean ± SD) in the y axis, against the scout times at 2, 3, 4 and 10 min in the x axis, indicating that TI increases linearly with time (r = 0.61, P < 0.0001)

SNR

The SNR showed a binomial curve being highest at 5–6 min and 10 min, and lowest at 7–9 min (Fig. 7). Counterintuitively, a lower SNR in this study corresponds to more optimal nulling, indicating that optimal nulling occurred at 7–9 min. Linear regression showed that lower SNR significantly correlated with higher kappa values (r = 0.22, P = 0.017), indicating better agreement for images with better nulling (Fig. 8).
https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig7_HTML.gif
Fig. 7

Nonlinear regression of magnitude image SNR (mean ± SD) per minute (y axis) of imaging from 5–10 min (x axis) showing a bimodal distribution with the lowest and optimal SNR occurring at 7–9 min

https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2276-z/MediaObjects/247_2011_2276_Fig8_HTML.gif
Fig. 8

Linear regression analysis of SNR (y axis) with interobserver agreement (i.e. linear-weighted Kappa, x axis) shows that better agreement (higher kappa) was associated with lower SNR and thus more optimal nulling

Discussion

This study shows that optimal nulling for late enhancement imaging can be performed with sufficient quality imaging as early as 7 min in children under general anesthesia with free-breathing.

Scout images

Our results showed that scout images at 2–4 min were of sufficient quality to yield satisfactory interobserver agreement, allowing determination of TI at an earlier time than reported in the adult literature. Performing a scout at 2 min may seem early, however, there was no difference in interobserver variability or grading score from scout images at 2, 3 and 4 min. Thus, if earlier scout images are used to determine TI, the knowledge that TI increases with time must be taken into account and increased accordingly for subsequent PSIR images. Although the kappa score was lower at 4 min, it was the best nulled image and thus should be used more practically in clinical protocols.

PSIR images

The PSIR images at 5–6 min showed a low grading score and poor agreement, indicating that suboptimal image quality most likely causes more variability in grading. Since we did not account for manual adjustment of window leveling to maximize the contrast, observers may have adjusted the windowing level more during these poorer images leading to higher variability. These results suggest that 5–6 min after gadolinium is still too early to start late enhancement imaging.

Inversion time

The use of phase-sensitive detection removes the need for precisely nulling the myocardium and offers the capability to remove off-resonance effects, thus minimizing background noise [6]. Phase-sensitive detection is beneficial in children as it eliminates the need for breath-holds and enables the use of an approximate TI resulting in more consistent image quality [5]. TI is affected by body weight, blood volume and heart rate; optimal TI also varies with the dynamic process of gadolinium washout. It has been demonstrated that the plasma concentration of gadolinium decreases exponentially with time following contrast administration [18]. With the application of PSIR, there is controversy as to whether TI should be adjusted during the course of scanning. While Kim et al. [15] suggested that TI does not need to be adjusted if imaging is performed within 5 min, Elgetti et al. [3] recommend that the TI may need to be readjusted several times in the course of an examination.

The linear increase in the optimal TI myocardial nulling over 10 min and known faster metabolism in younger children implies that the TI should be increased frequently during the course of an examination, which is often challenging in this time-sensitive situation [3]. When only assessing partial‐to‐good nulling PSIR images (grades 3–4), we found a 0–25 ms increase in TI per minute, which may serve as a guide for incremental TI increase during scanning.

Timing and SNR

Adult studies have shown that LGE images may be acquired 10–30 min after gadolinium [2, 18]. In animal studies, LGE imaging at 7 min overestimated the infarct size by 20%, but when performed at 25 min, correct identification of infarct size occurred [19]. The cause of this was postulated to be due to gadolinium accumulating in the peri-infarct zone where there is tissue edema, but this should not apply to children as the pathology is typically nonischemic. Unlike in most other applications, decreased SNR in the context of nulled myocardium paradoxically reflects improved nulling of myocardium, since no signal represents complete nulling. Our findings of bimodal distribution of the SNR with time suggests that waiting >10 min to perform late enhancement imaging is associated with suboptimal SNR. Our SNR results showing the greatest nulling at 7–9 min coincides with the highest agreement among observers at 7 min. Decreased SNR and thus improved nulling also correlated with kappa values indicating that agreement was higher at times of better nulling. These results support the observation that optimal nulling in anesthetized children is achieved approximately 7 min after gadolinium administration, within a narrow time frame, and thus protocols in this age group should be timed with this in mind.

The optimal time for LGE imaging observed in this study was earlier than that reported for adults [2, 4, 10, 11, 20]. These results are consistent with the findings of Baker et al. [21], who found significant age-related changes in elimination half-life of gadolinium, specifically a significantly increased elimination half-life in the younger age group. LGE is based on the retention of gadolinium chelates in the extracellular space of infarct related to enlargement of the interstitial space [22, 23]. Gadolinium chelates wash out from infarcted myocardium more slowly than from normal tissue [6, 24], or may be more visible because of their higher volume of distribution in the interstitium of scar [25], resulting in different T1 values for normal and infarcted tissue. Children tend to have relatively faster heart rates and cardiac outputs, as well as greater volume of distribution for gadolinium than adults, so it is possible that the wash-in and washout time constants reported by Kim et al. [24] may be faster in children, resulting in faster times to steady state [25].

Limitations

The study is limited by a small sample size. Due to a magnet upgrade during the course of this study, we were unable to recruit new subjects under the same conditions. Despite the fact that our patient selection criteria were expected to yield a low incidence of positive late enhancement, we had one patient with positive findings. Given that the same protocol was applied in all patients, we thought it unlikely to have false-negatives. Due to clinical indications for performing LGE, we could not perform scout imaging every minute; however, future studies may be performed to address this. Furthermore, multiple differing sequences with different manufacturers exist for examining LGE and thus differing techniques may follow different timing parameters; therefore, our results cannot be generalized to all types of magnets.

Conclusion

Our results confirm our hypothesis that when performing late gadolinium imaging in anesthetised children, the optimal nulling time occurs earlier than described in adult protocols. We propose a guide to imaging protocol in this population with determination of TI from scout images as early as 4 min, followed by PSIR images at 7 min, with the TI adjusted frequently during the study.

Copyright information

© Springer-Verlag 2011