Systolic modified Look–Locker inversion recovery myocardial T1 mapping improves the accuracy of T1 and extracellular volume fraction measurements of patients with high heart rate or atrial fibrillation
Image data for T1 mapping are generally acquired during mid-diastole period. However, T1 mapping tends to fail for patients with high heart rate or atrial fibrillation because of short or irregular R-R interval. Focusing on the evidence that the timing of systole is more stable than that of diastole from the R wave, we compared systolic T1 mapping with conventional diastolic T1 mapping for all participants (n = 58) by visual scoring of T1 calculation error. The systolic scores were significantly better than the diastolic scores (p < 0.05). This advantage of the systolic scores was confirmed selectively for patients with atrial fibrillation (p < 0.05, n = 19). The successful number of nonrigid image registration alignment for extracellular volume fraction (ECV) analysis also increased significantly for systolic images compared with diastolic images (p < 0.05). Thus, systolic T1 mapping improves the accuracy of T1 values and ECV analysis.
Hinweise
The original version of this article was revised due to a retrospective Open Access Cancellation.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
Late gadolinium enhancement (LGE) is good at detecting focal myocardial fibrosis but less effective in detecting diffuse myocardial fibrosis. Fifty-nine percent of dilated cardiomyopathy is not contrasted with delayed enhancement [1]. However, T1 mapping is clinically useful for quantitative evaluation of diffuse myocardial fibrosis [2‐5]. In addition, myocardial characteristics, such as edema and focal and diffuse fibrosis, can be quantitatively evaluated with extracellular volume fraction (ECV) analysis, and the ECV can be calculated from both precontrast and postcontrast T1 maps [6‐8]. The modified Look–Locker inversion recovery (MOLLI) sequence for measuring T1 values is widely useful because of its high precision [9, 10]. In the MOLLI sequence, several images are acquired, while T1 is changed in the process of recovering the signal after longitudinal magnetization by the inversion recovery pulse. With the MOLLI sequence, we can calculate T1 values for each pixel from the signal intensity of the acquired data. The MOLLI sequence can obtain stable T1 values through time-based data acquisition [11]. T1 mapping is generally acquired during the mid-diastole phase, when left ventricular volume change is minimal, as the optimal cardiac phase for patients with low heart rate. However, irregularities among patients with high heart rate and atrial fibrillation are commonly observed on cardiac magnetic resonance (CMR) imaging, and these patients have irregular and short R-R intervals. Therefore, diastolic T1 mapping cannot accommodate irregular and short R-R intervals and cannot be acquired in the optimal cardiac phase. This would increase calculation errors resulting from movements, and thus, the correct T1 values could not be obtained in the past. In some pilot studies, systolic T1 mapping is performed when the myocardium is the thickest, and few calculation errors resulting from movements are equally or more appropriate than diastolic T1 mapping in the quantitative evaluation of the myocardium because partial volume effects are reduced, and motion artifacts are minimized during tachycardia [12‐15]. Therefore, we visually evaluated calculation errors resulting from movements (motion artifacts during tachycardia) using the T1 mapping confidence map (error map) in the magnetic resonance system. Traditionally, during diastole, the alignment of both precontrast and postcontrast auto-analysis non-rigid image registration for ECV is mostly unsuccessful. Non-rigid image registration alignment is standard for every workstation's ECV analysis software, and the success of non-rigid image registration alignment is highly dependent on the parallel shift amount, rotation shift amount, and scaling deformation amount. If alignment fails, the left ventricular myocardium in three slices of the left ventricular base, middle, and apex with both precontrast and postcontrast images needs to be manually reextracted from the endocardial and epicardial boundaries. This re-extraction reduces the accuracy of the ECV analysis. On the contrary, successfully by auto-analysis, non-rigid image registration alignment improves the accuracy of ECV analysis because the subendocardial, mid-wall, and epicardial segment of the left ventricular myocardium matched precontrast and postcontrast images. In this study, we compared the conventional diastole and systole in T1 mapping and ECV.
2 Materials and methods
2.1 Study population
We recruited 20 healthy volunteers (10 men and 10 women; mean age, 45 ± 12 years) and 38 patients (23 men and 15 women; mean age, 61 ± 12 years) without contraindications to CMR imaging. Twenty healthy volunteers underwent non-contrast examinations. Thirty-eight patients were administered contrast agents, and T1 maps were obtained from both precontrast and postcontrast images; further, the ECV analysis was also performed. None of the healthy volunteers had any evidence of or risk factors for cardiovascular disease. Patients were recruited consecutively and had diagnoses of atrial fibrillation (n = 19), dilated cardiomyopathy (n = 4), ischemic cardiomyopathy (n = 2), hypertrophic cardiomyopathy (n = 2), old myocardial infarction (n = 4), cardiac sarcoidosis (n = 3), and other diseases (n = 4). The study protocol was approved by our institutional review board. The study was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants before the CMR imaging. A detailed explanation of the contrast agents was provided to each participant.
Anzeige
2.2 Determination of the MOLLI sequence sampling scheme
The MOLLI sequence is acquired with both precontrast and postcontrast T1 maps by time-based data acquisition. The time-based MOLLI sequence has fixed acquisition and pauses (recovery time) (Fig. 1). Precontrast T1 mapping is set to acquisition time (5 s) + pause time (3 s) + acquisition time (3 s). Postcontrast T1 mapping is set to acquisition time (4 s) + pause time (1 s) + acquisition time (3 s) + pause time (1 s) + acquisition time (2 s) [16]. Because the R-R interval is short at high heart rates, a fixed recovery pause time (precontrast pause time of 3 s, postcontrast pause time of 1 s) can fully recover longitudinal magnetization before the next inversion pulse is applied. We adopted a time-based MOLLI sequence that is independent of heartbeat, from low to high heartbeats.
Fig. 1
T1 mapping scan using the modified Look–Locker inversion recovery sequence. An example of precontrast scanning protocol for a patient with heart rate of 60 bpm (R-R interval = 1000 ms)
×
2.3 Trigger delay in diastole and systole
Acquisition timing was visually determined in short-axis cine images; when the left ventricular area was at its maximum, the delay time for diastolic data acquisition was calculated. Subsequently, when the left ventricular area was minimal, the delay time for systolic data acquisition was calculated [17].
2.4 CMR image acquisition
CMR imaging was performed with a 1.5-T magnetic resonance system (Ingenia, release 5.3.1.3, and dS Torso receiver coil; Philips Healthcare, Amsterdam, the Netherlands). With electrocardiographic gating, data for all sequences were acquired while participants held their breath. Cine images, based on electrocardiography-gated steady-state free precession imaging, were acquired in the short-axis view, and in four-, two-, and three-chamber views, and were reconstructed for 24 phases/heartbeat. The cine sequence parameters were as follows: echo time (TE), 1.16 ms; repetition time (TR), 2.3 ms; flip angle (FA), 60°; slice thickness, 10 mm; field of view, 380 × 380 mm2; in-plane resolution, 1.70 × 1.41 mm2; acquisition matrix, 224 × 236; and sensitivity encoding (SENSE) phase direction factor of 2.5. LGE sequence parameters with three-dimensional inversion recovery T1-turbo field echo were as follows: TE, 2.3 ms; TR, 4.7 ms; FA, 15°; slice thickness, 10 mm; field of view, 380 × 380 mm2; in-plane resolution, 1.48 × 1.65 mm2; acquisition matrix, 256 × 231; SENSE phase direction factor of 2.5; SENSE slice direction factor of 1.5; linear profile order; and radial turbo direction. LGE images were acquired 10–15 min after an injection of 0.1 mmol/kg gadobutrol (Gadovist, Bayer, Germany). T1 mapping sequence parameters were as follows: TE, 0.88 ms; TR, 1.97 ms; FA, 35°; slice thickness, 10 mm; field of view, 300 × 300 mm2; in-plane resolution, 2.0 × 2.0 mm2; acquisition matrix, 152 × 150; SENSE phase direction factor of 2.0; and linear profile order. T1 mapping images were acquired in basal, midventricular, and apical short-axis planes during diastole and systole before and 15 min after an injection of 0.1 mmol/kg gadobutrol.
2.5 Image analysis
2.5.1 Left ventricular functions
All image datasets were transferred to a workstation (Intelli Space Portal, version 7.0.2.20700; Philips Healthcare). Left ventricular ejection fraction (LVEF), end-diastolic volume (EDV), and end-systolic volumes (ESV) were evaluated. The LVEF was calculated as follows:
LVEF (%) = (EDV–ESV)/EDV × 100.
2.5.2 T1 mapping
T1 mapping images obtained from all participants were transferred to a workstation (Ziostation2, T1 map, version 2.4.2.3; Ziosoft, Inc., Tokyo, Japan). To measure T1 values, the T1 mapping confidence map was used to manually set a region of interest (ROI) of 5 mm2 at the middle of the septum in the mid-wall of the left ventricle (Fig. 2). The confidence map is displayed as an error when the standard deviation deviates significantly for each TI pixel collected in the T1 recovery curve. Therefore, the confidence map will show the calculation errors resulting from movements as blackout pixels. The ROI did not include the pixel of calculation errors resulting from movements; it was set in a reliable range [18].
Fig. 2
An example of T1 confidence map. A region of interest area of 5 mm2 at the middle of the septum in the mid-wall of the left ventricular was placed manually excluding the blackout pixels that represent calculation errors of T1 value resulting from movements
×
2.5.3 Visual evaluation
Visual evaluation was conducted on the precontrast T1 mapping confidence map for all participants. Figure 3 shows the visual evaluation method. The blackout pixels represent the calculation errors of the T1 value. Scoring was performed based on the number of blackout pixels in the heart area. If the number of pixels indicating calculation errors was more than 50, the score was set to 1. If it was between 20 and 50, the score was 2. If it was less than 20 pixels, the score was 3. An experienced reader with more than 5 years of cardiac magnetic resonance imaging radiology technician performed a visual evaluation of the confidence map calculation error. Visual evaluation was calculated separately for all participants, for patients with atrial fibrillation alone, and for each heart rate.
Fig. 3
Visual evaluation scoring examples of T1 confidence map. The blackout pixels represent calculation errors of T1 value. The scoring was performed by the number of the blackout pixels in the heart area: score 1 (> 50 pixels), score 2 (20–50 pixels) and score 3 (< 20 pixels)
×
2.5.4 ECV
In 38 patients, the ECV was calculated from the left ventricular myocardial T1 values, and the left ventricular cavity blood T1 values acquired pre- and post-contrast. ECV was calibrated with the blood hematocrit level at a workstation for an offline analysis (Ziostation2, T1 map, version 2.4.2.3) [19]. The blood hematocrit level is the result of a blood sample taken the day before CMR. The ECV was calculated as
Using the Bull's-Eye plot of ECV, we also calculated the average value of five segments of the left ventricular myocardial septum collected during diastole and systole (Fig. 4). The left ventricular septum was chosen because the myocardium is thicker than the anterior wall, lateral wall, and posterior wall. These facilitate segments of endocardial, mid-wall, and epicardial. In addition, the ECV was analyzed on an offline analysis workstation, auto-analysis non-rigid image registration alignment with both precontrast and postcontrast images was performed (Fig. 5), and the number of successful cases by auto-analysis was counted. If the alignment failed, the left ventricular myocardium in three slices of the left ventricular base, middle, and apex with both precontrast and postcontrast images was manually reextracted from the endocardial and epicardial boundaries.
Fig. 4
Bull's-Eye plot of extracellular volume fraction. The five values in the red circle are of the segments in the left ventricular myocardial septum
Fig. 5
An example of auto-analysis precontrast and postcontrast non-rigid image registration alignment for extracellular volume fraction (ECV) analysis. a and b are the same patient, a was success alignment and collected in systole, and b was unsuccess alignment (arrow) and collected in diastole. If auto-analysis precontrast and postcontrast non-rigid image registration alignments failed, the left ventricular myocardium in the left ventricular base, middle, and apex planes of the left ventricle was manually re-extracted to extract the endocardial and epicardial borders. Successful positioning of the non-rigid registration suggested improved accuracy of ECV because the left ventricular myocardium matched precontrast and postcontrast images.
×
×
2.6 Statistical analyses
Data are expressed as the means ± standard deviations. The normality of the diastolic and systolic T1 data in healthy volunteers and patients was assessed using the Kolmogorov–Smirnov test. Because of the normality of the data, the Pearson correlation coefficient (r) was used for the analysis, and the data were compared using paired t-test and one-way analysis of variance. The normality of the visual evaluation data in all participants and that of ECV data was assessed using the Kolmogorov–Smirnov test. Visual evaluation data and ECV data did not strictly follow a Gaussian distribution. Therefore, the statistical tests used were non-parametric, and the Kruskal–Wallis test was used. Statistical significance was defined as a p-value < 0.05. All statistical analyses were performed using R software, version 3.4.1 [20].
Anzeige
3 Results
3.1 Clinical characteristics of study participants
Table 1 shows the demographic data, heart rate, and heart rate fluctuation during scanning and left ventricular functional indices of healthy volunteers and patients.
Table 1
Characteristics and left-ventricle functional indices of healthy volunteers and patients
Healthy volunteers
Patients
(n = 20)
(n = 38)
Sex (man), n (%)
10 (50%)
23 (61%)
Age (years)
45.2 ± 11.9
61.4 ± 12.3
Height (cm)
163.9 ± 7.6
161.9 ± 7.6
Weight (kg)
61.8 ± 7.0
58.8 ± 10.7
BMI (kg/m2)
21.8 ± 3.6
23.58 ± 4.1
Heart rate (bpm)
65.4 ± 4.4
68.9 ± 11.2
HR fluctuation (bpm)
8.7 ± 3.7
16.4 ± 7.4
LVEF (%)
61.29 ± 7.58
48.63 ± 18.28
LVEDV (mL)
128.7 ± 23.5
87.2 ± 16.5
LVESV (mL)
58.1 ± 12.8
48.1 ± 21.4
Data are expressed as means ± standard deviations
BMI body mass index, HR heart rate, LV left ventricle, EF ejection fraction, EDV end-diastolic volume, ESV end-systolic volume
3.2 Comparison between diastolic and systolic T1 values
Left ventricular myocardial T1 values during diastole and systole were strongly and positively correlated in healthy volunteers (r = 0.94, p < 0.05) (Fig. 6). Table 2 lists left ventricular myocardial T1 values in diastole and systole of the healthy volunteers and patients. Left ventricular myocardial T1 systolic values were significantly lower than T1 diastolic values in both healthy volunteers (diastole 998.7 ± 17.1 ms vs systole 988.7 ± 24.9 ms; p < 0.05) and patients (diastole 1014.8 ± 45.9 ms vs systole 1006.9 ± 45.4 ms; p < 0.05).
Fig. 6
Scatter plot of the diastolic and systolic T1 values in 20 healthy volunteers
Table 2
Result of the LV–T1 values between diastole and systole (ms)
Diastole
Systole
p-value
Patients (n = 38)
1014.8 ± 45.9
1006.9 ± 45.4
p < 0.05
Healthy volunteers (n = 20)
998.7 ± 17.1
988.7 ± 24.9
p < 0.05
Total (n = 58)
1008.7 ± 36.0
1000.0 ± 37.8
p < 0.05
Data expressed as mean ± standard deviation
×
3.3 Visual evaluation
According to the visual evaluation of the diastolic and systolic T1 mapping confidence map images, in all participants, the systolic score was significantly higher than the diastolic score (diastole 2.03 ± 0.51 vs systole 2.25 ± 0.57; p < 0.05) (Table 3). In patients with atrial fibrillation, the systolic score was also significantly higher than the diastolic score (diastole 1.63 ± 0.53 vs systole 2.18 ± 0.59; p < 0.05) (Table 3). T1 mapping visual evaluation for each heart rate also revealed that when the heart rates were 40–60 bpm, the diastolic score was higher than the systolic score, whereas at heart rates of 61–120 bpm, the systolic score was higher than the diastolic score (Fig. 7).
Table 3
Visual evaluation of all participants and AF patients only
Diastole
Systole
p-value
All participants (n = 58)
2.03 ± 0.51
2.25 ± 0.57
p < 0.05
AF patients only (n = 19)
1.63 ± 0.53
2.18 ± 0.59
p < 0.05
Data expressed as mean ± standard deviation
Fig. 7
Visual evaluation of precontrast T1 mapping for each heart rate in all participants. The vertical axis represents the visual evaluation score of myocardial diastolic and systolic precontrast T1 mapping; the horizontal axis represents the presented heart rate during scan. The number of all participants at different heart rate levels is presented; 40–60 bpm (n = 17), 61–80 bpm (n = 23), 81–100 bpm (n = 14), and 101–120 bpm (n = 4)
×
3.4 ECV
ECV showed significantly lower systolic scores and T1 values. As a supplement, we included cases in which auto-analysis failed and manual analysis was performed for reasons of low success during diastole (diastole 0.34 ± 0.05 vs systole 0.33 ± 0.05; p < 0.05) (Fig. 8). Diastolic and systolic ECV were strongly and positively correlated in patients (r = 0.98, p < 0.05) (Fig. 9). Auto-analysis non-rigid image registration alignment for ECV analysis in low–high heart rate and atrial fibrillation cases was significantly more successful during systole, and systolic ECV showed improved accuracy (heart rate 40–70 bpm, diastole 7 vs systole 12; p < 0.05) (heart rate 71–120 bpm, diastole 2 vs systole 22; p < 0.05) (Fig. 10).
Fig. 8
Comparison of extracellular volume fraction between diastole and systole in 38 patient
Fig. 9
Scatter plot for extracellular volume fraction of diastole and systole in 38 patients
Fig. 10
Evaluation of extracellular volume fraction between diastole and systole in 38 patients. Number of successful cases with auto-analysis non-rigid image registration alignment was counted. Thirty-eight patients were divided in two groups: heart rate 40–70 bpm (n = 15) and heart rate 71–120 bpm (n = 23). All patients with atrial fibrillation (n = 19) are included in heart rate 71–120 bpm group (n = 23)
×
×
×
4 Discussion
In this study, diastolic and systolic T1 values were significantly different but strongly correlated in the normal myocardium of 20 healthy volunteers. In addition, systolic T1 values were lower than diastolic values [14]. The reason for this is that during diastole, the left ventricular myocardial thickness is lower and is affected by the partial volume effect of adjacent tissues, and includes blood within a set ROI, which may have resulted in slightly higher diastolic T1 values. In all participants and patients with atrial fibrillation, systolic T1 mapping had better visual evaluation with significant difference. A low heart rate of 40–60 bpm showed good visual evaluation even in diastolic T1 mapping because of the optimal phase at mid-diastole [15]. At a high heart rate, the motion-free time in diastole is shortened more than in systole. In patients with atrial fibrillation, although the R-R interval varies in each cardiac cycle, the variation in timing of systolic acquisition is smaller than that of diastolic acquisition [11, 15]. Therefore, systole acquisition resulted in more evaluable images [21, 22]. As with T1 mapping, ECV was significantly lower during systole than during diastole, and the systolic values showed a strong positive correlation with the diastolic values. These findings concur with those of some studies [14‐17]. Auto-analysis precontrast and postcontrast non-rigid image registration alignment for ECV analysis in both the heart rate 40–70 bpm group, and heart rate 71–120 bpm and atrial fibrillation groups was significantly more successful during systole. The success of non-rigid image registration alignment is highly dependent on the parallel shift amount, rotation shift amount, and scaling deformation amount. In systole, the left ventricular myocardium maintains its thickness and the partial volume effect is reduced, and calculation errors resulting from movements are reduced. Accordingly, successfully by auto-analysis, non-rigid image registration alignment improves the accuracy of ECV analysis because the subendocardial, mid-wall, and epicardial segment of the left ventricular myocardium matched precontrast and postcontrast images. A limitation of this study was the small study population. Previous consensus statements suggested that 20 is the sample size required to obtain a power of normal values for T1 mapping [23]. Therefore, the control data obtained from the 20 healthy volunteers in this study are probably reliable. Moreover, during diastole and systole, the acquisition section for T1 mapping may appear different because of the myocardium strain. In the future, further studies on not only the short-axis section but also the two- and four-chamber sections will be necessary.
5 Conclusion
Systolic MOLLI myocardial T1 mapping improves the accuracy of T1 values and ECV analysis by reducing calculation errors resulting from movements. It is applicable to patients with low–high heart rates and atrial fibrillation.
Acknowledgements
This manuscript was partly supported by Akiyoshi Ohtsuka Fellowship of the Japanese Society of Radiological Technology for improvement in English expression of a draft version of the manuscript.
Compliance with ethical standards
Conflict of Interest
The authors declare that they have no conflict of interest.
Anzeige
Research involving human participants
All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional Review Board and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This article does not contain any studies performed with animals.
Informed consent
Informed consent was obtained from all individual participants included in the study.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Mit e.Med Radiologie erhalten Sie Zugang zu CME-Fortbildungen des Fachgebietes Radiologie, den Premium-Inhalten der radiologischen Fachzeitschriften, inklusive einer gedruckten Radiologie-Zeitschrift Ihrer Wahl.
Systolic modified Look–Locker inversion recovery myocardial T1 mapping improves the accuracy of T1 and extracellular volume fraction measurements of patients with high heart rate or atrial fibrillation
