Published online Jul 06, 2022.
https://doi.org/10.22468/cvia.2022.00024
Splenic T1-Mapping for Predicting Adenosine Stress Adequacy in Cardiac Magnetic Resonance Myocardial Perfusion Imaging: A Validation and Reproducibility Study
,1
Catherine Ming-mun Yeung,1
Pak-ki Yam,1
Pan-pan Ng,1
Boris Chun-kei Chow,1
Jeanie Betsy Chiang,1
Jonan Chun-yin Lee,1
Kenneth Kai-yat Cheung,2
and Ming-yen Ng3
Abstract
Objective
Splenic switch-off (SSO) sign has been utilized as a surrogate marker of adequate stress but can only be assessed after first-pass perfusion imaging. A study previously reported that drop in T1spleen ≥30 ms during adenosine infusion predicts presence of SSO, but this finding has not been externally validated. This study aimed to prospectively validate whether drop in T1spleen ≥30 ms is a reliable marker of SSO and hence adequate stress, and to assess reproducibility of T1spleen measurements.
Materials and Methods
Data of fifty consecutive patients undergoing stress cardiac magnetic resonance were prospectively collected. Native T1-maps were acquired at rest and at 2.5 min after adenosine infusion in short axis slices, followed by perfusion images at 3 min. To measure T1spleen pre- and post-adenosine infusion, regions of interest were manually placed to include most splenic tissue. Adenosine stress adequacy was evaluated by visual SSO assessment and semi-quantitative splenic perfusion analysis.
Results
A significant association was found between a drop in T1spleen of ≥30 ms and SSO response (p<0.001). There was excellent correlation between SSO response and semiquantitative perfusion change in spleen (rho=0.847, p<0.001). Inter-observer and intra-observer agreement for measurement of ΔT1spleen values were excellent, with intra-class correlation coefficients of 0.987 and 0.995, respectively. By receiver-operating characteristic analysis, the optimal cut-off value of ΔT1spleen for predicting presence of SSO was -28 ms, with area under the curve=0.76 (p=0.002).
Conclusion
Splenic T1-mapping is accurate and reproducible for predicting SSO, potentially allowing optimization of adenosine dosage for adequate stress.
INTRODUCTION
Coronary artery disease (CAD) is the leading cause of death worldwide [1, 2]. Cardiac stress testing evaluates the hemodynamic significance of coronary artery stenosis [3]. Adequate vasodilator stress is a prerequisite for stress perfusion cardiac magnetic resonance (CMR) to prevent false negative results. Yet a recent study [4] reported that nearly 30% of patients exhibited inadequate stress. Stress testing traditionally relied on hemodynamic parameters (i.e., >10 bpm increase in heart rate and/or >10 mm Hg drop in systolic blood pressure) during adenosine infusion to determine adequate stress had occurred. However, these parameters correlated poorly with coronary vasodilation and increased myocardial blood flow (MBF) as a response to adenosine infusion [4]. Splenic switch-off (SSO) sign has gained increasing recognition as a more accurate surrogate marker for stress adequacy in CMR [5, 6, 7]. Its physiological mechanism was postulated to be related to a reduction of splenic blood volume due to reactive vasoconstriction by adenosine-induced hypotension [8], causing visually reduced splenic perfusion during stress when compared to rest [9]. Failed SSO was found to be significantly more common in patients with false-negative stress CMR findings than among those with true-negative findings [5]. Yet a major limitation is that the sign can only be observed after perfusion imaging, leaving no room for further dose optimization. This limitation has thus prompted research of a reliable marker that can be assessed before contrast administration.
Myocardial T1 mapping is routinely acquired for myocardial tissue characterization in some centers. Visceral T1 relaxation time is sensitive to changes in tissue blood volume, infiltration and fibrosis [10, 11, 12]. The spleen is typically visible on myocardial T1-maps and can be assessed without additional planning. Adenosine causes splenic vasoconstriction with reduction of splenic blood volume [5], and thus a decrease in the T1spleen value. A drop in T1spleen of ≥30 ms during adenosine infusion was shown to be predictive of subsequent presence of SSO in a retrospective single centre study [13], but this has not been externally validated. Our study aimed to validate whether T1spleen was a reliable tool to predict the occurrence of SSO and hence adequate stress in local clinical practice, as well as to assess the reproducibility of T1spleen measurement.
MATERIALS AND METHODS
Study population
Fifty consecutive patients referred to our department for known or suspected CAD for stress CMR examination were recruited between July and October 2020. The majority of the patients (78%) had history of ischemic heart disease and stress perfusion CMR was performed for assessment of myocardial perfusion and viability. The study protocol was approved by Kowloon Central/Kowloon East Cluster Research Ethics Committee, Hospital Authority, Hong Kong (approval number: KC/KE-20-0364-ER-1).
Cardiac magnetic resonance protocol
All patients avoided caffeine intake for ≥24 hours before the examination. All the CMR scans were performed at our 1.5T MRI scanner (Magnetom Aera; Siemens Healthcare, Erlangen, Germany). T1 mapping was done using the Modified Look-Locker Inversion recovery (MOLLI) prototype sequence, which requires 11-heartbeats breath holds per acquisition. Acquisition of native T1-maps was performed at rest in short axis (apical, mid-ventricular, and basal) slices (Fig. 1). At 2.5 min after starting adenosine intravenous infusion (140 µg/kg/min, 3 min), stress T1-map was performed in the short axis slice where spleen was best visualized. This was then followed by routine perfusion scan at 3 min.
Fig. 1
Native T1-maps in short axis slices. A: Basal. B: Mid-ventricular. C: Apical.
Analysis of splenic T1-mapping
All the T1-maps were analyzed by two independent radiologists (W.F.F.Y., 2 years of radiology experience and 6 months of CMR experience; L.J.C.Y., 9 years of radiology experience and 6 years of CMR experience) who were blinded to splenic perfusion images. T1-maps were excluded if the spleen was not well visualized in any of the short axis slices (2 out of 50 patients, 4%). To measure T1spleen pre- and post-stress, regions of interest (ROIs) were manually placed to include most of the splenic tissue while avoiding partial volume effects from large splenic blood vessels and borders with neighboring tissues (Fig. 2). The change in T1spleen in response to adenosine stress was measured by calculating the difference between stress T1spleen and rest T1spleen. The inter-observer reliability and intra-observer reliability of T1spleen measurement were assessed for reproducibility of the technique. The intra-observer reliability was assessed with measurements repeated by radiologist (F.F.Y.W.) 4 weeks after initial measurements.
Fig. 2
T1spleen measurement with region of interest (ROI) placed to include most of the spleen while avoiding partial volume effects from large splenic blood vessels and borders with neighboring tissues.
Assessment of splenic perfusion
Adenosine stress adequacy was analyzed by radiologist blinded to T1-maps. It was evaluated by both visual assessment of SSO response and semi-quantitative splenic perfusion analysis. Splenic perfusion was first assessed visually by grading the perfusion images as presence or absence of SSO response (Fig. 3). The presence of SSO was defined as the visual attenuation of splenic perfusion during adenosine stress compared with rest [5]. To assess semi-quantitative splenic perfusion analysis, splenic ROIs were placed on rest and stress perfusion images with manual correction for motion and artefacts to create curves showing changes of average splenic signal intensity (SI) over time (50–60 s) (Fig. 4). Splenic perfusion was estimated as the difference between baseline SI and maximum SI during first-pass perfusion. Adenosine-induced change in T1spleen was calculated as follow:
Fig. 3
The splenic switch off sign refers to visually reduced perfusion of the spleen (arrows) during adequate stress (A) when compared to rest (B).
Fig. 4
Semiquantitative splenic perfusion analysis with splenic regions of interest placed (red line) on stress (A) and rest (B) perfusion images to create curves showing changes of average splenic signal intensity over time.
Statistical analysis
Data are reported as means and standard deviations. p<0.05 denotes statistical significance.
The distribution of the quantitative data was first assessed for normality by using the Shapiro–Wilk test. Since the data were not parametrically distributed, the analysis of differences in characteristics between groups (e.g., ΔT1spleen in male vs. female patients) was tested using Wilcoxon rank sum test, and analysis of differences in characteristics within individuals (e.g., rest T1spleen vs. stress T1spleen) was tested using Wilcoxon signed rank test. Correlations between two variables (e.g., visual SSO response vs. ΔSIspleen) were assessed using Spearman’s rank correlation coefficient (rho). Fisher’s exact test was used to determine if there is non-random association between two categorical variables (e.g., SSO response and drop in T1spleen ≥30 ms). The inter-observer reliability and intra-observer reliability of measurement of ΔT1spleen were assessed using the intra-class correlation coefficient (ICC) reporting 95% confidence intervals (CIs) and the Bland–Altman plot. The performance of ΔT1spleen for predicting SSO responses was assessed using receiver-operating characteristic (ROC) curve, reporting area under the curve (AUC±standard error), as well as sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV) with 95% CIs.
RESULTS
Characteristics of study subject
Subject characteristics are summarized in Table 1. All patients experienced at least one of the three traditional markers for assessing stress adequacy, which include >10 bpm increase in heart rate, >10 mm Hg drop in systolic blood pressure, and the presence of adenosine-related symptoms. The spleen was visualized in 96% of our patients. In majority (75%) of the patients, the spleen was best visualized in the mid-ventricular slice.
Table 1
Subject characteristics
Rest and stress T1spleen
The mean rest T1spleen was 1056±87 ms and the mean stress T1spleen was 1004±87 ms. The mean ΔT1spleen during adenosine stress was -51±54 ms. The mean ΔT1spleen in patients with SSO response was -71±45 ms and the mean ΔT1spleen in patients without SSO response was -21±55 ms.
ΔT1spleen was not significantly affected by sex (male vs. female, -53±52 ms vs. -48±60 ms; p=0.674) or age (rho=0.07; p=0.635; range, 14–86 years) of the patients.
T1spleen inter-observer and intra-observer reliability
The inter-observer variability for ΔT1spleen was -31 to +42 ms. The intra-observer variability was within -32 to +16 ms for ΔT1spleen. Both showed excellent correlation as evidenced by ICC of 0.987 (95% CI, 0.977 to 0.993) and 0.995 (95% CI, 0.991 to 0.997), respectively. The Bland–Altman plots are shown in Fig. 5.
Fig. 5
Bland–Altman plots of inter-observer agreement (A) and intra-observer agreement (B) of T1spleen measurement.
Assessment of splenic perfusion
Visual SSO response was present in 60% of our patients. In patients with SSO response, peak splenic perfusion SI reduced significantly with stress compared to rest (mean ΔSIspleen [%]=-67%). In patients without SSO response, peak splenic perfusion SI did not demonstrate a significant drop with stress compared to rest (mean ΔSIspleen [%]=-14%). There was excellent correlation between visual SSO response and semi-quantitative splenic perfusion change (rho=0.847, p<0.001), suggesting that a visual assessment of SSO is a valid alternative in assessing splenic perfusion in clinical practice.
Associations between SSO response, ΔT1spleen and traditional hemodynamic markers. In patients with SSO response, there was a significant difference in T1spleen pre-stress and T1spleen post-stress with mean ΔT1spleen=-71 ms, p<0.001 (example shown in Fig. 6). In patients without SSO response, there was no significant difference in T1spleen pre-stress and T1spleen post-stress with mean ΔT1spleen=-21 ms, p=0.184 (example shown in Fig. 7).
Fig. 6
Example of patient with splenic switch-off response showing ΔT1spleen=-56 ms. A: T1spleen pre-stress=1074 ms. B: T1spleen post-stress=1018 ms. ROI, region of interest.
Fig. 7
Example of patient without SSO response showing ΔT1spleen=+17 ms. A: T1spleen pre-stress=974 ms. B: T1spleen post-stress=991 ms. ROI, region of interest.
In patients with ≥30 ms drop in T1spleen post-stress, 82% of them demonstrated SSO response, whereas in patients with <30 ms drop in T1spleen post-stress, only 13% of them had SSO response.
A significant association was found between SSO response and a drop in T1spleen of ≥30 ms (p<0.001). However, no significant association was found between SSO response and traditional hemodynamic markers, including >10 bpm increase in heart rate (p=0.193), >10 mm Hg drop in systolic blood pressure during stress (p>0.999) and presence of adenosine-related symptoms (p=0.766). Results suggested that splenic T1-mapping is a much more reliable marker of SSO response.
Receiver-operating characteristic analysis of ΔT1spleen for predicting splenic switch-off response
ROC analysis using SSO response as reference standard yielded AUC of 0.76±0.08 (p=0.002, Fig. 8). The optimal cut-off value of ΔT1spleen for predicting the presence of SSO was -28 ms, which was similar to the cut-off value yielded by Liu et al. [13]’s study, with sensitivity of 93% (95% CI, 77% to 99%,), specificity of 68% (95% CI, 43% to 87%), accuracy of 83% (95% CI, 70% to 93%), PPV of 82% (95% CI, 70% to 90%) and NPV of 87% (95% CI, 63% to 96%).
Fig. 8
Receiver-operating characteristic curve of Δ T1spleen for predicting SSO response. A drop in T1spleen of ≥28 ms during stress predicted SSO response (area under the curve=0.76, p=0.002) with sensitivity (93%), specificity (68%), accuracy (83%), positive predictive value (82%), and negative predictive value (87%). SSO, splenic switch-off.
DISCUSSION
Our results showed a significantly greater reduction in T1spleen in patients with adequate adenosine stress (as defined by the presence of SSO response), when compared to those with failed adenosine stress (as defined by the absence of SSO response), demonstrating that T1spleen of ≥30 ms was predictive of SSO response. The ΔT1spleen was independent of the age and sex of the patients, and therefore its usage was applicable to all the patients encountered in clinical practice. The splenic T1 mapping values were consistently reproducible, and its accurate measurement did not require an experienced CMR imager.
To the best of our knowledge, this is the first published study to assess the performance of splenic T1 mapping by Liu et al. [13]. Our results validated the usage of the ΔT1spleen cutoff of -30 ms suggested by Liu et al. [13], even in a different patient population and using a different T1 mapping sequence (i.e., MOLLI in our study vs. shortened MOLLI (ShMOLLI) in Liu et al. [13]). Although a MOLLI sequence requires a longer breath-hold to acquire as compared to ShMOLLI, MOLLI sequence is more readily available in most CMR centers, since ShMOLLI is only available in a single MRI vendor. Our results showed that splenic T1 mapping using MOLLI sequence could also be used in predicting SSO. In our study, we performed stress T1 mapping at 2.5 min of adenosine infusion rather than at 4 min; this finding suggested that splenic T1 maps could be performed earlier and still give good indication of stress adequacy.
Adenosine is commonly used for achieving myocardial stress in CMR. Since caffeine is an adenosine receptor antagonist, it will cause inadequate stress. In our center, all patients scheduled for stress CMR are advised to avoid caffeine intake within 24 hours prior to the CMR, otherwise CMR will be rescheduled. According to Kuijpers et al.’s [14] study, myocardial rest-stress T1-mapping was proposed as an alternative for patients with recent caffeine intake. However, only patients with normal stress perfusion CMR results were included in their study, because myocardial T1 values were known to be significantly affected by underlying myocardial disease. As a result, myocardial T1-mapping might only be applicable to patients without myocardial disease. In our center, stress CMR was mainly used for investigation of myocardial ischemia or infarction in suspected individuals, thus most patients in our study were clinically suspected to have abnormal stress perfusion CMR. For our patient population, we believe splenic T1 mapping would be more reliable compared with myocardial T1 mapping, as long as patients refrain from caffeine intake within 24 hours prior to the examination. Moreover, it will be difficult to assess whether myocardial T1 mapping can be used reliably prior to perfusion imaging, since whether a patient has abnormal stress perfusion CMR can only be interpreted retrospectively after imaging. If inadequate stress was identified retrospectively after the CMR, the only option would be to repeat the MRI examination later. In contrast, splenic T1-mapping could benefit patients regardless of the presence of myocardial pathologies. It also allows stress adequacy to be reliably assessed and adenosine dose to be optimized before acquisition of perfusion imaging. Therefore, it ensures optimal stress to be achieved, which is a prerequisite for stress perfusion CMR to prevent false negative results.
Our study had several limitations. First, like Liu et al. [13], we compared ΔT1spleen with SSO response, which is only a surrogate marker for adenosine stress adequacy. In the ideal situation, the gold standard for assessing stress-related coronary hemodynamics would be a real-time intravascular pressure-wire Doppler measurement. However, this gold standard is invasive, technically demanding and impractical in daily clinical practice. Therefore, SSO is the most widely used predictor of stress adequacy due to its feasibility in daily clinical practice. In addition, the use of SSO as a noninvasive surrogate marker has gained acceptance in major clinical studies as a marker of stress adequacy [5] and has recently been shown to correlate with MBF in myocardial perfusion positron emission tomography (PET) in a hybrid PET/CMR scanner [6]. Second, our splenic T1 findings were not correlated with catheter coronary angiography and fractional flow reserve to look for false negative results of stress CMR. Third, our study was a single center study with all our patients being scanned on 1.5 T scanners, but nonetheless showed that previous findings by Liu et al. [13] can be applied in our centre. Finally, the spleen was not visible for a small proportion of patients (4%). A dedicated imaging plane for splenic T1-mapping might be required in these patients at the expense of additional scanning time.
Future directions
Our validation study showed that ≥30 ms drop in T1spleen post-stress is a reproducible and acceptable marker of SSO response, and hence stress adequacy, before CMR first-pass perfusion imaging is acquired. With these findings, one can consider adding T1spleen as another marker of adequate stress. Assessment of splenic T1-mapping takes less than 1 minute while awaiting adenosine to achieve optimal effects, and therefore comes with minimal time penalty. It can practically guide up-titrations of adenosine dosage for optimization of stress response before contrast administration. As T1 mapping is increasingly available worldwide, this new approach could benefit patients by ensuring that adequate stress is achieved. Our study results have set the stage for further prospective studies to be performed, and to examine whether the use of T1spleen will translate to a reduction in false negative stress CMR scans.
In conclusion, we have validated splenic T1-mapping to be an accurate and reproducible technique to predict the occurrence of SSO before contrast administration, potentially allowing optimization of adenosine dosage for adequate stress. It can be feasibly included in stress CMR protocols to guide adenosine infusion without additional scanning time.
Conflicts of Interest:The authors have no potential conflicts of interest to disclose.
Author Contributions:
Conceptualization: Fiona Fong-ying Wan, Jonan Chun-yin Lee.
Data curation: Fiona Fong-ying Wan, Catherine Ming-mun Yeung, Pak-ki Yam, Jonan Chun-yin Lee.
Formal analysis: Fiona Fong-ying Wan, Jonan Chun-yin Lee.
Investigation: Pan-pan Ng.
Methodology: Boris Chun-kei Chow.
Project administration: Fiona Fong-ying Wan, Jonan Chun-yin Lee, Ming-yen Ng.
Resources: Fiona Fong-ying Wan, Jonan Chun-yin Lee, Kenneth Kai-yat Cheung, Ming-yen Ng.
Software: Jeanie Betsy Chiang.
Supervision: Fiona Fong-ying Wan, Jonan Chun-yin Lee.
Validation: Fiona Fong-ying Wan, Jonan Chun-yin Lee.
Visualization: Fiona Fong-ying Wan, Catherine Ming-mun Yeung, Pak-ki Yam, Jonan Chun-yin Lee.
Writing—original draft: Fiona Fong-ying Wan.
Writing—review & editing: Jonan Chun-yin Lee, Ming-yen Ng.
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
Acknowledgments
None
References
-
World Health Organization. Global health estimates 2016: deaths by cause, age, sex, by country and by region, 2000-2016. Geneva: World Health Organization; 2018.
-
-
Kidambi A, Sourbron S, Maredia N, Motwani M, Brown JM, Nixon J, et al. Factors associated with false-negative cardiovascular magnetic resonance perfusion studies: a clinical evaluation of magnetic resonance imaging in coronary artery disease (CE-MARC) substudy. J Magn Reson Imaging 2016;43:566–573.
-
-
Patriki D, von Felten E, Bakula A, Giannopoulos AA, Kamani CH, Schwyzer M, et al. Splenic switch-off as a predictor for coronary adenosine response: validation against 13N-ammonia during co-injection myocardial perfusion imaging on a hybrid PET/CMR scanner. J Cardiovasc Magn Reson 2021;23:3
-
MeSH Terms
Figures
1 / 8
Tables
1 / 1
