Two-center clinical validation and quantitative assessment of respiratory triggered retrospectively cardiac gated balanced-SSFP cine cardiovascular magnetic resonance imaging in adults

In patients with cardiovascular disease, ventricular volume, mass, and ejection fraction (EF) are powerful predictors of prognosis [1‐5]. In several studies of individuals with normal and abnormal ventricles, cardiovascular magnetic resonance (CMR) has been used to evaluate biventricular volume, mass, and EF with high accuracy and low inter- and intra-observer variability [6‐9]. For the accurate measurement of ventricular functional indices in CMR, it is necessary to have: a) sufficient and consistent blood-to-myocardium contrast throughout the cardiac cycle to reliably define endocardial contour, b) adequate spatio-temporal resolution to accurately capture end-diastole and end-systole, and c) immunity to cardiorespiratory motion and pulsatile blood flow. In routine clinical CMR practice, cine balanced steady-state free precession (bSSFP) is a sequence of choice for evaluation of ventricular function [10] as it offers the highest signal-to-noise ratio (SNR) per unit time of all other CMR imaging sequences along with a T₂/T₁-weighted image contrast [11]. In order to obtain diagnostic image quality, the bSSFP sequence mandates the shortest possible repetition time (TR) to minimize banding artifacts [11], high flip angle for adequate blood to myocardial contrast [12], and uninterrupted radiofrequency (RF) excitations to maintain the magnetization at steady-state (SS) to elude contrast variability [13] and artifacts [14, 15]. In addition, cine bSSFP acquisition employs retrospective cardiac gating to obtain the entire cardiac cycle along with prospective arrhythmia rejection. In clinical practice breath-holds (BH) are used to mitigate respiratory motion artifacts. Respiratory suspension for sufficient duration is not feasible in sedated patients and patients with impaired breath-holding capacity. Therefore, it would be of clinical interest to obtain cine bSSFP images without the constraint of breath holding while retaining the necessary features described above.

Approaches to combat artifacts introduced by respiratory motion in cine CMR imaging fall in three broad categories: (1) rapid imaging with data undersampling [16‐18]; (2) respiratory motion synchronized data acquisition using signals from an external respiratory bellows or RF navigator positioned over the lung-diaphragm interface [19‐23]; and (3) self-gated motion compensation methods [24‐26]. Combinations of data undersampling and self-gating are being explored for respiratory motion-state cine imaging [27]. Some of these approaches require contrast administration [18, 24, 28], or prospective cardiac gating [16, 17] and undersampling methods may require considerably longer reconstruction times and/or advanced hardware capabilities [16, 18, 24, 25, 27]. The dual RF navigator gated approach was limited to partial acquisition of the cardiac cycle [19]. In contrast, respiratory synchronized acquisition methods using bellows or diaphragmatic RF navigators permit retrospective cardiac gating and real-time reconstruction of bSSFP cine sequences, which are used in clinical practice without imposing any hardware or reconstruction burden [20‐23]. Furthermore, these respiratory motion synchronized techniques allow the flexibility to be seamlessly combined with undersampling strategies such as compressed sensing that may become clinically available in the future.

We have previously demonstrated respiratory synchronized bSSFP cine acquisitions using both respiratory bellows and diaphragmatic RF navigator, in a healthy cohort study [20]. These implementations require SS to be reached before the acquisition of cardiac gated cine data to ensure uniform signal and contrast throughout the cardiac cycle. The bellows based respiratory triggered (RT) implementation was validated in a pediatric population [22]. Similar approaches using respiratory bellows and diaphragmatic RF navigator for respiratory synchronized bSSFP cine acquisition have also been reported [21, 23]. However, these implementations use the available time between the respiratory trigger and the subsequent R-top for SS preparation. This can cause artifacts because the variable time interval may not be long enough to attain SS in certain respiratory cycles [21]. In this manuscript we describe a two-center clinical validation of a bellows based RT-bSSFP in an adult population with rigorous quantitative assessment of the temporal stability of the myocardial signal as well as blood-to-myocardial contrast.

The results from the two-center clinical study show that RT-bSSFP yielded diagnostic image quality cine images in all 60 patients encompassing a wide range of heart rates (43–116 bpm), respiratory rates (12–20 bpm), and ventricular indices (LV end-diastolic volume 84–397 mL, LV end-systolic volume 35–325 mL, LVEF 18–71%) with spatial, temporal, and contrast resolutions that were comparable to BH-bSSFP. To the best of our knowledge, this is the first study that has performed clinical validation of the free breathing cine bSSFP imaging in a large cohort of adult patients. The bias and limits of agreement between RT and BH sequences for LV and RV volumetric indices and LV mass were comparable to inter-observer variability in both RT and BH sequences. An interesting observation was that while the LV volumes were slightly overestimated (mean bias of  2–3 mL) with BH sequence compared to RT sequence, RV volumes were slightly underestimated (mean bias of 2–3 mL). Although such small differences in ventricular volumes may not be clinically significant, it may be worth investigating if the intra-thoracic pressure differences between respiratory suspended state and free breathing could be a contributing factor.

There are several key findings from the study that are worth noting. Firstly, conventional respiratory triggering methods for circumventing the breath-holding in bSSFP imaging that do not account for signal intensity modulation during the approach to SS can introduce significant artifacts as demonstrated in this study. Interestingly, even in BH bSSFP acquisitions, through-plane motion of the base toward the apex of the ventricle during systole introduces quantifiable temporal signal heterogeneity in the myocardium. Compared to other previously published respiratory triggered bSSFP cine CMR implementations [21, 23], RT sequence implementation [20, 22] mandates attainment of M_ss prior to cardiac gated data acquisition. Metrics to quantify the temporal stability of the signal demonstrated that the RT sequence attained SS in synchrony with the physiologic cycle that is comparable to BH bSSFP. Importantly, the proposed acquisition technique and image quality evaluation metric have the potential to be used more broadly in other clinical applications. Also, it is worth noting that RT-bSSFP cine sequence is inherently compatible with prospective arrhythmia rejection and retrospective cardiac gating that are typically used with BH techniques, allowing the acquisition of a complete cardiac cycle. Furthermore, the cardiorespiratory synchronization approach can be seamlessly combined with other techniques such as non-cartesian sampling, sparse sampling, and self-navigation.

Secondly, bSSFP cine acquisition is feasible with both respiratory triggering using bellows and respiratory triggering with slice tracking using diaphragmatic RF navigators [20]. One advantage of using diaphragmatic RF navigators is the ability to track slice location in real time from the navigator information. However, the relatively long duration (at least one RR interval) between the slice position measurement from RF navigators and the completion of cine data acquisition may diminish this potential benefit. Furthermore, the quality of RF navigators may be compromised in patients with significant iron overload [23]. In contrast to RF navigators, respiratory bellows do not provide a direct measure of the lung-diaphragmatic interface. However, respiratory bellows provide the ability to monitor respiratory motion continuously and independent of CMR data acquisition without incurring additional RF dose.

Thirdly, respiratory triggered cine bSSFP sequences have several strengths. In contrast to continuous multi-NSA sequences which are often used to obtain cine bSSFP images in sedated patients or those with compromised respiratory function [22], the RF duty cycle in these sequences remains at about 50%, mitigating the specific energy deposition concerns and permits data acquisition over a longer scan duration. Freedom from the breath-hold constraint further allows one to attain bSSFP cine images with higher spatial resolution, temporal resolution, or coverage. Therefore, this technique has the potential to broaden the scope of clinical applications for cine bSSFP sequences, such as assessing diastolic function by obtaining a high temporal resolution dataset.

Another advantage of the RT bSSFP cine sequence is that it would reduce the potential for errors in estimation of ventricular chamber volumes due to inconsistent breath-holds. Although CMR is considered the gold standard for the estimation of ventricular chamber volumes, inconsistencies in the level of the diaphragm across breath-holds for each slice can result in under and/or over sampling of the myocardium in the direction of the ventricular long axis. This too may lead to variability in calculated volumes and EF, especially sampling-dependent inclusion or exclusion of the basal slice in end-systole [30]. This problem would be significantly diminished with free-breathing acquisitions [19].

Although we made no attempt to include only patients with compromised respiratory function, our clinical results suggest that the reduction in image quality of RT bSSFP was only small compared to the image quality attainable with the BH bSSFP technique. Notably, in our previous study, the image quality obtained with the RT-SSFP technique was better than that obtained with the multiple signal averaging that is routinely used in sedated pediatric patients [22]. In this two-center study in adults, our findings with respect to LV mass and ventricular volumetric results obtained with the RT and BH sequences are in line with the inter-observer variability.

The current implementation of RT-SSFP cine sequence has two key limitations. First, image quality of RT-bSSFP, while good, was inferior to that of the BH-bSSFP sequence. In 10/60 subjects, RT-SSFP image quality scores were moderate (still diagnostic) and we speculate that this reduction in image quality may be partly because of inconsistent respiratory patterns, such as when a patient starts inspiration before data acquisition from the expiratory trigger has completed, or respiratory drift over the course of an examination. The extent of the image quality degradation is dependent on the location of the acquisition segment with respect to center of the k-space. The image quality of the RT sequence could be improved by imposing an additional prospective constraint for rejecting data that were acquired during such respiratory inconsistencies at the cost of increasing scan time. A second limitation of the RT-bSSFP sequence is that it takes on an average 55% more time than the conventional cine BH-bSSFP sequence. For a cardiac cine bSSFP technique that requires n shots for a complete data acquisition, the RT-bSSFP sequence takes the time required for n respiratory cycles, whereas the BH-bSSFP sequence takes the time required for n + 1 heartbeats (1 extra beat to drive to M_SS). This is because the RT sequence, in its current implementation, acquires data for a single cardiac cycle after each respiratory trigger. Often, the expiratory phase of the respiratory cycle is longer than one cardiac cycle and may permit data acquisition for several cardiac cycles, which would reduce the acquisition time of the RT sequence. Although this was not considered in this study, data acquisition over consecutive cardiac cycles during expiration with prospective constraint for rejecting the data are an important area of future investigation [31] as respiratory bellows permits continuous and simultaneous tracking of the respiratory motion.

In conclusion, we have shown that the RT sequence is a robust, free-breathing alternative to BH-bSSFP for evaluating ventricular function in patients with impaired breath-holding capacity and/or arrhythmia. The RT sequence yields cine images with contrast and spatio-temporal resolutions that are identical to those attained by using the BH sequence, albeit with a modest time penalty and a small reduction in image quality. We believe that the elimination of the BH constraint on a cine bSSFP sequence has the potential to improve the evaluation of ventricular function in subjects with compromised respiratory capacity with small reduction in image quality. Moreover, this would allow one to acquire high temporal resolution cine images necessary for the assessment of transient phenomena, such as isovolumic relaxation time, as well as the ability to acquire multi-phase, high-resolution anatomic images, such as coronary MR angiography.