Feedback of the diaphragm position during CMR has been shown to improve navigator efficiency in adults [
5,
6]. This study explored how the use of a feedback game (FG) affects navigator efficiency compared to traditional no-feedback (NF) acquisitions
in children. The results of the study showed that, compared to NF, using the FG resulted in 1) substantially improved navigator efficiency (from 33 to 58 %); 2) slightly improved SNR; 3) slightly higher mean heart rate; and 4) slightly lower global strain magnitudes. Importantly, these results were not affected by the use of an off-scanner training protocol, which suggests that lengthy, robust training (11 min in our protocol) does not need to be a part of the clinical/imaging workflow for this interface.
Navigator efficiency
Navigator efficiency was improved from 33 to 58 % by using a FG in children (Fig.
3a). This increase in navigator efficiency led to a 43 % reduction in the number of heartbeats required to complete a scan. Studies have shown that feedback of the diaphragm position during CMR results in a more reproducible breath-hold position [
5,
8,
21], which can lead to improved navigator efficiency. Previous CMR studies have reported that NF navigator efficiencies can vary from 20 to 45 % in adults [
4‐
7], and we found a comparable NF navigator efficiency of 33 % in children using a conservative dual-navigator acceptance strategy. Visual feedback of the diaphragm position has been shown to improve end-expiratory navigator efficiency from 45 to 56 % [
6] and from 42 to 71 % with the addition of supplemental oxygen [
5] leading to a 20 % and 41 % reduction in the number of required heartbeats, respectively. With the use of the FG, we found a slightly better improvement of navigator efficiency from 33 to 58 % in children without the use of supplemental oxygen. Average navigator efficiency was not correlated with age (Fig.
3b). Therefore, children ages eight and older should be able to utilize the FG to effectively improve navigator efficiency compared to conventional NF acquisitions.
Extensive off-scanner training using an MRI-simulator was not necessary to achieve the observed improvement in navigator efficiency using the FG. Instead, the subjects with minimal training immediately prior to data acquisition had equivalent efficiency as their extensively-trained counterparts. While this finding might suggest that the chest wall excursion-based training method was ineffective, it is more likely that the intuitive interface design was easy to learn and therefore the children did not require much training. Importantly, the two 30-beat practice scans provided some degree of training in both cases, which is intuitively necessary. Future efforts can optimize that practice time to provide the needed minimal training in the most efficient manner.
Heart rate
A potential negative finding of this study was the slight increase in heart rate observed with the use of the feedback game. To be clear, this difference did not represent an increase in heart rate variability—as evidenced by the comparable standard deviation values—but simply a higher baseline value. Such findings are not unprecedented, as a previous CMR study found a mean heart rate increase of 5 beats/min with use of navigator feedback in adults (compared to our 3 beats/min), and similarly no differences in heart rate variability between NF and navigator feedback [
5]. A likely reason for this difference is the longer breath-holds performed during the FG, which could have increased the heart rate, compared to relaxed breathing during NF. Another mechanism could be related to stimulation and adrenaline associated with playing the game, compared to the relaxed, passive state associated with NF.
The importance and implications of this potential heart rate difference likely depends on the imaging application. While it may mean very little for purely anatomic evaluations, functional measures, such as strains, may be affected by changing loading conditions and contractility [
22]. To counteract such effects, if undesirable, patients could be coached to relax when playing the game and to not be too competitive. The design of the game could be modified to enforce such behavior; for example, by programmatically requiring the participant to inhale/exhale after a fixed period of time, or instructing him/her to periodically take a series of relaxed breaths between cycles of breath-holding.
Clinical implications
Importantly, the equipment needed to utilize the FG is minimal and does not directly integrate into an imaging sequence; it connects externally to the scanner user interface. Due to the minimal equipment needed and non-invasive connection to the MRI scanner, we anticipate that the FG system can be easily adopted at research and clinical sites that perform CMR navigator gating, especially in children.
Since navigator efficiency can be increased from 33 to 58 %, leading to reduced acquisition times, use of the FG can help improve the clinical feasibility of advanced imaging techniques. While reducing the acquisition time would likely be the most common use of increased navigator efficiency from the FG, the saved time could be allocated to improve image spatial or temporal resolution [
5]. Importantly, pre-scan training was not necessary for navigator efficiency improvement with our system, so clinical and research sites would not have to invest in an MRI simulator environment or spend significant time training children. Navigator feedback has been shown to reduce acquisition time in adults [
5], thus, the use of the FG will likely work well in adults also.
Since we only acquired DENSE images for this study, the specific findings are only definitively relevant for DENSE. However, it is reasonable to expect that these findings are generalizable to many other CMR acquisitions that utilize a respiratory navigator. Possible exceptions include higher resolution applications, such as coronary MR angiography, which may be more sensitive to registration issues. Further study is needed to test this technique for these applications.
Comparison with previous work
A previous study presented a respiratory biofeedback game and continuously adaptive windowing strategy (CLAWS) to increase navigator efficiency of imaging the thoracic aorta. The authors reported an increase in efficiency in that study from 45 to 56 % in adults [
6], which represents a smaller magnitude of improvement (25 % vs. 11 %) but a similar end result (58 % vs. 56 %) compared to our study. Although the two studies are similar, there are distinct differences in design. Most notably, the previous study was in adults; whereas we exclusively focused on children, based on their limited ability to breath-hold and thus potentially greater need for respiratory navigated sequences. Additionally, the previous study modified their pulse sequence to allow acquisition of multiple navigator echoes, likely providing a smoother game experience. We did not modify our cine pulse sequence in our evaluation—we had a single navigator echo per cardiac cycle—in order to ensure more general clinical applicability. Collectively, these studies demonstrate the potential utility of user-friendly interfaces for improving efficiency and image quality of cardiovascular imaging sequences using a respiratory navigator in a broad array of patients.
Limitations
This study used a dual-navigator strategy when performing image acquisition. Dual-navigator strategies have stricter data acceptance criteria compared to previously used single-navigator strategies [
1], and, given the same imaging parameters, will likely result in lower navigator efficiencies. However, a previous study using a single-navigator strategy with navigator feedback reported similar navigator efficiency results compared to our study. Therefore, the use of the FG with a single-navigator strategy will likely have similar results to this study except that both NF and FG acquisitions may have improved navigator efficiency compared to a dual-navigator strategy.
The respiratory navigator gating sequence used in this study only measured the diaphragm position once per cardiac cycle. This low refresh rate can make fine control of the diaphragm position more challenging, especially for participants who may have lower heart rates. Increasing the number of navigator echoes per cardiac cycle could therefore improve performance, but such modifications may not be possible for all sequences, as is the case for DENSE. Furthermore, even with this limitation, we still found substantial improvement in navigator efficiency when using the FG compared to NF acquisitions.
Due to the randomization of the participants into the trained and untrained groups, there was no attempt to balance age between groups. Therefore, the average trained participant was about 2 years older than the average untrained participant. We found that there was no difference in FG navigator efficiency between trained and untrained participants. Even though there was an age difference between trained and untrained groups, there was no correlation between age and navigator efficiency with the FG (Fig.
3b); thus, the results of the study apply to all children aged eight to eighteen.
In order to accurately assess the NF navigator efficiency as it would be in the clinical setting, we did not want to influence the children’s natural breathing pattern. In particular, we did not want the breathing pattern performed during the FG acquisitions to influence the NF breathing pattern. Therefore, NF acquisitions were always performed before FG acquisitions. Since the order of NF and FG acquisitions was not randomized, this may have affected the results as participants may have become more comfortable as they spent more time in the MRI scanner. However, performing this randomization likely would have resulted in similar conclusions and we feel that it was important to accurately measure the navigator efficiency of the NF acquisitions.
Due to the potential for patient movement or erratic breathing patterns, we utilized a stoppage criterion to attempt to maintain a 30 min protocol length. We observed eleven cases which satisfied stoppage criterion and four cases of patient movement (one which also satisfied stoppage criterion). In these participants, we estimated navigator efficiency, SNR, and heart rate from fewer acquisitions than the remaining participants. However, since we used all of the data that we did acquire for each participant, the computed values are appropriate.
The two 30-heartbeat practice scans were not included in the computation and analysis of navigator efficiency for the FG technique. Their inclusion would only slightly decrease the reported gains in efficiency (for example, if we used the FG to acquire 300 heart beats of actual data, the reduction in scan time would change minimally from 43 to 37 % after accounting for the two practice scans); however, it must be noted that the selection of those practice parameters was arbitrary and not optimized. In reality, less training is likely required to familiarize the subject with the interface, so factoring this specific training design into the analysis is not critical.
We performed this study in children with no significant past medical history. While we did attempt to recruit from a broad clinical population using recruitment services at our Center for Clinical and Translational Science, the population we ultimately studied may not be entirely representative of a standard pediatric clinical population that would routinely undergo cardiac MRI. For example, approximately 25 % of patients with tetralogy of Fallot may have learning and behavioral difficulties [
23], which may impair their ability to benefit from the feedback game. It is therefore reasonable to expect that the true benefit of the feedback game in a standard clinical population will be smaller than what was measured in the current study, but still better than what can be expected without the use of feedback. Even if only half of the patients benefit to the extent shown in the current study, the overall navigator efficiency for the clinical population as a whole would still increase from 33 % efficiency to 46 % efficiency (a 38 % relative benefit). Future research will seek to evaluate this in further detail as we implement the feedback game during routine clinical workflows.