Table 1.
Summary of past studies exploring the benefits of MB acquisition and their main findings.
| Publication | |||||||
|---|---|---|---|---|---|---|---|
| Technique |
Number of subjects |
Rest/Task (~scan time) |
Main results | ||||
| Field strength | MB factor(s) | Voxel size (mm3) | TRs compared | ||||
| Feinberg et al. (2010) | |||||||
| SIR, MB | 3 | Rest (10 min) | Increased peak functional sensitivity. | ||||
| 3 T | 1, 2, 3 | 3 × 3 × 3 | 2.0s, 0.8s, 0.4s | ||||
| Koopmans et al. (2012) | |||||||
| MB | 6 | Rest (15 min) | Exquisite localization to grey matter. | ||||
| 7 T | 1, 4 | 1.5 × 1.5 × 1.6 | 7.4s, 1.8s | ||||
| (X.-H. Liao et al., 2013) | |||||||
| MB | 11 | Rest (6 min) | Validity of multiband rs-fMRI to reliably detect functional hubs. | ||||
| 3 T | 1, 4 | 3 × 3 × 3 | 2.5s, 0.6s | ||||
| Boubela et al. (2013) | |||||||
| SIR, MB | 10 | Rest (6 min) | Resting-state networks like the default-mode network in frequencies above 0.25 Hz. | ||||
| 3 T | 4 | 2.4 × 1.9 × 3.5 | 0.3s | ||||
| Griffanti et al. (2014) | |||||||
| MB | 76 | Rest (10 min) | With optimal cleaning procedures, functional connectivity results from accelerated data were statistically comparable or significantly better. | ||||
| 3 T | 1, 6 | 3 × 3 × 3 2 × 2 × 2 | 3.0s, 1.3s | ||||
| Tong et al. (2014) | |||||||
| MB | 5 | Rest (6 & 10 min) | Many voxels are highly correlated with pulsation regressors or its temporally shifted version. | ||||
| 3 T | 6 | 3 × 3 × 3 | 0.4s | ||||
| Tong et al. (2014b) | |||||||
| MB | 9 | Rest (6 min) | Spatial distributions of different physiological processes are distinct. | ||||
| 3 T | 6 | 3 × 3 × 3 | 0.4s | ||||
| Tong et al. (2014a) | |||||||
| MB | 7 | Rest (6 & 10 min) | Systemic oscillations pervade the BOLD signal; Temporal traces evolve as the blood propagates though the brain; They can be effectively extracted via a recursive procedure and used to derive the cerebral circulation map. | ||||
| 3 T | 6 | 3 × 3 × 3 | 0.4s | ||||
| Kalcher et al. (2014) | |||||||
| SIR, MB | 20 + 20 | Rest (6 min) | Correlations between resting-state signal fluctuations of distant brain regions even at high frequencies, which can be measured using low-TR fMRI. In the high-TR data, loss of specificity of measured fluctuations leads to lower sensitivity in detecting functional connectivity. | ||||
| 3 T | 1, 4 | 1.5 × 1.5 × 3 2.4 × 1.9 × 3.5 | 1.8s, 0.3s | ||||
| Olafsson et al. (2015) | |||||||
| ME, MB, 1.33-fold phase encode acceleration | 12 | Rest (10 min) | ME-ICA identifies significantly more BOLD-like components in the MESMS data as compared to data acquired with a conventional multi-echo single-slice acquisition. | ||||
| 3 T | 1, 3 | 3.7 × 3.7 × 4 | 2.6s, 0.87s | ||||
| Gohel & Biswal (2015) | |||||||
| MB | 21 | Rest (10 min) | Functional integration between brain regions at rest occurs over multiple frequency bands. | ||||
| 3 T | 4 | 3 × 3 × 3 | 0.6s | ||||
| (X. Liao et al., 2015) | |||||||
| MB | 11 | Rest (10 min) | Economical, efficient, and flexible characteristics of dynamic functional coordination in large-scale human brain networks during rest, and their relationship with underlying structural connectivity. | ||||
| 3 T | 4 | 3 × 3 × 3 | 0.6s | ||||
| Preibisch et al. (2015) | |||||||
| MB, 2-fold in-plane sensitivity encoding acceleration | 20 | Rest (7 min) | MB factor of 2 only causes negligible SNR decrease but reveals common RSN with increased sensitivity and stability. Further MB factor increase produced random artifacts that may affect interpretation of RSNs under common scanning conditions. | ||||
| 3 T | 1, 2, 3, 4 | 3 × 3 × 3 | 2.0s, 1.0s, 0.7s, 0.5s | ||||
| Smith-Collins et al. (2015) | |||||||
| MB | 21 | Rest (5 min) | Rapid rs-fMRI acquisition in neonates, and adoption of an extended frequency range for analysis allows identification of a substantial proportion of signal power residing above 0.2 Hz. | ||||
| 3 T | 1, 3 | 2.5 × 2.5 × 2.5 | 1.7s, 0.9s | ||||
| Kalcher et al. (2015) | |||||||
| MB | 15 | Rest (7 min) | Graph clustering based method for identifying venous voxels has a high specificity and additional advantages of being computed in the same voxel grid as the fMRI dataset itself and not needing any additional data beyond what is usually acquired in standard fMRI experiments. | ||||
| 3 T | 8 | 1.7 × 1.7 × 2 | 0.3s | ||||
| Thanh Vu et al. (2016) | |||||||
| MB, 2-3-fold in-plane phase encoding acceleration | 24 | Rest (15 min) | High resolution images acquired at 7 T provide increased functional contrast to noise ratios with significantly less partial volume effects and more distinct spatial features. | ||||
| 3 T vs 7 T | 3, 5, 8 | 0.9 × 0.9 × 0.9 1.2 × 1.2 × 1.2 1.5 × 1.5 × 1.5 1.6 × 1.6 × 1.6 2 × 2 × 2 | 0.7s, 1.3s 1.9s, 3.7 | ||||
| Reynaud, Jorge, Gruetter, Marques, & van der Zwaag (2017) | |||||||
| MB, 2D vs. 3D EPI | 8 | Rest (5 min) | After physiological noise correction, 2D- and 3D-accelerated sequences provide similar performances at high fields, both in terms of tSNR and resting state network identification and characterization. | ||||
| 7 T | 1, 6, 8 | 2 × 2 × 2 | 3.3s, 0.6s, 0.4s | ||||
| Cohen et al. (2017b) | |||||||
| MB | 10 | Rest (7 min) | Sensitivity and specificity increases and reproducibility either increases or does not change for the MB compared to the single band acquisitions. The MB scans also show improved grey matter/white matter contrast compared to the single band scans. The local functional connectivity density and global functional connectivity density patterns remain similar across MB and single band scans and confined predominantly to grey matter. A strong spatial correlation of functional connectivity density between MB and single band scans is observed indicating the two acquisitions provide similar information. | ||||
| 3 T | 8 | 2 × 2 × 2 3.5 × 3.5 × 3.5 |
0.8s, 2.0s | ||||
| Golestani et al. (2017) | |||||||
| MB | 12 | Rest (12 min) | Physiological noise characteristics differ between SMS-EPI and regular EPI, with cardiac pulsatility contributing more to noise in regular EPI data but low-frequency heart rate variability contributing more to SMS-EPI. Signficant slice-group bias was observed in the functional connectivity density maps derived from SMS-EPI data. Making appropriate corrections for physiological noise is likely more important for SMS-EPI than for regular EPI acquisitions. | ||||
| 3 T | 1, 3 | 3.4 × 3.4 × 4.6 | 0.3s | ||||
| Smitha et al. (2018) | |||||||
| MB+2-fold in plane acceleration | 9 | Rest (7.4 min) | Negligible differences between the conventional-rsfMRI and MB rsfMRI acquisitions on the computed graph theoretic measures. MB-rsfMRI may be used as a time reducing acquisition technique that enables mapping of functional connectivity with similar outcome as conventional rs-fMRI in healthy subjects. | ||||
| 3 T | 1, 3 | 3.3 × 3.3 × 3.2 | 3.0s, 0.9s | ||||
| Boubela et al. (2014) | |||||||
| MB | 10 | Rest (7 min) & Task | Fast scanning may help to identify and eliminate physiologic components, increasing tSNR and functional contrast. | ||||
| 3 T | 8 | 1.7 × 1.7 × 2 | 0.3s | ||||
| Boyacioğlu et al. (2015) | |||||||
| ME, MB, 3-fold in plane GRAPPA | 11 | Rest (5 min) & Task | After noise correction, the detection of rs-networks improves with more non-artefactual independent components being observed. Additional activation clusters for task data are discovered for MBME data (increased sensitivity) whereas existing rs-networks become more localized (improved spatial specificity). | ||||
| 7 T | 1, 3 | 3.5 × 3.5 × 3.5 | 2.2s, 0.7s | ||||
| Shah et al. (2016) | |||||||
| SIR, MB | 476 | Rest (15 min) & Task | Longer scan times are needed to acquire data on single subjects for information on connections between specific ROIs. Longer scans may be facilitated by acquisition during task paradigms, which will systematically affect functional connectivity but may preserve individual differences in connectivity on top of task modulations. | ||||
| 3 T | Not specified | 2 × 2 × 2 | 0.7s | ||||
| Demetriou et al. (2018) | |||||||
| MB, in plane acceleration factor = 2 | 10 + 14 | Rest (6 min) & Task | Strong benefits of the multiband protocols on results derived from resting-state data, but more varied effects on results from the task paradigms. Multiband protocols were superior when Multi-Voxel Pattern Analysis was used to interrogate the faces/places data, but showed less benefit in conventional General Linear Model analyses of the same data. In general, ROI-derived measures of statistical effects benefitted only modestly from higher sampling resolution. | ||||
| 3 T | 1, 2, 3, 4, 6 | 3 × 3 × 3 | 2.0s, 1.0s, 0.7s, 0.5s, 0.3s | ||||
| Moeller et al. (2010) | |||||||
| MB, under sampling factor of 4 in PE direction | 3 | Task | Task/stimulus-induced signal changes and temporal signal behavior under basal conditions were comparable for multiband and standard single-band excitation and longer pulse repetition times. | ||||
| 7 T | 4 | 2 × 2 × 2 1 × 1 × 2 |
1.2s, 1.5s | ||||
| Boyacioğlu et al. (2014) | |||||||
| MB (GE vs SE), 3-fold in-plane acceleration, PINS | 6 | Task | GE-EPI shows higher efficiency and higher CNR in most brain areas; GE EPI was able to detect robust activation near air/tissue interfaces such due to reduced intra-voxel dephasing because of the thin slices used and high in-plane resolution. | ||||
| 7 T | 3 | 1.5 × 1.5 × 1.3 | 1.4s, 2.0s | ||||
| Chu & Noll (2016) | |||||||
| MB, coil compression | 5 | Task | Method to compress and reconstruct concentric ring SMS data improves preservation of functional activation over standard coil compression methods. | ||||
| 3 T | 1, 3 | Slice thickness 3 | 2.0s, 0.6s | ||||
| De Martino et al. (2015) | |||||||
| MB | 6 | Task | Reducing the length of the scanner noise results in stronger functional responses. | ||||
| 3 T | 1, 2 | 2 × 2 × 2 | 3.0s | ||||
| (L. Chen et al., 2015) | |||||||
| SIR, MB | 7 | Task | Low acceleration factors (N ≤ 6), setting SIR = 1 and varying MB alone yielded the best results in all evaluation metrics, while at acceleration N = 8 the results were mixed using both S = 1 and S = 2 sequences. | ||||
| 3 T | 1, 2, 4, 6, 8, 10, 12, 14, 16 | 2.5 × 2.5 × 3 | 4.0s–0.2s | ||||
| Sahib et al. (2016) | |||||||
| MB | 15 | Task | Colored noise in event-related fMRI obtained at short TRs originates mainly from neural sources and calls for more sophisticated correction of serial autocorrelations which cannot be achieved with standard methods relying on AR(1)+w models with globally fixed AR coefficients. | ||||
| 3 T | 1, 2, 4, 5, 8, 10 | 3 × 3 × 3 | 2.6s–0.3s | ||||
| Todd et al. (2016) | |||||||
| MB, 2-fold in-plane GRAPPA acceleration | 10 | Task | Imaging protocols using an acceleration factor of MB 2 × GRAPPA 2 can be confidently used for high-resolution whole-brain imaging to improve BOLD sensitivity with very low probability for false-positive activation due to slice leakage. Imaging protocols using higher acceleration factors (MB 3 or MB 4 × GRAPPA 2) can likely provide even greater gains in sensitivity but should be carefully optimized to minimize the possibility of false activations. | ||||
| 3 T | 1, 2, 4, 6 | 1.5 × 1.5 × 1.5 | 6.6s, 3.3s, 1.6s, 1.1s | ||||
| Kim et al. (2016) | |||||||
| ME, MB, Thin-slice summation | 10 | Task | The SMSME-thin imaging technique enhanced the temporal-signal-to-noise ratio and functional activation at high susceptibility regions of the brain. | ||||
| 3 T | 5 | Slice thickness 4 vs. 1 | 2.5s | ||||
| Vu et al. (2016) | |||||||
| MB | 4 | Task | Substantial word timing information can be extracted using fast TRs, with diminishing benefits beyond TRs of 1000 ms. | ||||
| 3 T & 7 T | 6, 7 | Slice thickness 3 at 3 T and 2.5 at 7 T | 0.5s | ||||
| Bollmann et al. (2018) | |||||||
| MB, 2 or 3-fold GRAPPA acceleration | 10 | Task | Commonly used noise models, such as the AR(1) model, are inadequate for modelling serial correlations in fMRI using sub-second TRs. Rather, physiological noise modelling in combination with advanced pre-whitening schemes enable valid inference in single-subject analysis using fast fMRI sequences. | ||||
| 7 T | 4, 3 | 2.5 × 2.5 × 2.5 1.3 × 1.3 × 1.3 |
0.6s, 2.0s | ||||
| Todd et al. (2017) | |||||||
| MB, 12% in phase acceleration | 10 | Task | Lower g-factor noise area of V1 shows significant improvements at higher SMS factors; the moderate-level g-factor noise area of the para-hippocampal place area shows only a trend of improvement; and the high g-factor noise area of the ventral-medial pre-frontal cortex shows a trend of declining t-scores at higher SMS factors. This spatial variability suggests that the optimal SMS factor for fMRI studies is region dependent. SMS accelerations of 4x (conservative) to 8x (aggressive) for most studies and a more conservative acceleration of 2x for studies interested in anterior midline regions is recommended. | ||||
| 3 T | 1, 2, 4, 8 | 3 × 3 × 2.5 | 2.8s, 1.4s, 0.7s, 0.4s | ||||
| Kiss et al. (2018) | |||||||
| MB | 21 | Task | ~4 min of the scan time with 1 Hz (TR = 1000 ms) sampling rate and ~2 min scanning at ~ 2.5 Hz (TR = 410 ms) sampling rate provide similar localization sensitivity and selectivity to that obtained with 11-min session at conventional, 0.5 Hz (TR = 2000 ms) sampling rate. | ||||
| 3 T | 4, 6 | 3 × 3 × 3 | 2.0s, 1.0s, 0.4s | ||||
| McDowell & Carmichael (2018) | |||||||
| MB | 10 | Task | Modest TR reductions (to 1000 ± 200 ms) optimally improved event related fMRI performance independent of design frequency. Autoregressive models with a local as opposed to global fit performed better, while low order autoregressive models were sufficient at the optimal TR. | ||||
| 3 T | 1, 2, 3, 4 | 2.5 × 2.5 × 2.5 | 2.5s, 1.2s, 0.8s, 0.4s | ||||
| Sahib et al. (2018) | |||||||
| MB | 15 | Task | At a conventional TR of 2.6 s, Functional Connectivity Degree (FCD) values were marginal compared to FCD values using sub-seconds TRs achievable with multiband (MB) fMRI. | ||||
| 3 T | 1, 2, 4, 8 | 3 × 3 × 3 | 2.6s, 1.3s, 0.7s, 0.3s | ||||
| Su et al. (2018) | |||||||
| MB, 2-fold GRAPPA acceleration | 20 | Task | Accelerated gradient echo (GRE) sequence combining simultaneous multislice excitation (SMS) with echo-shifting technique for high spatial resolution BOLD fMRI has potential for high spatial resolution fMRI at ultra-high field because of its sufficient BOLD sensitivity as well as improved acquisition speed over conventional GRE-based techniques. | ||||
| 7 T | 5, 1 | 1 × 1 × 2.5 | 3s | ||||
| Risk et al. (2018) | |||||||
| MB | 98 | Task | When data were smoothed, we found evidence of slice leakage in some, but not all, subjects. We also found evidence of SMS noise amplification in unprocessed task and processed resting-state HCP data. | ||||
| 3 T | 8 | 2 × 2 × 2 | 0.7s | ||||
| Corbin et al. (2018) | |||||||
| MB+12% phase-over sampling | 10 | Task | The “FAST” model implemented in SPM is used with a well-controlled number of parameters, it can successfully prewhiten 80% of grey matter voxels even with volume repetition times as short as 0.35 s. Temporal signal-to-noise ratio can be augmented to account for the temporal correlations in the time series. | ||||
| 3 T | 1, 2, 4, 8 | 3 × 3 × 2.5 | 2.8s, 1.4s, 0.7s, 0.35s | ||||
| Current Study | |||||||
| MB, 1.5 or 2-fold in-plane phase encoding acceleration | 23 | Task & Pseudo resting state | (i) Sequences with different acceleration factors are able to detect the brain networks involved in task processing. (ii) Group level t-statistics improves with faster scanning. (iii) However it cannot compensate for the effects of larger voxel sizes, sample sizes or total scan duration. (iv) This is true for both task and resting-state analyses. | ||||
| 3 T | 1, 2, 4 | 2.7 × 2.7 × 2.7 3 × 3 × 3.3 |
2.4s, 2.0s, 1.2s, 0.7s, 0.6s | ||||