- Email: [email protected]
Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for human fMRI
Author’s Accepted Manuscript Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for human fMRI An T. Vu, Alex Beckett, Kawin Setsompop, David A. Feinberg www.elsevier.com
PII: DOI: Reference:
S1053-8119(17)30105-2 http://dx.doi.org/10.1016/j.neuroimage.2017.02.001 YNIMG13786
To appear in: NeuroImage Received date: 16 September 2016 Revised date: 19 January 2017 Accepted date: 1 February 2017 Cite this article as: An T. Vu, Alex Beckett, Kawin Setsompop and David A. Feinberg, Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for human fMRI, NeuroImage, http://dx.doi.org/10.1016/j.neuroimage.2017.02.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDERSMS) for human fMRI An T. Vu1,2,3*, Alex Beckett2,3, Kawin Setsompop4, David A. Feinberg2,3 1
San Francisco VA Health Care System, Center for Imaging of Neurodegenerative Diseases, San Francisco, CA 2
University of California, Berkeley, Berkeley, CA
3
Advanced MRI Technologies, Sebastopol, CA
4
Martinos Center for Biomedical Imaging, Massachusetts General Hospital Charlestown, MA *
To whom correspondence should be addressed: Center for Imaging of Neurodegenerative Diseases, 4150 Clement Street, Building 13, (114M Mail Stop), San Francisco, CA 94121. Tel.: +408 636 8375. [email protected] Abstract High isotropic resolution fMRI is challenging primarily due to long repetition times (TR) and insufficient SNR, especially at lower field strengths. Recently, Simultaneous MultiSlice (SMS) imaging with blipped-CAIPI has substantially reduced scan time and improved SNR efficiency of fMRI. Similarly, super-resolution techniques utilizing subvoxel spatial shifts in the slice direction have increased both resolution and SNR efficiency. Here we demonstrate the synergistic combination of SLIce Dithered Enhanced Resolution (SLIDER) and SMS for high-resolution, high-SNR whole brain fMRI in comparison to standard resolution fMRI data as well as high-resolution data. With SLIDER-SMS, high spatial frequency information is recovered (unaliased) even in absence of super-resolution deblurring algorithms. Additionally we find that BOLD CNR (as measured by t-value in a visual checkerboard paradigm) is improved by as much as 100% relative to traditionally acquired high- resolution data. Using this gain in CNR, we are able to obtain unprecedented nominally isotropic resolutions at 3T (0.66 mm) and 7T (0.45 mm). Keywords: 3T; 7T; High-resolution; fMRI; SLIDER; super resolution
Abbreviations HR, high resolution (isotropic); HRb, HR, blurred along the slice axis in post processing using a 2-slice sliding window to mimic SLIDER factor 2 acquisition.; HRb_db, HRb with the SLIDER deblurring step; SLIDER, slice dithered enhanced resolution technique (factor 2 in this study); SLIDER-XD, SLIDER without the deblurring step.
Introduction: Advancements in functional magnetic resonance imaging (fMRI) have pushed the limits of achievable spatial resolution enabling sub-millimeter imaging of human brain function down to the columnar level (Yacoub et al., 2007, 2008; De Martino et al., 2015; Nasr et al., 2016; Zimmermann et al., 2011) and laminar level (De Martino et al., 2015; Harel et al., 2006; Muckli et al., 2015; Olman et al., 2012; Polimeni et al., 2010; Zhao et al., 2006; Zimmermann et al., 2011). These studies were all performed at ultra-high field (7T) since sub-millimeter fMRI is generally regarded as not feasible at lower field strengths due to SNR and BOLD CNR constraints. However, given that 7T scanners are relatively few in the human research setting, there is strong motivation to further develop techniques to enable sub-millimeter functional imaging at lower, more readily available field strengths. Of course such developments would also be useful at ultra-high field strengths as well given that the field has yet to achieve routine human fMRI at the spatial scale of more invasive techniques such as optical imaging (Lu et al., 2010).
Super-resolution techniques utilizing sub-voxel spatial shifts in the slice direction
(Greenspan et al., 2002; Greenspan 2008; Setsompop et al., 2015) have shown great promise for increasing both resolution and SNR efficiency but are not without their challenges. Super-resolution techniques like SLIce Dithered Enhanced Resolution (SLIDER; Setsompop et al., 2015) improve resolution in two ways. First, SLIDER increases the Nyquist sampling frequency by acquiring S sets of high SNR, lowresolution, thick slice data (e.g. 1x1xSth mm3) with Sth/S mm shifts between each set along the slice direction (where S is the SLIDER factor and Sth is the thickness of the excited slice). Because slices are spatially encoded, in contrast to the frequency and phase encoded dimensions (Mayer and Vrscay 2007), this form of oversampling resolves and unaliases additional high spatial frequency information that is otherwise lost when using traditional inter-slice spacing equal to the slice thickness. However, this high spatial frequency information is dampened relative to lower spatial frequencies due to the point spread function (PSF) of the thicker slices used. Thus, the second way SLIDER improves resolution is by applying a deblurring algorithm along the slice axis. Given the slice profile (e.g. PSF) is known, the deblurring algorithm can estimate what the underlying data would be had a thinner slice, of thickness Sth/S mm, been used.
Because SLIDER acquires S sets of low-resolution, anisotropic thick slice data to more finely sample the image volume along the slice axis, the total volume acquisition time is S times that of the original low-resolution data. Analogous to conventional thin slice, isotropic acquisitions, imaging volume times can be very long (>> 3000 seconds) given it can take well over 100 slices to cover the brain at sub-millimeter resolution. This is
undesirable due to the relatively low SNR efficiency (Feinberg et al., 2010; Smith et al., 2013; Xu et al., 2013) and loss of high temporal frequency BOLD information (Chang et al., 2013; Chen et al., 2015; Vu et al., 2016). Fortunately, recent advancements in Simultaneous Multi-Slice (SMS) imaging (Larkman et al., 2001; Moeller et al., 2010) with blipped-CAIPI (Setsompop et al., 2012), in combination with in-plane undersampling techniques (Sodickson and Manning, 1997; Pruessmann et al., 1999; Griswold et al., 2002; Griswold et al., 2006; Talagala et al., 2013; Uğurbil et al., 2013; Polimeni et al., 2016) have substantially improved scan time and SNR efficiency (Ugurbil et al., 2013; Vu et al., 2015; Vu et al., In Press). Thus the combination of SLIDER and SMS should provide suitable temporal resolution and SNR efficiency for mesoscale imaging at both 3T and 7T.
Besides long imaging volume times, another common concern regarding super-resolution techniques is the residual blurring and noise amplification resulting from post-acquisition deblurring (Lam, 2003; Otazo et al., 2009). In the absence of noise, the deblurring component of super-resolution techniques can perfectly reconstruct high-resolution images from multiple low-resolution images. However, in practice, noise levels in the images can be quite significant leading to high frequency noise amplification and banding artifacts. To mitigate the effect of noise, prior studies have resorted to regularization of the deblurring algorithm, which effectively trades noise amplification for residual blurring. How this tradeoff manifests in the context of SLIDER fMRI and subsequently generated functional activation maps remains unclear. In some SLIDER fMRI applications, the increase in resolved high spatial frequency information may be
sufficient without deblurring and may warrant forgoing the post-acquisition deblurring of individual image volumes to achieve maximal BOLD CNR. More specifically, the benefits of deblurring may be limited in the context of fMRI, particularly for highresolution sub-mm acquisitions, given that the vascular PSF is estimated to be on the order of ~1.5-4 mm (Engel et al., 1997; Parkes et al., 2005; Shmuel et al., 2007) and thus would dominate blurring effects due to SLIDER. Fortunately, contrast subtraction techniques ubiquitous to analysis of fMRI data are known to serve as specificity or PSF enhancing operations (Yacoub et al., 2007, 2008; Muckli et al., 2015). This concept is analogous to super resolution techniques used in optical imaging (e.g. spectral precision distance microscopy (Lemmer et al., 2008)) where detailed structural information can be obtained far beyond the diffraction limit by selectively activating spatially sparse subsets of objects at any one time and precisely locating the positions of these objects. Similarly in fMRI, individual columns or layers can be sparsely activated in space at different time points – allowing for their precise location and organizational structure to be determined even in the presence of relatively broad PSFs.
Here, we evaluated the complimentary combination of slice dithering and SMS for highresolution whole brain fMRI. Specifically, we compared SLIDER-SMS to SMS alone at low-resolution, thick slice acquisition (1.25x1.25x2.5 mm3) as well as to thin slice, high isotropic resolution acquisition (1.25x1.25x1.25 mm3) using the following metrics: data smoothness, k-space energy at various spatial frequency bins (along the slice axis), tSNR, and BOLD CNR. The effect of the deblurring post-processing step with various levels of regularization was also investigated and revealed that SLIDER without deblurring (which
we refer to as SLIDER-XD) may actually be preferred for optimal BOLD CNR while still retaining significantly more high-spatial frequency information than conventional lowresolution, anisotropic acquisitions. Using these gains in CNR, we show it is feasible to perform fMRI experiments using 0.65 mm isotropic nominal resolution data acquired at 3T and 0.45 mm data acquired at 7T.
Methods Data were acquired from four healthy subjects on a Siemens 3T Trio using the standard 32 ch head coil. During the 2D gradient echo EPI fMRI scans, subjects viewed three 96 s runs of flashing checkerboard stimulus (30 sec period) for each scan protocol (SMS-5 and SLIDER-2 SMS-5). Imaging parameters were: 1.25 mm isotropic (nominal; 2.5 mm excitation thickness for SLIDER=2); FOV = 210x210x137.5 mm3; PF = 6/8; TE = 45 ms; TR = 3000 ms (1500 ms per dithered volume); Flip angle = 84° (72° for SLIDER=2); PE direction= AP; echo spacing = 0.88 ms; axial oblique slices; SMS=5; FOV/3 shift with sliceGrappa reconstruction (Setsompop et al., 2012); and no in-plane under sampling.
Figure 1 left shows the traditional interleaved SLIDER acquisition scheme where one half of a slice has an effective TR of twice that of the other half. Although this acquisition scheme worked well in prior studies where the repetition times were long (TR>>T1; which mitigates T1 relaxation and spin history effects) it is particularly problematic for fMRI applications where shorter TRs are desirable. One early SLIDER fMRI study (Peeters et al., 2004) took advantage of the shorter T1 at 1.5T and a longer TR (>3000 ms) but was only able to acquire 2 mm isotropic data, which is now considered fairly
standard. Furthermore, results of this study were unconvincing and difficult to interpret given the 3 mm smoothing kernel applied. In our study presented here, we mitigated the T1 relaxation differences between sub-slices by using an ascending slice order (Figure 1 right) instead of a long TR with interleaving slices. This reduced the difference in effective TR between sub-slices to just ~100 ms (the time of a single slice acquisition). In this case, the effective TR for all sub-slices is then approximately 1/S times the total volume imaging time; enabling further SNR optimization through the use of the Ernst flip angle.
To assess whether or not gains in SNR and CNR were simply due to the increased PSF of SLIDER along the slice axis, as a control, we generated a “High res blurred” (HRb) condition by taking the SMS-5 “High res” (HR) data (Figure 1 right; ascending slice order) and performing a 2-slice moving-window average (mimicking the SLIDER acquisition in post-processing). For deblurring/reconstruction, the analytical solution was based on the excited slice profiles. More specifically, the Toeplitz matrix was used as the forward model (T) and Tickhnov regularization was used to calculate the inverse model (Tinv) with λ ranging from 0 to 50% of the largest eigenvalue of T. In Matlab: Tinv = (V*E/(E*E+eye(size(E))*max(diag(E))*λ)*U'); where [U E V] = svd(T). Note larger λ’s result in greater residual blurring. For simplicity, the Toeplitz matrix used here assumed perfect, rectangular slice profiles. Simulations revealed that this assumption was reasonable (not shown) and that the effect of slice crosstalk due to imperfect slice profiles with our ascending SLIDER slice order resulted in ~5% narrower PSF in all but the first acquired slice. Deblurring using a Toeplitz matrix accounting for the imperfect slice
profiles in this case gave qualitatively the same results albeit with slightly worse tSNR and BOLD CNR values.
Imaging parameters for the 0.65 mm isotropic nominal 3T data were: SLIDER=2; FOV = 120x120x39 mm; PF = 6/8; TE = 45 ms; TR = 4000 ms (2000 ms per dithered volume); Flip angle = 80°; PE direction= HF; echo spacing = 1.52 ms; coronal slices; SMS=2; GRAPPA=2; FOV/3 shift. This high-resolution protocol was used for two 9 minute scans of ocular dominance column (ODC) mapping in primary visual cortex using commercially available red/blue anaglyphs and the stimulus paradigm denoted in (Cheng et al., 2001; repetitions of 6 s blanks with 12 s left or right eye stimulation). A GLM analysis was used to estimate responses to each eye, and a statistical map was generated by comparing responses to stimulation for each eye, and thresholded by the response to stimulation by either/both eyes (F,p<0.001).
An additional pilot dataset was acquired on a Siemens 7T MAGNETOM scanner with standard 70 mT/m body gradients and a custom 8 ch surface coil (Virtumed, LLC) with 4 cm loop diameter (SNR gains described in Beckett et al., 2016). Imaging parameters for the 7T data were: 0.45 mm isotropic (nominal; 0.9 mm excitation thickness for SLIDER=2); FOV = 90x81x23.4 mm3; PF = 5/8; TE = 27.8 ms; TR = 3000 ms (1500 ms per dithered volume); Flip angle = 80° (67° for SLIDER=2); PE direction= HF; coronal slices; SMS=2; GRAPPA=2. For this experiment, one subject viewed 3 minute runs of a flickering (8 Hz) checkerboard paradigm presented in alternating blocks of 12 s on and 12 s off.
All experiments used a high-resolution projector (Avotec, Inc.) to display visual stimuli. In order to preserve resolution properties across acquisition conditions, smoothing and motion correction steps were not performed. Head motion was minimized by recruiting expert subjects with informed consent, by keeping individual scan times short and by padding subjects’ heads well inside the head coils. Motion traces (provided by MrTools; Stanford, CA) confirmed that subject motion was within 0.5 mm across the entire 18 min ODC dataset. It is noted that future studies pursuing sub-mm imaging, regardless of subject population, would benefit from improved head (micro)-motion correction strategies (Yan et al 2013) as well as head restraint technologies. The 3T experimental protocols were approved by the Committee for the Protection of Human Subjects at UC Berkeley. The 7T protocols were approved by and the UC San Francisco and the Veterans Administrations Committees on Human Research.
Results & Discussion: Figure 2 shows the coronal cross section (of the 1.25 mm axially acquired slices) of a representative subject’s low-resolution (thick slice), HR (thin slice), HRb, SLIDER-XD, and SLIDER data deblurred with various λ values. Since HRb was blurred in post processing, the original HR image is recovered perfectly without regularization (λ=0). However, as with most super-resolution techniques in the presence of noise, deblurring SLIDER without regularization results in high spatial frequency noise amplification artifacts. SLIDER achieves best results with modest regularization (λ~0.1). Note that
even without deblurring, SLIDER-XD provides additional detail due to unaliasing of high spatial frequency information (for additional examples see: Figures S1-S2). Figure 3 shows BOLD CNR (t-values) for the corresponding Figure 2 datasets. SLIDER-XD yields substantially stronger BOLD CNR than both HR and HRb while SLIDER (with deblurring) has more spatial detail than the low-resolution data with similar BOLD CNR as HRb.
The amount of slice blur as a function of λ is plotted in Figure 4 left (quantified as the spatial correlation between the time averaged brain volume and the same volume shifted one slice down). The black dashed line denotes the λ (~0.08) where the effective SLIDER slice resolution matches that of HR. However, note that this is not true resolution as this is simply where residual blurring cancels out the high frequency noise artifacts. The true resolution including residual blurring but without noise artifacts can be determined from the cyan curve depicting the regularized deblurring of HRb data (which corresponds to ~1.75 mm at λ=0.08, using linear approximation between the HR (1.25 mm) and HRb (2.5 mm) curve values). Error bars are SEM across subjects. Mean t-values were calculated as the average t-value across an ROI of voxels activated (p<0.05; uncorrected) across all acquisition conditions. These values are plotted as a function of λ in Figure 4 right. HRb has a square-root of two increase over HR while SLIDER with λ~0.08 results in t-values at similar levels to HRb. This square-root of S gain is consistent with analogous Hadamard encoded techniques (Saritas et al 2014) as well as generalized SLIDER techniques (Setsompop et al 2016); which can be thought of as 3D or volume encoded imaging. Importantly, without deblurring, SLIDER-XD results in over double
the t-values as HR. This reflects not only the improved SNR of SLIDER (i.e. thick slice acquisition) but also the superiority of blurring through acquisition (as opposed to in post-processing) given that signal increases but not thermal noise in this case. Importantly, this also reflects SLIDER’s improved sensitivity to BOLD susceptibility changes (Lai and Glover 1998) given that tSNR improvement only accounts for about half of our observed gain in BOLD CNR (see Figure S3). Given these results and to reduce confounds of residual blurring and high-frequency artifacts, the subsequent analysis focuses on the use of SLIDER-XD.
To determine whether the higher slice blur values of SLIDER-XD depicted in Figure 4 reflect enhanced low spatial frequency information (as opposed to reduced high spatial frequency information), we also calculated the tSNR of the k-space time series (Figure 5). The middle row of Figure 5 shows the k-space tSNR normalized to each dataset’s DC value at the center. While it is clear that SLIDER-XD acquisition results in reduced high spatial frequency energy (relative to low spatial frequencies), the amount is certainly more than zero which is the amount in the non-dithered, low-resolution (1.25x1.25x2.5 mm3) data. When viewing all the acquisition conditions at the same scale (Figure 5 bottom row), it can be appreciated that the tSNR in lower k-space regions of SLIDERXD are higher than that of the HR data while the high k-space regions of SLIDER-XD and HR data are comparable.
To quantify these results, we calculated the average k-space tSNR in the lowest 50%, middle 25%, and highest 25% slice-axis k-space frequency regions (normalized to the
lowest 50% k-space region of the HR data; Figure 6). Notably, SLIDER-XD has significantly higher k-space tSNR (p<0.05, paired-T(2)) in both low and mid k-space frequencies and similar high k-space tSNR compared to HR and HRb (Figure 6 right). This shows that the SNR benefits of thick slice excitation in SLIDER-XD can offset the relative dampening of higher spatial frequencies. This is consistent with prior literature demonstrating that with enough CNR, even methods with moderate amounts of blurring (e.g. in the PE direction due to T2* decay) can yield maps of columnar level structure (Yacoub et al., 2007, 2008). However, it should be noted that, analogous to post acquisition smoothing kernels (White et al., 2001; Geissler et al., 2005), for optimal detectability of features of interest without deblurring, the SLIDER factor should provide slices thick enough to cover functionally homologous anatomic regions, yet thin enough not to partial volume together functionally distinct regions.
Utilizing these gains in BOLD CNR, we are able to obtain unprecedented nominally isotropic resolutions at 3T (0.65 mm) and 7T (0.45 mm). Figure 7 left shows preliminary results from one subject using the 0.65 mm SLIDER-XD acquisition at 3T (for additional SLIDER-XD ODC data, including cross-session reproducibility, see Feinberg et al In Review). The axial cross section of coronal acquired EPI data is shown at top, variance explained by task regressors (F-value) is shown at middle, and the eye preference map is shown at the bottom (orange for left eye, blue for right eye, purple for no eye preference). As expected, these maps contain substantially more spatial detail compared to the eye preference maps generated from the individual, low slice resolution (i.e. no SLIDER), dithers of the SLIDER-XD dataset; Figure S4). Figure 7 right shows the results from the
same subject using the 0.45 mm SLIDER-XD acquisition at 7T in a simple visual checkerboard paradigm. Thresholded t-values are shown overlaid onto the axial cross section of a T2 weighted anatomical image. Impressively, the 3 min SLIDER-XD acquisition results in BOLD CNR equivalent to four averages (12 min) of the nonSLIDER data with comparable spatial detail.
Importantly, SLIDER reminds us of the fact that inter-slice spacing (i.e. sampling frequency) can be optimized independently of slice thickness (i.e. PSF). For example, in a simulated case of imaging human ODCs which are approximately 1 mm per eye and thus of 2 mm period (Adams et al 2007; Yacoub et al 2007), Figure 8 left shows that a slice thickness of 1 mm is optimal for maximum CNR between columns. However, Figure 8 right shows that use of conventional slice spacing equal to the slice thickness, in this case, results in sampling exactly at the Nyquist rate (1 sample per mm) which is prone to head position dependent aliasing, poor reproducibility and/or poor structural fidelity. Increasing sampling to 1.25 times the Nyquist rate reduces head position dependent aliasing but the problem of different effective TRs within sub-slices becomes problematic again. By using SLIDER-XD and thus increasing sampling to twice the Nyquist rate (inter-slice spacing of 0.5 mm), head position dependent aliasing is eliminated while effective TRs within sub-slices are homogenous such that the Ernst flip angle can be used for optimal SNR (Figure 1 right). Note that for SLIDER-XD, higher SLIDER factors (>2) will not provide additional spatial resolution given that most aliasing in the slice direction is already eliminated at SLIDER factor 2. Future studies attempting higher SLIDER factors for fMRI should pay special attention to SNR
tradeoffs given the shorter effective relaxation times, the increase in instabilities due to variation in effective TR across sub-slices and the increase in susceptibility dropout due to the thickness of the slices excited.
In a field where the norm is to set inter-slice spacing equal to slice thickness, we find that this practice results in unnecessary loss of high spatial frequency information and often results in inadequate sampling (at or very close to the Nyquist frequency) of structures of interest (e.g. ODCs or orientation columns). This prevents studies from realizing the limits of the Rayleigh Criterion for a given slice profile. To improve isotropic sampling frequencies, prior studies have attempted to increase isotropic resolutions to where CNR become almost unbearably low. Given the results of our study, we recommend optimizing both slice thickness and inter-slice spacing for the underlying structure of interest for improved CNR efficiency.
Conclusions: We evaluated the synergistic combination of SLIDER and SMS in high-resolution whole brain fMRI and found that SLIDER-SMS can generate very high-resolution, high CNR fMRI data at both 3T and 7T. The BOLD CNR of SLIDER was significantly greater than that of HR and even HRb due to the linear relationship between voxel volume and SNR (as opposed to square-root when blurring in post processing) as well as the greater BOLD sensitivity with thicker slices. We also found it advantageous to forgo post-acquisition deblurring of individual image volumes in fMRI as this avoided noise amplification and banding artifacts, while resolving higher spatial frequency information with similar
BOLD CNR relative to low-resolution, thick slice acquisitions. Future use of SLIDER for fMRI may enable robust sub-millimeter imaging at 3T as well as even higher spatial resolution for mesoscale fMRI investigations at 7T.
Figure 1. Slice acquisition schemes for conventional and SLIDER acquisitions. Left) When SLIDER-2 slices are acquired interleaved (i.e. replicating and shifting the traditional low-resolution acquisition by half a slice thickness), sub-slices (denoted by the green box and separated by the dashed line) will experience different effective TRs. Right) By acquiring SLIDER-2 slices in ascending order (i.e. taking the traditional highresolution acquisition and doubling the slice thickness), the difference in effective TRs is minimized. Figure 2. Posterior occipital lobe coronal cross section (of the 1.25 mm axially acquired slices) of a representative subject’s low-resolution, HR, HRb, SLIDER-XD and SLIDER data deblurred with various λ values. Figure 3. BOLD CNR (t-values) for the corresponding Figure 2 datasets, showing activation to a visual checkerboard paradigm. T-values are shown going from black to red to white, white being the highest values. Figure 4. The effect of deblurring regularization parameter λ on left) residual slice blurring and right) average t-values. Error bars are SEM across subjects. Figure 5. Resolution enhancement in SLIDER-XD. Top row) Average EPI cross sections with zoomed-in subpanel for HR, SLIDER-XD, and low-resolution data. Middle row) Log tSNR of the k-space time series, normalizing center of k-space to 1 for each dataset. Bottom row) Same as middle row but using the same color scale across all datasets. Figure 6. Average k-space tSNR in the lowest 50%, middle 25%, and highest 25% sliceaxis k-space frequency regions (normalized to the lowest 50% k-space region of the HR data. Error bars are SEM across subjects. Figure 7. Ultrahigh-resolution fMRI examples. Left) 3T ocular dominance mapping experiment (0.65 mm SLIDER-XD). Total acquisition time: 18 min. F-values are shown in the middle row going from black to red to white, white being the highest values. ODC eye preference map is shown in the bottom row, with the following color scale: orange left eye preference, purple: no eye preference, blue: right eye preference. Right) F-values from the 7T visual checkerboard experiment (0.45 mm). Total acquisition time for a
single run average: 3 min. Lowest values are in blue-green, highest values are dark red. Figure 8. Simulation for optimizing slice thickness and inter-slice spacing for an ODClike structure (2 mm period square wave). Left) Signal intensity of ODC-like structure after being convolved (e.g. infinite sampling) with PSF of various slice thicknesses. The slice thickness of 1 mm matching the width of a single column resulted in the largest contrast between columns. The slice thickness of 2 mm was worst and resulted in a phase reversal of the underlying structure. The dashed black line depicts the underlying 2 mm period square wave. Right) Signal intensity and spline curve reconstruction of ODC-like structure as a function of sampling frequency (i.e. inter-slice spacing) and sampling grid phase using a slice thickness of 1 mm. The solid black line depicts the underlying 2 mm period square wave after convolution with the 1 mm slice. Top row) At twice Nyquist (e.g. SLIDER-XD; sampling every 0.5 mm), the reconstructed ODC-like structure does not depend on the sampling grid phase. Differently phased sampling grids are denoted by marker color/shape while corresponding spline reconstructions are denoted by line color. Middle rows) Sampling close to the Nyquist frequency (e.g. conventional sampling interval equal to the slice thickness; every ~1.0 mm), results in significant phase dependent aliasing. In many cases the estimated centers of the columns are inaccurate and their detectability (e.g. CNR) is greatly reduced. Bottom row) Sampling at below the Nyquist frequency results in aliasing which varies strongly with sampling phase, resulting in various low frequency ODC-like patterns that are nothing like the true underlying ODC structure. Thus for resolving structures such as ocular dominance columns (which are roughly ~1 mm per column), it is recommended to use a sampling grid significantly finer than 1 mm (e.g. SLIDER-2 at ~0.5 mm nominal isotropic). Figure S1. Phantom example of SLIDER un-aliasing high-frequency information along the slice axis when dithered sampling frequency equals twice the Nyquist frequency. A phantom with 16 mm period structure was built using 8 mm Lego® pieces and filled with water (regions outlined in red). EPI images were acquired at 8 mm isotropic resolution: left) dither #1 with the image sampling grid in phase with the phantom and right) dither #2 out of phase with the phantom. Combining both dithers provides the SLIDER sampling acquisition scheme enabling robust detection of high spatial frequency anatomical features. Conventional sampling schemes (i.e. without dithering) in this case would result in sampling exactly at the Nyquist frequency and generate results ranging from dither #1 to dither #2 depending on the phase alignment between the sampling grid and structure of interest. This phase dependent aliasing is problematic when sampling close to the Nyquist frequency and impacts reproducibility when the object (e.g. patient head) moves from one scan to the next. Figure S2. Frequency space explanation of sampling phase dependent aliasing. For an object of frequency extent [-B B], aliasing can be avoided if the sampling frequency (fs) is greater than 2B. For a single dither of Figure S1, fs = 2B which results in aliasing depending on the sampling grid positioning. Top row) Diagram for dither #1 sampling scheme where high frequency structure is resolved when the sampling grid is in phase (e.g. cosine). Bottom row) Diagram for dither #2 sampling scheme where high frequency structure cancels out when sampling grid is out of phase (e.g. sine). Blue circles indicate
net frequency signal after sampling is performed in the space domain and aliasing is accounted for. Figure S3. Effect of acquisition scheme on left) mean tSNR and right) mean t-value relative to HR acquisition. HRb results in the expected ~40% increase in both tSNR and t-value. SLIDER-XD results in a 50% increase in tSNR (less than the theoretical 100% due to T1 relaxation effects) and a 100% increase in t-value. This suggests the additional 50% of t-value improvement is due to the increased sensitivity of thicker slice acquisitions to BOLD susceptibility changes. Deblurring here is performed with λ=0.08. Figure S4. Individual slice dithers of ODC eye preference map (same data as Figure 7). Top) SLIDER-XD with the following color scale: orange left eye preference, purple: no eye preference, blue: right eye preference. Middle) No SLIDER (Dither #1; 0.65x0.65x1.3 mm3). Bottom) No SLIDER (Dither #2). Note, the SLIDER-XD map is formed by interleaving slices from dither #1 and dither #2.
Acknowledgements: NIH BRAIN Initiative grant - 1R24MH106096
References Adams D, Sincich L, Horton J (2007) Complete pattern of ocular dominance columns in human primary visual cortex. The Journal of Neuroscience 27:10391-10403. Beckett A, Vu AT, Keil B, Setsompop, Wald LL, Schillack S, Feinberg DA (2016) Assessment of coil arrays with small loop diameter at 7T for micron-scale resolution fMRI of human neocortex. Proc ISMRM.
Chang WT, Nummenmaa A, Witzel T, Ahveninen J, Huang S, Tsai KW, Chu YH, Polimeni JR, Belliveau JW, Lin FH (2013) Whole-head rapid fMRI acquisition using echo-shifted magnetic resonance inverse imaging. Neuroimage 78:325-338. Chen L, A TV, Xu J, Moeller S, Ugurbil K, Yacoub E, Feinberg DA (2015) Evaluation of highly accelerated simultaneous multi-slice EPI for fMRI. Neuroimage 104:452-459. Cheng K, Waggoner RA, Tanaka K (2001) Human ocular dominance columns as revealed by high-field functional magnetic resonance imaging. Neuron 32:359-374. De Martino F, Moerel M, Ugurbil K, Goebel R, Yacoub E, Formisano E (2015) Frequency preference and attention effects across cortical depths in the human primary auditory cortex. Proceedings of the National Academy of Sciences of the United States of America 112:16036-16041. Engel, S.A., Glover, G.H., Wandell, B.A., 1997. Retinotopic organization in human visual cortex and the spatial precision of functional MRI. Cereb. Cortex 7, 181–192. doi:10.1093/cercor/7.2.181 Feinberg DA, Moeller S, Smith SM, Auerbach E, Ramanna S, Gunther M, Glasser MF, Miller KL, Ugurbil K, Yacoub E (2010) Multiplexed echo planar imaging for sub-second whole brain FMRI and fast diffusion imaging. PloS one 5:e15710. Feinberg DA, Vu AT, Beckett A (In Review) Pushing the limits of ultra-high resolution human
brain imaging with SMS-EPI demonstrated for columnar level fMRI. Neuroimage. Geissler A, Lanzenberger R, Barth M, Tahamtan AR, Milakara D, Gartus A, Beisteiner R (2005) Influence of fMRI smoothing procedures on replicability of fine scale motor localization. Neuroimage 24:323-331. Greenspan H (2009) Super-resolution in medical imaging. The Computer Journal 52:43-63. Greenspan H, Oz G, Kiryati N, Peled S (2002) MRI inter-slice reconstruction using superresolution. Magnetic resonance imaging 20:437-446. Griswold MA, Jakob PM, Heidemann RM, Nittka M, Jellus V, Wang J, Kiefer B, Haase A (2002) Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magnetic resonance in medicine 47:1202-1210. Harel N, Lin J, Moeller S, Ugurbil K, Yacoub E (2006) Combined imaging-histological study of cortical laminar specificity of fMRI signals. Neuroimage 29:879-887. Lam EY (2003) Noise in superresolution reconstruction. Optics letters 28:2234-2236. Larkman DJ, Hajnal JV, Herlihy AH, Coutts GA, Young IR, Ehnholm G (2001) Use of multicoil arrays for separation of signal from multiple slices simultaneously excited. Journal of magnetic resonance imaging 13:313-317. Lai S, Glover G (1998) Three-dimensional spiral fMRI technique: a comparison with 2D spiral acquisition. Magn Reson Med 39:68–78. Lemmer P, Gunkel M, Baddeley D, Kaufmann R, Urich A, Weiland Y, Reymann J, Muller P, Hausmann M, Cremer C (2008) SPDM: light microscopy with single-molecule resolution at the nanoscale. Applied Physics B 93:1-12. Lu HD, Chen G, Tanigawa H, Roe AW (2010) A motion direction map in macaque V2. Neuron 68:1002-1013. Mayer GS, Vrscay ER (2007) Measuring information gain for frequency-encoded superresolution MRI. Magnetic resonance imaging 25:1058-1069. Moeller S, Yacoub E, Olman CA, Auerbach E, Strupp J, Harel N, Ugurbil K (2010) Multiband multislice GE-EPI at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fMRI. Magnetic resonance in medicine 63:1144-1153. Muckli L, De Martino F, Vizioli L, Petro LS, Smith FW, Ugurbil K, Goebel R, Yacoub E (2015) Contextual Feedback to Superficial Layers of V1. Current biology : CB 25:2690-2695. Nasr S, Polimeni JR, Tootell RB (2016) Interdigitated Color- and Disparity-Selective Columns within Human Visual Cortical Areas V2 and V3. The Journal of neuroscience : the official journal of the Society for Neuroscience 36:1841-1857. Olman CA, Harel N, Feinberg DA, He S, Zhang P, Ugurbil K, Yacoub E (2012) Layer-specific fMRI reflects different neuronal computations at different depths in human V1. PloS one 7:e32536. Otazo R, Lin FH, Wiggins G, Jordan R, Sodickson D, Posse S (2009) Superresolution parallel magnetic resonance imaging: application to functional and spectroscopic imaging. Neuroimage 47:220-230. Parkes, L.M., Schwarzbach, J.V., Bouts, A.A., Deckers, R. h R., Pullens, P., Kerskens, C.M., Norris, D.G., 2005. Quantifying the spatial resolution of the gradient echo and spin echo BOLD response at 3 Tesla. Magn. Reson. Med. 54, 1465–1472. doi:10.1002/mrm.20712 Peeters R, Kornprobst P, Nikolova M, Sunaert S, Vieville T, Malandain G, Deriche R, Faugeras O, Ng M, Hecke P (2004) The use of super-resolution techniques to reduce slice
thickness in functional MRI. International Journal if Imaging Systems and Technology 14:131-138. Polimeni JR, Bhat H, Witzel T, Benner T, Feiweier T, Inati SJ, Renvall V, Heberlein K, Wald LL (2016) Reducing sensitivity losses due to respiration and motion in accelerated echo planar imaging by reordering the autocalibration data acquisition. Magnetic resonance in medicine 75:665-679. Polimeni JR, Fischl B, Greve DN, Wald LL (2010) Laminar analysis of 7T BOLD using an imposed spatial activation pattern in human V1. Neuroimage 52:1334-1346. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P (1999) SENSE: sensitivity encoding for fast MRI. Magnetic resonance in medicine 42:952-962. Saritas EU, Lee D, Cukur T, Shankaranarayanan A, Nishimura DG (2014) Hadamard slice encoding for reduced-FOV diffusion-weighted imaging. Magnetic resonance in medicine 72:1277-1290. Setsompop K, Bilgic B, Nummenmaa A, Fan Q, Cauley S, Huang S, Chatnuntawech I, Yogesh R, Witzel T, Wald L (2015) SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for high resolution (700 um) diffusion imaging of the human brain. In: ISMRM Toronto, Onterio, Canada. Setsompop K, Gagoski BA, Polimeni JR, Witzel T, Wedeen VJ, Wald LL (2012) Blippedcontrolled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty. Magnetic resonance in medicine 67:1210-1224. Setsompop K, Stockmann K, Qiuyun F, Witzel T, Wald LL (2016) Generalized SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (gSlider-SMS) to increase volume encoding, SNR and partition profile fidelity in high-resolution diffusion imaging. In: ISMRM Singapore. Shmuel, A., Yacoub, E., Chaimow, D., Logothetis, N.K., Ugurbil, K., 2007. Spatio-temporal point-spread function of fMRI signal in human gray matter at 7 Tesla. NeuroImage 35, 539–552. doi:10.1016/j.neuroimage.2006.12.030 Smith SM, Beckmann CF, Andersson J, Auerbach EJ, Bijsterbosch J, Douaud G, Duff E, Feinberg DA, Griffanti L, Harms MP, Kelly M, Laumann T, Miller KL, Moeller S, Petersen S, Power J, Salimi-Khorshidi G, Snyder AZ, Vu AT, Woolrich MW, Xu J, Yacoub E, Ugurbil K, Van Essen DC, Glasser MF, Consortium WU-MH (2013) Restingstate fMRI in the Human Connectome Project. Neuroimage 80:144-168. Sodickson DK, Manning WJ (1997) Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magnetic resonance in medicine 38:591603. Talagala S, Sarlls J, Inati S (2013) Improved temporal SNR of accelerated EPI using a FLASH based GRAPPA reference scan. In: ISMRM Salt Lake City, Utah, USA. Ugurbil K, Xu J, Auerbach EJ, Moeller S, Vu AT, Duarte-Carvajalino JM, Lenglet C, Wu X, Schmitter S, Van de Moortele PF, Strupp J, Sapiro G, De Martino F, Wang D, Harel N, Garwood M, Chen L, Feinberg DA, Smith SM, Miller KL, Sotiropoulos SN, Jbabdi S, Andersson JL, Behrens TE, Glasser MF, Van Essen DC, Yacoub E, Consortium WU-MH (2013) Pushing spatial and temporal resolution for functional and diffusion MRI in the Human Connectome Project. Neuroimage 80:80-104. Van Essen DC, Smith SM, Barch DM, Behrens TE, Yacoub E, Ugurbil K, Consortium WU-MH (2013) The WU-Minn Human Connectome Project: an overview. Neuroimage 80:62-79. Vu AT, Auerbach E, Lenglet C, Moeller S, Sotiropoulos SN, Jbabdi S, Andersson J, Yacoub E,
Ugurbil K (2015) High resolution whole brain diffusion imaging at 7T for the Human Connectome Project. Neuroimage 122:318-331. Vu AT, Jamison K, Glasser M, Smith S, Coalson T, Moeller S, Auerbach E, Ugurbil K, Yacoub E (In Press) Tradeoffs in pushing the spatial resolution of fMRI for the 7 T Human Connectome Project. Neuroimage. Vu AT, Phillips J, Kay K, Phillips M, Johnson M, Shinkareva S, Tubridy S, Millin R, Grossman M, Gureckis T, Bhattacharyya R, Yacoub E (2016) Using precise word timing information improves decoding accuracy in a multiband-accelerated multimodal reading experiment. Cognitive Neuropsychology 33:265-275. White T, O'Leary D, Magnotta V, Arndt S, Flaum M, Andreasen NC (2001) Anatomic and functional variability: the effects of filter size in group fMRI data analysis. Neuroimage 13:577-588. Xu J, Moeller S, Auerbach EJ, Strupp J, Smith SM, Feinberg DA, Yacoub E, Ugurbil K (2013) Evaluation of slice accelerations using multiband echo planar imaging at 3 T. Neuroimage 83:991-1001. Yacoub E, Harel N, Ugurbil K (2008) High-field fMRI unveils orientation columns in humans. Proceedings of the National Academy of Sciences of the United States of America 105:10607-10612. Yacoub E, Shmuel A, Logothetis N, Ugurbil K (2007) Robust detection of ocular dominance columns in humans using Hahn Spin Echo BOLD functional MRI at 7 Tesla. Neuroimage 37:1161-1177. Yan C-G, Cheung B, Kelly C, Colcombe S, Craddock RC, Di Martino A, Li Q, Zuo X-N, Castellanos FX, Milham MP (2013) A comprehensive assessment of regional variation in the impact of head micromovements on functional connectomics. Neuroimage 76C:183201. Zhao F, Wang P, Hendrich K, Ugurbil K, Kim SG (2006) Cortical layer-dependent BOLD and CBV responses measured by spin-echo and gradient-echo fMRI: insights into hemodynamic regulation. Neuroimage 30:1149-1160. Zimmermann J, Goebel R, De Martino F, van de Moortele PF, Feinberg D, Adriany G, Chaimow D, Shmuel A, Ugurbil K, Yacoub E (2011) Mapping the organization of axis of motion selective features in human area MT using high-field fMRI. PloS one 6:e28716. Uludag K, Ugurbil K (2015) Physiology and Physics of the fMRI Signal. In: fMRI: From Nuclear Spins to Brain Functions, pp 163-213 US: Springer.
N/2+2
2
N=# slices
N+N/2+2
N+2
N+N/2+1
N+1
SLIDER-2 interleaved
N/2+2
2
N/2+1
1
Traditional low resolution
N/2+1
1
Figure 1
4
3
2
1
4
3
2
1
N+4
N+3
N+2
N+1
SLIDER-2 ascending
N+4
N+3
N+2
N+1
Traditional high resolution
High res
Figure 2
Deblurred (λ=0)
Deblurred (λ=0.10)
High res blurred
SLIDER-XD
slice axis
readout axis
Low res
Deblurred (λ=0, PSF=1.25mm)
Deblurred (λ=0.1, PSF=1.8mm)
High res
Figure 3
High res blurred
SLIDER-XD
slice axis
readout axis
Low res
slice blur (relative to HR)
0.92 0
0.94
0.96
0.98
1.02
1.04
1.06
0.05
0.1
Figure 4
0.15
0.2 0.25 0.3 regularization factor
0.35
0.4
0.45
0
0
0.05
0.1
0.15
0.2 0.25 0.3 regularization factor
0.35
0.4
0.45
HRb_db SLIDER
SLIDER-XD
HRb
HR
HRb_db
0.5
0.5
1
1.5
2
2.5
SLIDER
SLIDER-XD
HRb
HR
PSF=1.25 mm
PSF=2.50 mm
mean t-value (relative to HR)
0.5
High res SLIDER-XD
2D FFT, accounting for SNR differences
2D FFT
Mean EPI cross section
Figure 5
slice axis
readout axis
3T, 1.25 mm iso
Low res
Figure 6
slice axis
0
low k-space
mid k-space
high k-space
Relative tSNR
readout axis
1 High res
High res blurred
SLIDER-XD
L
R
Figure 7
ODC map
F-value
3T, 18 min experiment
EPI cross section
slice axis
readout axis
No SLIDER x 4 avgs
SLIDER-XD
7T, 3 min experiment
No SLIDER
Figure 8
PII: DOI: Reference:
S1053-8119(17)30105-2 http://dx.doi.org/10.1016/j.neuroimage.2017.02.001 YNIMG13786
To appear in: NeuroImage Received date: 16 September 2016 Revised date: 19 January 2017 Accepted date: 1 February 2017 Cite this article as: An T. Vu, Alex Beckett, Kawin Setsompop and David A. Feinberg, Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for human fMRI, NeuroImage, http://dx.doi.org/10.1016/j.neuroimage.2017.02.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDERSMS) for human fMRI An T. Vu1,2,3*, Alex Beckett2,3, Kawin Setsompop4, David A. Feinberg2,3 1
San Francisco VA Health Care System, Center for Imaging of Neurodegenerative Diseases, San Francisco, CA 2
University of California, Berkeley, Berkeley, CA
3
Advanced MRI Technologies, Sebastopol, CA
4
Martinos Center for Biomedical Imaging, Massachusetts General Hospital Charlestown, MA *
To whom correspondence should be addressed: Center for Imaging of Neurodegenerative Diseases, 4150 Clement Street, Building 13, (114M Mail Stop), San Francisco, CA 94121. Tel.: +408 636 8375. [email protected] Abstract High isotropic resolution fMRI is challenging primarily due to long repetition times (TR) and insufficient SNR, especially at lower field strengths. Recently, Simultaneous MultiSlice (SMS) imaging with blipped-CAIPI has substantially reduced scan time and improved SNR efficiency of fMRI. Similarly, super-resolution techniques utilizing subvoxel spatial shifts in the slice direction have increased both resolution and SNR efficiency. Here we demonstrate the synergistic combination of SLIce Dithered Enhanced Resolution (SLIDER) and SMS for high-resolution, high-SNR whole brain fMRI in comparison to standard resolution fMRI data as well as high-resolution data. With SLIDER-SMS, high spatial frequency information is recovered (unaliased) even in absence of super-resolution deblurring algorithms. Additionally we find that BOLD CNR (as measured by t-value in a visual checkerboard paradigm) is improved by as much as 100% relative to traditionally acquired high- resolution data. Using this gain in CNR, we are able to obtain unprecedented nominally isotropic resolutions at 3T (0.66 mm) and 7T (0.45 mm). Keywords: 3T; 7T; High-resolution; fMRI; SLIDER; super resolution
Abbreviations HR, high resolution (isotropic); HRb, HR, blurred along the slice axis in post processing using a 2-slice sliding window to mimic SLIDER factor 2 acquisition.; HRb_db, HRb with the SLIDER deblurring step; SLIDER, slice dithered enhanced resolution technique (factor 2 in this study); SLIDER-XD, SLIDER without the deblurring step.
Introduction: Advancements in functional magnetic resonance imaging (fMRI) have pushed the limits of achievable spatial resolution enabling sub-millimeter imaging of human brain function down to the columnar level (Yacoub et al., 2007, 2008; De Martino et al., 2015; Nasr et al., 2016; Zimmermann et al., 2011) and laminar level (De Martino et al., 2015; Harel et al., 2006; Muckli et al., 2015; Olman et al., 2012; Polimeni et al., 2010; Zhao et al., 2006; Zimmermann et al., 2011). These studies were all performed at ultra-high field (7T) since sub-millimeter fMRI is generally regarded as not feasible at lower field strengths due to SNR and BOLD CNR constraints. However, given that 7T scanners are relatively few in the human research setting, there is strong motivation to further develop techniques to enable sub-millimeter functional imaging at lower, more readily available field strengths. Of course such developments would also be useful at ultra-high field strengths as well given that the field has yet to achieve routine human fMRI at the spatial scale of more invasive techniques such as optical imaging (Lu et al., 2010).
Super-resolution techniques utilizing sub-voxel spatial shifts in the slice direction
(Greenspan et al., 2002; Greenspan 2008; Setsompop et al., 2015) have shown great promise for increasing both resolution and SNR efficiency but are not without their challenges. Super-resolution techniques like SLIce Dithered Enhanced Resolution (SLIDER; Setsompop et al., 2015) improve resolution in two ways. First, SLIDER increases the Nyquist sampling frequency by acquiring S sets of high SNR, lowresolution, thick slice data (e.g. 1x1xSth mm3) with Sth/S mm shifts between each set along the slice direction (where S is the SLIDER factor and Sth is the thickness of the excited slice). Because slices are spatially encoded, in contrast to the frequency and phase encoded dimensions (Mayer and Vrscay 2007), this form of oversampling resolves and unaliases additional high spatial frequency information that is otherwise lost when using traditional inter-slice spacing equal to the slice thickness. However, this high spatial frequency information is dampened relative to lower spatial frequencies due to the point spread function (PSF) of the thicker slices used. Thus, the second way SLIDER improves resolution is by applying a deblurring algorithm along the slice axis. Given the slice profile (e.g. PSF) is known, the deblurring algorithm can estimate what the underlying data would be had a thinner slice, of thickness Sth/S mm, been used.
Because SLIDER acquires S sets of low-resolution, anisotropic thick slice data to more finely sample the image volume along the slice axis, the total volume acquisition time is S times that of the original low-resolution data. Analogous to conventional thin slice, isotropic acquisitions, imaging volume times can be very long (>> 3000 seconds) given it can take well over 100 slices to cover the brain at sub-millimeter resolution. This is
undesirable due to the relatively low SNR efficiency (Feinberg et al., 2010; Smith et al., 2013; Xu et al., 2013) and loss of high temporal frequency BOLD information (Chang et al., 2013; Chen et al., 2015; Vu et al., 2016). Fortunately, recent advancements in Simultaneous Multi-Slice (SMS) imaging (Larkman et al., 2001; Moeller et al., 2010) with blipped-CAIPI (Setsompop et al., 2012), in combination with in-plane undersampling techniques (Sodickson and Manning, 1997; Pruessmann et al., 1999; Griswold et al., 2002; Griswold et al., 2006; Talagala et al., 2013; Uğurbil et al., 2013; Polimeni et al., 2016) have substantially improved scan time and SNR efficiency (Ugurbil et al., 2013; Vu et al., 2015; Vu et al., In Press). Thus the combination of SLIDER and SMS should provide suitable temporal resolution and SNR efficiency for mesoscale imaging at both 3T and 7T.
Besides long imaging volume times, another common concern regarding super-resolution techniques is the residual blurring and noise amplification resulting from post-acquisition deblurring (Lam, 2003; Otazo et al., 2009). In the absence of noise, the deblurring component of super-resolution techniques can perfectly reconstruct high-resolution images from multiple low-resolution images. However, in practice, noise levels in the images can be quite significant leading to high frequency noise amplification and banding artifacts. To mitigate the effect of noise, prior studies have resorted to regularization of the deblurring algorithm, which effectively trades noise amplification for residual blurring. How this tradeoff manifests in the context of SLIDER fMRI and subsequently generated functional activation maps remains unclear. In some SLIDER fMRI applications, the increase in resolved high spatial frequency information may be
sufficient without deblurring and may warrant forgoing the post-acquisition deblurring of individual image volumes to achieve maximal BOLD CNR. More specifically, the benefits of deblurring may be limited in the context of fMRI, particularly for highresolution sub-mm acquisitions, given that the vascular PSF is estimated to be on the order of ~1.5-4 mm (Engel et al., 1997; Parkes et al., 2005; Shmuel et al., 2007) and thus would dominate blurring effects due to SLIDER. Fortunately, contrast subtraction techniques ubiquitous to analysis of fMRI data are known to serve as specificity or PSF enhancing operations (Yacoub et al., 2007, 2008; Muckli et al., 2015). This concept is analogous to super resolution techniques used in optical imaging (e.g. spectral precision distance microscopy (Lemmer et al., 2008)) where detailed structural information can be obtained far beyond the diffraction limit by selectively activating spatially sparse subsets of objects at any one time and precisely locating the positions of these objects. Similarly in fMRI, individual columns or layers can be sparsely activated in space at different time points – allowing for their precise location and organizational structure to be determined even in the presence of relatively broad PSFs.
Here, we evaluated the complimentary combination of slice dithering and SMS for highresolution whole brain fMRI. Specifically, we compared SLIDER-SMS to SMS alone at low-resolution, thick slice acquisition (1.25x1.25x2.5 mm3) as well as to thin slice, high isotropic resolution acquisition (1.25x1.25x1.25 mm3) using the following metrics: data smoothness, k-space energy at various spatial frequency bins (along the slice axis), tSNR, and BOLD CNR. The effect of the deblurring post-processing step with various levels of regularization was also investigated and revealed that SLIDER without deblurring (which
we refer to as SLIDER-XD) may actually be preferred for optimal BOLD CNR while still retaining significantly more high-spatial frequency information than conventional lowresolution, anisotropic acquisitions. Using these gains in CNR, we show it is feasible to perform fMRI experiments using 0.65 mm isotropic nominal resolution data acquired at 3T and 0.45 mm data acquired at 7T.
Methods Data were acquired from four healthy subjects on a Siemens 3T Trio using the standard 32 ch head coil. During the 2D gradient echo EPI fMRI scans, subjects viewed three 96 s runs of flashing checkerboard stimulus (30 sec period) for each scan protocol (SMS-5 and SLIDER-2 SMS-5). Imaging parameters were: 1.25 mm isotropic (nominal; 2.5 mm excitation thickness for SLIDER=2); FOV = 210x210x137.5 mm3; PF = 6/8; TE = 45 ms; TR = 3000 ms (1500 ms per dithered volume); Flip angle = 84° (72° for SLIDER=2); PE direction= AP; echo spacing = 0.88 ms; axial oblique slices; SMS=5; FOV/3 shift with sliceGrappa reconstruction (Setsompop et al., 2012); and no in-plane under sampling.
Figure 1 left shows the traditional interleaved SLIDER acquisition scheme where one half of a slice has an effective TR of twice that of the other half. Although this acquisition scheme worked well in prior studies where the repetition times were long (TR>>T1; which mitigates T1 relaxation and spin history effects) it is particularly problematic for fMRI applications where shorter TRs are desirable. One early SLIDER fMRI study (Peeters et al., 2004) took advantage of the shorter T1 at 1.5T and a longer TR (>3000 ms) but was only able to acquire 2 mm isotropic data, which is now considered fairly
standard. Furthermore, results of this study were unconvincing and difficult to interpret given the 3 mm smoothing kernel applied. In our study presented here, we mitigated the T1 relaxation differences between sub-slices by using an ascending slice order (Figure 1 right) instead of a long TR with interleaving slices. This reduced the difference in effective TR between sub-slices to just ~100 ms (the time of a single slice acquisition). In this case, the effective TR for all sub-slices is then approximately 1/S times the total volume imaging time; enabling further SNR optimization through the use of the Ernst flip angle.
To assess whether or not gains in SNR and CNR were simply due to the increased PSF of SLIDER along the slice axis, as a control, we generated a “High res blurred” (HRb) condition by taking the SMS-5 “High res” (HR) data (Figure 1 right; ascending slice order) and performing a 2-slice moving-window average (mimicking the SLIDER acquisition in post-processing). For deblurring/reconstruction, the analytical solution was based on the excited slice profiles. More specifically, the Toeplitz matrix was used as the forward model (T) and Tickhnov regularization was used to calculate the inverse model (Tinv) with λ ranging from 0 to 50% of the largest eigenvalue of T. In Matlab: Tinv = (V*E/(E*E+eye(size(E))*max(diag(E))*λ)*U'); where [U E V] = svd(T). Note larger λ’s result in greater residual blurring. For simplicity, the Toeplitz matrix used here assumed perfect, rectangular slice profiles. Simulations revealed that this assumption was reasonable (not shown) and that the effect of slice crosstalk due to imperfect slice profiles with our ascending SLIDER slice order resulted in ~5% narrower PSF in all but the first acquired slice. Deblurring using a Toeplitz matrix accounting for the imperfect slice
profiles in this case gave qualitatively the same results albeit with slightly worse tSNR and BOLD CNR values.
Imaging parameters for the 0.65 mm isotropic nominal 3T data were: SLIDER=2; FOV = 120x120x39 mm; PF = 6/8; TE = 45 ms; TR = 4000 ms (2000 ms per dithered volume); Flip angle = 80°; PE direction= HF; echo spacing = 1.52 ms; coronal slices; SMS=2; GRAPPA=2; FOV/3 shift. This high-resolution protocol was used for two 9 minute scans of ocular dominance column (ODC) mapping in primary visual cortex using commercially available red/blue anaglyphs and the stimulus paradigm denoted in (Cheng et al., 2001; repetitions of 6 s blanks with 12 s left or right eye stimulation). A GLM analysis was used to estimate responses to each eye, and a statistical map was generated by comparing responses to stimulation for each eye, and thresholded by the response to stimulation by either/both eyes (F,p<0.001).
An additional pilot dataset was acquired on a Siemens 7T MAGNETOM scanner with standard 70 mT/m body gradients and a custom 8 ch surface coil (Virtumed, LLC) with 4 cm loop diameter (SNR gains described in Beckett et al., 2016). Imaging parameters for the 7T data were: 0.45 mm isotropic (nominal; 0.9 mm excitation thickness for SLIDER=2); FOV = 90x81x23.4 mm3; PF = 5/8; TE = 27.8 ms; TR = 3000 ms (1500 ms per dithered volume); Flip angle = 80° (67° for SLIDER=2); PE direction= HF; coronal slices; SMS=2; GRAPPA=2. For this experiment, one subject viewed 3 minute runs of a flickering (8 Hz) checkerboard paradigm presented in alternating blocks of 12 s on and 12 s off.
All experiments used a high-resolution projector (Avotec, Inc.) to display visual stimuli. In order to preserve resolution properties across acquisition conditions, smoothing and motion correction steps were not performed. Head motion was minimized by recruiting expert subjects with informed consent, by keeping individual scan times short and by padding subjects’ heads well inside the head coils. Motion traces (provided by MrTools; Stanford, CA) confirmed that subject motion was within 0.5 mm across the entire 18 min ODC dataset. It is noted that future studies pursuing sub-mm imaging, regardless of subject population, would benefit from improved head (micro)-motion correction strategies (Yan et al 2013) as well as head restraint technologies. The 3T experimental protocols were approved by the Committee for the Protection of Human Subjects at UC Berkeley. The 7T protocols were approved by and the UC San Francisco and the Veterans Administrations Committees on Human Research.
Results & Discussion: Figure 2 shows the coronal cross section (of the 1.25 mm axially acquired slices) of a representative subject’s low-resolution (thick slice), HR (thin slice), HRb, SLIDER-XD, and SLIDER data deblurred with various λ values. Since HRb was blurred in post processing, the original HR image is recovered perfectly without regularization (λ=0). However, as with most super-resolution techniques in the presence of noise, deblurring SLIDER without regularization results in high spatial frequency noise amplification artifacts. SLIDER achieves best results with modest regularization (λ~0.1). Note that
even without deblurring, SLIDER-XD provides additional detail due to unaliasing of high spatial frequency information (for additional examples see: Figures S1-S2). Figure 3 shows BOLD CNR (t-values) for the corresponding Figure 2 datasets. SLIDER-XD yields substantially stronger BOLD CNR than both HR and HRb while SLIDER (with deblurring) has more spatial detail than the low-resolution data with similar BOLD CNR as HRb.
The amount of slice blur as a function of λ is plotted in Figure 4 left (quantified as the spatial correlation between the time averaged brain volume and the same volume shifted one slice down). The black dashed line denotes the λ (~0.08) where the effective SLIDER slice resolution matches that of HR. However, note that this is not true resolution as this is simply where residual blurring cancels out the high frequency noise artifacts. The true resolution including residual blurring but without noise artifacts can be determined from the cyan curve depicting the regularized deblurring of HRb data (which corresponds to ~1.75 mm at λ=0.08, using linear approximation between the HR (1.25 mm) and HRb (2.5 mm) curve values). Error bars are SEM across subjects. Mean t-values were calculated as the average t-value across an ROI of voxels activated (p<0.05; uncorrected) across all acquisition conditions. These values are plotted as a function of λ in Figure 4 right. HRb has a square-root of two increase over HR while SLIDER with λ~0.08 results in t-values at similar levels to HRb. This square-root of S gain is consistent with analogous Hadamard encoded techniques (Saritas et al 2014) as well as generalized SLIDER techniques (Setsompop et al 2016); which can be thought of as 3D or volume encoded imaging. Importantly, without deblurring, SLIDER-XD results in over double
the t-values as HR. This reflects not only the improved SNR of SLIDER (i.e. thick slice acquisition) but also the superiority of blurring through acquisition (as opposed to in post-processing) given that signal increases but not thermal noise in this case. Importantly, this also reflects SLIDER’s improved sensitivity to BOLD susceptibility changes (Lai and Glover 1998) given that tSNR improvement only accounts for about half of our observed gain in BOLD CNR (see Figure S3). Given these results and to reduce confounds of residual blurring and high-frequency artifacts, the subsequent analysis focuses on the use of SLIDER-XD.
To determine whether the higher slice blur values of SLIDER-XD depicted in Figure 4 reflect enhanced low spatial frequency information (as opposed to reduced high spatial frequency information), we also calculated the tSNR of the k-space time series (Figure 5). The middle row of Figure 5 shows the k-space tSNR normalized to each dataset’s DC value at the center. While it is clear that SLIDER-XD acquisition results in reduced high spatial frequency energy (relative to low spatial frequencies), the amount is certainly more than zero which is the amount in the non-dithered, low-resolution (1.25x1.25x2.5 mm3) data. When viewing all the acquisition conditions at the same scale (Figure 5 bottom row), it can be appreciated that the tSNR in lower k-space regions of SLIDERXD are higher than that of the HR data while the high k-space regions of SLIDER-XD and HR data are comparable.
To quantify these results, we calculated the average k-space tSNR in the lowest 50%, middle 25%, and highest 25% slice-axis k-space frequency regions (normalized to the
lowest 50% k-space region of the HR data; Figure 6). Notably, SLIDER-XD has significantly higher k-space tSNR (p<0.05, paired-T(2)) in both low and mid k-space frequencies and similar high k-space tSNR compared to HR and HRb (Figure 6 right). This shows that the SNR benefits of thick slice excitation in SLIDER-XD can offset the relative dampening of higher spatial frequencies. This is consistent with prior literature demonstrating that with enough CNR, even methods with moderate amounts of blurring (e.g. in the PE direction due to T2* decay) can yield maps of columnar level structure (Yacoub et al., 2007, 2008). However, it should be noted that, analogous to post acquisition smoothing kernels (White et al., 2001; Geissler et al., 2005), for optimal detectability of features of interest without deblurring, the SLIDER factor should provide slices thick enough to cover functionally homologous anatomic regions, yet thin enough not to partial volume together functionally distinct regions.
Utilizing these gains in BOLD CNR, we are able to obtain unprecedented nominally isotropic resolutions at 3T (0.65 mm) and 7T (0.45 mm). Figure 7 left shows preliminary results from one subject using the 0.65 mm SLIDER-XD acquisition at 3T (for additional SLIDER-XD ODC data, including cross-session reproducibility, see Feinberg et al In Review). The axial cross section of coronal acquired EPI data is shown at top, variance explained by task regressors (F-value) is shown at middle, and the eye preference map is shown at the bottom (orange for left eye, blue for right eye, purple for no eye preference). As expected, these maps contain substantially more spatial detail compared to the eye preference maps generated from the individual, low slice resolution (i.e. no SLIDER), dithers of the SLIDER-XD dataset; Figure S4). Figure 7 right shows the results from the
same subject using the 0.45 mm SLIDER-XD acquisition at 7T in a simple visual checkerboard paradigm. Thresholded t-values are shown overlaid onto the axial cross section of a T2 weighted anatomical image. Impressively, the 3 min SLIDER-XD acquisition results in BOLD CNR equivalent to four averages (12 min) of the nonSLIDER data with comparable spatial detail.
Importantly, SLIDER reminds us of the fact that inter-slice spacing (i.e. sampling frequency) can be optimized independently of slice thickness (i.e. PSF). For example, in a simulated case of imaging human ODCs which are approximately 1 mm per eye and thus of 2 mm period (Adams et al 2007; Yacoub et al 2007), Figure 8 left shows that a slice thickness of 1 mm is optimal for maximum CNR between columns. However, Figure 8 right shows that use of conventional slice spacing equal to the slice thickness, in this case, results in sampling exactly at the Nyquist rate (1 sample per mm) which is prone to head position dependent aliasing, poor reproducibility and/or poor structural fidelity. Increasing sampling to 1.25 times the Nyquist rate reduces head position dependent aliasing but the problem of different effective TRs within sub-slices becomes problematic again. By using SLIDER-XD and thus increasing sampling to twice the Nyquist rate (inter-slice spacing of 0.5 mm), head position dependent aliasing is eliminated while effective TRs within sub-slices are homogenous such that the Ernst flip angle can be used for optimal SNR (Figure 1 right). Note that for SLIDER-XD, higher SLIDER factors (>2) will not provide additional spatial resolution given that most aliasing in the slice direction is already eliminated at SLIDER factor 2. Future studies attempting higher SLIDER factors for fMRI should pay special attention to SNR
tradeoffs given the shorter effective relaxation times, the increase in instabilities due to variation in effective TR across sub-slices and the increase in susceptibility dropout due to the thickness of the slices excited.
In a field where the norm is to set inter-slice spacing equal to slice thickness, we find that this practice results in unnecessary loss of high spatial frequency information and often results in inadequate sampling (at or very close to the Nyquist frequency) of structures of interest (e.g. ODCs or orientation columns). This prevents studies from realizing the limits of the Rayleigh Criterion for a given slice profile. To improve isotropic sampling frequencies, prior studies have attempted to increase isotropic resolutions to where CNR become almost unbearably low. Given the results of our study, we recommend optimizing both slice thickness and inter-slice spacing for the underlying structure of interest for improved CNR efficiency.
Conclusions: We evaluated the synergistic combination of SLIDER and SMS in high-resolution whole brain fMRI and found that SLIDER-SMS can generate very high-resolution, high CNR fMRI data at both 3T and 7T. The BOLD CNR of SLIDER was significantly greater than that of HR and even HRb due to the linear relationship between voxel volume and SNR (as opposed to square-root when blurring in post processing) as well as the greater BOLD sensitivity with thicker slices. We also found it advantageous to forgo post-acquisition deblurring of individual image volumes in fMRI as this avoided noise amplification and banding artifacts, while resolving higher spatial frequency information with similar
BOLD CNR relative to low-resolution, thick slice acquisitions. Future use of SLIDER for fMRI may enable robust sub-millimeter imaging at 3T as well as even higher spatial resolution for mesoscale fMRI investigations at 7T.
Figure 1. Slice acquisition schemes for conventional and SLIDER acquisitions. Left) When SLIDER-2 slices are acquired interleaved (i.e. replicating and shifting the traditional low-resolution acquisition by half a slice thickness), sub-slices (denoted by the green box and separated by the dashed line) will experience different effective TRs. Right) By acquiring SLIDER-2 slices in ascending order (i.e. taking the traditional highresolution acquisition and doubling the slice thickness), the difference in effective TRs is minimized. Figure 2. Posterior occipital lobe coronal cross section (of the 1.25 mm axially acquired slices) of a representative subject’s low-resolution, HR, HRb, SLIDER-XD and SLIDER data deblurred with various λ values. Figure 3. BOLD CNR (t-values) for the corresponding Figure 2 datasets, showing activation to a visual checkerboard paradigm. T-values are shown going from black to red to white, white being the highest values. Figure 4. The effect of deblurring regularization parameter λ on left) residual slice blurring and right) average t-values. Error bars are SEM across subjects. Figure 5. Resolution enhancement in SLIDER-XD. Top row) Average EPI cross sections with zoomed-in subpanel for HR, SLIDER-XD, and low-resolution data. Middle row) Log tSNR of the k-space time series, normalizing center of k-space to 1 for each dataset. Bottom row) Same as middle row but using the same color scale across all datasets. Figure 6. Average k-space tSNR in the lowest 50%, middle 25%, and highest 25% sliceaxis k-space frequency regions (normalized to the lowest 50% k-space region of the HR data. Error bars are SEM across subjects. Figure 7. Ultrahigh-resolution fMRI examples. Left) 3T ocular dominance mapping experiment (0.65 mm SLIDER-XD). Total acquisition time: 18 min. F-values are shown in the middle row going from black to red to white, white being the highest values. ODC eye preference map is shown in the bottom row, with the following color scale: orange left eye preference, purple: no eye preference, blue: right eye preference. Right) F-values from the 7T visual checkerboard experiment (0.45 mm). Total acquisition time for a
single run average: 3 min. Lowest values are in blue-green, highest values are dark red. Figure 8. Simulation for optimizing slice thickness and inter-slice spacing for an ODClike structure (2 mm period square wave). Left) Signal intensity of ODC-like structure after being convolved (e.g. infinite sampling) with PSF of various slice thicknesses. The slice thickness of 1 mm matching the width of a single column resulted in the largest contrast between columns. The slice thickness of 2 mm was worst and resulted in a phase reversal of the underlying structure. The dashed black line depicts the underlying 2 mm period square wave. Right) Signal intensity and spline curve reconstruction of ODC-like structure as a function of sampling frequency (i.e. inter-slice spacing) and sampling grid phase using a slice thickness of 1 mm. The solid black line depicts the underlying 2 mm period square wave after convolution with the 1 mm slice. Top row) At twice Nyquist (e.g. SLIDER-XD; sampling every 0.5 mm), the reconstructed ODC-like structure does not depend on the sampling grid phase. Differently phased sampling grids are denoted by marker color/shape while corresponding spline reconstructions are denoted by line color. Middle rows) Sampling close to the Nyquist frequency (e.g. conventional sampling interval equal to the slice thickness; every ~1.0 mm), results in significant phase dependent aliasing. In many cases the estimated centers of the columns are inaccurate and their detectability (e.g. CNR) is greatly reduced. Bottom row) Sampling at below the Nyquist frequency results in aliasing which varies strongly with sampling phase, resulting in various low frequency ODC-like patterns that are nothing like the true underlying ODC structure. Thus for resolving structures such as ocular dominance columns (which are roughly ~1 mm per column), it is recommended to use a sampling grid significantly finer than 1 mm (e.g. SLIDER-2 at ~0.5 mm nominal isotropic). Figure S1. Phantom example of SLIDER un-aliasing high-frequency information along the slice axis when dithered sampling frequency equals twice the Nyquist frequency. A phantom with 16 mm period structure was built using 8 mm Lego® pieces and filled with water (regions outlined in red). EPI images were acquired at 8 mm isotropic resolution: left) dither #1 with the image sampling grid in phase with the phantom and right) dither #2 out of phase with the phantom. Combining both dithers provides the SLIDER sampling acquisition scheme enabling robust detection of high spatial frequency anatomical features. Conventional sampling schemes (i.e. without dithering) in this case would result in sampling exactly at the Nyquist frequency and generate results ranging from dither #1 to dither #2 depending on the phase alignment between the sampling grid and structure of interest. This phase dependent aliasing is problematic when sampling close to the Nyquist frequency and impacts reproducibility when the object (e.g. patient head) moves from one scan to the next. Figure S2. Frequency space explanation of sampling phase dependent aliasing. For an object of frequency extent [-B B], aliasing can be avoided if the sampling frequency (fs) is greater than 2B. For a single dither of Figure S1, fs = 2B which results in aliasing depending on the sampling grid positioning. Top row) Diagram for dither #1 sampling scheme where high frequency structure is resolved when the sampling grid is in phase (e.g. cosine). Bottom row) Diagram for dither #2 sampling scheme where high frequency structure cancels out when sampling grid is out of phase (e.g. sine). Blue circles indicate
net frequency signal after sampling is performed in the space domain and aliasing is accounted for. Figure S3. Effect of acquisition scheme on left) mean tSNR and right) mean t-value relative to HR acquisition. HRb results in the expected ~40% increase in both tSNR and t-value. SLIDER-XD results in a 50% increase in tSNR (less than the theoretical 100% due to T1 relaxation effects) and a 100% increase in t-value. This suggests the additional 50% of t-value improvement is due to the increased sensitivity of thicker slice acquisitions to BOLD susceptibility changes. Deblurring here is performed with λ=0.08. Figure S4. Individual slice dithers of ODC eye preference map (same data as Figure 7). Top) SLIDER-XD with the following color scale: orange left eye preference, purple: no eye preference, blue: right eye preference. Middle) No SLIDER (Dither #1; 0.65x0.65x1.3 mm3). Bottom) No SLIDER (Dither #2). Note, the SLIDER-XD map is formed by interleaving slices from dither #1 and dither #2.
Acknowledgements: NIH BRAIN Initiative grant - 1R24MH106096
References Adams D, Sincich L, Horton J (2007) Complete pattern of ocular dominance columns in human primary visual cortex. The Journal of Neuroscience 27:10391-10403. Beckett A, Vu AT, Keil B, Setsompop, Wald LL, Schillack S, Feinberg DA (2016) Assessment of coil arrays with small loop diameter at 7T for micron-scale resolution fMRI of human neocortex. Proc ISMRM.
Chang WT, Nummenmaa A, Witzel T, Ahveninen J, Huang S, Tsai KW, Chu YH, Polimeni JR, Belliveau JW, Lin FH (2013) Whole-head rapid fMRI acquisition using echo-shifted magnetic resonance inverse imaging. Neuroimage 78:325-338. Chen L, A TV, Xu J, Moeller S, Ugurbil K, Yacoub E, Feinberg DA (2015) Evaluation of highly accelerated simultaneous multi-slice EPI for fMRI. Neuroimage 104:452-459. Cheng K, Waggoner RA, Tanaka K (2001) Human ocular dominance columns as revealed by high-field functional magnetic resonance imaging. Neuron 32:359-374. De Martino F, Moerel M, Ugurbil K, Goebel R, Yacoub E, Formisano E (2015) Frequency preference and attention effects across cortical depths in the human primary auditory cortex. Proceedings of the National Academy of Sciences of the United States of America 112:16036-16041. Engel, S.A., Glover, G.H., Wandell, B.A., 1997. Retinotopic organization in human visual cortex and the spatial precision of functional MRI. Cereb. Cortex 7, 181–192. doi:10.1093/cercor/7.2.181 Feinberg DA, Moeller S, Smith SM, Auerbach E, Ramanna S, Gunther M, Glasser MF, Miller KL, Ugurbil K, Yacoub E (2010) Multiplexed echo planar imaging for sub-second whole brain FMRI and fast diffusion imaging. PloS one 5:e15710. Feinberg DA, Vu AT, Beckett A (In Review) Pushing the limits of ultra-high resolution human
brain imaging with SMS-EPI demonstrated for columnar level fMRI. Neuroimage. Geissler A, Lanzenberger R, Barth M, Tahamtan AR, Milakara D, Gartus A, Beisteiner R (2005) Influence of fMRI smoothing procedures on replicability of fine scale motor localization. Neuroimage 24:323-331. Greenspan H (2009) Super-resolution in medical imaging. The Computer Journal 52:43-63. Greenspan H, Oz G, Kiryati N, Peled S (2002) MRI inter-slice reconstruction using superresolution. Magnetic resonance imaging 20:437-446. Griswold MA, Jakob PM, Heidemann RM, Nittka M, Jellus V, Wang J, Kiefer B, Haase A (2002) Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magnetic resonance in medicine 47:1202-1210. Harel N, Lin J, Moeller S, Ugurbil K, Yacoub E (2006) Combined imaging-histological study of cortical laminar specificity of fMRI signals. Neuroimage 29:879-887. Lam EY (2003) Noise in superresolution reconstruction. Optics letters 28:2234-2236. Larkman DJ, Hajnal JV, Herlihy AH, Coutts GA, Young IR, Ehnholm G (2001) Use of multicoil arrays for separation of signal from multiple slices simultaneously excited. Journal of magnetic resonance imaging 13:313-317. Lai S, Glover G (1998) Three-dimensional spiral fMRI technique: a comparison with 2D spiral acquisition. Magn Reson Med 39:68–78. Lemmer P, Gunkel M, Baddeley D, Kaufmann R, Urich A, Weiland Y, Reymann J, Muller P, Hausmann M, Cremer C (2008) SPDM: light microscopy with single-molecule resolution at the nanoscale. Applied Physics B 93:1-12. Lu HD, Chen G, Tanigawa H, Roe AW (2010) A motion direction map in macaque V2. Neuron 68:1002-1013. Mayer GS, Vrscay ER (2007) Measuring information gain for frequency-encoded superresolution MRI. Magnetic resonance imaging 25:1058-1069. Moeller S, Yacoub E, Olman CA, Auerbach E, Strupp J, Harel N, Ugurbil K (2010) Multiband multislice GE-EPI at 7 tesla, with 16-fold acceleration using partial parallel imaging with application to high spatial and temporal whole-brain fMRI. Magnetic resonance in medicine 63:1144-1153. Muckli L, De Martino F, Vizioli L, Petro LS, Smith FW, Ugurbil K, Goebel R, Yacoub E (2015) Contextual Feedback to Superficial Layers of V1. Current biology : CB 25:2690-2695. Nasr S, Polimeni JR, Tootell RB (2016) Interdigitated Color- and Disparity-Selective Columns within Human Visual Cortical Areas V2 and V3. The Journal of neuroscience : the official journal of the Society for Neuroscience 36:1841-1857. Olman CA, Harel N, Feinberg DA, He S, Zhang P, Ugurbil K, Yacoub E (2012) Layer-specific fMRI reflects different neuronal computations at different depths in human V1. PloS one 7:e32536. Otazo R, Lin FH, Wiggins G, Jordan R, Sodickson D, Posse S (2009) Superresolution parallel magnetic resonance imaging: application to functional and spectroscopic imaging. Neuroimage 47:220-230. Parkes, L.M., Schwarzbach, J.V., Bouts, A.A., Deckers, R. h R., Pullens, P., Kerskens, C.M., Norris, D.G., 2005. Quantifying the spatial resolution of the gradient echo and spin echo BOLD response at 3 Tesla. Magn. Reson. Med. 54, 1465–1472. doi:10.1002/mrm.20712 Peeters R, Kornprobst P, Nikolova M, Sunaert S, Vieville T, Malandain G, Deriche R, Faugeras O, Ng M, Hecke P (2004) The use of super-resolution techniques to reduce slice
thickness in functional MRI. International Journal if Imaging Systems and Technology 14:131-138. Polimeni JR, Bhat H, Witzel T, Benner T, Feiweier T, Inati SJ, Renvall V, Heberlein K, Wald LL (2016) Reducing sensitivity losses due to respiration and motion in accelerated echo planar imaging by reordering the autocalibration data acquisition. Magnetic resonance in medicine 75:665-679. Polimeni JR, Fischl B, Greve DN, Wald LL (2010) Laminar analysis of 7T BOLD using an imposed spatial activation pattern in human V1. Neuroimage 52:1334-1346. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P (1999) SENSE: sensitivity encoding for fast MRI. Magnetic resonance in medicine 42:952-962. Saritas EU, Lee D, Cukur T, Shankaranarayanan A, Nishimura DG (2014) Hadamard slice encoding for reduced-FOV diffusion-weighted imaging. Magnetic resonance in medicine 72:1277-1290. Setsompop K, Bilgic B, Nummenmaa A, Fan Q, Cauley S, Huang S, Chatnuntawech I, Yogesh R, Witzel T, Wald L (2015) SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for high resolution (700 um) diffusion imaging of the human brain. In: ISMRM Toronto, Onterio, Canada. Setsompop K, Gagoski BA, Polimeni JR, Witzel T, Wedeen VJ, Wald LL (2012) Blippedcontrolled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty. Magnetic resonance in medicine 67:1210-1224. Setsompop K, Stockmann K, Qiuyun F, Witzel T, Wald LL (2016) Generalized SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (gSlider-SMS) to increase volume encoding, SNR and partition profile fidelity in high-resolution diffusion imaging. In: ISMRM Singapore. Shmuel, A., Yacoub, E., Chaimow, D., Logothetis, N.K., Ugurbil, K., 2007. Spatio-temporal point-spread function of fMRI signal in human gray matter at 7 Tesla. NeuroImage 35, 539–552. doi:10.1016/j.neuroimage.2006.12.030 Smith SM, Beckmann CF, Andersson J, Auerbach EJ, Bijsterbosch J, Douaud G, Duff E, Feinberg DA, Griffanti L, Harms MP, Kelly M, Laumann T, Miller KL, Moeller S, Petersen S, Power J, Salimi-Khorshidi G, Snyder AZ, Vu AT, Woolrich MW, Xu J, Yacoub E, Ugurbil K, Van Essen DC, Glasser MF, Consortium WU-MH (2013) Restingstate fMRI in the Human Connectome Project. Neuroimage 80:144-168. Sodickson DK, Manning WJ (1997) Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magnetic resonance in medicine 38:591603. Talagala S, Sarlls J, Inati S (2013) Improved temporal SNR of accelerated EPI using a FLASH based GRAPPA reference scan. In: ISMRM Salt Lake City, Utah, USA. Ugurbil K, Xu J, Auerbach EJ, Moeller S, Vu AT, Duarte-Carvajalino JM, Lenglet C, Wu X, Schmitter S, Van de Moortele PF, Strupp J, Sapiro G, De Martino F, Wang D, Harel N, Garwood M, Chen L, Feinberg DA, Smith SM, Miller KL, Sotiropoulos SN, Jbabdi S, Andersson JL, Behrens TE, Glasser MF, Van Essen DC, Yacoub E, Consortium WU-MH (2013) Pushing spatial and temporal resolution for functional and diffusion MRI in the Human Connectome Project. Neuroimage 80:80-104. Van Essen DC, Smith SM, Barch DM, Behrens TE, Yacoub E, Ugurbil K, Consortium WU-MH (2013) The WU-Minn Human Connectome Project: an overview. Neuroimage 80:62-79. Vu AT, Auerbach E, Lenglet C, Moeller S, Sotiropoulos SN, Jbabdi S, Andersson J, Yacoub E,
Ugurbil K (2015) High resolution whole brain diffusion imaging at 7T for the Human Connectome Project. Neuroimage 122:318-331. Vu AT, Jamison K, Glasser M, Smith S, Coalson T, Moeller S, Auerbach E, Ugurbil K, Yacoub E (In Press) Tradeoffs in pushing the spatial resolution of fMRI for the 7 T Human Connectome Project. Neuroimage. Vu AT, Phillips J, Kay K, Phillips M, Johnson M, Shinkareva S, Tubridy S, Millin R, Grossman M, Gureckis T, Bhattacharyya R, Yacoub E (2016) Using precise word timing information improves decoding accuracy in a multiband-accelerated multimodal reading experiment. Cognitive Neuropsychology 33:265-275. White T, O'Leary D, Magnotta V, Arndt S, Flaum M, Andreasen NC (2001) Anatomic and functional variability: the effects of filter size in group fMRI data analysis. Neuroimage 13:577-588. Xu J, Moeller S, Auerbach EJ, Strupp J, Smith SM, Feinberg DA, Yacoub E, Ugurbil K (2013) Evaluation of slice accelerations using multiband echo planar imaging at 3 T. Neuroimage 83:991-1001. Yacoub E, Harel N, Ugurbil K (2008) High-field fMRI unveils orientation columns in humans. Proceedings of the National Academy of Sciences of the United States of America 105:10607-10612. Yacoub E, Shmuel A, Logothetis N, Ugurbil K (2007) Robust detection of ocular dominance columns in humans using Hahn Spin Echo BOLD functional MRI at 7 Tesla. Neuroimage 37:1161-1177. Yan C-G, Cheung B, Kelly C, Colcombe S, Craddock RC, Di Martino A, Li Q, Zuo X-N, Castellanos FX, Milham MP (2013) A comprehensive assessment of regional variation in the impact of head micromovements on functional connectomics. Neuroimage 76C:183201. Zhao F, Wang P, Hendrich K, Ugurbil K, Kim SG (2006) Cortical layer-dependent BOLD and CBV responses measured by spin-echo and gradient-echo fMRI: insights into hemodynamic regulation. Neuroimage 30:1149-1160. Zimmermann J, Goebel R, De Martino F, van de Moortele PF, Feinberg D, Adriany G, Chaimow D, Shmuel A, Ugurbil K, Yacoub E (2011) Mapping the organization of axis of motion selective features in human area MT using high-field fMRI. PloS one 6:e28716. Uludag K, Ugurbil K (2015) Physiology and Physics of the fMRI Signal. In: fMRI: From Nuclear Spins to Brain Functions, pp 163-213 US: Springer.
N/2+2
2
N=# slices
N+N/2+2
N+2
N+N/2+1
N+1
SLIDER-2 interleaved
N/2+2
2
N/2+1
1
Traditional low resolution
N/2+1
1
Figure 1
4
3
2
1
4
3
2
1
N+4
N+3
N+2
N+1
SLIDER-2 ascending
N+4
N+3
N+2
N+1
Traditional high resolution
High res
Figure 2
Deblurred (λ=0)
Deblurred (λ=0.10)
High res blurred
SLIDER-XD
slice axis
readout axis
Low res
Deblurred (λ=0, PSF=1.25mm)
Deblurred (λ=0.1, PSF=1.8mm)
High res
Figure 3
High res blurred
SLIDER-XD
slice axis
readout axis
Low res
slice blur (relative to HR)
0.92 0
0.94
0.96
0.98
1.02
1.04
1.06
0.05
0.1
Figure 4
0.15
0.2 0.25 0.3 regularization factor
0.35
0.4
0.45
0
0
0.05
0.1
0.15
0.2 0.25 0.3 regularization factor
0.35
0.4
0.45
HRb_db SLIDER
SLIDER-XD
HRb
HR
HRb_db
0.5
0.5
1
1.5
2
2.5
SLIDER
SLIDER-XD
HRb
HR
PSF=1.25 mm
PSF=2.50 mm
mean t-value (relative to HR)
0.5
High res SLIDER-XD
2D FFT, accounting for SNR differences
2D FFT
Mean EPI cross section
Figure 5
slice axis
readout axis
3T, 1.25 mm iso
Low res
Figure 6
slice axis
0
low k-space
mid k-space
high k-space
Relative tSNR
readout axis
1 High res
High res blurred
SLIDER-XD
L
R
Figure 7
ODC map
F-value
3T, 18 min experiment
EPI cross section
slice axis
readout axis
No SLIDER x 4 avgs
SLIDER-XD
7T, 3 min experiment
No SLIDER
Figure 8