BOLD fMRI at 9.4T with 3D stack-of-spirals readouts

BOLD fMRI at 9.4T with 3D stack-of-spirals readouts Alejandro Monreal-Madrigal1, Denizhan Kurban1, Desmond H.Y. Tse2, Dimo Ivanov1, Nicolas Boulant3, Benedikt A. Poser1 1Maastricht Brain Imaging Centre, Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, The Netherlands; 2Scannexus BV, Maastricht, The Netherlands; 3University Paris-Saclay, CEA, CNRS, BAOBAB, NeuroSpin, Gif-sur-Yvette, France Abstract In this work we investigate the use of spiral readouts for sub-millimeter BOLD fMRI at 9.4T, and verify simulations of the BOLD PSF with functional experiments. We confirm that a TE < T2* can be employed for spiral-out readouts without compromising BOLD sensitivity. The shorter TE provides a reduced TR, improvement in tSNR and minimizes off-resonance effects. We conclude that spiral-in readouts are mostly useful for lower resolutions at UHF, but that segmented spiral-out readouts can play an important role in mesoscopic BOLD fMRI at ultra high fields (> 7T). 1. INTRODUCTION Functional MRI (fMRI) is an MRI modality that non-invasively measures brain activity. It was first introduced as Blood Oxygenation Level Dependent (BOLD) [1] contrast, following the observation that changes in blood deoxyhemoglobin content during neuronal activation alter the MRI signal. Typical spatial and temporal resolutions for human fMRI experiments have been on the order of a couple of millimeters and seconds, respectively. The desire for higher resolutions in fMRI has been one of the main drivers of ultra-high field MRI [2–4], increasing the field strength provides a nearly quadratic increase in SNR [5]. Gradient coils with higher gradient amplitudes and slew rates are also being developed [6, 7], providing significant speed-up in echo-planar and spiral readouts. Additionally, receive coils with high channel count RF receive arrays [8] are being developed, allowing higher parallel acceleration by k-space undersampling. These hardware advances together with new sequence developments have made it possible to achieve sub-millimeter resolutions and sub-second acquisitions. These tools can be used by neuroscientists to explore the human brain at high temporal and spatial resolutions with unprecedented detail. BOLD fMRI is an indirect measure of neuronal activation, it measures the Hemodynamic Response Function (HRF) which is a combination of different physiological processes: Cerebral Blood Flow (CBF), Cerebral Blood Volume (CBV) and CMRO2 (oxygen consumption) [9]. Non-BOLD contrasts have been proposed to measure those effects more directly, such as calibrated fMRI to measure CMRO2 [10, 11], Arterial Spin Labeling (ASL) [12, 13] for CBF, and Vascular Space Occupancy (VASO) [14, 15] for CBV. At Ultra High Field (UHF) however, these pose different implementation challenges, for example, ASL and VASO make use of RF preparation modules that increase the minimum achievable repetition time (TR), and require a high use of RF power for the saturation and inversion pulses [2]. For these reasons, BOLD contrast is still the most widely used in fMRI studies also at UHF. In functional imaging, small changes in T2*, caused by the local change in deoxyhemoglobin concentration that accompanies neuronal activation manifest and thus the highest BOLD signal difference between activation and rest is observed when TE ~ T2* of the gray matter [16]. The most commonly used readout strategy for BOLD fMRI is Cartesian echo-planar imaging (EPI) [17], which achieves high sampling efficiency and enables volume repetition times on the order of seconds at the typically desired resolutions for fMRI applications. In the past decade, the incorporation of CAIPIRINHA [18] acceleration into 2D simultaneous multi-slice [19–21] and 3D EPI [22–24] has improved EPI and further increased its neuroscientific use. Alternatives to EPI for fast imaging are non-Cartesian readouts, of which the most common are spiral and radial [25, 26]. Other non-Cartesian readouts that have been proposed for fMRI can be found in [27–29]. Spiral readouts with full k-space sampling were introduced for BOLD fMRI at 3T [30–32]. Following advances in spiral reconstruction and off-resonance correction [33, 34], spiral imaging has been revisited at 7T, by Engel et al. [35] and for 2D BOLD fMRI by Kasper et al. [36]. Recently, spiral readouts have also been used for balanced SSFP BOLD in work by Valsala et al. [37], in which the authors exploited the flexibility to choose TE to improve the functional contrast. Spirals have also been used for different non-BOLD contrasts such as ASL [38] and VASO [39], where a short echo time is desired to improve the inherent low SNR of these techniques. Since these types of readouts do not fall into a Cartesian grid, the use of the Fast Fourier Transform is not possible; the reconstruction is then typically performed offline and it is a computationally expensive process. The simplest design of a spiral readout follows an Archimedean spiral. In this case, an entire k-space plane (2D) is acquired after an excitation (single-shot). Each plane can also be acquired using more excitations (multi-shot) which shortens each readout and reduces vulnerability to B0 offresonance, however at the cost of increased volume TR. The k-space can be transversed from the center to the edge (spiral-out) or starting from the edge and moving inwards (spiral-in). A combination of both (spiral-out-in) acquires two echos per excitation. The center of the k-space can be sampled more densely than the outer part (variable density spirals). Stacking spiral planes in the partition direction produces a 3D stack-of-spirals pattern. To improve the undersampling behaviour, k-space planes can be rotated [40] . All these parameters affect the contrast, acquisition duration and overall image quality and careful selection of them is important for a successful implementation of spiral imaging. The widespread use of spiral readouts has been limited due to practical challenges in their implementation. One of them has been the need to account for trajectory deviations as well as eddy currents, which requires knowledge of the trajectory executed by the scanner. The trajectory can be measured with imaging techniques [41], by use of special hardware such as field cameras [34], or measurement of the Gradient Impulse Response Function [42] which can subsequently be applied to predict the achieved gradient waveform. Another challenge is the reconstruction process, which typically makes use of the computationally demanding Non-Uniform Fourier Transform NUFFT [43], and hence often impractical to implement on the scanner's own image reconstruction hardware. Finally, B0 off-resonance results in spiral image blurring with a 2D point spread that is much harder to correct for than the 1D phase-encoding distortion in EPI, which can be addressed with straightforward post-processing (e.g. FSL top-up [44]). The continuously changing phase encoding direction and rate of k-space traversal in spirals cause a complex phase accrual and blurring that need to be addressed during the image reconstruction [45], considerably adding extra computational demands and reconstruction times. To avoid off-resonance effects due to B0 field inhomogeneities when using long spiral readouts, an off-resonance frequency term needs to be included in the SENSE signal model [46]: 𝑆𝛾(𝑡) = ∫𝑉𝑐𝛾(𝑟)𝑚(𝑟)𝑒−𝑖𝑘(𝑡)𝑟𝑒−𝑖𝛥𝜔0(𝑟)𝑡𝑑𝑉 (1) where s(t) is the k-space data from each receiver channel ɣ, c(r) is the complex spatial sensitivity of the ɣ coil, m(r) is the magnetization, k(t) is the k-space trajectory, and the term Δω0 is the angular off-resonance frequency, proportional to field inhomogeneities. Different methods for off-resonance correction exist, such as time segmented [47], multi-frequency interpolation [48], and gridding approaches [49]. Equation 1 can be discretized and a linear system of equations be created with the k-space data: 𝑠 = 𝐸𝑚 , 𝐸 = 𝑐𝛾(𝑟)𝑒−𝑖𝑘(𝑡)𝑟𝑒−𝑖𝛥𝜔0(𝑟)𝑡 (2) This system of equations can be solved iteratively with a regularized least-squares cost function: 𝑚 = 𝑎𝑟𝑔𝑚𝑖𝑛1 2||𝐸𝑚 − 𝑠||2 2+𝛽 2𝑅(𝑚), (3) where β is the regularization parameter and R(m) is a regularization function that is used to ensure that the algorithm converges to a stable solution, common regularization functions are L1 norm, L2 norm and total variation (TV) [50]. Different algorithms exist to solve equation 3, some of the most widely used are CG-SENSE [33], ADMM [51], and FISTA [52]. In BOLD fMRI the highest contrast between activation and rest is observed at a TE approximately equal to T2*. Engel et al. [53] proposed the concept of the BOLD point spread function (PSF) to better characterize the impact of the readout and TE on BOLD functional experiments. Using simulations, they showed that a TE ~ T2* achieves the highest BOLD sensitivity with spiral readouts. In MR Fourier encoding, the point spread function (PSF) is the inverse Fourier of the k-space filter H(k): 𝑃𝑆𝐹(𝑟) = Ƒ−1(𝐻(𝑘)) (4) with 𝑘 = [𝑘𝑥,𝑘𝑦,𝑘𝑧]𝑇 . The k-space filter H(k) comprises the sampling of k-space at discrete points on a finite support and a weighting function: 𝐻(𝑘) = 𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔(𝑘) ⋅ 𝐻𝑤𝑒𝑖𝑔ℎ𝑡(𝑘) (5) In gradient echo experiments, a mono-exponential signal decay weighting 𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝑇2∗is imposed. In a BOLD experiment, the measure of interest is the small differences arising from temporal changes in T2*. The BOLD weight filter 𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝐵𝑂𝐿𝐷 is the derivative of the T2* weight filter: 𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝑇2∗(𝑘) = 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗ ,𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝐵𝑂𝐿𝐷 (𝑘) = 𝜕𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝑇2∗(𝑘) 𝜕𝑇2∗=𝑡(𝑘) 𝑇2 ∗2 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗ (6) BOLD imaging is then characterized by a differential PSF: 𝑃𝑆𝐹𝐵𝑂𝐿𝐷(𝑥,𝑇2 ∗,𝛥𝑇2 ∗) = 𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗ +𝛥𝑇2 ∗) − 𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗) = 𝜕𝑃𝑆𝐹𝐺𝐸(𝑥,𝑇2∗) 𝜕𝑇2∗𝛥𝑇2 ∗ = Ƒ−1(𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔𝐻𝑤𝑒𝑖𝑔ℎ𝑡 𝐵𝑂𝐿𝐷 )𝛥𝑇2 ∗ = Ƒ−1(𝐻𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔(𝑘)𝑡(𝑘) 𝑇2 ∗2 𝑒𝑥𝑝−𝑡(𝑘)/𝑇2∗)𝛥𝑇2 ∗ (7) In this work we combine some of the previously mentioned hardware and sequence advances. We use a 3D stack-of-spiral sequence for BOLD fMRI on a 9.4T scanner, equipped with high-performance head-only gradients for more efficient k-space coverage. We experimentally verify the BOLD point spread function simulations by Engel et al. [53] for spiral readouts and further investigate the benefits of spiral readouts for sub-millimeter fMRI at UHF. We also provide the complete pipeline for sequence, reconstruction and analysis for spiral fMRI data. The goal of this project was to better characterize the tradeoffs of spiral BOLD at laminar resolution and gain insights into the challenges when aiming for yet higher resolutions (< 0.6 mm) fMRI and field strengths (> 9.4T) as envisaged by the AROMA project. With further shortening of T2* and increases in B0 inhomogeneities single-shot EPI will also become increasingly challenging at high resolution with the prohibitively long readout precluding suitably short TE, unless using segmentation which comes at TR and tSNR penalties. 2. METHODS Simulations of the BOLD point spread function were performed for a range of T2* values (T2*= 20-26 ms) to determine a suitable TE for 0.6 mm isotropic resolution BOLD fMRI with dual-shot spiral-out (DS-SO) and dual-shot spiral-in (DS-SI) readouts. In previous work we concluded that highresolution single-shot spiral-out images at 9.4T suffer severe off-resonance effects and are not usable [54], for this reason we focus on dual-shot spiral readouts in this work. Echo times ranging from 2 to 30 ms were used for the spiral-out simulations, and 30 to 50 ms for spiral-in. Figure 1: BOLD imaging filter (HBOLD) schematic depiction and main components of the BOLD point spread function from a spiral-out readout. (a) k-space vs HBOLD and cross-section. (b) Real part of the BOLD PSF and its sections. The main lobe is divided into nominal main lobe (nominal voxel width) and residual main lobe. Side lobes include both the negative and positive ones. The real part of the PSF was divided in three different sections: nominal main lobe, residual main lobe and side lobes (Figure 1b), which were used to characterize each PSF similarly to [53, 55]. BOLD resolution (FWHM of main lobe), sensitivity (integral over the nominal main lobe), and specificity (relation of the integral over nominal main lobe and the integral of the absolute value of the residual main lobe and side lobes) were used as metrics: 𝑟𝑒𝑠𝐵𝑂𝐿𝐷 = 𝐹𝑊𝐻𝑀𝑃𝑆𝐹 , 𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 = ∑𝑛𝑜𝑚.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒 , (8)𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 = ∑ 𝑛𝑜𝑚.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒 ∑|𝑠𝑖𝑑𝑒 𝑙𝑜𝑏𝑒𝑠| +∑| 𝑟𝑒𝑠.𝑚𝑎𝑖𝑛 𝑙𝑜𝑏𝑒| Signal-to-noise ratio estimates were also obtained, the SNR of an image is proportional to: 𝑆𝑁𝑅 ∝ 𝑥𝑦𝑧√𝑛𝑠ℎ𝑜𝑡𝑠𝑇𝑎𝑐𝑞 𝑓(𝑇𝑅,𝑇𝐸) (9) 𝑓(𝑇𝑅,𝑇𝐸) ∝ (1−𝐸)𝑠𝑖𝑛(𝛼) 1−𝑐𝑜𝑠(𝛼)𝐸 𝑒 −𝑇𝐸 𝑇2 ∗, 𝐸 = 𝑒−𝑇𝑅 𝑇1 (10) where x, y, z are the voxel dimensions, nshots is the number of interleaves (lines or segments per plane), Tacq is the duration of the acquisition readout and f(TR,TE) is a function describing the dependence on the relaxation characteristics of the tissue for gradient echo. Data were acquired on a Siemens 9.4T scanner equipped with a 16Tx-31Rx head coil [56] and an AC-84II head-only gradient with 80 mT/m peak amplitude and 330 T/m/s slew rate. When using spiral readouts covering a circular k-space, an area similar to that of a square k-space should be used to achieve the desired resolution. In this work, we used a k-space radius which is 1.13 times larger than the nominal one as proposed in [57]. To do so, an in-plane resnom=0.54x0.54x0.60 mm3 was used, resulting in an effective resolution reseff=0.6 mm isotropic. The spiral readouts were designed with a variable density 𝛼=1.3, in-plane under-sampling Rxy=3.3 and a 120° rotation of segments between planes. An optimal gradient design algorithm [58] was used to achieve the desired k-space trajectory in the shortest time with Gmax=50 mT/m, SRmax=250 T/m/s and BW=312 kHz. A total of nine healthy volunteers were scanned after providing written consent following the protocols of the local ethics committee. Four volunteers were scanned to validate the BOLD point spread function simulations. Five volunteers were later scanned to assess the relative utility of spiral-in and spiral-out readouts. To address the B1+ inhomogeneities in the visual cortex, we used pre-calculated RF shim settings (phase-only), previously obtained as an average of several subjects. To validate the BOLD PSF simulations, a sequence with a dual-shot spiral-out (DS-SO) stackof-spirals readout and two different nominal echo times (6 and 12 ms) was implemented in Pulseq [59]. The echo time of 6 ms was selected as a good compromise between the different quality metrics and the 12 ms one since simulations suggest it gives the highest sensitivity at T2*=22 ms, expected value of GM at 9.4T [60]. Budde et al. [61] reported an average GM T2* of 28.3 ± 6.8 ms at 9.4T, obtained from ME-GRE scans with 0.35x0.35x2.00 mm3 resolution. Plane partitions were interleaved between echo times resulting in effectively simultaneous acquisition of the fMRI BOLD response at the two echo times. Fat saturation was applied before each excitation. The sequence furthermore contained a 1 ms long navigator module consisting of two 0.3 ms non-phase encoded FID 0.4 ms apart, applied at TE=2 ms. The following parameters were used: FOV=140x140x18 mm3, 30 kz partitions with linear encoding order, TE1/TE2 6/12 ms, TRshot=45 ms, TRvolume=2.7 s,TRpair=5.41 s, FA=15° and BWTP=25. This resulted in a readout duration of 27 ms per shot. The sequence used RF spoiling with quadratic phase increments and gradient spoiling in three axes. A schematic depiction of this sequence can be found in figure 2a. Seven 15 minutes functional runs were acquired (three volunteers underwent 2 runs, and another one 1 run). A full-screen flickering (approx. 8Hz) black and white radial checkerboard programmed in Psychopy [62] was used as stimuli. Each block consisted of 7 TRpair of rest (38 s) followed by 7 TRs of activation (38 s), effectively acquiring 14 volumes in each block (7 per echo time); each run consisted of 154 volumes per TE. The block design, stimuli and slice positioning can be found in figure 2e. To further investigate the performance of different spiral readouts, three volunteers were scanned with a dual-shot spiral-out (DS-SO) and dual-shot spiral-in (DS-SI) sequences. Most of the parameters were the same as described above, except for FOV=140x140x24 mm3 and 40 kz partitions. The DS-SO had a TE=6 ms, TRshot=48 ms, TRvolume=3.88 s, FA=11° and a readout duration of 27 ms (per shot) whereas the DS-SI had a TE=31 ms, TRshot=47 ms, TRvolume=3.76 s, FA=11° and a readout duration of 27 ms (per shot). The TE=6 ms ultimately was chosen following the simulations which show that it is a reasonable compromise between the different metrics used in this work. To test the acquisition of a larger FOV, two additional volunteers were scanned with the DSSO sequence as described, and in addition with doubled FOVpartition (48 mm instead of 24 mm), but with Rxy=3, Rz=2 to achieve a similar TR and 8% phase oversampling to avoid foldover in the center slices; all other parameters were kept constant. Since a coronal view provides limited sensitivity changes in the AP direction, a kz blip was applied between shots to improve the aliasing pattern in the partition direction. This resulted in a CAIPIRINHA-like [18] pattern with shift Δ=1 along kz, effectively resulting in an undersampling of Rxy=6 and Rz=1. A schematic depiction of these sequences and k-space trajectories can be found in figure 2 b,c,d. Each functional run lasted approximately 12 minutes, with the block design consisting of 8 TRs of rest followed by 8 TRs of activation. For DS-SO and DS-SI , this resulted in 31 s and 30 s block duration, respectively, and 192 volumes were acquired for each run. The same full-screen flickering checkerboard was used for the BOLD PSF experiments. A 3D low-resolution fully-sampled multi-echo GRE acquisition with segmented spiral-out readout was acquired before each functional run and used to calculate sensitivity and B0 maps. This sequence was FOV matched to the fMRI scans and used the following parameters: resnom=1.1x1.1x0.6 mm3, TE1/TE2/TE3=2.3/6.5/10.7 ms, TR=62 ms (for SAR restrictions), FA=11° and BWTP=25. Spiral parameters: 32 shots, Rxyz=1 , variable density 𝛼=1, resulting in a readout duration of 3.2 ms (per shot) and total scan duration of 1.5 and 3 minutes for the 24 mm and 48 mm partition FOV, respectively. Figure 2: Schematic depiction of sequences, k-space trajectory, stimuli and slice positioning. (a) Sequence used for the BOLD PSF experiments. The partitions of both TEs are acquired interleaved (e.g. TE1 (blue): 1st shot, 1st partition -> TE1 (blue): 2nd shot, 1st partition -> TE2 (purple): 1st shot, 1st partition -> TE2 (purple): 2nd shot, 1st partition, etc). A fat saturation module is used before the first excitation of each partition. The DS-SO (b) and DS-SI (c) sequence used for the spiral-in and spiral-out experiments. A fat saturation module is used before each excitation. (d) schematic depiction of the k-space trajectories used in this work, for both the FOVpartition=24 mm (Rz=1) and the FOVpartition=48 mm (Rz=2), for the latter, a blip is applied after each shot to achieve CAIPIRINHA-like k-space coverage in the partition (kz) direction. (e) FOV orientation, and paradigm block-design using a flicker checkerboard. Image reconstruction was performed offline using a pipeline based on the one presented in [39], using Neurodesk [63]. The first echo of the low-resolution ME-GRE scan was used to calculate the sensitivity maps using the ESPiRIT [64] implementation in MRIReco.jl [65]. These maps were then compressed to 16 virtual channels [66]. The three echoes were used to calculate the off-resonance map using the regularized field map estimation [67] as implemented in MRIFieldmaps.jl. The GIRF predicted trajectory and k0 modulations were estimated from the nominal trajectory using previously acquired gradient impulse response function (GIRF) data following the method described in [68]. Dynamic off-resonance correction in k-space (DORK) [69] was applied to the raw data to correct for scanner drift. The phase of the first FID navigator of each volume was compared to the second one of the fMRI timeseries and any phase difference was removed from the raw data. The DORK corrected functional raw data was merged with the GIRF predicted trajectory into an MRD file [70]. The sensitivity maps, off-resonance maps and MRD file were the input to the reconstruction performed in MRIReco.jl using an ADMM solver and “L1” regularization, including a density compensation function [71] and multi-frequency interpolation B0 correction [48]. The number of iterations and stopping criteria was set to 30 and 1.2x10-7, respectively. The number of bins was selected as 4𝛥𝜔𝑚𝑎𝑥𝑇 𝜋, where 𝛥𝜔𝑚𝑎𝑥 is the maximum (absolute) off-resonance frequency and 𝑇 is the readout duration, resulting in 16 bins for the 27 ms long readout. Reconstruction was performed on a dedicated server running Ubuntu 22.04, with 1 TB of memory and two Intel Xeon Scalable Processor “Skylake” Gold 6140 with 18 cores each, for a total of 36 cores. Figure 3 shows an overview of the acquisition, reconstruction and analysis pipeline. The data analysis pipeline was implemented in Neurodesk [63] and consisted of: (1) Motion correction using AFNI’s 3dAllineate command [72], the time series was registered to the second volume of the acquisition. (2) High-pass temporal filter using FSL with an FWHM of the duration of rest/activity blocks. (3) Mean timeseries and tSNR were computed as quality metrics using AFNI. (4) Activation maps were computed using a General Linear Model (GLM) and clustered using AFNI’s 3dDeconvolve and 3dclust commands. BOLD sensitivity in terms of BOLD contrast to temporal noise ratio (tCNR) was calculated for each voxel average block as tCNR = Δs/σn, with Δs = mean BOLD signal change (mean activation - mean rest) and σn = standard deviation of GLM residual. No spatial smoothing was applied. Figure 3: Overview of the 3D stack-of-spirals acquisitions, reconstruction, and analysis pipeline. Sequence: the Pulseq sequences are generated in MATLAB; nominal trajectories and parameter dictionaries are saved. The nominal trajectories are used to obtain the GIRF predicted ones. Fully sampled, highly segmented low-resolution GRE-ME and spiral functional data are acquired; both data sets are converted into MRD files. Reconstruction: The DORK-corrected spiral raw data is merged with the GIRF predicted trajectory and other parameters into an MRD file. Sensitivity and B0 maps are calculated from the low-resolution ME-GRE scan. The MRD file, sensitivity and B0 maps are the input to the reconstruction in MRIReco.jl. Analysis: Rigid motion correction is performed with AFNI, high-pass filter in FSL and the statistical analysis is done with AFNI. The Reconstruction and analysis is done in Neurodesk. Declaration of competing interest The authors have no competing interests to report. Funding This work has received financial support from the European Union Horizon 2020 Research and Innovation program under grant agreement no. 885876 (AROMA). Supplementary material Supplementary figure 1: 0.8 mm isotropic data. (a) mean timeseries, (b) effective tSNR and (c) activation maps. Single-shot spiral-out images present severe off-resonance effects even after correction, dual-shot acquisitions have good image quality. For this resolution, the dual-shot spiral-in acquisition has good tSNR and activation, these results differ from the 0.6 mm isotropic data. Additional details can be found in [54]. REFERENCES 1. Ogawa S, Tank DW, Menon R, Ellermann JM, Kim SG, Merkle H, Ugurbil K (1992) Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci 89:5951–5955. 2. Ivanov D, De Martino F, Formisano E, Fritz FJ, Goebel R, Huber L, Kashyap S, Kemper VG, Kurban D, Roebroeck A, Sengupta S, Sorger B, Tse DHY, Uludağ K, Wiggins CJ, Poser BA (2023) Magnetic resonance imaging at 9.4 T: the Maastricht journey. Magn Reson Mater Phys Biol Med 36:159–173. 3. Sadeghi-Tarakameh A, DelaBarre L, Lagore RL, Torrado-Carvajal A, Wu X, Grant A, Adriany G, Metzger GJ, Van de Moortele P-F, Ugurbil K, Atalar E, Eryaman Y (2020) In vivo human head MRI at 10.5T: A radiofrequency safety study and preliminary imaging results. Magn Reson Med 84:484–496. 4. Boulant N, Quettier L, Aubert G, Amadon A, Belorgey J, Berriaud C, Bonnelye C, Bredy Ph, Chazel E, Dilasser G, Dubois O, Giacomini E, Gilgrass G, Gras V, Guihard Q, Jannot V, Juster FP, Lannou H, Leprêtre F, Lerman C, Le Ster C, Luong M, Mauconduit F, Molinié F, Nunio F, Scola L, Sinanna A, Touzery R, Védrine P, Vignaud A, the Iseult Consortium (2023) Commissioning of the Iseult CEA 11.7 T whole-body MRI: current status, gradient–magnet interaction tests and first imaging experience. Magn Reson Mater Phys Biol Med 36:175–189. 5. Le Ster C, Grant A, Van de Moortele P-F, Monreal-Madrigal A, Adriany G, Vignaud A, Mauconduit F, Rabrait-Lerman C, Poser BA, Uğurbil K, Boulant N (2022) Magnetic field strength dependent SNR gain at the center of a spherical phantom and up to 11.7T. Magn Reson Med 88:2131–2138. 6. Feinberg DA, Beckett AJS, Vu AT, Stockmann J, Huber L, Ma S, Ahn S, Setsompop K, Cao X, Park S, Liu C, Wald LL, Polimeni JR, Mareyam A, Gruber B, Stirnberg R, Liao C, Yacoub E, Davids M, Bell P, Rummert E, Koehler M, Potthast A, Gonzalez-Insua I, Stocker S, Gunamony S, Dietz P (2023) Next-generation MRI scanner designed for ultra-high-resolution human brain imaging at 7 Tesla. Nat Methods 20:2048–2057. 7. Huang SY, Witzel T, Keil B, Scholz A, Davids M, Dietz P, Rummert E, Ramb R, Kirsch JE, Yendiki A, Fan Q, Tian Q, Ramos-Llordén G, Lee H-H, Nummenmaa A, Bilgic B, Setsompop K, Wang F, Avram AV, Komlosh M, Benjamini D, Magdoom KN, Pathak S, Schneider W, Novikov DS, Fieremans E, Tounekti S, Mekkaoui C, Augustinack J, Berger D, Shapson-Coe A, Lichtman J, Basser PJ, Wald LL, Rosen BR (2021) Connectome 2.0: Developing the next-generation ultrahigh gradient strength human MRI scanner for bridging studies of the micro-, mesoand macro-connectome. NeuroImage 243:118530. 8. Wiggins GC, Polimeni JR, Potthast A, Schmitt M, Alagappan V, Wald LL (2009) 96-Channel receiveonly head coil for 3 Tesla: Design optimization and evaluation. Magn Reson Med 62:754– 762. 9. Uludağ K (2023) Physiological modeling of the BOLD signal and implications for effective connectivity: A primer. NeuroImage 277:120249. 10. Davis TL, Kwong KK, Weisskoff RM, Rosen BR (1998) Calibrated functional MRI: Mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci 95:1834–1839. 11. Hoge RD, Atkinson J, Gill B, Crelier GR, Marrett S, Pike GB (1999) Investigation of BOLD signal dependence on cerebral blood flow and oxygen consumption: The deoxyhemoglobin dilution model. Magn Reson Med 42:849–863. 12. Alsop DC, Detre JA, Golay X, Günther M, Hendrikse J, Hernandez-Garcia L, Lu H, MacIntosh BJ, Parkes LM, Smits M, van Osch MJP, Wang DJJ, Wong EC, Zaharchuk G (2015) Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 73:102–116. 13. Calamante F, Thomas DL, Pell GS, Wiersma J, Turner R (1999) Measuring Cerebral Blood Flow Using Magnetic Resonance Imaging Techniques. J Cereb Blood Flow Metab 19:701–735. 14. Lu H, Golay X, Pekar JJ, van Zijl PCM (2003) Functional magnetic resonance imaging based on changes in vascular space occupancy. Magn Reson Med 50:263–274. 15. Lu H, Hua J, van Zijl PCM (2013) Noninvasive functional imaging of cerebral blood volume with vascular-space-occupancy (VASO). NMR Biomed 26:932–948. 16. Bandettini PA, Wong EC, Hinks RS, Tikofsky RS, Hyde JS (1992) Time course EPI of human brain function during task activation. Magn Reson Med 25:390–397. 17. Mansfield P (1977) Multi-planar image formation using NMR spin echoes. J Phys C Solid State Phys 10:L55. 18. Breuer FA, Blaimer M, Mueller MF, Seiberlich N, Heidemann RM, Griswold MA, Jakob PM (2006) Controlled aliasing in volumetric parallel imaging (2D CAIPIRINHA). Magn Reson Med 55:549–556. 19. Setsompop K, Gagoski BA, Polimeni JR, Witzel T, Wedeen VJ, Wald LL (2012) Blipped-controlled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g - factor penalty: Blipped-CAIPI for Simultaneous Multislice EPI. Magn Reson Med 67:1210– 1224. 20. Feinberg DA, Setsompop K (2013) Ultra-fast MRI of the human brain with simultaneous multislice imaging. J Magn Reson 229:90–100. 21. Barth M, Breuer F, Koopmans PJ, Norris DG, Poser BA (2016) Simultaneous multislice (SMS) imaging techniques. Magn Reson Med 75:63–81. 22. Narsude M, Gallichan D, van der Zwaag W, Gruetter R, Marques JP (2016) Three-dimensional echo planar imaging with controlled aliasing: A sequence for high temporal resolution functional MRI: 3D-EPI-CAIPI: A Sequence for High Temporal Resolution fMRI. Magn Reson Med 75:2350–2361. 23. Poser BA, Kemper VG, Ivanov D, Kannengiesser SA, Uludag K, Barth M (2013) CAIPIRINHAaccelerated 3D EPI for high temporal and/or spatial resolution EPI acquisitions. Proc. ESMRMB Congr. Toulouse Fr. p 226 24. Stirnberg R, Stöcker T (2021) Segmented K-space blipped-controlled aliasing in parallel imaging for high spatiotemporal resolution EPI. Magn Reson Med 85:1540–1551. 25. Delattre BMA, Heidemann RM, Crowe LA, Vallée J-P, Hyacinthe J-N (2010) Spiral demystified. Magn Reson Imaging 28:862–881. 26. Feng L (2022) Golden-Angle Radial MRI: Basics, Advances, and Applications. J Magn Reson Imaging 56:45–62. 27. Zahneisen B, Grotz T, Lee KJ, Ohlendorf S, Reisert M, Zaitsev M, Hennig J (2011) Threedimensional MR-encephalography: Fast volumetric brain imaging using rosette trajectories. Magn Reson Med 65:1260–1268. 28. Assländer J, Zahneisen B, Hugger T, Reisert M, Lee H-L, LeVan P, Hennig J (2013) Single shot whole brain imaging using spherical stack of spirals trajectories. NeuroImage 73:59–70. 29. Lazarus C, Weiss P, Chauffert N, Mauconduit F, Bottlaender M, Vignaud A, Ciuciu P SPARKLING: Novel Non-Cartesian Sampling Schemes for Accelerated 2D Anatomical Imaging at 7T Using Compressed Sensing. 30. Lai S, Glover GH (1998) Three-dimensional spiral fMRI technique: A comparison with 2D spiral acquisition. Magn Reson Med 39:68–78. 31. Glover GH, Law CS (2001) Spiral-in/out BOLD fMRI for increased SNR and reduced susceptibility artifacts. Magn Reson Med 46:515–522. 32. Hu Y, Glover GH (2007) Three-dimensional spiral technique for high-resolution functional MRI. Magn Reson Med 58:947–951. 33. Pruessmann KP, Weiger M, Börnert P, Boesiger P (2001) Advances in sensitivity encoding with arbitrary k -space trajectories: SENSE With Arbitrary k -Space Trajectories. Magn Reson Med 46:638–651. 34. Barmet C, Zanche ND, Pruessmann KP (2008) Spatiotemporal magnetic field monitoring for MR. Magn Reson Med 60:187–197. 35. Engel M, Kasper L, Barmet C, Schmid T, Vionnet L, Wilm B, Pruessmann KP (2018) Single‐shot spiral imaging at 7 T. Magn Reson Med 80:1836–1846. 36. Kasper L, Engel M, Heinzle J, Mueller-Schrader M, Graedel NN, Reber J, Schmid T, Barmet C, Wilm BJ, Stephan KE, Pruessmann KP (2022) Advances in spiral fMRI: A high-resolution study with single-shot acquisition. NeuroImage 246:118738. 37. Iyyappan Valsala P, Veldmann M, Bosch D, Scheffler K, Ehses P Submillimeter balanced SSFP BOLD–functional MRI accelerated with 3D stack-of-spirals at 9.4 T. Magn Reson Med. doi: 10.1002/mrm.30064 38. Kurban D, Ivanov D, Kashyap S, Huber L, Liberman G, Poser BA (2022) Concurrent CBF and BOLD fMRI with dual-echo spiral simultaneous multi-slice acquisitions at 7T. NeuroImage 247:118820. 39. Monreal-Madrigal A, Kurban D, Huber L, Ivanov D, Boulant N, Poser BA (2024) Combining the benefits of 3D acquisitions and spiral readouts for VASO fMRI at UHF. Imaging Neurosci 2:1– 14. 40. Deng W, Zahneisen B, Stenger VA (2016) Rotated stack-of-spirals partial acquisition for rapid volumetric parallel MRI: Stack-of-Spirals, CAIPIRINHA, GRAPPA. Magn Reson Med 76:127– 135. 41. Duyn JH, Yang Y, Frank JA, van der Veen JW (1998) Simple Correction Method fork-Space Trajectory Deviations in MRI. J Magn Reson 132:150–153. 42. Vannesjo SJ, Haeberlin M, Kasper L, Pavan M, Wilm BJ, Barmet C, Pruessmann KP (2013) Gradient system characterization by impulse response measurements with a dynamic field camera: Gradient System Characterization with a Dynamic Field Camera. Magn Reson Med 69:583– 593. 43. Fessler JA, Sutton BP (2003) Nonuniform fast Fourier transforms using min-max interpolation. IEEE Trans Signal Process 51:560–574. 44. Andersson JLR, Skare S, Ashburner J (2003) How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. NeuroImage 20:870–888. 45. Haskell MW, Nielsen J-F, Noll DC (2022) Off-resonance artifact correction for magnetic resonance imaging: a review. 46. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P (1999) SENSE: Sensitivity encoding for fast MRI. Magn Reson Med 42:952–962. 47. Noll DC, Meyer CH, Pauly JM, Nishimura DG, Macovski A (1991) A homogeneity correction method for magnetic resonance imaging with time-varying gradients. IEEE Trans Med Imaging 10:629–637. 48. Man L-C, Pauly JM, Macovski A (1997) Multifrequency interpolation for fast off-resonance correction. Magn Reson Med 37:785–792. 49. Knopp T, Eggers H, Dahnke H, Prestin Jü, Senegas J (2009) Iterative Off-Resonance and Signal Decay Estimation and Correction for Multi-Echo MRI. IEEE Trans Med Imaging 28:394–404. 50. Fessler JA (2010) MODEL-BASED IMAGE RECONSTRUCTION FOR MRI. IEEE Signal Process Mag 27:81–89. 51. Boyd S, Parikh N, Chu E, Peleato B, Eckstein J (2011) Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers. Found Trends® Mach Learn 3:1–122. 52. Beck A, Teboulle M (2009) A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems. SIAM J Imaging Sci 2:183–202. 53. Engel M, Kasper L, Pruessmann KP (2022) BOLD PSF: Impact of k-space sampling on T2* contrast. Proc Jt Annu Meet ISMRM-ESMRMB abstract:2446. 54. (2024) Book of Abstracts ESMRMB 2024 Online 40th Annual Scientific Meeting 2–5 October 2024. Magn Reson Mater Phys Biol Med 37:1–781. 55. Feizollah S, Tardif CL (2023) High-resolution diffusion-weighted imaging at 7 Tesla: Single-shot readout trajectories and their impact on signal-to-noise ratio, spatial resolution and accuracy. NeuroImage 274:120159. 56. Shajan G, Kozlov M, Hoffmann J, Turner R, Scheffler K, Pohmann R (2014) A 16-channel dual-row transmit array in combination with a 31-element receive array for human brain imaging at 9.4 T. Magn Reson Med 71:870–879. 57. Pipe JG, Zwart NR (2014) Spiral trajectory design: A flexible numerical algorithm and base analytical equations. Magn Reson Med 71:278–285. 58. Lustig M, Kim S-J, Pauly JM (2008) A fast method for designing time-optimal gradient waveforms for arbitrary k-space trajectories. IEEE Trans Med Imaging 27:866–873. 59. Layton KJ, Kroboth S, Jia F, Littin S, Yu H, Leupold J, Nielsen J-F, Stöcker T, Zaitsev M (2017) Pulseq: A rapid and hardware-independent pulse sequence prototyping framework: Rapid Hardware-Independent Pulse Sequence Prototyping. Magn Reson Med 77:1544–1552. 60. Norris DG, Ladd ME (2023) Chapter 37 - New horizons: Human MRI at extremely high field strengths. In: Markenroth Bloch K, Guye M, Poser BA (eds) Adv. Magn. Reson. Technol. Appl. Academic Press, pp 577–592 61. Budde J, Shajan G, Hoffmann J, Uğurbil K, Pohmann R (2011) Human imaging at 9.4 T using T2*-, phase-, and susceptibility-weighted contrast. Magn Reson Med 65:544–550. 62. Peirce JW (2007) PsychoPy—Psychophysics software in Python. J Neurosci Methods 162:8–13. 63. Renton AI, Dao TT, Johnstone T, Civier O, Sullivan RP, White DJ, Lyons P, Slade BM, Abbott DF, Amos TJ, Bollmann S, Botting A, Campbell MEJ, Chang J, Close TG, Dörig M, Eckstein K, Egan GF, Evas S, Flandin G, Garner KG, Garrido MI, Ghosh SS, Grignard M, Halchenko YO, Hannan AJ, Heinsfeld AS, Huber L, Hughes ME, Kaczmarzyk JR, Kasper L, Kuhlmann L, Lou K, MantillaRamos Y-J, Mattingley JB, Meier ML, Morris J, Narayanan A, Pestilli F, Puce A, Ribeiro FL, Rogasch NC, Rorden C, Schira MM, Shaw TB, Sowman PF, Spitz G, Stewart AW, Ye X, Zhu JD, Narayanan A, Bollmann S (2024) Neurodesk: an accessible, flexible and portable data analysis environment for reproducible neuroimaging. Nat Methods 1–5. 64. Uecker M, Lai P, Murphy MJ, Virtue P, Elad M, Paeuly JM, Vasanawala SS, Lustig M (2014) ESPIRiT-an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. Magn Reson Med 71:990–1001. 65. Knopp T, Grosser M (2021) MRIReco.jl: An MRI reconstruction framework written in Julia. Magn Reson Med 86:1633–1646. 66. Huang F, Vijayakumar S, Li Y, Hertel S, Duensing GR (2008) A software channel compression technique for faster reconstruction with many channels. Magn Reson Imaging 26:133–141. 67. Lin CY, Fessler JA (2020) Efficient Regularized Field Map Estimation in 3D MRI. IEEE Trans Comput Imaging 6:1451–1458. 68. Correction of B0 eddy current effects in spiral MRI - Robison - 2019 - Magnetic Resonance in Medicine - Wiley Online Library. https://onlinelibrary.wiley.com/doi/full/10.1002/mrm.27583. Accessed 27 Mar 2023 69. Pfeuffer J, Van de Moortele P-F, Ugurbil K, Hu X, Glover GH (2002) Correction of physiologically induced global off-resonance effects in dynamic echo-planar and spiral functional imaging. Magn Reson Med 47:344–353. 70. Inati SJ, Naegele JD, Zwart NR, Roopchansingh V, Lizak MJ, Hansen DC, Liu C, Atkinson D, Kellman P, Kozerke S, Xue H, Campbell‐Washburn AE, Sørensen TS, Hansen MS (2017) ISMRM Raw data format: A proposed standard for MRI raw datasets. Magn Reson Med 77:411–421. 71. Pipe JG, Menon P (1999) Sampling density compensation in MRI: Rationale and an iterative numerical solution. Magn Reson Med 41:179–186. 72. Cox RW (1996) AFNI: Software for Analysis and Visualization of Functional Magnetic Resonance Neuroimages. Comput Biomed Res 29:162–173. 73. Gao J-H, Liu H-L (2012) Inflow effects on functional MRI. NeuroImage 62:1035–1039. 74. Renvall V, Nangini C, Hari R (2014) All that glitters is not BOLD: inconsistencies in functional MRI. Sci Rep 4:3920. 75. Quantitative assessment of blood inflow effects in functional MRI signals - Gao - 1996 - Magnetic Resonance in Medicine - Wiley Online Library. https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.1910360219. Accessed 17 Feb 2025 76. Huang F, Vijayakumar S, Li Y, Hertel S, Duensing GR (2008) A software channel compression technique for faster reconstruction with many channels. Magn Reson Imaging 26:133–141. 77. Dietrich BE, Brunner DO, Wilm BJ, Barmet C, Gross S, Kasper L, Haeberlin M, Schmid T, Vannesjo SJ, Pruessmann KP (2016) A field camera for MR sequence monitoring and system analysis. Magn Reson Med 75:1831–1840. 78. Zhao B, Setsompop K, Adalsteinsson E, Gagoski B, Ye H, Ma D, Jiang Y, Ellen Grant P, Griswold MA, Wald LL (2018) Improved magnetic resonance fingerprinting reconstruction with lowrank and subspace modeling. Magn Reson Med 79:933–942. 79. Cao X, Liao C, Iyer SS, Wang Z, Zhou Z, Dai E, Liberman G, Dong Z, Gong T, He H, Zhong J, Bilgic B, Setsompop K (2022) Optimized multi-axis spiral projection MR fingerprinting with subspace reconstruction for rapid whole-brain high-isotropic-resolution quantitative imaging. Magn Reson Med 88:133–150.