Fast 19F spectroscopic imaging with pseudo‐spiral k‐space sampling

Fluorine MRI is finding wider acceptance in theranostics applications where imaging of 19F hotspots of fluorinated contrast material is central. The essence of such applications is to capture ghosting‐artifact‐free images of the inherently low MR response under clinically viable conditions. To serve this purpose, this work introduces the balanced spiral spectroscopic imaging (BaSSI) sequence, which is implemented on a 3.0 T clinical scanner and is capable of generating 19F hotspot images in an efficient manner. The sequence utilizes an all‐phase‐encoded pseudo‐spiral k‐space trajectory, enabling the acquisition of broadband (80 ppm) fluorine spectra free from chemical shift ghosting. BaSSI can acquire a 64 × 64 image with 1 mm × 1 mm voxels in just 14 s, significantly outperforming typical MRSI sequences used in 1H or 31P imaging. The study employed in silico characterization to verify essential design choices such as the excitation pulse, as well as to identify the boundaries of the parameter space explored for optimization. BaSSI's performance was further benchmarked against the 3D ultrashort‐echo‐time balanced steady‐state free precession (3D UTE BSSFP) sequence, a well established method used in 19F MRI, in vitro. Both sequences underwent extensive optimization through exploration of a wide parameter space on a small phantom containing 10 μL of non‐diluted bulk perfluorooctylbromide (PFOB) prior to comparative experiments. Subsequent to optimization, BaSSI and 3D UTE BSSFP were employed to capture images of small non‐diluted bulk PFOB samples (0.10 and 0.05 μL), with variations in the number of signal averages, and thus the total scan time, in order to assess the detection sensitivities of the sequences. In these experiments, the detection sensitivity was evaluated using the Rose criterion (Rc), which provides a quantitative metric for assessing object visibility. The study further demonstrated BaSSI's utility as a (pre)clinical tool through postmortem imaging of polymer microspheres filled with PFOB in a BALB/c mouse. Anatomic localization of 19F hotspots was achieved by denoising raw data obtained with BaSSI using a filter based on the Rose criterion. These data were then successfully registered to 1H anatomical images. BaSSI demonstrated superior detection sensitivity in the benchmarking analysis, achieving Rc values approximately twice as high as those obtained with the 3D UTE BSSFP method. The technique successfully facilitated imaging and precise localization of 19F hotspots in postmortem experiments. However, it is important to highlight that imaging 10 mM PFOB in small mice postmortem, utilizing a 48 × 48 × 48 3D scan, demanded a substantial scan time of 1 h and 45 min. Further studies will explore accelerated imaging techniques, such as compressed sensing, to enhance BaSSI's clinical utility.

Fluorine MRI is finding wider acceptance in theranostics applications where imaging of 19 F hotspots of fluorinated contrast material is central.The essence of such applications is to capture ghosting-artifact-free images of the inherently low MR response under clinically viable conditions.To serve this purpose, this work introduces the balanced spiral spectroscopic imaging (BaSSI) sequence, which is implemented on a 3.0 T clinical scanner and is capable of generating 19 F hotspot images in an efficient manner.The sequence utilizes an all-phase-encoded pseudo-spiral k-space trajectory, enabling the acquisition of broadband (80 ppm) fluorine spectra free from chemical shift ghosting.BaSSI can acquire a 64 Â 64 image with 1 mm Â 1 mm voxels in just 14 s, significantly outperforming typical MRSI sequences used in 1 H or 31 P imaging.
The study employed in silico characterization to verify essential design choices such as the excitation pulse, as well as to identify the boundaries of the parameter space explored for optimization.BaSSI's performance was further benchmarked against the 3D ultrashort-echo-time balanced steady-state free precession (3D UTE BSSFP) sequence, a well established method used in 19 F MRI, in vitro.Both sequences underwent extensive optimization through exploration of a wide parameter space on a small phantom containing 10 μL of non-diluted bulk perfluorooctylbromide (PFOB) prior to comparative experiments.Subsequent to optimization, BaSSI and 3D UTE BSSFP were employed to capture images of small non-diluted bulk PFOB samples (0.10 and 0.05 μL), with variations in the number of signal averages, and thus the total scan time, in order to assess the detection sensitivities of the sequences.In these experiments, the detection sensitivity was evaluated using the Rose criterion (R c ), which provides a quantitative metric for assessing object visibility.The study further demonstrated BaSSI's utility as a (pre)clinical tool through postmortem imaging of polymer microspheres filled with PFOB in a BALB/c mouse.Anatomic localization of 19 F hotspots was achieved by denoising raw data obtained with BaSSI using a filter based on the Rose criterion.These data were then successfully registered to 1

H
Abbreviations: 3D UTE BSSFP, 3D ultrashort-echo-time balanced steady-state free precession; BaSSI, balanced spiral spectroscopic imaging; FA, flip angle; FID, free induction decay; FOV, field of view; F-uTSI, fluorine ultrafast turbo spectroscopic imaging; N SA , number of signal averages; PFC, perfluorocarbon; PFOB, perfluorooctylbromide; PFPE, perfluoropolyether; PLGA, polylacticco-glycolic acid; R c , Rose criterion; SAR, specific absorption rate; SNR, signal-to-noise ratio; T acq , time of acquisition-acquisition time of FID.anatomical images.BaSSI demonstrated superior detection sensitivity in the benchmarking analysis, achieving R c values approximately twice as high as those obtained with the 3D UTE BSSFP method.The technique successfully facilitated imaging and precise localization of 19 F hotspots in postmortem experiments.However, it is important to highlight that imaging 10 mM PFOB in small mice postmortem, utilizing a 48 Â 48 Â 48 3D scan, demanded a substantial scan time of 1 h and 45 min.Further studies will explore accelerated imaging techniques, such as compressed sensing, to enhance BaSSI's clinical utility.4][15][16] Fluorine MRI offers a high degree of intrinsic specificity, as it relies on the physiological absence of fluorine, resulting in the generation of so-called "hotspot" images on an empty background comprised of noise, enabling direct localization and quantification of 19 F signal.

19
F MRI poses several technical challenges.One of the main limitations of this technique is the limited amount of fluorine that can be administered in vivo, which results in inherently low MR signal levels. 17While high-field imaging, typically greater than 6.0 T, can partially overcome this limitation, it is not a feasible solution for clinical applications, which are typically conducted at lower field strengths of 1.5 or 3.0 T. Low signal levels can result in images of low diagnostic value or prolonged scan times.Additionally, clinically viable PFCs, such as perfluorooctylbromide (PFOB), exhibit complex NMR properties, such as spectra with multiple resonance peaks spread over a large chemical shift range and short apparent T 2 relaxation times. 12,18These short T 2 relaxation times further reduce MR signal levels, and the complexity of the 19 F spectra can make it difficult to accurately localize and quantify signals in the images due to ghosting caused by chemical shift. 17method to mitigate the effects of chemical shift ghosting in 19 F MRI is to use molecular structures with simpler NMR spectra, such as perfluoro-crown-ether, perfluoropolyethers (PFPEs), or their derivatives. 3,19However, as PFCs with single NMR peaks and proven biocompatibility are not yet available, the ghosting issue is addressed by adjusting sequence parameters to select a single resonance peak.This is done by exploiting differences in relaxation properties of the peaks 20 or by applying spectrally selective RF pulses that excite a single peak or a group of peaks over a narrow range. 21The ghosting artifact is also addressed by treating it as a multipoint Dixon or spatial deconvolution problem, [22][23][24][25][26] or by applying chemical shift encoding in a multi-echo acquisition scheme. 27While all these methods have been shown to be effective, their applicability is limited by the NMR properties of the contrast material (e.g., distance between resonance lines) as well as the RF bandwidth and gradient strength of the MR system.
An alternative avenue for conducting fluorine MRI independently of the chosen 19 F probe and ensuring the creation of chemical shift artifact-free images is through MRSI.9][30] This technique generates a series of spatially resolved spectra, convertible into images by integrating each spectrum over a specified peak area. 29 demonstrated by Bolo et al. in pharmacokinetics experiments with Cartesian acquisition grids in vivo 31,32 and later by Yildirim et al. in a number of studies employing the multi-spin-echo based fluorine ultrafast turbo spectroscopic imaging (F-uTSI) sequence with radial k-space trajectories both in vitro and in vivo, [33][34][35][36][37] MRSI proves advantageous in acquiring ghosting-free images of virtually any 19 F probe at clinically relevant magnetic fields.However, the point-by-point acquisition approach dictated by the use of all-phase encoding gradients significantly increases acquisition time, rendering spectroscopic imaging less efficient than techniques such as RARE or 3D ultrashort-echo-time balanced steady-state free precession (3D UTE BSSFP) in terms of overall efficiency and clinical viability. 21,38 this study, we introduce a novel 19 F MRSI sequence, termed "balanced spiral spectroscopic imaging (BaSSI)," designed to acquire artifact-free 19 F images of fluorinated imaging agents with high sensitivity and efficiency on a clinical scanner operating at 3.0 T. BaSSI leverages the insights gained from the F-uTSI sequence, aiming to enhance imaging efficiency and bring 19 F MRSI closer to clinical practice.
We assess BaSSI's performance by measuring tiny amounts of PFOB and comparing it with the 3D UTE BSSFP method in vitro.The paper concludes by demonstrating the anatomical localization of 19 F hotspots through postmortem imaging of polymer microspheres filled with PFOB in BALB/c mice.

| Experimental setup
All experiments were conducted using a 3.0 T clinical MR system (Achieva, Philips Healthcare, Best, The Netherlands), equipped with a 5 cm 1 H/ 19 F RF coil designed for small loads, for example, mice that weigh 25-50 g, and a high-performance gradient system capable of delivering a maximum gradient strength of 40 mT/m and a maximum slew rate of 235 T/m/s.The setup included a custom-designed coil (Rapid Biomedical, Rimpar, Germany) and sample holders fabricated using additive methods.The design of the setup ensured accurate positioning of the specimen at the iso-center of the scanner and the RF center of the 19 F coil, to maximize signal-to-noise ratio (SNR) and image quality by minimizing the impact of gradient nonlinearity and inhomogeneity of B 0 and B 1 over the imaging volume.
The accurate and reproducible positioning of the imaging volume improved consistency and comparability of the results of separate experiments conducted with different samples at separate times.The setup also improved efficiency in post-processing and allowed for automated analysis, as it simplified the selection of slices and ROIs for signal and noise calculations.
In all experiments, the setup was tuned and matched to maintain low RF reflection levels (<0.1%) for optimal and consistent RF deposition, ensuring the reliability and accuracy of the results obtained.

| The sequence
The BaSSI sequence is a free induction decay (FID) method that acquires the MR response using an all-phase-encoded sampling scheme.The core sequence, which is a spoiler-free, non-spatially and spectrally selective derivative of the CE-FAST sequence, 30 comprises a broadband excitation pulse, a set of phase encoding and rewinding gradients, and an acquisition window (Figure 1A).The sequence traces the k-space by means of a 2D pseudo-spiral trajectory (Figure 1B), which can be extended to a stack of spirals for 3D imaging (Figure 1C).Spectral properties of BaSSI are optimized to match the NMR properties of PFCs, specifically the dispersed nature of their 19 F spectra.The RF pulse used for excitation is a short block pulse that can achieve a broad bandwidth (>10 kHz).The duration of the pulse, and thus the effective excitation bandwidth, is controlled by the excitation flip angle (FA) and varies between 6.7 and 73.5 μs.After excitation and phase encoding, the FID signal is acquired over a window of 1 ms that starts at acquisition time T acq .The signal is sampled with a rate of 64 and 1 kHz spectral resolution.

| In silico assessment of BaSSI
Numerical simulations were performed to verify design choices, for example, excitation pulse shape, and to gain insight into the transient and steady-state behavior of the BaSSI.These simulations were also instrumental in exploring the parameter space boundaries for subsequent optimization experiments.In all simulations, the Bloch equations were solved in the time domain using an explicit Runge-Kutta 2,3 pair implementation in MATLAB (MathWorks, Natick, MA, USA) 39 for a numerically synthesized spin system mimicking PFOB with seven resonance lines (Figure 2A) and a digital phantom image comprising a single point on an empty background (Figure 2B).The chemical shift positions and intensities of NMR peaks of the synthetic spin system, as well as the relaxation properties, were derived from measurements and literature. 12,21As shown in Figure 2C, the simulations were set to calculate the response of the synthetic spin system over a chemical shift window of 80 ppm, with a resolution of 0.05 ppm.The time step of the solver was set to 15.6 μs to achieve reasonable computation times while maintaining adequate temporal resolution.
To study the transient and steady-state behavior of the sequence and investigate the PSF of the pseudo-spiral trajectory, simulations were run a 64 Â 64 k-space grid with dummy iterations varying from 0 to 504, and 4096 iterations to encode k-space.

| Optimization
Optimization of the BaSSI sequence involved exploring a large parameter space that included the time of acquisition (T acq ), repetition time (T R ), FA, and number of signal averages (N SA ).The boundaries of the parameter space were established through in silico experiments.Additionally, hardware parameters such as the maximum B 1 and power settings of the RF coil, gradient strength and slew rate, as well as safety limitations including specific absorption rate (SAR) and peripheral nerve stimulation, were optimized in accordance with Edition 3.1 of IEC 60601-2-33. 40e aim of the optimization process was to minimize signal loss caused by relaxation and destructive interference of peak ensembles over narrow chemical shift ranges, specifically the CF 2 peaks of PFOB, while maximizing imaging efficiency through the use of short T R times.To achieve this, a phantom containing 10 μL of non-diluted bulk PFOB was imaged with a spatial resolution of 1 mm Â 1 mm on a 64 Â 64 kspace grid.
Optimal gradient settings and B 1 of the experimental setup were identified by sweeping the maximum B 1 of the coil over a range of 50-55 μT, and the gradient strength and gradient slew rate over ranges of 31-35 mT/m and 200-230 T/m/s, respectively.T acq was varied from the shortest possible value of 0.615 ms to 0.8 ms with increments of 1 μs.FA and T R were optimized simultaneously by varying the FA of the CF 2 group from 5 to 55 in steps of 5 , and T R over a range of 2.5 to 5 ms with steps of 0.1 ms.T R was further extended to 1 s with larger steps to identify the maximum SNR that can be reached under full relaxation conditions.

| Optimization of 3D UTE BSSFP
The 3D UTE BSSFP sequence was optimized by exploring a comprehensive parameter space using a phantom containing 10 μL of non-diluted bulk PFOB.The echo time (T E ) was varied from the minimum possible value of 0.115 to 0.2 ms.The T R and FA were optimized simultaneously by varying T R over a range of 2 to 5 ms in steps of 0.1 ms, and FA from 5 to 55 .The offset of excitation was set to the CF 2 group of PFOB, and the pixel bandwidth was set to 1 kHz to achieve SNR for a spatial resolution of 1 mm Â 1 mm Â 1 mm over a field of view (FOV) of 64 mm Â 64 mm Â 64 mm.The data were acquired on an isotropic grid of 64 Â 64 Â 64, and subsequently re-gridded to 80 Â 80 Â 80 during reconstruction.The slice with the highest SNR was used as input for optimization and benchmarking, allowing for a thorough exploration of the parameter space and optimization of the 3D UTE BSSFP sequence for imaging PFOB.
F I G U R E 2 A, Synthetic spin system used in numerical simulations.The spin system is designed to mimic the NMR response of PFOB.B, Input image used to simulate PSF of BaSSI.C, Spectral and temporal simulation parameters used to simulate 2D BaSSI on the synthetic spin system.

| Quantification of SNR
The SNR was calculated using the following expression: where hI obj i and hnoisei represent the mean signal and mean noise of the magnitude image, respectively.The factor 0.8 was used to derive the standard deviation of the complex signal from the mean of the magnitude signal. 41r optimization measurements with large samples, the mean intensity of an ROI located in the sample was used as the signal, whereas for smaller samples the brightest voxel was used as the signal to calculate SNR.For all sample sizes, the mean of an artifact-free ROI outside the sample was used as the noise estimate.
For both BaSSI and 3D UTE BSSFP, SNR values were scaled against the overall maximum values obtained in the respective optimization experiments.

| Benchmarking for detection sensitivity and the Rose criterion
The performance of the 2D BaSSI sequence was compared against that of 3D UTE BSSFP sequence on phantoms containing minute amounts of non-diluted bulk PFOB (0.05 and 0.1 μL).These phantoms featured extremely small objects, with maximum dimensions of just 1 mm in diameter and height.The exact placement of the setup ensured that the signal was contained within a cubic volume measuring 1 mm Â1 mm Â 1 mm.This precision allowed for a direct comparison between a 2D sequence and a 3D sequence, without any noticeable disadvantage to either of the techniques.
The detection performances of the sequences were quantified using the Rose criterion, 42 which measures object visibility.The Rose criterion states that for two objects to be visually separated the difference in their signals should be significantly larger than the noise governing the experiment.This criterion, also referred to as the SNR threshold and denoted by R c , is calculated by dividing the signal difference by the noise.With hI obj i and hnoisei representing the mean signal and mean noise, respectively, the SNR threshold for magnitude images is given by the following expression: For detection with sufficient confidence R c is required to be between 3 and 5, and for detection with high confidence (≥97.5%)R c needs to be at least 4.
Both sequences were executed with effective voxel sizes of 1 mm Â 1 mm Â 1 mm, over an FOV of 64 mm in all directions.Since the objects were very small, the voxel with the highest intensity was used to calculate R c .

| Postmortem imaging of PFOB-filled microbubbles
Polylactic-co-glycolic acid (PLGA) microspheres with mean diameter of 2.4 μm and filled with PFOB were injected into dead mice and imaged using the 3D BaSSI sequence.The goal of these experiments was to anatomically locate 19 F hotspots.They were conducted with a 200 μL microsphere solution containing a PFOB concentration of 10 mM.The positioning of the mice was first surveyed using low-resolution fastgradient-echo multi-slice imaging for 1 H and 2D-BaSSI for 19 F along the three imaging axes.Anatomic images were then obtained with a voxel size of 0.5 mm Â 0.5 mm and slice thickness of 1 mm. 19F hotspot images were obtained using 3D BaSSI (stack of spirals) imaging, with a voxel size of 1 mm Â 1 mm Â 1 mm on a 48 Â 48 Â 48 grid.
To obtain sufficient SNR, N SA was set to 16, resulting in a scan time of 2.15 min per slice and about 1 h 45 min in total. 19F images were filtered by eliminating voxels that did not meet the Rose criterion, to identify voxels with detectable 19 F content.The R c used in the filter was varied from 0.1 to 10.0 to identify the optimal setting that produced clean hotspot images free of false positives, which manifested as spots outside of the expected target region.To anatomically locate the 19 F hotspots, the 19 F and 1 H images were registered by means of the offset and angulation parameters of the acquisitions.

| Preparation of contrast material and phantoms
Non-diluted bulk PFOB phantoms used in the optimization and benchmarking were prepared by pipetting the required volume into centrifuge tubes with a 0.1 mL capacity and weighing the amount of PFOB using a microbalance (Figure 3A).The 10 μL sample used in sequence optimization exhibited a significant height, extending to several millimeters.In contrast, the samples of 0.10 and 0.05 μL used in sequence benchmarking had a height of less than 1 mm, allowing the signal to be confined within a single slice, for both 2D and 3D acquisitions (Figure 3B).
Polymer microspheres encapsulating PFOB used in postmortem imaging were prepared using the protocol described by Pisani et al. 43 In brief, 100 mg of PLGA was dissolved in 4 mL of dichloromethane under sonication.Then, 60 μL of PFOB was added to the solution and stirred for 10 min at room temperature.Afterwards, the organic solution was poured into 20 mL of 1.5% (w/v) sodium cholate aqueous solution.The mixture was emulsified for 1 min at 24 000 rpm and 0 C with an Ultraturrax-T18 disperser (IKA-Werke, Staufen, Germany), and subsequently stirred for F I G U R E 4 A, Simulations verified that a short block pulse excites the CF 2 group with sufficient homogeneity.B, In full sequence simulations, BaSSI exhibited strong transient behavior with rapid signal decay, reaching a steady state in about 500 iterations.C, PSF obtained with no dummy iterations showed maximum blurring of 4% that drops rapidly to 0%, whereas PSF with 500 dummy iterations was found to be sharper, with virtually no blurring (<0.025%).Note: Central points with highest intensity are not included in the scaling.D, Efficiency of BaSSI decreases with increasing number of dummies.
water and mixed with 20 mL of 1% (w/v) polyvinyl alcohol solution.The mixture was centrifuged (2000 rpm, 4 C, 10 min) to allow the microbubbles to sediment at the bottom of the container.After the supernatant was extracted, the precipitated particles were redispersed in 5 mL of double distilled water by vortexing at room temperature.The process yielded polymer bubbles with a mean diameter of 2.4 ± 1.5 μm as determined with a Coulter counter (Figure 3C).

| In silico characterization of BaSSI
The outcomes of numerical simulations for BaSSI confirmed the efficacy of utilizing a short-duration block pulse in exciting the spin-system, particularly the CF 2 group.Notably, this excitation exhibited a high degree of homogeneity, with minimal FA variation observed along the chemical shift axis (Figure 4A).
F I G U R E 5 Optimizing BaSSI.A, T acq variation resulted in slow change of signal for all FA.B, Optimal T acq was determined by fitting a fifthdegree polynomial to acquired data.C, Varying T R resulted in SNR curves with multiple peaks and maximum at 3.5 ms for the majority of FAs.
Furthermore, these simulations unveiled BaSSI's transient behavior.Driven primarily by rapid T 2 * decay, the signal rapidly decreased to around 33% of its maximum intensity within approximately 200 iterations of the core sequence.This decay continued until the signal reached a stable, steady-state level at approximately 32% of the maximum signal intensity, a point it reached after approximately 500 iterations (Figure 4B).
PSF analysis, as illustrated in Figure 4C, indicated that in cases without any dummy scans there was an approximately 4% blurring effect in the immediate vicinity of the PSF center.However, this blurring was reduced to less than 1% over a span of five voxels.Notably, as the number of dummy scans increased, the PSF exhibited enhanced sharpness.Specifically, for acquisitions with 100 dummy scans blurring remained below 1%, and with 500 dummy scans it dropped below 0.1% across the entire image.
The inclusion of dummy scans resulted in an incremental total scan time increase of approximately 0.35 s per 100 dummy scans.In the case of 500 dummy scans, this time increment amounted to 1.75 s, equivalent to 12% of the total scan time for a 64 Â 64 scan with N SA = 1.Additionally, it was observed that the use of dummy scans led to a signal level reduction of about 3.5% at 500 dummy scans.
The combination of increased scan time and decreased signal translated to an overall efficiency loss of approximately 10%, where efficiency is calculated as signal Â time ½ (Figure 4D).Taking into consideration this decline in signal efficiency, the decision was made to forgo dummy scans in the experimental phase of the work, prioritizing the maximization of detection sensitivity over spatial resolution.

| Optimization of hardware settings and BaSSI sequence parameters
BaSSI performed optimally with maximum B 1 set to 52 μT, and gradient strength and slew rate set to 33 mT/m and 225 T/m/s, respectively.
Varying T acq was observed to cause SNR to vary slowly (Figure 5A), with an optimum at 0.675 ms, identified by a fifth-degree polynomial fit (Figure 5B).Varying T R resulted in SNR curves consisting of multiple peaks reaching a maximum at around T R = 3.5 ms for an acquisition with T acq = 0.675 ms (Figure 5C).
When T R and FA were varied together, several regions with high SNR were observed.Among these, the region with FA 20-35 and T R 3.3-4 ms was found to be the most viable, as it had a clear maximum with a peak at FA/T R = 30 /3.5 ms, and was well within the boundaries of normal operating mode in terms of SAR (Figure 6A).
As a result, T acq /T R was set to 0.675 ms/3.5 ms, and the FA was set to 30 for the optimal configuration of BaSSI.With optimized settings, the duration of a 64 Â 64 2D-BaSSI scan with 1 mm Â 1 mm voxel size was found to be 14 s for N SA = 1.Scan time was observed to scale linearly with N SA .

| Optimization of sequence parameters of 3D UTE BSSFP
In optimization experiments of the 3D UTE BSSFP sequence for the CF 2 group of PFOB, the optimal T E value was found to be 0.145 ms, yielding an SNR of approximately 90% of the overall maximum.
F I G U R E 6 A, Optimizing BaSSI by varying T R and FA together yielded an optimum at FA/T R = 30 /3.5 ms with SAR value allowing the system to stay in normal operating mode.B, Optimizing 3D UTE BSSFP by varying T R and FA together yielded an optimum at FA/T R = 20 /2.5 ms with SAR value allowing the system to stay in normal operating mode.
Combined optimization of T R and FA showed that with T R set to 2.5 ms the sequence delivered good performance over a range of FAs (15 to 45 ).Out of these, the T R /FA combination of 2.5 ms/20 was found to be optimal and therefore was chosen as the setting for a 64 Â 64 Â 64 isotropic acquisition with 1 kHz pixel bandwidth (Figure 6B).
With optimal settings, a 64 Â 64 Â 64 scan was found to take about 25 s for N SA = 1.Scan time was observed to scale linearly with N SA up to N SA = 8.For higher N SA values, however, scan time per acquisition increased due to duty-cycle limitations of the gradient system.

| Benchmarking results
Benchmark results were obtained using a 0.10 μL PFOB sample.As depicted in Figure 7A, the 3D UTE BSSFP sequence was able to achieve an R c In the measurements with a smaller sample containing 0.05 μL PFOB, the 2D-BaSSI sequence was able to detect and localize the 19 F hotspot with an R c of 5.4, using N SA = 8 and scan time = 2 min (Figure 7C).
The performance of BaSSI in acquiring 19 F spectra of such small samples is also presented in Figure 7D.As seen, the 19 F spectrum was clearly identifiable for the measurements with an R c of approximately 5, even though the hotspots in the corresponding images were barely visible.

| Anatomic localization with postmortem imaging
Figure 8A shows color-inverted raw 19 F images of PFOB microspheres accumulated in the peritoneal cavity of a previously euthanized BALB/c mouse.These images were denoised by means of an R c -based filter eliminating pixels that do not satisfy the detection criteria.The effectiveness of R c -based filtering was found to be proportional to the R c value.The cleanest images, that is, images with no hotspots outside of the expected ROIs, were obtained with R c = 5.7 (Figure 8B).Subsequent to filtering, hotspot images were registered to 1 H images, using geometric metadata of acquisitions, to achieve accurate anatomic localization, as shown in Figure 8C.

| DISCUSSION AND CONCLUSIONS
This work has demonstrated BaSSI as a viable method to acquire artifact-free 19 F data for virtually any fluorine bearing contrast material by means of spectroscopic imaging.The sequence was implemented and optimized on a 3.0 T clinical scanner equipped with a preclinical RF coil and setup.In vitro experiments revealed BaSSI's superior ability to detect and image minute amounts of PFOB compared with its imaging counterparts, such as 3D UTE BSSFP with radial readout.
BaSSI achieved T R times that were significantly shorter than those observed in spectroscopic imaging sequences used for 1 H or 31 P spectroscopy.This feature was primarily enabled by the NMR properties of PFCs, which typically exhibit multiple peaks distributed across wide chemical shift ranges.To take advantage of these properties, the sequence employed a very short block pulse with a duration in the range of 6.7-73.5 μs.
As the simulation results demonstrated, such a short pulse allowed us to excite a broad spectral range of approximately 20 kHz with sufficient homogeneity, particularly over the CF 2 group.Subsequent to excitation, the generated FID signal was sampled with a 64 kHz sampling rate and a 1 kHz resolution over a 1 ms acquisition window, allowing the core sequence to attain a T R as short as 2.5 ms.While the acquisition parameters allowed the signal originating from the CF 2 group to be sampled sufficiently, they also enabled the acquisition of the entire PFOB spectrum and the resolution of all three resonance groups (CF 2 Br, CF 3 , and CF 2 ), if there was sufficient SNR.
To acquire the CF 2 signal most efficiently, it was crucial to minimize the time gap between excitation and the start of the acquisition window to prevent dephasing and destructive interference of CF 2 resonance lines.To achieve this, BaSSI used short phase encoding gradients, made possible by a gradient strength of 33 mT/m and slew rate of 225 m/T/s, and initiated acquisition in just 0.62 ms following excitation.Optimization experiments have shown that a time of 0.675 ms provided the best signal level, along with a T R of 3.5 ms and FA of 30 , resulting in a signal level that is about 95% of what would be obtained with a longer T R allowing for full relaxation.
BaSSI is an all-phase-encoded sequence that traces a 2D k-space grid in discrete steps defined by incremental changes in the gradients.
Numerical simulations have shown that BaSSI has a well defined PSF with little blurring despite being run under transient conditions.The sequence uses a pseudo-spiral trajectory that starts at the center of k-space and extends toward the edge by revolving around the center and sampling each discrete point on its path.While the trajectory ensures that the k-space is fully sampled, its spiral shape allows the transient characteristics of the signal to be distributed in all directions, minimizing widening of the PSF, and thus blurring in images.
BaSSI's ability to detect small amounts of contrast material and image small hotspots was assessed by benchmarking it against a 3D UTE BSSFP sequence, which is a widely used 19 F imaging technique. 21Detection performance of 19 F sequences is typically assessed by the detection sensitivity metric, which characterizes how the SNR scales with amount of contrast material and total experiment time.This study, however, employs the Rose criterion (R c ) to assess detection sensitivity, as it was thought to be more fitting for the purpose of detecting small objects in a noise background.
For similar scan times, BaSSI demonstrate higher detection sensitivities than the 3D UTE SSFP sequence.In the 2 min scans of a 0.10 μL PFOB sample, BaSSI achieved an R c of 11, whereas 3D UTE SSFP barely passed the detection threshold with an R c of 5.4.Furthermore, BaSSI was able to detect the hotspot of a smaller sample containing 0.05 μL PFOB, sample in about 2 min with N SA = 8.However, it was not possible to conduct a similar measurement with 3D UTE SSFP as the system kept shutting down due to overheating of the gradients.
The study clearly demonstrates that BaSSI is capable of detecting minuscule amounts of 19 F contrast material under clinically viable conditions, at least as effectively as widely used imaging techniques, by means of the Rose criterion of object visibility.
In addition to using the Rose criterion in assessing detection performance, this study also demonstrates the utility of a properly set R c as an effective tool for denoising, which operates through the process of elimination; that is, voxels that do not satisfy the SNR threshold set by the desired and thus do not contain sufficient 19 F signal are eliminated.This results in hotspot images that are free from false positives.However, in the process, voxels with low levels of 19 F signal may also be eliminated.In a fully blind experiment without prior knowledge of target locations, strict Rose criterion values can be employed to detect and image 19 F hot spots with high specificity.However, when a priori information on target locations is available, more lenient R c values can be applied to minimize loss of voxels with low amounts of 19  Even though BaSSI has proved to be effective in acquiring tiny amounts of fluorinated contrast material, our results come with limitations.
For clinical translation, and to propose BaSSI as a clinical tool, the work needs to be extended with live animal studies to be conducted under regulatory rules introducing two main constraints: (i) amount of contrast material that can be administered, and (ii) total experiment duration per subject.These two constraints are in conflict with each other, as reducing contrast material will require the total scan time to be increased to reach sufficient SNR, and this will pose a technical challenge for BaSSI.BaSSI is a fast sequence, particularly in comparison to conventional spectroscopic imaging methods, and its 2D acquisition has proven to be more sensitive than imaging sequences.However, BaSSI is still on the slow side for a full 3D scan.As shown in the anatomical localization experiments, a 48 Â 48 Â 48 acquisition with N SA = 16 requires 1.5 h to complete.In general, such scan durations are not uncommon in 19 F MRI, even at high magnetic fields.This is particularly due to the miniscule amounts of contrast material that are available for signal generation.In the work of Mastropietro et al., for instance, a 32 Â 32 RARE imaging sequence was run for 2 h to image small phantoms of potassium hexafluorophosphate (KPF 6 ) in a 7.0 T scanner. 20Similarly, a 64 Â 64 Â 64 RARE scan performed by Zhong et al. required slightly more than an hour at 11.7 T to acquire signal of small amounts of a PFPE-based contrast agent. 44The additional challenge of BaSSI with respect to the methods used in aforementioned studies, and other works using imaging sequences, is that it is a spectroscopic imaging method that encodes the k-space grid in a discrete manner.This means that BaSSI samples one grid point per T R , and acquires a full k-space by means of a stacked spirals trajectory that goes through each and every point in the grid.This point-by-point acquisition scheme inherently increases total scan time.An all-phase-encoded CSI sequence similar to BaSSI was also used in the work of Kampf et al., 45 which took about 37 min to sample a k-space of 48 Â 48 Â 70 points with N SA = 1 at 7.0 T. Such a scan time would correspond to about an hour on a 3.0 T scanner by linear approximation, demonstrating that scan times of several hours are to be expected in fluorine spectroscopic imaging, and pose a practical impediment for clinical applications.
To overcome the limitation posed by long scan times, alternative k-space sampling strategies could be considered.For instance, methods such as compressed sensing that exploit the sparse nature of 19 F data [44][45][46][47][48] have been demonstrated to reduce total scan time of fluorine scans.Among these, the works of Kampf et al. provide a basis for 3D compressed sense for fully phase-encoded chemical shift imaging, including reconstruction strategies to reduce artifacts, that is directly relevant for BaSSI. 45,46However, the majority of compressed sensing studies for 19 F MRI have been conducted at field strengths higher than 3.0 T, circumventing the challenges constituted by low SNR to a substantial extent, and thereby satisfy a critical prerequisite for compressed sensing to achieve significant compression factors.For low-SNR applications at 3.0 T, methods such as variable density acquisition where the central region of k-space is acquired more densely with respect ti the outer regions, as demonstrated by Schoormans et al. 49 in 1 H carotid imaging, and by Darçot et al. 48in 19 F imaging of PFPE emulsions, may prove useful in finding an optimal balance between detection sensitivity, precision of localization and scan time.
In conclusion, BASSI is an efficient and robust sequence that is capable of generating ghosting-artifact-free images of 19 F contrast materials by means of an all-phase-encoded pseudo-spiral k-space trajectory.The efficacy of the technique was demonstrated by in vitro experiments conducted on minute samples of PFOB, a compound exhibiting complex NMR characteristics, and by means of postmortem imaging where polymer microbubbles filled with the same contrast material were used.The work was conducted on a clinical 3.0 T scanner equipped with a specially designed 1 H/ 19 F coil and showed that BaSSI has the potential to play a key role in bringing 19 F MRI to clinical settings.

19 F
MRI, fast MRSI, fluorine, pseudo-spiral k-space trajectory, spectroscopic imaging 1 | INTRODUCTION 19 The core sequence of BaSSI.B, Pseudo-spiral k-space trajectory acquiring a Cartesian k-space.C, Stack of spirals for 3D acquisition.

3 h
at room temperature to allow for evaporation of the organic solvent.The solvent-free suspension was filled up to 20 mL with double distilled F I G U R E 3 A, Amount of contrast material in non-diluted PFOB phantoms used for sequence optimization and benchmarking.B, Sample height and position w.r.t.imaging volume.Smaller samples are confined within a slice of 1 mm.C, Optical microscope images of PLGA microspheres carrying PFOB payload.

of 4 . 5 ,
satisfying the Rose criterion for detection, with N SA = 4 corresponding to a scan time of 2 min.In comparison, the 2D-BaSSI sequence was able to achieve an R c of 11 for a similar scan time of 2 min, with N SA = 8, exhibiting better detection sensitivity than 3D UTE BSSFP.This was also visible in the obtained images.BaSSI succeeded in producing images in which it was possible to identify the19 F hotspot in a noisy background in as little as 28 s with N SA = 2, yielding an R c of 5.3 (Figure7B).F I G U R E 7 Benchmarking detection sensitivities with 0.10 μL PFOB sample.A, 3D UTE BSSFP detected 0.10 μL PFOB in 2 min of scan time (N SA = 4) by achieving an R c of 4.5.B, 2D-BaSSI detected the 0.10 μL PFOB sample in 28 s (N SA = 2) by achieving an R c of 5.3.C, 2D-BaSSI was able to detect 0.05 μL PFOB in 2 min (N SA = 16) by achieving an R c of 5.4.D, Spectroscopic content of 19 F hotspots obtained with 0.05 μL PFOB.Resonance lines start to become visible even if the R c values of corresponding images barely satisfy the Rose criterion.
based filtering and anatomic localization of fluorine hotspots.R c -based filtering is performed by removing pixels that do not satisfy the given R c threshold.The images were inverted to demonstrate R c -based filtering more effectively.A, Unfiltered images of 19 F hotspots.B, Filtered images obtained with R c = 5.7.C, 19 F hotspots obtained with R c = 5.7 registered to 1 H anatomical images.
F content.In the postmortem anatomic localization experiments R c of 5.7 was used to denoise images obtained with PFOB bearing microbubbles.The denoising process yielded clean images free from spots outside of the expected accumulation sites, and therefore improved the specificity of anatomic localization.The focus of this work was to present BaSSI as an effective tool for19 F MRI on a clinical scanner.This has been achieved by means of in vitro experiments, where BaSSI was shown to be more sensitive than established methods, and by postmortem imaging, where accurate localization of fluorine hotspots with high specificity co-registered with an anatomical 1 H image was demonstrated.The reason to conduct anatomic localization in dead subjects was twofold: to avoid excessive use of live animals, and to minimize the set of parameters.Postmortem conditions allowed us to adjust the amount of contrast material freely, and perform data acquisition in the absence of physiological factors, such as respiratory and cardiac motion.