DeepFLAIR: A neural network approach to mitigate signal and contrast loss in temporal lobes at 7 Tesla FLAIR images

BACKGROUND AND PURPOSE

Higher magnetic field strength introduces stronger magnetic field inhomogeneities in the brain, especially within temporal lobes, leading to image artifacts. Particularly, T2-weighted fluid-attenuated inversion recovery (FLAIR) images can be affected by these artifacts. Here, we aimed to improve the FLAIR image quality in temporal lobe regions through image processing of multiple contrast images via machine learning using a neural network.

METHODS

Thirteen drug-resistant MR-negative epilepsy patients (age 29.2 ± 9.4y, 5 females) were scanned on a 7 T MRI scanner. Magnetization-prepared (MP2RAGE) and saturation-prepared with 2 rapid gradient echoes, multi-echo gradient echo with four echo times, and the FLAIR sequence were acquired. A voxel-wise neural network was trained on extratemporal-lobe voxels from the acquired structural scans to generate a new FLAIR-like image (i.e., deepFLAIR) with reduced temporal lobe inhomogeneities. The deepFLAIR was evaluated in temporal lobes through signal-to-noise (SNR), contrast-to-noise (CNR) ratio, the sharpness of the gray-white matter boundary and joint-histogram analysis. Saliency mapping demonstrated the importance of each input image per voxel.

RESULTS

SNR and CNR in both gray and white matter were significantly increased (p < 0.05) in the deepFLAIR's temporal ROIs, compared to the FLAIR. The gray-white matter boundary sharpness was either preserved or improved in 10/13 right-sided temporal regions and was found significantly increased in the ROIs. Multiple image contrasts were influential for the deepFLAIR reconstruction with the MP2RAGE second inversion image being the most important.

CONCLUSIONS

The deepFLAIR network showed promise to restore the FLAIR signal and reduce contrast attenuation in temporal lobe areas. This may yield a valuable tool, especially when artifact-free FLAIR images are not available.

Pitfalls of Using T2-contrast Enhancement Techniques in 3D-FLAIR to Detect Endolymphatic Hydrops

PURPOSE

To determine whether T2-contrast enhancement techniques can be used to diagnose endolymphatic hydrops, we compared fluid signal artifacts with and without T2-contrast enhancement techniques in 3D fluid-attenuated inversion recovery (3D-FLAIR).

METHODS

We prepared a custom-made phantom consisting of eight tubes half-filled with saline. Images were obtained using four 3D-FLAIR: without T2-contrast enhancement (Normal), with non-selective T2-inversion recovery (T2-IR), and two with non-selective T2 preparation IR (T2-prep). Scans were performed with and without rice covering the phantom to simulate minimal and severe B0-inhomogeneity conditions. The average signal intensity (SI) values of eight saline tubes were compared between the four sequences and between each other. Comparisons were performed for all measurement slices and the central 10 slices. The images using T2-contrast enhancement technique were obtained from a volunteer and a patient suspected of Meniere's disease.

RESULTS

The Normal sequence SI for all slices was significantly lower than that for the other sequences, with smaller standard deviation (SD) and no outliers. Several outliers were detected in the other sequences. The SDs and outliers were larger without rice than with rice. When the central 10 slices with rice, the T2-IR had a significantly higher SI with more outliers compared with the Normal sequence. The T2-prep had no outliers and SIs that were comparable to those of the Normal sequence. However, without rice, the T2-IR and T2-prep sequences had significantly higher SIs with outliers and larger SDs compared to the Normal sequence. In the corresponding images, the Normal sequence achieved excellent fluid suppression, whereas the T2-IR and T2-prep sequences showed high-signal artifacts. Imperfect fluid suppressions were observed in the volunteer image and the endolymphatic hydrops on the post-gadolinium image differed in size and shape in the non-injected T2-IR in the patient image.

CONCLUSION

T2-contrast enhancement techniques should be used with caution in 3D-FLAIR for diagnosing endolymphatic hydrops.

The effects of RF coils and SAR supervision strategies for clinically applicable nonselective parallel-transmit pulses at 7 T

PURPOSE

To investigate the effects of using different parallel-transmit (pTx) head coils and specific absorption rate (SAR) supervision strategies on pTx pulse design for ultrahigh-field MRI using a 3D-MPRAGE sequence.

METHODS

The PTx universal pulses (UPs) and fast online-customized (FOCUS) pulses were designed with pre-acquired data sets (B₀ , B₁ ⁺ maps, specific absorption rate [SAR] supervision data) from two different 8 transmit/32 receive head coils on two 7T whole-body MR systems. For one coil, the SAR supervision model consisted of per-channel RF power limits. In the other coil, SAR estimations were done with both per-channel RF power limits as well as virtual observation points (VOPs) derived from electromagnetic field (EMF) simulations using three virtual human body models at three different positions. All pulses were made for nonselective excitation and inversion and evaluated on 132 B₀ , B₁ ⁺ , and SAR supervision datasets obtained with one coil and 12 from the other. At both sites, 3 subjects were examined using MPRAGE sequences that used UP/FOCUS pulses generated for both coils.

RESULTS

For some subjects, the UPs underperformed when simulated on a different coil from which they were derived, whereas FOCUS pulses still showed acceptable performance in that case. FOCUS inversion pulses outperformed adiabatic pulses when scaled to the same local SAR level. For the self-built coil, the use of VOPs showed reliable overestimation compared with the ground-truth EMF simulations, predicting about 52% lower local SAR for inversion pulses compared with per-channel power limits.

CONCLUSION

FOCUS inversion pulses offer a low-SAR alternative to adiabatic pulses and benefit from using EMF-based VOPs for SAR estimation.

Ultra-high field MRI: parallel-transmit arrays and RF pulse design

This paper reviews the field of multiple or parallel radiofrequency (RF) transmission for magnetic resonance imaging (MRI). Currently the use of ultra-high field (UHF) MRI at 7 tesla and above is gaining popularity, yet faces challenges with non-uniformity of the RF field and higher RF power deposition. Since its introduction in the early 2000s, parallel transmission (pTx) has been recognized as a powerful tool for accelerating spatially selective RF pulses and combating the challenges associated with RF inhomogeneity at UHF. We provide a survey of the types of dedicated RF coils used commonly for pTx and the important modeling of the coil behavior by electromagnetic (EM) field simulations. We also discuss the additional safety considerations involved with pTx such as the specific absorption rate (SAR) and how to manage them. We then describe the application of pTx with RF pulse design, including a practical guide to popular methods. Finally, we conclude with a description of the current and future prospects for pTx, particularly its potential for routine clinical use.

Universal pulses for homogeneous excitation using single channel coils

PURPOSE

Universal Pulses (UPs) are excitation pulses that reduce the flip angle inhomogeneity in high field MRI systems without subject-specific optimization, originally developed for parallel transmit (PTX) systems at 7 T. We investigated the potential benefits of UPs for single channel (SC) transmit systems at 3 T, which are widely used for clinical and research imaging, and for which flip angle inhomogeneity can still be problematic.

METHODS

SC-UPs were designed using a spiral nonselective k-space trajectory for brain imaging at 3 T using transmit field maps (B1+) and off-resonance maps (B0) acquired on two different scanner types: a 'standard' single channel transmit system and a system with a PTX body coil. The effect of training group size was investigated using data (200 subjects) from the standard system. The PTX system was used to compare SC-UPs to PTX-UPs (15 subjects). In two additional subjects, prospective imaging using SC-UP was studied.

RESULTS

Average flip angle homogeneity error fell from 9.5 ± 0.5 % for 'default' excitation to 3.0 ± 0.6 % using SC-UPs trained over 50 subjects. Performance of the UPs was found to steadily improve as training group size increased, but stabilized after ~15 subjects. On the PTX-enabled system, SC-UPs again outperformed default excitation in simulations (4.8 ± 0.6 % error versus 10.6 ± 0.8 % respectively) though greater homogenization could be achieved with PTX-UPs (3.9 ± 0.6 %) and personalized pulses (SC-PP 3.6 ± 1.0 %, PTX-PP 2.9 ± 0.6 %). MP-RAGE imaging using SC-UP resulted in greater separation between grey and white matter signal intensities than default excitation.

CONCLUSIONS

SC-UPs can improve excitation homogeneity in standard 3 T systems without further calibration and could be used instead of a default excitation pulse for nonselective neuroimaging at 3 T.

OTUP workflow: target specific optimization of the transmit k-space trajectory for flexible universal parallel transmit RF pulse design

PURPOSE

To optimize transmit k-space trajectories for a wide range of excitation targets and to design "universal pTx RF pulses" based on these trajectories.

METHODS

Transmit k-space trajectories (stack of spirals and SPINS) were optimized to best match different excitation targets using the parameters of the analytical equations of spirals and SPINS. The performances of RF pulses designed based on optimized and non-optimized trajectories were compared. The optimized trajectories were utilized for universal pulse design. The universal pulse performances were compared with subject specific tailored pulse performances. The OTUP workflow (optimization of transmit k-space trajectories and universal pulse calculation) was tested on three test target excitation patterns. For one target (local excitation of a central area in the human brain) the pulses were tested in vivo at 9.4 T.

RESULTS

The workflow produced appropriate transmit k-space trajectories for each test target. Utilization of an optimized trajectory was crucial for the pulse performance. Using unsuited trajectories diminished the performance. It was possible to create target specific universal pulses. However, not every test target is equally well suited for universal pulse design. There was no significant difference in the in vivo performance between subject specific tailored pulses and a universal pulse at 9.4 T.

CONCLUSIONS

The proposed workflow further exploited and improved the universal pulse concept by combining it with gradient trajectory optimization for stack of spirals and SPINS. It emphasized the importance of a well suited trajectory for pTx RF pulse design. Universal and tailored pulses performed with a sufficient degree of similarity in simulations and a high degree of similarity in vivo. The implemented OTUP workflow and the B₀ /B₁⁺ map data from 18 subjects measured at 9.4 T are available as open source (https://github.com/ole1965/workflow_OTUP.git).

Can 7T MPRAGE match MP2RAGE for gray-white matter contrast?

Ultra-High Field (UHF) MRI provides a significant increase in Signal-to-Noise Ratio (SNR) and gains in contrast weighting in several functional and structural acquisitions. Unfortunately, an increase in field strength also induces non-uniformities in the transmit field (B1+) that can compromise image contrast non-uniformly. The MPRAGE is one of the most common T1 weighted (T1w) image acquisitions for structural imaging. It provides excellent contrast between gray and white matter and is widely used for brain segmentation. At 7T, the signal non-uniformities tend to complicate this and therefore, the self-bias-field corrected MP2RAGE is often used there. In both MPRAGE and MP2RAGE, more homogeneous image contrast can be achieved with adiabatic pulses, like the TR-FOCI inversion pulse, or special pulse design on parallel transmission systems, like Universal Pulses (UP). In the present study, we investigate different strategies to improve the bias-field for MPRAGE at 7T, comparing the contrast and GM/WM segmentability against MP2RAGE. The higher temporal efficiency of MPRAGE combined with the potential of the user-friendly UPs was the primary motivation for this MPRAGE-MP2RAGE comparison. We acquired MPRAGE data in six volunteers, adding a k-space shutter to reduce scan time, a kt-point UP approach for homogeneous signal excitation, and a TR-FOCI pulse for homogeneous inversion. Our results show remarkable signal contrast improvement throughout the brain, including regions of low B1⁺ such as the cerebellum. The improvements in the MPRAGE were largest following the introduction of the UPs. In addition to the CNR, both SNR and GM/WM segmentability were also assessed. Among the MPRAGEs, the combined strategy (UP + TR-FOCI) yielded highest SNR and showed highest spatial similarity between GM segments to the MP2RAGE. Interestingly, the distance between gray and white matter peaks in the intensity histograms did not increase, as better pulses and higher SNR especially benefitted the (cerebellar) gray matter. Overall, the gray-white matter contrast from MP2RAGE is higher, with higher CNR and higher intensity peak distances, even when scaled to scan time. Hence, the extra acquisition time for MP2RAGE is justified by the improved segmentability.

Local excitation universal parallel transmit pulses at 9.4T

PURPOSE

To demonstrate that the concept of "universal pTx pulses" is applicable to local excitation applications.

METHODS

A database of B₀ / B 1 + maps from eight different subjects was acquired at 9.4T. Based on these maps, universal pulses that aim at local excitation of the visual cortex area in the human brain (with a flip angle of 90° or 7°) were calculated. The remaining brain regions should not experience any excitation. The pulses were designed with an extension of the "spatial domain method." A 2D and a 3D target excitation pattern were tested, respectively. The pulse performance was examined on non-database subjects by Bloch simulations and in vivo at 9.4T using a GRE anatomical MRI and a presaturated TurboFLASH B 1 + mapping sequence.

RESULTS

The calculated universal pulses show excellent performance in simulations and in vivo on subjects that were not contained in the design database. The visual cortex region is excited, while the desired non-excitation areas produce the only minimal signal. In simulations, the pulses with 3D target pattern show a lack of excitation uniformity in the visual cortex region; however, in vivo, this inhomogeneity can be deemed acceptable. A reduced field of view application of the universal pulse design concept was performed successfully.

CONCLUSIONS

The proposed design approach creates universal local excitation pulses for a flip angle of 7° and 90°, respectively. Providing universal pTx pulses for local excitation applications prospectively abandons the need for time-consuming subject-specific B₀ / B 1 + mapping and pTx-pulse calculation during the scan session.

Simultaneous optimization of radio frequency and gradient waveforms with exact Hessians and slew rate constraints applied to kT-points excitation

We consider an excitation pulse with piecewise constant gradient trajectories and radio frequency (RF) waveforms such that the solution of the Bloch equations without relaxation terms can be represented by rotations. Based on this analytic solution we formulate a non-linear program for finding sub-pulse durations, gradient strengths, and complex RF voltages which minimize the deviation between the achieved and desired magnetization. We develop explicit expressions for the first and second order derivatives of the objective function. We extend the non-linear program to precisely account for gradient slew rate constraints. Using an interior point solver we apply the developed theory to simultaneously optimize the positions of k_T-points, their associated RF voltages and durations.

Fast online-customized (FOCUS) parallel transmission pulses: A combination of universal pulses and individual optimization

PURPOSE

To mitigate spatial flip angle (FA) variations under strict specific absorption rate (SAR) constraints for ultra-high field MRI using a combination of universal parallel transmit (pTx) pulses and fast subject-specific optimization.

METHODS

Data sets consisting of B₀ , B 1 + maps, and virtual observation point (VOP) data were acquired from 72 subjects (study groups of 48/12 healthy Europeans/Asians and 12 Europeans with pathological or incidental findings) using an 8Tx/32Rx head coil on a 7T whole-body MR system. Combined optimization values (COV) were defined as combination of spiral-nonselective (SPINS) trajectory parameters and an energy regularization weight. A set of COV was optimized universally by simulating the individual RF pulse optimizations of 12 training data sets (healthy Europeans). Subsequently, corresponding universal pulses (UPs) were calculated. Using COV and UPs, individually optimized pulses (IOPs) were calculated during the sequence preparation phase (maximum 15 s). Two different UPs and IOPs were evaluated by calculating their normalized root-mean-square error (NRMSE) of the FA and SAR in simulations of all data sets. Seven additional subjects were examined using an MPRAGE sequence that uses the designed pTx excitation pulses and a conventional adiabatic inversion.

RESULTS

All pTx pulses resulted in decreased mean NRMSE compared to a circularly polarized (CP) pulse (CP = ~28%, UPs = ~17%, and IOPs = ~12%). UPs and IOPs improved homogeneity for all subjects. Differences in NRMSE between study groups were much lower than differences between different pulse types.

CONCLUSION

UPs can be used to generate fast online-customized (FOCUS) pulses gaining lower NRMSE and/or lower SAR values.

Improving FLAIR SAR efficiency at 7T by adaptive tailoring of adiabatic pulse power through deep learning B 1 + estimation

PURPOSE

The purpose of this study is to demonstrate a method for specific absorption rate (SAR) reduction for 2D T₂ -FLAIR MRI sequences at 7 T by predicting the required adiabatic radiofrequency (RF) pulse power and scaling the RF amplitude in a slice-wise fashion.

METHODS

We used a time-resampled frequency-offset corrected inversion (TR-FOCI) adiabatic pulse for spin inversion in a T₂ -FLAIR sequence to improve B 1 + homogeneity and calculated the pulse power required for adiabaticity slice-by-slice to minimize the SAR. Drawing on the implicit B 1 + inhomogeneity in a standard localizer scan, we acquired 3D AutoAlign localizers and SA2RAGE B 1 + maps in 28 volunteers. Then, we trained a convolutional neural network (CNN) to estimate the B 1 + profile from the localizers and calculated pulse scale factors for each slice. We assessed the predicted B 1 + profiles and the effect of scaled pulse amplitudes on the FLAIR inversion efficiency in oblique transverse, sagittal, and coronal orientations.

RESULTS

The predicted B 1 + amplitude maps matched the measured ones with a mean difference of 9.5% across all slices and participants. The slice-by-slice scaling of the TR-FOCI inversion pulse was most effective in oblique transverse orientation and resulted in a 1 min and 30 s reduction in SAR induced delay time while delivering identical image quality.

CONCLUSION

We propose a SAR reduction technique based on the estimation of B 1 + profiles from standard localizer scans using a CNN and show that scaling the inversion pulse power slice-by-slice for FLAIR sequences at 7T reduces SAR and scan time without compromising image quality.

Universal nonselective excitation and refocusing pulses with improved robustness to off-resonance for Magnetic Resonance Imaging at 7 Tesla with parallel transmission

PURPOSE

In MRI at ultra-high field, the k T -point and spiral nonselective (SPINS) pulse design techniques can be advantageously combined with the parallel transmission (pTX) and universal pulse techniques to create uniform excitation in a calibration-free manner. However, in these approaches, pulse duration is typically increased as compared to standard hard pulses, and excitation quality in regions exhibiting large resonance frequency offsets often suffer. This limitation is inherent to structure of k T -point or SPINS pulse, and likely can be mitigated using parameterization-free pulse design approaches.

METHODS

The Gradient Ascent Pulse Engineering (GRAPE) algorithm was used to design parameterization-free RF and magnetic field gradient (MFG) waveforms for creating 8 ∘ excitation, up to 105 ∘ scalable refocusing and inversion, nonselectively across the brain. Simulations were performed to provide flip angle normalized root-mean-squares error (FA-NRMSE) estimations for the 8 ∘ and the 180 ∘ k T -point, SPINS, and GRAPE pulses. GRAPE pulses were tested experimentally with anatomical head scans at 7T.

RESULTS

As compared to k T -points and SPINS, GRAPE provided substantial improvement of excitation, refocusing, and inversion quality at off-resonance while at least preserving the same global FA-NRMSE performance. As compared to k T -points, GRAPE allowed for a substantial reduction of the pulse duration for the 8 ∘ excitation and the 105 ∘ refocusing.

CONCLUSIONS

Parameterization-free universal nonselective pTX-pulses were successfully computed using GRAPE. Performance gains as compared to k T -points were validated numerically and experimentally for three imaging protocols. In its current implementation, the computational burden of GRAPE limits its use to applications where pulse computations are not subject to time constraints.

Improving PCASL at ultra-high field using a VERSE-guided parallel transmission strategy

Purpose

To improve the labeling efficiency of pseudo‐continuous arterial spin labeling (PCASL) at 7T using parallel transmission (pTx).

Methods

Five healthy subjects were scanned on an 8‐channel‐transmit 7T human MRI scanner. Time‐of‐flight (TOF) angiography was acquired to identify regions of interest (ROIs) around the 4 major feeding arteries to the brain, and B1+ and B₀ maps were acquired in the labeling plane for tagging pulse design. Complex weights of the labeling pulses for each of the 8 transmit channels were calculated to produce a homogenous radiofrequency (RF) ‐shimmed labeling across the ROIs. Variable‐Rate Selective Excitation (VERSE) pulses were also implemented as a part of the labeling pulse train. Whole‐brain perfusion‐weighted images were acquired under conditions of RF shimming, VERSE with RF shimming, and standard circularly polarized (CP) mode. The same subjects were scanned on a 3T scanner for comparison.

Results

In simulation, VERSE with RF shimming improved the flip‐angles across the ROIs in the labeling plane by 90% compared with CP mode. VERSE with RF shimming improved the temporal signal‐to‐noise ratio by 375% compared with CP mode, but did not outperform a matched 3T sequence with a matched flip‐angle.

Conclusion

We have demonstrated improved PCASL tagging at 7T using VERSE with RF shimming on a commercial head coil under conservative SAR limits at 7T. However, improvements of 7T over 3T may require strategies with less conservative SAR restrictions.

Designing parallel transmit head coil arrays based on radiofrequency pulse performance

PURPOSE

A new approach to design parallel transmit (pTx) head arrays is proposed that integrates transmit radiofrequency pulse designs with electromagnetic modeling of array coil elements.

THEORY AND METHODS

An approach to design pTx head arrays is proposed that finds optimal groupings of a large number of coils into a small number of channels. An algorithm is proposed to extend array-compressed parallel transmit pulse design by adding the ability to optimally select and prune coil elements, in addition to optimizing compression weights. The performance of the method is demonstrated in simulations of dynamic multislice shimming of the human brain in axial, coronal, and sagittal directions, and of reduced field-of-view excitation targeting the human occipital lobe, with simulated electromagnetic field maps from a group of 5 human head models at 7T.

RESULTS

For both dynamic multislice shimming and reduced field-of-view excitation, the method successfully designed pTx arrays that simultaneously achieved in general 15% lower mean excitation errors with 20% lower SDs, along with 20% lower mean global averaged specific absorption rate and 50% lower SD than previously reported pTx head array designs.

CONCLUSION

With the proposed optimal coil element selection algorithm, the array-compressed parallel transmit pulse design can be extended to design pTx transmit head arrays with joint consideration of the fields within the sample and the radiofrequency pulse. The pTx arrays from such an approach achieved higher transmit excitation accuracy, lower radiofrequency heating in subjects, and more robust performance across subjects compared with previously reported pTx head arrays with the same number of channels.