Simultaneous T2‐weighted real‐time MRI of two orthogonal slices

MR guidance is used during therapy to detect and compensate for lesion motion. T2‐weighted MRI often has a superior lesion contrast in comparison to T1‐weighted real‐time imaging. The purpose of this work was to design a fast T2‐weighted sequence capable of simultaneously acquiring two orthogonal slices, enabling real‐time tracking of lesions.


INTRODUCTION
The accuracy of therapeutic procedures such as radiotherapy (RT), 1 percutaneous needle biopsies, 2 or cryoablation 3 can be severely reduced in the presence of respiratory, peristaltic, or cardiac motion.5][6] Ideally, imaging would provide the exact target location in real time to guide the needle or probe in percutaneous needle biopsies and cryoablations, or for monitoring organ motion in RT, 7 to ensure that the tumor stays within the defined margins.Because of its widespread availability and high spatial and temporal resolution, CT imaging is the current clinical gold standard for image guidance.Unfortunately, CT lacks soft-tissue contrast and uses ionizing radiation, making it unsuitable for prolonged and continuous real-time imaging. 8MRI, on the other hand, offers an excellent soft-tissue contrast without the use of ionizing radiation. 8][16][17] Dynamic MRI sequences with a high spatiotemporal resolution are used commonly for real-time motion monitoring, such as T 1 -weighted spoiled gradient echo sequences and the T 2 /T 1 -weighted balanced SSFP (bSSFP) 8 sequence.9][20] In renal cancer, for example, T 2 -weighted MRI sequences were used before treatment for positioning and tumor targeting in MR-guided radiotherapy (MRgRT), 21,22 and for targeting or planning of the needle trajectory in MRI-guided cryoablation 23 and percutaneous renal mass biopsies. 24uring each of these procedures, however, kidney motion was monitored using a dynamic T 2 /T 1 -weighted bSSFP acquisition, which offers a high temporal resolution.Unfortunately, bSSFP images show banding artifacts, which, if they cover the target region, could make automatic lesion tracking challenging.Furthermore, the limitations of FLASH and bSSFP sequences and the inapt temporal resolution of conventional T 2 -weighted MR sequences for real-time MR guidance, can be countered by using the pre-excitation refocused steady-state sequence 25,26 (SSFP-Echo).
In addition to a T 2 -weighted contrast, ideally, 3D imaging would be required to assess the complex organ motion.The temporal resolution of 3D MRI, however, is currently insufficient for real-time motion monitoring, 27 even if acceleration techniques such as parallel imaging and undersampling are applied.To achieve a temporal resolution of several frames per second, MRgRT studies applied 2D single-slice cine imaging. 28,29s tumor motion can occur both in plane and out of plane, 28,29 single-slice cine imaging can be inadequate to monitor the entire motion range.Hence, to enable the visualization of out-of-plane motion, orthogonal 2D cine imaging was used, where the slices were acquired sequentially, [30][31][32] resulting in a time shift between the reconstructed slices.
Recently, a truly simultaneous acquisition of two orthogonal slices was introduced with the crushed rephased orthogonal slice selection (CROSS) sequence. 33Initially designed for proton resonance frequency temperature mapping, the CROSS sequence uses the long TEs of 10-20 ms required to resolve proton resonance frequency changes, to encode a second orthogonal slice.Building on the CROSS acquisition scheme, simultaneous orthogonal plane imaging 34 was developed specifically for MRgRT by minimizing TE, removing crusher gradients, and using multiplexed data acquisition to reduce total acquisition time (TA).
In this work, we present a sequence that combines a pre-excitation refocused steady-state (SSFP-Echo) sequence with an orthogonal acquisition scheme.The sequence (Ortho-SSFP-Echo) is capable of rapidly acquiring two orthogonal T 2 -weighted MR images simultaneously.In both phantom and in vivo experiments, the Ortho-SSFP-Echo sequence is used to track the motion of predetermined targets in a time series.

Pulse sequence design
The Ortho-SFFP-Echo sequence is based on the principles of a single-slice pre-excitation refocused steady-state sequence (SSFP-Echo). 25,26Characteristics of such a sequence are (1) a fully balanced phase-encoding (PE) gradient and (2) rephasing of the slice selection (SS) gradient before application of the next RF pulse. 35To reach steady state, the sequence uses consecutive phase coherent RF pulses and a TR shorter than the T 2 relaxation time. 35In steady state sequences, two signals are created: the SSFP-FID (or S + ) signal occurring after the RF pulse, and the SSFP-Echo (or S − ) signal before the next pulse.The SSFP-Echo is conceptually a spin echo (i.e., the preceding RF pulse acts as a 180 • pulse); thus, the S − signal has a T 2 -weighted contrast. 36or simultaneous acquisition of two orthogonal SSFP-Echo slices, the phase-encoding (PE) and slice-select (SS) directions of the two slices are interchanged. 33,34,37,38he RF pulse of Slice 1 is applied conventionally, and the RF pulse of Slice 2 is applied in the PE direction of Slice 1 (SS2/PE1) (Figure 1).The readout (RO) direction is shared by both slices.Because the SSFP-Echo occurs in the subsequent TR, the effective TE (TE eff ) is TE eff = TE + TR (Figure 1).In the Ortho-SFFP-Echo sequence, the same TE eff is used to ensure that both slices have identical contrast.
Several sequence-design constraints must be met for an interleaved orthogonal slice acquisition.To acquire the echo in the subsequent TR, the accumulated zeroth gradient moments (M 0 ) between the RF pulse at t RF1 and the echo at t RF1 + TE eff in the subsequent TR have to be zero in both the SS and RO directions, and in PE direction it has to correspond to the moment of the current PE line (M 0,PE ): This principle applies analogously to the second pulse with swapped PE and SS directions: To fulfill the conditions in Eqs. ( 1) and ( 2) so that a spin echo is generated in the subsequent TR for each slice, the gradient moments must be balanced.Therefore, the sequence timing was broken down into four sections (Figure 1): between the RF pulses (pre), between the second RF pulse and the center of ADC1 (common), between the center of the two ADCs (inter), and after ADC2 and the first RF pulse (post).For Slice 1, the echo condition in Eq. ( 1) for RO and SS directions can be written as Note that the moments of the four time intervals before the second occurrence of RF Pulse 1 have a negative sign because of the 180 • refocusing effect (spin-echo condition).From Eq. (3), it follows that M 0,inter = −M 0,post .A similar relationship can be written for the second slice: In the SS1 direction, the gradient moments M 0,post and M 0,inter are zero; therefore, M 0,pre must also vanish, which can be realized with an additional gradient corresponding to half the moment of SS1 (Figure 1).
To acquire only the SSFP-Echo signal and thus generate a T 2 -weighted contrast, the FID in both SS1/PE2 and SS2/PE1 directions must be dephased during the acquisition (ADC1 and ADC2) in the first TR.This is achieved by adding crusher gradients in the common region, which are completely refocused by the second RF pulse for both slices and hence do not contribute to the zeroth gradient moment (Eqs.[3] and [4]).
To ensure that only the signals from Slice 1 and Slice 2 are refocused during ADC1 and ADC2, respectively, the signals are separated in the RO direction.This is achieved by adding a dephasing gradient, M 0,pre , which only acts on the magnetization in Slice 1.
Finally, to reduce the encoding time, adjacent gradients were merged by overlapping the gradient timing; this is not shown in Figure 1 to increase the interpretability of the diagram.

Flow compensation
The Ortho-SSFP-Echo sequence is inherently sensitive to motion-related dephasing due to the long TE eff . 39This motion sensitivity can partially be mitigated with the introduction of flow compensation (i.e., balancing of the first gradient moments in the shared RO direction).More specifically, the gradients in pre, inter, and post have already been determined by Eq. ( 5), and altering the gradients in the common region does not influence the total zeroth moment.Therefore, the common region can be used for flow compensation.A perfect balancing of the gradient moments is not possible due to the application of the M 0,pre gradient, which separates the slices in the zeroth moment balancing.Hence, the ideal M 0,common gradient was determined first for each slice individually by balancing the first moment between the RF pulse excitation and acquisition (Eqs.[6] and [7]).
The average of the ideal flow compensation moments for each individual slice was then applied in the RO direction as the M 0,common gradient.This allowed the first moment to be reduced by 76% at the echo, when compared with a non-flow-compensated sequence.

F I G U R E 1
Ortho-SSFP-Echo pulse sequence diagram.The sequence extends from t RF to TE eff , where the echoes are acquired.Components of Slice 1 are colored light blue, whereas Slice 2 is represented by dark blue.The slice-selection direction of Slice 2 is the phase-encoding direction of Slice 1 and both slices share the same readout direction.Crusher gradients are characterized by dashed diagonal lines, and the gradients added to balance the zeroth and first moments in the readout direction are shaded in dark and light green, respectively.The sequence is divided into four sections for gradient moment calculations: pre, common, inter, and post.The sequence satisfies the conditions for generating a spin echo in the readout and slice-selection directions for both slices.

Phantom experiments
The Ortho-SSFP-Echo sequence was implemented on a clinical 3T whole-body MR system (MAGNETOM Prisma; Siemens Healthineers, Erlangen, Germany).For signal reception, a body array and a spine array coil were applied.A purpose-built abdominal breathing phantom (ACMIT, Wiener Neustadt, Austria) was used to simulate breathing motion via pneumatic actuation to mimic an RT or interventional MRI scenario.The breathing phantom (Figure S1) contains artificial organs (kidney, liver, pancreas) with embedded target lesions made of different silicone types; this allows the lesions to be distinguished from the organ tissue.The phantom is driven via a ventilation unit in the control room, which is connected to the main air supply using flexible compressed air hoses.The lesion centroid was tracked in MATLAB (v.2022b; MathWorks, Natick, MA, USA) using deformable registration in postprocessing.First, a mid-breathing-cycle frame was selected as the reference image, on which the lesion was manually contoured.Then, the reference image was deformably registered to all other frames using the MAT-LAB function imregdeform.Each of the resulting deformation fields were then applied to the initial contour, which generated a new contour for each frame.To determine the time course of the lesion motion, the centroid of each contour was calculated.Identical steps were performed on the reference GRE data set.The Ortho-SSFP-Echo sequence was compared with the reference GRE scan in the head-foot (H-F) direction by interpolating the centroid time course onto the same grid and calculating the Pearson correlation coefficient.The kidney was contoured on a reference image, to which all other images in the time series were deformably registered (imregdeform).The centroid of each deformed contour was calculated to determine the time course of the kidney motion.To validate the breathing motion captured by the Ortho-SSFP-Echo sequence, the motion tracking data were interpolated and compared with respiratory belt measurements.The degree of agreement in the H-F direction was assessed using the Pearson correlation coefficient.

RESULTS
Figure 2 demonstrates the RO flow compensation, in vivo and under free-breathing conditions.Without flow compensation, the signal is largely dephased.However, by using flow compensation, minimal dephasing is observed.
In the abdominal phantom, Ortho-SSFP-Echo effectively resolved the simulated breathing motion, with a TA of 0.47 s for both slices (Figure 3).Using Ortho-SSFP-Echo (Videos S1 and S2), the lesion was tracked over a range of 9.7, 2.  images using the SSIM for Volunteer 1 and 2 (out of 5).The resulting SSIM over all 5 volunteers was 0.99969 for the coronal slice and 0.99928 for the sagittal slice (Figures S3-S5).Breathing-induced kidney motion was observed in 5 volunteers using Ortho-SSFP-Echo, with a temporal resolution of 0.45 s (Figure 5, Figures S6-S8, Videos S4 and S5).The Pearson correlation coefficient between the respiratory belt and motion in the H-F direction varied across the volunteers as follows: 0.90 and 0.91 (Volunteer 1), 0.85 and 0.87 (Volunteer 2), 0.91 and 0.92 (Volunteer 3), 0.85 and 0.89 (Volunteer 4), and 0.86 and 0.86 (Volunteer 5).

DISCUSSION
In this work, the Ortho-SSFP-Echo sequence rapidly acquired two T 2 -weighted orthogonal slices simultaneously.The sequence uses the T 2 -weighted SSFP-Echo signal, which is beneficial for distinguishing lesions from surrounding healthy tissue, such as in radiotherapy, cryoablation, or percutaneous needle biopsies.
To demonstrate the sequence, we conducted a phantom experiment to monitor the breathing-induced motion of an embedded artificial target lesion in the liver.In postprocessing steps, the deformable registration allowed for semi-automatic retrospective lesion tracking.The sequence was compared with a reference GRE sequence in the H-F direction, and a high Pearson correlation coefficient was obtained indicating a strong agreement between the motion patterns captured by both sequences.The breathing phantom predominantly creates motion in the H-F direction, resulting in minimal motion (∼2 mm) observed in the A-P and L-R directions.The observed 1-2 mm motion in the A-P and L-R directions may be attributed to the inherent uncertainty associated with the deformable registration method used for tracking.
To validate the Ortho-SSFP-Echo sequence in vivo, a respiratory belt was used as an external measure of motion.The respiratory belt demonstrated a high Pearson correlation coefficient in the H-F direction, confirming the sequence's ability to resolve breathing-induced motion.However, the respiratory belt cannot provide information regarding motion in the R-L and A-P directions.
In conventional interleaved acquisitions of orthogonal slices, motion can occur within the time between the two image acquisitions, inhibiting target tracking.Hence, while the Ortho-SSFP-Echo sequence does not offer a speed advantage over its conventional interleaved counterpart, it does provide a truly simultaneous acquisition, allowing accurate detection of in-plane and out-of-plane motion, which is crucial for radiotherapy.
The Ortho-SSFP-Echo sequence experiences partial signal dephasing within the slice.The effect is enhanced in the SSFP-Echo technique where TE eff is longer than in conventional steady-state sequences. 39This allows for more time for excited spins to move out of the excited slice, resulting in a reduction in signal during acquisition.The impact of dephasing in the Ortho-SSFP-Echo sequence was partially compensated by the introduction of flow-compensation gradients, and the dephasing did not interfere with target tracking.Increasing the slice thickness is another option to reduce the effects of dephasing, by increasing the relative amount of magnetization that stays within the slice; however, this also increases the slice overlap saturation band.Although it is uncommon to treat with radiotherapy or biopsy very small lesions, if the slice thickness was significantly increased (> 10 mm), the band could impede the view of the target.In comparison, the commonly used T 2 /T 1 -weighted bSSFP sequence shows banding artifacts that can obstruct the view of the target, needle, or probe.
In in vivo applications, breathing motion was visualized in two orthogonal slices, with the target at the intersection of the coronal and sagittal slices.When the volunteer is in a supine position, the breathing motion occurs primarily in the H-F direction, which is the RO direction in a coronal slice with a R-L PE direction.Under these assumptions, the flow compensation was integrated in the RO direction.
The current sequence implementation visualizes breathing motion at a sufficient frame rate (∼2 Hz) for cryoablation and needle guidance, which involve slow needle movements.For RT applications, this may be insufficient; a higher imaging frame rate (> 4 Hz) 40,41 is recommended, so that the radiation beam can be adjusted to the exact target location in real time.To achieve this, the matrix could be reduced to 128 × 128, without affecting target tracking (Figure S2, Video S3).Additionally, even lower image quality could potentially be used in RT applications, in which the high-quality diagnostic MR and CT planning images with the corresponding RT plan could be combined with Ortho-SSFP-Echo to adjust the beam.
In the slice overlap region, in addition to the anticipated saturation band, a small bright artifact occurs.This artifact is likely a combination of stimulated echoes from the two orthogonal slices.The artifact is not severe, although it can be shifted by FOV/2 through RF phase cycling. 42While the artifact does not affect tracking, further simulation of this signal is recommended.

CONCLUSIONS
The Ortho-SSFP-Echo sequence could resolve breathing-induced motion of an artificial lesion and the kidney simultaneously in two orthogonal slices.The T 2 -weighted sequence has a TA of 0.45 s, and its application in real-time or interventional MRI has the potential to enhance lesion visibility and improve the accuracy of MR-guided procedures.

A
time series was acquired with the Ortho-SSFP-Echo sequence in 5 healthy volunteers.Reference sequential SSFP-Echo images were also acquired as a ground truth for comparison of the structural similarity, using the MATLAB function SSIM.Both sequences used the same parameters:  = 30 • , TR/TE eff = 6.58/9.86ms, BW = 1300 Hz/px, slice thickness = 6.5 mm, FOV = 380 × 380 mm, matrix = 192 × 192, partial Fourier = 6/8, phase resolution = 80%, GRAPPA = 2, flow compensation = yes, and TA = 0.45 s (per two slices in the orthogonal version and per single slice for the reference).Volunteer experiments were approved by the institutional review board (Ethik-Kommission) of the University Medical Center Freiburg (No. 160/2000).Informed written consent was obtained before imaging.Quantification of the kidney motion for all volunteers was completed by performing the same postprocessing steps used in the phantom lesion-tracking experiments.
Figure2demonstrates the RO flow compensation, in vivo and under free-breathing conditions.Without flow compensation, the signal is largely dephased.However, by using flow compensation, minimal dephasing is observed.In the abdominal phantom, Ortho-SSFP-Echo effectively resolved the simulated breathing motion, with a TA of 0.47 s for both slices (Figure3).Using Ortho-SSFP-Echo (Videos S1 and S2), the lesion was tracked over a range of 9.7, 2.3, and 2.6 mm in the H-F, right-left (R-L), and anterior-posterior (A-P) directions, respectively.The motion patterns between the GRE and Ortho-SSFP-Echo

Figure 4
provides an in vivo contrast comparison between Ortho-SSFP-Echo and the single-slice SSFP-Echo F I G U R E 5 (A,C) In vivo orthogonal images of Volunteers 1 and 2: coronal (A) and sagittal (C) images of the kidney were acquired with the Ortho-SSFP-Echo sequence.The right kidney has been contoured and the corresponding centroids are illustrated with a cross.(B,D) Kidney displacement caused by breathing-induced motion in the head-foot (H-F), right-left (R-L), and anterior-posterior (A-P) directions are shown for the orthogonal slices.The breathing motion indicated by the respiratory belt is displayed alongside the H-F motion.The Pearson correlation coefficient between the respiratory belt and each (coronal and sagittal) slice was 0.91 and 0.92, respectively, for Volunteer 1, and 0.85 and 0.87 for Volunteer 2. Videos S4 and S5 show the in vivo time series of the Ortho-SSFP-Echo, with the kidney contoured.
-Echo sequence.The right kidney has been contoured, and the corresponding centroids are illustrated with a cross.(C) Kidney displacement caused by breathing-induced motion in the head-foot (H-F), right-left (R-L), and anterior-posterior (A-P) directions are shown for the orthogonal slices.The breathing motion indicated by the respiratory belt is displayed alongside the H-F motion.Figure S7.Volunteer 4. In vivo orthogonal images: coronal (A) and sagittal (B) of the kidney, acquired with the Ortho-SSFP-Echo sequence.The right kidney has been contoured, and the corresponding centroids are illustrated with a cross.(C) Kidney displacement caused by breathing-induced motion in the head-foot (H-F), right-left (R-L), and anterior-posterior (A-P) directions are shown for the orthogonal slices.The breathing motion indicated by the respiratory belt is displayed alongside the H-F motion.Figure S8.Volunteer 5.In vivo orthogonal images: coronal (A) and sagittal (B) of the kidney, acquired with the Ortho-SSFP-Echo sequence.The right kidney has been contoured, and the corresponding centroids are illustrated with a cross.(C) Kidney displacement caused by breathing-induced motion in the head-foot (H-F), right-left (R-L), and anterior-posterior (A-P) directions are shown for the orthogonal slices.The breathing motion indicated by the respiratory belt is displayed alongside the H-F motion.