B₀ field homogeneity recommendations, specifications, and measurement units for MRI in radiation therapy

Abstract

Purpose:

The purpose is: (a) Relate magnetic resonance imaging (MRI) quality recommendations for radiation therapy (RT) to B₀ field homogeneity; (b) Evaluate manufacturer specifications of B₀ homogeneity for 34 commercial whole-body MRI systems based on the MRI quality recommendations and RT application; (c) Measure field homogeneity in five commercial MRI systems and one commercial MRI-Linac used in RT and compare the results with their B₀ homogeneity specifications.

Methods:

Magnetic resonance imaging quality recommendations for spatial integrity, image blurring, fat saturation, and null banding in RT were developed based on the literature. Guaranteed (maximum) and typical B₀ field homogeneity specifications for various diameter spherical volumes (DSVs) were provided by GE, Philips, Siemens, and Canon. For each system, the DSV that conforms to each MRI quality recommendation and anatomical RT application was estimated based on the manufacturer specifications. B₀ field homogeneity was measured on six MRI systems including Philips (1.5 T), Siemens (1.5 and 3 T), and ViewRay MRI (0.35 T) systems using 24 and 35 cm DSV spherical phantoms. Two measurement techniques were used: (a) MRI using phase contrast field mapping to measure peak-to-peak (pk-pk), volume root mean square (VRMS), and standard deviation (SD); and (b) Magnetic resonance (MR) spectroscopy by acquiring a volumetric free induction decay (FID) to measure full width at half maximum (FWHM). The measurements were used to assess: (a) conformance with the manufacturer specifications; and (b) the relationship between the various field homogeneity measurement units. Measurements were made with and without gradient shimming (gradshim) or second-order active shimming. Multiple comparisons, analysis of variance (ANOVA), and Pearson correlations were performed to assess the dependence of pk-pk, VRMS, SD, and FWHM measurements of field homogeneity on shim volume, level of shim, and MRI system.

Results:

For a 40 cm DSV, the B₀ homogeneity specifications ranged from 0.35 to 5 ppm (median = 0.75 ppm) VRMS for 1.5 T systems and 0.2 to 1.4 ppm (median = 0.5 ppm) VRMS for 3 T systems. The usable DSVs ranged from 16 to 49 cm (median = 35 cm) based on the image quality recommendations and the manufacturer specifications. There was general compliance between the six measured field homogeneities and manufacturer specifications although signal dephasing was observed in two systems at < 35 cm DSV. The relationships between pk-pk, VRMS, SD, and FWHM varied based on MRI system, shim volume, and quality of shim. However, VRMS and SD measurements were highly correlated.

Conclusions:

The delineation of the diseased lesion from organs at risk is the main priority for RT. Therefore, field homogeneity performance for RT must minimize image blurring and image artifacts (null bands and signal dephasing) while optimizing spatial integrity and fat saturation. Based on the specifications and recommendations for field homogeneity, some MRI systems are not well suited to meet the strict demands of RT particularly for the large imaging volumes used in body MRI. VRMS and SD measurements of B₀ field homogeneity tend to be more stable and sensitive to field inhomogeneities in RT applications than pk-pk and FWHM.

1. INTRODUCTION

1.A. MRI in RT

Magnetic resonance imaging (MRI) serves several roles in radiation therapy (RT) including: (a) Diagnostic imaging; (b) Simulations for treatment planning; (c) MRI-guided radiation therapy (MR-IGRT); and (d) RT response assessment and monitoring. Diagnostic MRIs are typically performed in a Radiology department using clinical MRI systems with static magnetic fields (B₀) ranging from 0.2 to 7 T. The primary objective is diagnosis and delineation of diseased lesions from healthy tissues.

Magnetic resonance imaging simulations are typically performed in Radiation Oncology departments using clinical MRI systems with additional features adapted for RT including flat table top, external laser positioning system (ELPS), and pulse sequences optimized for RT treatment planning. Field strengths range typically range from 0.35 to 3 T.

Magnetic resonance imaging-guided RT can include MRI on rails or fully integrated systems like MRI-⁶⁰Co, MRI-Linac, and MR-guided high intensity focused ultrasound (MRgHIFU).¹ Commercial MR-Linac systems currently operate at field strengths 0.35 and 1.5 T. Any MRI system may be used to assess RT response.

Magnetic resonance imaging in RT often requires performance above and beyond diagnostic MRI.² For example, spatial integrity is a high priority for RT simulations and MR-IGRT. RT applications require a geometric accuracy of ≤2 mm while stereotactic radiosurgery (SRS) demands a geometric accuracy of ≤1 mm.^1,3 MR-IGRT using integrated hybrid systems like the MRI-Linac must maintain high MRI performance despite the presence of RF interference (RFI) sources (e.g., Linac or HIFU arrays) and quasistatic electromagnetic interference (EMI) related to the rotation of the gantry.⁴

1.B. B₀ field homogeneity and MRI quality

The foundation of MRI and RT quality control (QC) programs is the use of ground truth phantoms and measurement procedures to ensure good QC in vivo.^5-7 B₀ field homogeneity is critical to MRI QC and can affect spatial integrity, signal uniformity, and image artifacts. Published in 1986, AAPM Report No. 20 addresses site planning for MRI systems and factors that affect B₀ homogeneity including moving metal.⁸ The 2002 report from AAPM MR Task Group #9 (MR TG9) provides B₀ homogeneity criteria for MR spectroscopy (MRS).⁹ The 2002 report from NMR Task Group #8 addresses field homogeneity recommendations for functional MRI.¹⁰

According to the 2010 AAPM Report No. 100 Acceptance Testing and Quality Assurance Procedures for Magnetic Resonance Imaging Facilities: “For modern cylindrical superconducting magnets, possible inhomogeneity criteria over a 35-centimeter (cm) diameter spherical volume (DSV) are <0.5 parts per million (ppm) root-mean-square (RMS) for systems used for routine imaging, and <0.1 ppm RMS for systems used for ultrafast imaging (echo planar imaging, EPI) and spectroscopy applications. Unfortunately, MRI system vendors report their homogeneity specifications in various ways, using peak-to-peak (pk-pk) or volume RMS (VRMS) measures, for example, from a range of DSV values. This makes direct comparisons of field homogeneity specifications from different vendors rather difficult.”¹¹ To date, these statements have not been methodically evaluated for commercial MRI systems or MRI in RT. AAPM Report No. 100 does not state the rationale for the field homogeneity recommendations nor provide applicable references.

The fundamental question is: What value of field homogeneity is required for RT? Based on the literature, the recommended limits on B₀ inhomogeneity should ideally be based on the imaging application. In RT treatment planning, the imaging priorities are structural: (a) Delineation of the tumor or target volume, (b) Delineation of organs at risk (OARs), and (c) Delineation of fiducial markers or brachytherapy implants. Therefore, the priorities for MRI QC in RT are to limit spatial distortion, blurring, and obscuring artifacts, for example, null banding or poor fat saturation.

Functional MRI including diffusion and perfusion MRI (dynamic susceptibility contrast or dynamic contrast enhanced) acquisitions are commonly employed in RT for tumor detection, MRI simulation, and treatment response.^12,13 However, image fidelity is typically sacrificed to obtain the functional information (e.g., tissue diffusion or perfusion). MRS is not commonly employed in diagnostic or RT MRI protocols despite MRS’s utility for classification of tumors (e.g., glioma) and assessment of treatment response.¹⁴

Gradient-recalled echo (GRE) sequences are more vulnerable to the effects of B₀ field inhomogeneity than spin echo sequences. Yet, GRE sequences are integral to MRI simulation, MR-IGRT, and functional MRI.^15,16 For example, brain SRS relies on 3D T₁-weighted (T1W) spoiled gradient-recalled echo (SPGR) sequences.¹⁷ Body MRI frequently uses a T1W volumetric interpolated breath-hold examination (VIBE) GRE sequence.¹⁸ Prostate treatment planning and MR-IGRT and partial breast irradiation treatment planning use 3D Dixon GRE acquisitions to create pseudo CT maps and facilitate tissue segmentation.^19-21 Balanced steady-state free precession (bSSFP) is used in all aspects of MR-IGRT including simulation, patient setup/alignment, and real-time tracking.^22,23 Four-dimensional (4D) GRE MRI is being used in body RT treatment planning to reduce dose errors caused by organ motion.²⁴ Poor field homogeneity also causes peak broadening that degrades fat saturation used with both GRE and spin echo MRI, and affects the sensitivity of MRS.²⁵

As suggested by AAPM Report No. 100, the choice of measurement unit for field homogeneity is important to ensure clarity and enable comparisons.¹¹ Some performance specifications like MRS and susceptibility are field strength dependent so a relative unit like ppm is appropriate with:

$$ppm=\frac{\Delta B}{B_0}=\frac{\Delta f}{f_L}$$ (1)

where B₀ is the magnetic field strength, f_L is the Larmor (resonant) frequency, ΔB represents the magnetic field inhomogeneity or offset (e.g., chemical shift), and Δf represents the frequency variation or offset. Note: We will use ΔB and Δf interchangeably herein to refer to field inhomogeneity where Δf = γΔB/(2π) and γ is the gyromagnetic ratio (2.675 × 10⁸/(T–s) for ¹H).

However, many of the imaging specifications are independent of field strength since they depend on the receiver bandwidth (rBW), and thus, absolute units of T (ΔB) or Hz (Δf) are preferred.

Previous studies measured global (over the volume or slice of interest) and local (over the voxel of interest) ΔB in terms of standard deviation (SD), that is, σB₀^Global and σB₀^Local, respectively. The B₀ inhomogeneities were primarily caused by the variations in tissue magnetic susceptibility. Typical σB₀^Global values are 0.8 ppm for the brain and 1.6-3.2 ppm for the body with active shimming.^26-28 σB₀^Local values can range from 0 ppm in homogeneous tissue to 9 ppm at air/tissue boundaries and 10⁴-10¹¹ ppm near metal implants.²⁹ Pk-pk field inhomogeneity values of 1.6 ppm in the brain, 4.8 ppm in the breast, 5.6 ppm in the liver, and 4.8 ppm in the spinal cord were reported with active shimming.^27,30-32

Hence, there are four objectives that will be addressed herein. First, we will relate MRI quality recommendations for RT to B₀ field homogeneity for systems designed for diagnostic MRI, RT simulation, and MR-IGRT. Second, we will evaluate manufacturer specifications of B₀ field homogeneity for clinical MRI and MR-IGRT systems based on the MRI quality recommendations and anatomical RT applications. Third, we will compare phantom measurements of B₀ field homogeneity from five MRI systems and one MRI-Linac used in our RT clinical and research operations with their manufacturer specifications. Fourth, we will determine which field homogeneity measurement unit is best suited for RT.

2. MATERIAL AND METHODS

2.A. B₀ field homogeneity and MRI quality

Magnetic resonance imaging field homogeneity criteria were developed to address four concerns: (a) spatial integrity for brain SRS; (b) image blurring during large volume (body) treatment planning; (c) fat-water chemical shift effects on simulations and MR-IGRT associated with the prostate and breast; and (d) bSSFP null banding associated with large volume MR-IGRT (e.g., adaptive MR-IGRT³³).

2.A.1. Spatial integrity

The main causes of spatial integrity errors are gradient nonlinearities, eddy currents and k-space trajectory errors, and unsatisfactory distortion correction (e.g., gradwarp). However, B₀ field inhomogeneities including localized areas of high susceptibility like the sinuses can cause image distortion especially with GRE sequences. Geometric shifts (Δx) depend on the readout gradient (G_r) or receiver bandwidth (rBW):

$$\mid \Delta x\mid =\mid \frac{\Delta B}{G_r}\mid =\frac{d_x\cdot \gamma \mid \Delta B\mid }{2\pi \cdot rBW}=\frac{d_x\cdot \mid \Delta f\mid }{\delta \cdot f_L}$$ (2)

where d_x is the pixel dimension and ∣ΔB∣ is the field inhomogeneity across the voxel. Ideally, the pixel bandwidth should be larger than the fat-water chemical shift (δ = 3.4 ppm) to avoid chemical shift artifacts in gradient echo acquisitions. However, in brain SRS, pixel bandwidth ranges of 1.6-3.2 ppm/mm are common. To ensure Δx is <0.5 mm, ∣Δf∣ should be <0.9 ppm assuming d_x is 1 mm. Δf_VRMS should be <0.3 ppm to ensure all voxels are compliant (±3σ).

2.A.2. Blurring

Field inhomogeneities cause blurring. The contribution of ΔB to the pixel resolution is:³⁴

$$\Delta x_{blur}=\Delta x\frac{\alpha }{1-e^{-\alpha }}$$ (3)

$$\text{with}\alpha =\frac{t_s}{2}\left(\frac{\epsilon }{T_2}+\kappa \gamma \mid \Delta B\mid \right)$$ (4)

where Δx is the voxel dimension, t_s is the sampling time, κ is a geometric factor (κ = 1 is typically used), and ϵ = 1 for gradient echoes and 0 for spin echoes.^34,35 Assuming t_s < 5 ms, a local field inhomogeneity of <100 Hz is desirable to ensure Δx_blur < 2∙Δx and the lesion margin is not perceived to extend into the adjoining voxel.³⁶ Δf_VRMS should be <33 Hz to ensure all voxels are compliant (±3σ).

2.A.3. bSSFP null banding

In conventional bSSFP using RF pulses that have alternating (0, π) phase, null bands typically occur at the frequencies ±n/(2∙TR) where n is an odd integer. For a maximum TR of 4 ms, null bands would occur at ±n∙125 Hz off-resonance.³⁷ Therefore, null bands are most likely to occur near regions of high susceptibility (e.g., air/tissue interfaces and metal implants) or distal to isocenter. Δf_VRMS should be <42 Hz to ensure all voxels are compliant (±3σ).

2.A.4. Fat saturation

A field inhomogeneity of <0.5 ppm was recommended for effective fat saturation using frequency selective pre-saturation pulses associated with a wide range of anatomical targets.³⁸ This recommendation will also minimize phase errors that degrade 3D Dixon GRE tissue-segmented images used in prostate and breast treatment planning.

2.B. Clinical B₀ field homogeneity specifications

Field homogeneity specifications were provided by GE, Philips, Siemens, and Canon for their 1.5 and 3 T horizontal bore whole-body MRI systems. Maximum (or guaranteed) specifications were provided along with typical values based on an average from multiple (e.g., 10) installations. Typical values were calculated by the vendor based on field camera measurements of the magnet after superconducting (where available e.g., GE and ViewRay) and room-temperature passive shimming, and assume the first, and if applicable, second-order spherical harmonics are zeroed using active shimming. The actual field homogeneity of a system may vary from the typical values but should be less than or equal to the maximum specification.

The MRI manufacturers typically report their field homogeneity specifications in units of VRMS, defined as:³⁹

$$VRMS={\left[\frac{1}{V}\underset{V}{\overset{}{\int }}{\left[B_0\left(\stackrel{\rightharpoonup }{\mathrm{r}},\theta ,\phi \right)-B_0(0)\right]}^{2}dV\right]}^{1∕2}$$ (5)

where r is the radius, φ is the azimuthal angle, θ is the polar angle, and B₀(0) is measured at the center of the volume of interest (VOI).

Standard deviation (SD) is commonly used to measure field inhomogeneities (ΔB) and their effects, for example, on functional MRI (fMRI) and spectroscopy.^26,40 SD is defined using the mean B₀ (<B₀>) within the volume of interest and is typically measured in material (e.g., a phantom or in vivo):

$$SD={\left[\frac{1}{V}\underset{V}{\overset{}{\int }}{\left[B_0\left(\stackrel{\rightharpoonup }{\mathrm{r}}\right)-\langle B_0\left(\stackrel{\rightharpoonup }{\mathrm{r}}\right)\rangle \right]}^{2}dV\right]}^{1∕2}$$ (6)

RMS is synonymous with SD when the DC component of the field is subtracted.

ViewRay and Elekta report their field homogeneities for their MRI-Linacs in pk-pk and FWHM units. Pk-pk field homogeneity for a spherical volume was defined as:

$$pk-pk(r)=Max\left[B_0\left(\stackrel{\rightharpoonup }{\mathrm{r}}\le r\right)\right]-Min\left[B_0\left(\stackrel{\rightharpoonup }{\mathrm{r}}\le r\right)\right]$$ (7)

During MRI installation and shimming, the pk-pk measurements are typically derived directly from the maximum and minimum field camera measurements for the field camera’s DSV. For smaller DSV, pk-pk field homogeneity was derived from the calculated free space magnetic field.³⁹ For phantom measurements, pk-pk field homogeneity was derived from the MRI-based field map.

The FWHM for the free induction decay (FID) was calculated by fitting the spectrum to a Lorentzian (i.e., Cauchy distribution):

$$L_{pdf}(f)=\frac{\epsilon }{\pi \left((f-f_c{)}^2+{\epsilon }^2\right)}$$ (8)

where f_c is the peak’s center. The FWHM (in Hz) of the Cauchy distribution is:³⁴

$$FWH{M}_L=2\epsilon =\frac{1}{\pi T_2^{\ast }}$$ (9)

$$\frac{1}{T_2^{\ast }}=\frac{1}{T_2}+\kappa \gamma \mid \Delta B\mid =\frac{1}{T_2}+\frac{1}{T_2^{\prime }}$$ (10)

Variance is undefined for the Cauchy probability density function (pdf) so the relationship between FWHM and SD was not calculated using only spectroscopy. Mathematica's interpolation function (Wolfram Research Inc., Version 12.0) was also used to calculate FWHM since we found its fit to the spectroscopy data was superior to the Lorentzian fit function.

The field homogeneity for each MRI system was fit to polynomial functions (up to sixth order) and spline functions using Microsoft Excel. The minima of the spline and polynomial fits of the vendor’s DSV specifications were used to interpolate the system specifications to a 35 cm DSV and minimize the risk of a false negative for compliance with system specifications. The polynomial interpolations were used to calculate the maximum DSV that complied with each MRI quality recommendation.

2.C. Institutional B₀ field homogeneity measurements

Measurements of field homogeneity were conducted on six MRI systems used in our RT clinical and research operations to: (a) verify compliance with the manufacturer specifications; and (b) determine the best measurement units for RT QC. The MRI systems included a Philips Ingenia 1.5 T MRI Simulator (Version 5.3.1, last shimmed in August 2014) and a 0.35 T ViewRay MRIdian MRI-Linac system (Version VB19, last shimmed in October 2017) located in the Department of Radiation Oncology, three Siemens 3 T systems located in the Mallinckrodt Institute of Radiology Center for Clinical Imaging Research (CCIR): Vida (Version XA11A, last shimmed in August 2017), mMR (Version VE11P, last shimmed in July 2017), and Prisma (Version VE11C, last shimmed in January 2019), and a Siemens Espree 1.5 T (VB19, last shimmed in June 2010) located in the Center for Advanced Medicine Radiology suite. The Siemens MRIs were equipped with second-order active shims. The vendor’s field homogeneity (ppm) specifications were interpolated to 24 and 35 cm DSVs except the maxima was used to increase the likelihood our measurements were in compliance with the interpolated specifications.

Field homogeneity was measured on the ViewRay MRIdian MRI-Linac at gantry angles from 0 to 330° in increments of 30°. The 0.35 T ViewRay MRI-Linac system has six 227 kg steel shields that hold the linear accelerator components and are spaced 60° about the rotating gantry. Each Linac gantry angle creates a unique field homogeneity for the MRI subsystem. The data were used to test the hypothesis that there is a consistent relationship between different measurements units of MRI magnetic field homogeneity.

On the Philips 1.5 T, field maps were acquired using Philips’ 3D B₀ Map sequence [TE1/TE2/repetition time (TR): 2.86/7.41/30 ms, acquisition matrix: 128 × 128, acquired voxel size: 3x3x5 mm, reconstructed voxel size: 3 × 3 × 2.5 mm, 383 Hz/pixel, flip angle (FA): 60°, acquisition time (TA): 584 s]. On the Siemens MRIs, B₀ maps were acquired using Siemens’ T₁-weighted 2D dual GRE field map sequence (gre_field_mapping, TE1/TE2/TR: 5.00/7.46-9.76/800-1000 ms, slice thickness/gap: 5/1 mm, matrix: 128 × 128, Resolution: 3 × 3 × 5 mm, 260 Hz/pixel, FA: 90°, TA: ≤260 s). On the ViewRay MRIdian, B₀ maps were acquired using Siemens’ T₁-weighted 2D dual GRE field map sequence (gre_field_mapping, TE1/TE2/TR: 6/8.5-10/1000 ms, slice thickness/gap: 5/1 mm, matrix: 128 × 128, Resolution: 3 × 3 × 5 mm, 130 Hz/pixel, FA: 90°, TA: 254 s). TE2 was adjusted for the ViewRay MRIdian to minimize phase aliasing to ensure the entire range of frequencies can be unambiguously resolved. The 2D B₀ maps were acquired in the axial, sagittal, and coronal orientations per the ACR MRI QC Manual.⁴¹ The body coil was used for transmission and reception for all field mapping and signal-to-noise ratio (SNR) measurements. Measurements were conducted at room temperature (19-22°C).

For the Philips and ViewRay systems, measurements were made with first-order (X, Y, and Z) active (gradient) shimming (gradshim) turned on and then off. For the Siemens systems, we were unable to acquire field maps with the first-order and second-order (Z2, ZX, ZY, C2, and S2) active shim currents set to zero due to severe geometric distortion and signal dephasing. Thus, we obtained field maps with first and second-order active shims turned on (Full Shim), and then, with second-order active shim currents set to zero. Please note that other publications on this subject substitute “degree” for “order.”⁴⁰

Measurements were made using a 35 cm internal DSV phantom (HP Manufacturing, Cleveland, OH) filled with food-grade white mineral oil (UltraSource, UltraPro white mineral oil, CAS 8042-47-5, $5/L, molecular weight: 234 g/mol, density 0.86 g/cm³, viscosity: 11.5 mm²/s at 40°C, Supplier: Amazon). The low dielectric constant of the mineral oil (ε’ < 2.6) prevents dielectric artifacts from B₁ inhomogeneities at the radio frequencies (RF) used in this study.⁴² T₁, T₂, and the proton density of the mineral oil met the NEMA guidelines for phantom fill liquids as previously reported.⁴³ The NMR spectrum of the mineral oil was measured using a Pulsar 60 MHz NMR table-top spectrometer (Oxford Instruments) at 37°C.

Measurements were also made using two 24 cm DSV phantoms for comparison with the 35 cm DSV phantom, including: (a) Siemens spherical water phantom doped with 5 mM NiSO₄ for ≤1.5 T measurements; and (b) Siemens spherical mineral oil phantom for 3 T measurements to minimize dielectric effects. All of the phantoms were centered at isocenter for the measurements.

With the exception of the 1.5 T Ingenia, MRS measurements were made by acquiring a FID in the phantom [TR: 3 s, FA: 90°, 4 Averages, BW: 1260 Hz, 128 complex points] acquired using the body coil. FIDs were averaged and converted into a spectrum using a discrete fast Fourier transform (FFT). The institution’s Ingenia did not have spectroscopic sequences so the spectrum was exported from the manual adjustments menu (interactive f0 feature).

The Philips B₀ Map sequence generated unwrapped field maps in units of Hz after rescaling. For the ViewRay and Siemens MRIs, the field map images $S\left(\stackrel{\rightharpoonup }{r}\right)$ were converted into phase maps ${\phi }_{raw}\left(\stackrel{\rightharpoonup }{r}\right)$ with values [−π, π] using:

$${\phi }_{raw}\left(\stackrel{\rightharpoonup }{r}\right)=\pi \left(\frac{S\left(\stackrel{\rightharpoonup }{r}\right)-2048}{2048}\right)$$ (11)

where $\stackrel{\rightharpoonup }{\mathrm{r}}$ is the pixel location, and a pixel value range of [0, 4095]. FSL’s PRELUDE module (https://www.win.ox.ac.uk/) was used to unwrap the phase, where needed.^44,45 The unwrapped phase maps $\Delta \phi \left(\stackrel{\rightharpoonup }{r}\right)$ were converted into frequency distributions $\Delta f\left(\stackrel{\rightharpoonup }{r}\right)$ using:

$$\Delta f\left(\stackrel{\rightharpoonup }{r}\right)=\frac{\Delta \phi \left(\stackrel{\rightharpoonup }{r}\right)}{2\pi \cdot \Delta \text{TE}}$$ (12)

where ΔTE is the difference in echo times (TE2-TE1).

VRMS [using Eq. (5)], pk-pk, and SD_I were calculated from the field maps for DSVs ranging from 10 to 35 cm in the 35 cm DSV phantom, and for 10 to 24 cm in the 24 cm DSV phantoms.

2.D. Statistical analyses

Statistical analyses were performed using Microsoft Excel and Mathematica to assess the relationships, if any, between the various field homogeneity measurement units. For the 2D B₀ maps, the maxima between the three orthogonal B₀ maps was recorded and used in the analyses. Pearson correlation coefficients were calculated between the various measurement units using the phantom measurements. An analysis of variance (ANOVA) and multiple comparisons (Tukey’s test) were performed to determine the effects of shimming, DSV (24 vs 35 cm), and the MRI system on the various measurement units and their ratios.

Repeatability measurements were conducted on the 0.35 T ViewRay MRIdian system using the 24 and 35 DSV phantoms to investigate the effects of low SNR on measurement uncertainties. Ten B₀ maps were acquired at gantry angle 0° for each phantom with shimming. The results were compared with repeatability measurements conducted on the 1.5 T Ingenia and the 3 T mMR using the 24 cm DSV phantoms. SNR was measured on the six systems using both phantoms and calculated using $0.655\cdot \frac{\langle Signal\rangle }{{\sigma }_{background}}$⁴⁶. The SNR was measured from the B₀ map magnitude images (first echo time, central slice) and averaged over the acquisitions.

3. RESULTS

3.A. B₀ field homogeneity and MRI quality

The Δf_VRMS for each MRI quality recommendation and RT application is shown in Table I along with the assumptions. We also assumed that: (a) σB₀^Global and σB₀^Local were equivalent which occurs with good field homogeneity;²⁶ and (b) VRMS and σB₀ (SD) were equal.

Table I.

Δf_VRMS recommendations for different RT applications.

RT application	Vulnerability	Effect of ΔB0	Assumptions	ΔfVRMS
Brain SRS	Susceptibility and low rBW	Geometric distortion	3D T1W GRE rBW> 1.6 ppm/mm	<0.3 ppm
Body (large FOV) treatment planning	Long GRE readout time	Blurring	3D GRE (VIBE) ts < 5 ms	<33 Hz
Abdominal adaptive MR-IGRT	Susceptibility, large FOV	Null banding	2D & 3D bSSFP TR ≤ 4 ms	<42 Hz
Prostate and breast treatment planning and MR-IGRT	High-fat content, susceptibility, large FOV	Poor fat saturation or fat-water delineation	3D Dixon GRE 2D T2W with Fat Sat ∣δ∣ = 3.4 ppm	<0.5 ppm

ppm, Hz/Larmor frequency (MHz); δ, fat-water chemical shift; bSSFP, balanced steady-state free precession; FOV, field of view; GRE, gradient-recalled echo; MR-IGRT, MRI-guided radiation therapy; rBW, receiver bandwidth; SRS, stereotactic radiosurgery; TR, repetition time; t_s, readout sampling time; VIBE, volumetric interpolated breath-hold examination.

3.B. B₀ field homogeneity specifications

The vendor field homogeneity specifications are presented in the Tables S1-S7. The field homogeneity ranges and medians versus DSV for the MRI systems (excluding ViewRay) are summarized in the box plot (Fig. 1). Based on the interpolations, the Siemens Prisma and Trio are the only MRI systems from our list that satisfy the AAPM field homogeneity recommendation for ultrafast imaging and spectroscopy (<0.1 ppm RMS for a 35 cm DSV) based on their typical performance specifications. Also based on the interpolations, at least six systems did not typically meet the AAPM field homogeneity recommendation for routine MRI (<0.5 ppm RMS for a 35 cm DSV). The ViewRay MRIdian was excluded since its specifications are not in units of VRMS.

Fig. 1.

Box plots of VRMS field homogeneity (in ppm) derived from the vendor maximum specifications (Tables S1-S6) for the 34 clinical 1.5 T (top) and 3 T (bottom) MRI models. Note: Some models did not have specifications for DSVs > 40 cm resulting in a drop in the upper field homogeneity range at 45 cm DSV. We excluded elliptical volumes.

3.C. B₀ field homogeneity measurements

The results of the phantom field homogeneity measurements for the six MRI systems by DSV are compared in Fig. 2 with their maximum and typical system specifications (with the exception of ViewRay). The results are compared for different levels of active shimming. The measurements complied with the manufacturer specifications for the DSVs in which excessive signal dephasing was absent. Detailed results by DSV are included in the Tables S8-S11.

Fig. 2.

Comparison of field homogeneity measurements from 35 cm DSV phantom versus manufacturer specifications (where available). (a) ViewRay 0.35 T system measured at gantry angle 0°. (b) Philips Ingenia 1.5 T. (c) Siemens Espree 1.5 T. (d) Siemens mMR 3 T. (e) Siemens Prisma 3 T. (f) Siemens Vida 3 T. Note: ViewRay does not have a manufacturer specification for VRMS. Measurement data were limited to ≤ 24 cm DSV for the Espree and ≤ 30 cm DSV for the mMR due to signal dephasing associated with the field map sequences. The ViewRay and Philips systems do not have second-order active shims.

Table II presents the field inhomogeneity maxima for the 1.5 T Philips Ingenia and Siemens Espree. The Ingenia has only first-order active shimming (gradshim). The Espree has first and second-order shimming (named full shim herein). Table III presents the field inhomogeneity maxima for the 3 T Siemens Prisma, Vida, and mMR. Results are presented by measurement unit, phantom, and level of active shim. The field inhomogeneities increased with increasing DSV and reduced active shimming (no gradshim for the Ingenia, gradshim only for the other systems). For the Siemens 1.5 T Espree and 3 T mMR, signal dephasing associated with the field map echo times restricted the usable DSV to ≤24 and ≤30 cm, respectively (Fig. S1).

Table II.

Maxima from 1.5 T Philips Ingenia and Siemens Espree field homogeneity measurements for 24 and 35 cm DSV phantoms.

Unit	DSV (cm)	Ingeniab		Espree
		With gradshim	No gradshim	Full shim	Gradshim only
Field homogeneity (ppm)
pk-pk	24	2.214	2.576	3.337	4.035
	35	3.396	3.398	a	a
VRMS	24	0.104	0.244	0.163	0.392
	35	0.264	0.462	a	a
SD	24	0.104	0.244	0.163	0.392
	35	0.264	0.462	a	a
FWHM	24	0.264	0.972	0.096	0.552
	35	1.127	1.972	1.074	1.757

^aIndicates the measurement was affected by significant signal dephasing.

^bDoes not have second-order active shimming.

Table III.

Maxima from 3 T Siemens MRI field homogeneity measurements for 24 and 35 cm DSV phantoms.

Unit	DSV (cm)	Prisma		Vida		mMR
		Full shim	Gradshim only	Full shim	Gradshim only	Full shim	Gradshim only
Field homogeneity (ppm)
pk-pk	24	0.386	0.696	0.816	1.232	1.046	1.208
	35	2.378	2.075	3.268	3.296	a	a
VRMS	24	0.024	0.084	0.049	0.183	0.087	0.090
	35	0.067	0.136	0.150	0.366	a	a
SD	24	0.023	0.084	0.048	0.183	0.087	0.090
	35	0.067	0.136	0.150	0.366	a	a
FWHM	24	0.107	0.273	0.174	0.517	0.252	0.316
	35	0.158	0.317	0.237	1.079	0.379	0.521

^aIndicates the measurement was affected by significant signal dephasing.

Table IV presents the mean ratios of the field homogeneity measurement units from the 1.5 T Philips Ingenia and Siemens Espree. Table V presents the mean ratios of the field homogeneity measurement units from the 3 T Siemens Prisma, Vida, and mMR. Results are presented by measurement unit ratio, phantom, and level of active shim. Variations of ratios between different systems or phantom sizes indicates that the relationships between the measurement units are not constant.

Table IV.

Mean ratios of field homogeneity measurement units from 1.5 T MRIs for 24 and 35 cm DSV phantoms.

Ratio	DSV (cm)	Ingeniab		Espree
		With Gradshim	No Gradshim	Full shim	Gradshim only
pk-pk/VRMS	24	20.626	10.697	19.430	9.687
	35	12.805	7.365	a	a
pk-pk/FWHM	24	8.186	2.698	32.011	6.771
	35	3.000	1.723	a	a
VRMS/FWHM	24	0.397	0.252	1.642	0.699
	35	0.234	0.234	a	a

^aIndicates the measurement was affected by significant signal dephasing.

^bDoes not have second-order active shimming.

Table V.

Mean ratios of field homogeneity measurement units from 3 T Siemens MRIs for 24 and 35 cm DSV phantoms.

Ratio	DSV (cm)	Prisma		Vida		mMR
		Full shim	Gradshim only	Full shim	Gradshim only	Full shim	gradshim only
pk-pk/VRMS	24	16.332	7.809	14.093	6.550	12.473	13.213
	35	28.929	14.790	21.464	9.432	a	a
pk-pk/FWHM	24	3.488	2.271	3.754	2.216	3.971	3.710
	35	14.063	6.990	13.349	3.053	a	a
VRMS/FWHM	24	0.214	0.290	0.267	0.338	0.322	0.281
	35	0.485	0.432	0.622	0.324	a	a

^aIndicates the measurement was affected by significant signal dephasing.

The results of field homogeneity measurements for the 0.35 T MRI-Linac at various gantry angles are presented in Fig. 3 using the 24 and 35 cm DSV phantoms and the various measurement units. The measurement values vary with gantry angle especially without gradshim.

Fig. 3.

Comparison of field homogeneity measurements and units versus gantry angle in 0.35 T ViewRay MRI-Linac. The 35 cm DSV phantom data with (a) and without (b) gradient shimming. The 24 cm DSV phantom data with (c) and without (d) gradient shimming.

Table VI summarizes the mean ratios and correlation coefficients between different field homogeneity measurement units for the ViewRay MRIdian data acquired at various gantry angles. SD values and ratios were not shown since they were similar to VRMS. The 0.35 T results corroborate the findings from the 1.5 and 3 T measurements that the relationships between the measurement units were variable.

Table VI.

Comparison of measurement unit mean ratios and correlation coefficients from 24 and 35 cm DSV phantoms for 0.35 T ViewRay MRIdian MRI-Linac measurements for gantry angles from 0° to 330° in 30° increments.

Comparison	DSV	With gradient shimming			Without gradient shimming
		Mean ratio (SD)	Pearson coefficient	P value	Mean ratio (SD)	Pearson coefficient	P value
pk-pk vs VRMS	24	7.702 (1.093)	−0.638	0.023	5.871 (0.928)	0.946	<<0.001
	35	10.979 (2.347)	−0.004	0.991	7.397 (1.474)	0.692	0.010
pk-pk vs FWHM	24	3.288 (0.901)	−0.371	0.230	1.809 (0.867)	0.822	<0.001
	35	3.753 (0.704)	0.360	0.246	2.092 (0.761)	0.541	0.065
VRMS vs FWHM	24	0.424 (0.090)	0.707	0.008	0.304 (0.135)	0.931	<<0.001
	35	0.345 (0.033)	0.560	0.054	0.279 (0.071)	0.910	<<0.001

FWHM: Full width half maximum, pk-pk: phase map peak-to-peak, VRMS: phase map volume root mean square.

Statistical analysis results including correlations, ANOVA, and multiple comparisons are detailed in the Tables S12-S15. The statistical analyses indicated that: (a) The units SD and VRMS were the most sensitive to changes in field homogeneity; and (b) The relationships between different measurement units were variable.

Measurement repeatability SDs for the 0.35 T MRI-Linac, 1.5 T Ingenia, and 3 T mMR are presented in the Supplemental Materials (Fig. S2). The SNR at 0.35 T was 20 and 39, respectively, for the 24 and 35 cm DSV phantoms (Table S16). By comparison, the SNRs for the 1.5 and 3 T systems were >117. Yet, the repeatability for the 0.35 T was high especially for the SD and VRMS measurements (<1 Hz), and compared favorably with the repeatabilities of the 1.5 and 3 T systems.

The maximum DSV that meets each B₀ field homogeneity recommendation and RT anatomical application is listed by system in Tables S17-S21 for the 34 vendor systems. The systems and recommendations are combined in the box plot of Fig. 4. The DSVs were rounded to the nearest centimeter and are based on polynomial fits of the maximum (guaranteed) specifications provided by the manufacturers.

Fig. 4.

The box plot shows the distribution of the maximum DSV for the 34 MRI models that meet each Δf_VRMS recommendation. The maximum DSV was based on a polynomial interpolation of the manufacturers’ maximum (guaranteed) field homogeneity specification (excluding ViewRay). The results assume field homogeneity isotropy so actual values may be different due to anisotropy or local regions of increased inhomogeneity. The AAPM Report No. 100 recommendations for a 35 cm DSV (red line) for ultrafast MRI and MRS, and ^†routine MRI (same Δf_VRMS as prostate and breast RT) are shown for comparison.

4. DISCUSSION

4.A. B₀ field homogeneity and MRI quality

We proposed B₀ field homogeneity recommendations based on RT image quality priorities. The recommendations can be easily adapted to a site’s nominal pulse sequence or MRI protocol performance requirements, or updated as the technology improves. Based on the QC requirements and the manufacturer specifications, it is easy to estimate the usable fields of view and the utility of the prospective MRI system.

4.B. B₀ field homogeneity specifications

There was divergence between the field homogeneity recommendations of AAPM Report No. 100 and the manufacturer specifications. Based on the maximum (or guaranteed) specifications provided by the MRI manufacturers, none of the 34 commercial MRI systems meet < 0.1 ppm VRMS over a 35 cm DSV recommended for ultrafast imaging and spectroscopy. However, the Trio and Prisma meet this specification based on their typical performances. In addition, the guaranteed specifications for a third of the commercial MRI systems did not meet the recommendation of <0.5 ppm VRMS over a 35 cm DSV for routine imaging. Two systems could not meet this specification based on typical performance. The number may be higher since typical performance specifications were not provided for all systems.

4.C. B₀ field homogeneity measurements

Our B₀ field homogeneity VRMS measurements were consistent with the manufacturer system specifications with the exception of the 3 T MRIs for the 10 cm DSV (Table S10) and the Espree and mMR for large DSVs. One should not assume that the largest DSV in the specification represents the largest imaging field of view.

With the exception of VRMS and SD, no fixed relationship between field homogeneity measurement units (Pk-pk, VRMS, SD, and FWHM) was observed between MRI systems or between gantry angles on the MR-IGRT. For the phantom measurements, SD was more robust than VRMS since SD is defined using the average B₀ in the volume. For spectroscopy, FWHM is often calculated as 2.35·SD based on a Gaussian pdf.²⁶ In our measurements, FWHM/SD ranged from 0.61 to 4.67 and varied by DSV, MRI system, and use of shimming.

All of the major commercial manufacturers of diagnostic MRI systems use VRMS and assume full shimming for their specifications. Therefore, direct comparisons between the vendor specifications is straightforward. It would also be beneficial if the commercial MR-IGRT systems used VRMS specifications for field homogeneity.

A 2006 study compared field homogeneity measured on seven different MRI systems using the water linewidth (FWHM) within individual 15 × 15 × 40 mm voxels, phase-difference maps (pk-pk), and bandwidth-difference (pk-pk) measurements using a custom 22.6 cm DSV grid phantom.⁴⁷ The primary aim of their study was to demonstrate the bandwidth-difference method. Comparison of the three methods was challenging since they were not measured at the same DSV. Comparison of their results with our study is difficult since they used different MRI models and did not acquire a global FWHM.

4.D. B₀ field homogeneity and RT

B₀ field homogeneity typically benefits from a long magnet, small bore diameter, and large shim capacity. However, RT typically requires a large bore diameter, at least 70 cm given the trend toward large body habitus and the desire to position the tumor at isocenter for maximal field homogeneity and spatial uniformity.

Although patient comfort favors a short bore, MRI simulation and MR-IGRT systems used in RT must have high field homogeneity over a large field of view for body applications. Typical fields of view range from 24 to 36 cm in the longitudinal (head/foot) direction and >50 cm in the orthogonal directions.

3 T MRI systems provide roughly twice the SNR as 1.5 T so they are attractive candidates for MRI simulation in RT. The SNR can be traded for increased image resolution during image acquisition. Since B₀ homogeneity specifications are in ppm, a 3 T system with a VRMS of 0.5 ppm for a given DSV has equivalent field homogeneity performance as a 1.5 T system with a VRMS of 1 ppm for the same DSV. However, higher field strength results in higher specific absorption rate (SAR). Therefore, a 3 T MRI for simulations requires extra vigilance for RT given the prevalence of metal implants in our patient population (e.g., >70% of the patients receiving MRI simulations at our institution had metal in their bodies).

Local B₀ inhomogeneities at soft tissue interfaces with air, bone, or metal scale with field strength. Excluding implants, field inhomogeneities of up to ±3 ppm were reported in vivo after active shimming.^27,30,31,48 Increased shimming capability is required at high field. Thus, 3 T systems are equipped with second-order active shims.⁴⁹

We noted severe signal dephasing in the longitudinal (head/foot) direction on the Espree for a FOV > 24 cm and the mMR for an FOV > 30 cm due to the use of gradient echo sequences with TEs ≥ 5 ms. The use of larger coronal or sagittal FOVs for these systems may result in artifact and geometric distortion. Yet, we routinely perform MRI simulations for Gamma Knife SRS on the Siemens 1.5 T Espree using gradient echo sequences with echo times ≤ 8 ms. Geometric distortion in the brain is consistently ≤ 1 mm based on the Elekta stereotactic frame landmarks and co-registration of the brain MRIs with CT images. Although the Espree is no longer in production, MRI systems with similar field homogeneities are available. Such systems should be avoided for RT simulations of the body.

We have also successfully used the Siemens 3 T mMR for PET/MRI delineation of gynecologic and prostate tumor boundaries and assessment of ⁹⁰Y microsphere radioembolization dose.^50,51 However, the nominal 60 cm bore diameters of the commercial PET/MRIs may not accommodate many of our patients since ~40% of our patients are obese consistent with the National Health and Nutrition Examination Survey.^52,53

Of the institutional 70 cm diameter wide-bore systems we measured, the Siemens 3 T Vida and the Philips 1.5 T Ingenia had the highest field homogeneity for their respective field strengths. The Ingenia is our default MRI simulator and came equipped with RT features (flat table, external laser positioning, RT QA phantoms and sequences).

MRI systems with excellent field homogeneity specifications are desirable but not sufficient if the MRI system lacks the capabilities to adequately shim the field in the presence of a human body.⁵⁴ Regions of high susceptibility like tissue/air interfaces or metal implants will produce significant field inhomogeneities and may require effective shimming including active shims with high current capacities and higher order terms at high field strengths.⁵⁵ Dynamic shimming (e.g., Siemens BioMatrix CoilShim and Slice Adjust) now permits localized shimming based on slice or voxel.^56,57 The aim of dynamic shimming is to address inhomogeneities over single slices or regions that cannot be adequately minimized by conventional field shimming.⁵⁸ Dynamic shimming can be challenging because of the settling times that may be required after changing the shim value.

The Elekta Unity MR-IGRT (Stockholm) uses a Philips 1.5 T Ingenia MRI modified with a gap designed into the B₀ and gradient windings to minimize beam attenuation at the longitudinal center of the system. The 1.5 T Elekta Unity does not currently have field homogeneity specifications. However, pk-pk measurements of 0.3-1.6 ppm without shimming and 0.2-0.67 ppm with shimming were recently reported for a single 2D image plane at isocenter in a 40 cm cylindrical phantom across gantry angles of 0°–360° in 30° intervals.⁵⁹

If we compare the 0.35 T ViewRay MRIdian with the 1.5 T Philips Ingenia (as a surrogate for the Elekta Unity), their VRMS field homogeneities for a 35 cm DSV are similar (17 Hz). bSSFP used in tumor tracking during MR-IGRT has less null banding and higher contrast in vivo at 0.35 versus 1.5 T. Nevertheless, the choice between 0.35 versus 1.5 T for MR-IGRT may depend on other considerations including SNR, sequence acquisition times, SAR, secondary electron return effects, the quality of alternative (e.g., T₁ and T₂ weighted) imaging sequences, the desire to acquire ancillary (e.g., BOLD, diffusion weighted) imaging or MRS, and siting and system costs.⁶⁰

4.E. B₀ field homogeneity QC

Field homogeneity may change over time yet there are no relevant published studies. Systems are typically reshimmed after major service events (e.g., replacement of gradient coil or recovery from a quench). Our measurements were made months to years after the last shim of each system. However, we inspected and cleaned each bore to ensure there was no ferromagnetic debris that could degrade the field homogeneity.

Field homogeneity QC should be performed at commissioning, at least annually, and after major service or upgrades that can impact field homogeneity.^2,41,61 We recommend monthly field homogeneity QC measurements for MR-IGRT systems with rotating gantries. B₀ field maps are preferable since they provide spatial information that can be used to localize the cause of sudden changes in homogeneity (e.g., a loose passive shim tray in the gantry or a paper clip in the bore). More frequent QC may be required if screening practices are lax (street clothes and shoes allowed in the bore, missed hair pins) or construction is occurring nearby.

Field homogeneity measured with field maps typically require the involvement of a Medical or MRI Physicist to analyze. Several field map data analysis methods were described.^11,41 For systems that do not provide access to field maps or FID spectra, the bandwidth-difference method can be used to measure field homogeneity.⁴⁷

Ideally, B₀ field homogeneity is measured using a spherical phantom, field camera, or a volume that can be decomposed into spherical harmonics (e.g., Quasar MRID3D, London, Ontario). Such measurements enable comparison with the system specifications that can be used for pass/fail criteria. Some MRI vendors supply a spherical phantom with their system. Otherwise, a spherical phantom can be purchased from a phantom manufacturer (e.g., Magphan SMR100, Salem NY). Low dielectric (e.g., oil) phantoms are required for fields >1.5 T to minimize dielectric artifacts and the effects of B₁ inhomogeneities on phase maps.⁴¹ However, cylindrical or spatial integrity phantoms can be substituted.

Regardless of the shape and size of the phantom, a baseline for the B₀ field homogeneity and its variance (including noise, repeatability, and reproducibility) should be established based on the commissioning results. The tolerance should be based on variations measured over a time, ideally a year, since seasonal environmental fluctuations may affect structural steel near the MRI or MR-IGRT system. Periodic QC measurements can be compared to the baseline to determine if the system is out of specification. The baseline may need to be updated if major changes occur, for example, the magnet is reramped or reshimmed, the gradient drivers are replaced or recalibrated, or major MR-IGRT gantry components are replaced.

Daily and weekly B₀ field homogeneity checks, performed by the MRI Technologist or Simulation Therapist, were recommended for SRS.² Fatemi et al. used a tolerance and action level of 2% and 4% variations, respectively.² The checks can be performed using spatial integrity tests or measurements of QC phantom dimensions. Follow-up field homogeneity tests are warranted if the checks fail, or image quality degradation or artifacts are observed.

Active shims can compensate for nominal longitudinal changes in field homogeneity. Therefore, changes in B₀ field homogeneity can be detected by examining the active (especially linear) shim settings for a reproducible QC phantom setup and acquisition. Significant changes in the active shim values for a fixed setup are indicative of a change in either the B₀ field homogeneity or gradient (or higher-order shim) driver calibration. Depending on the MRI vendor, the active shim values are accessible either through the manual adjustments interface, pulse sequence parameter page, or listed in the DICOM header. The levels of active shimming can vary widely depending on the vendor, system, site, and phantom. Therefore, a baseline set of measurements is required.

4.F. Study limitations

B₀ field homogeneity measurements in this study were limited to MRI systems available at our institution. The systems represented a broad range of field homogeneity performance. Our aims were to demonstrate the ability to compare field homogeneity measurements with system specifications, and illuminate issues surrounding field homogeneity metrics (performance and measurement units) as they relate to RT.

Field homogeneity measurements can be vulnerable to noise, outliers, edge effects (magnetic susceptibilities). In our phantoms, regions of high susceptibility around the fill stem and equatorial seam produced localized inhomogeneities. We used a threshold mask to extract the spherical volumes from the phantom image and minimize artifacts related to susceptibility. Pk-pk measurements would tend to be underestimated by excessive threshold masking while VRMS, SD, and FWHM are more insensitive.

Field homogeneity may also be affected by the environment (e.g., mounting for the phantom and cleanliness of the bore). With the exception of the ViewRay MRIdian, the 35 cm DSV phantom could not be centered at isocenter using the patient table so it was mounted to the bottom of the bore using a cellulose and polypropylene platform. The 24 cm DSV phantom was mounted to the patient table for homogeneity measurements. By comparison, field camera measurements are made with the field camera mounted to the bore and patient table fully retracted.

The maximum DSVs calculated from the vendor specifications are vulnerable to several sources of error. First, they assume isotropy in field homogeneity. Yet, isotropy is impacted by the magnet length. Second, extrapolations beyond the largest specified DSV are prone to large errors. Such extrapolations were limited to the 1.5 T Philips Ingenia CX. In Sections 3.B. and 3.C., polynomial interpolations of the manufacturer field homogeneity specifications were used since we did not have access to the spherical harmonic coefficients for each system. The interpolations were also subject to error particularly in systems with large field inhomogeneities.

The NMR spectrum of both mineral oils had two peaks separated by 0.44 ppm, consistent with the literature (see Fig. S3).^62,63 At 3 T without second-order shimming, the two spectral peaks of the mineral oil overlapped thus requiring manual calculation of FWHM. At 1.5 T in the 35 cm DSV phantom, the peaks merged into a single broadened peak regardless of shimming so the resulting FWHM values were likely higher than if a water phantom was used.

Quantitative measurements made at 0.35 T were potentially vulnerable to noise. However, our repeatability results indicated that the SD of the measurements was comparable to 1.5 T (Fig. S2). The body coil was used for image reception in this study because: (a) The size and mounting of the 35 cm DSV phantom precluded the use of multichannel receiver coils; and (b) Vendor commissioning and service typically use the body coil for field homogeneity measurements. Multichannel coils may be used with appropriately sized phantoms as long as the effects of phase errors are addressed in phase map and SNR calculations.^64,65

Our study was limited to whole-body horizontal bore MRI systems with fields up through 3 T. Commercial 7 T MRIs are now available for clinical diagnostics. These systems typically come equipped with third-order active shims. However, they are typically limited to head and extremity imaging for clinical workflows. Thus, their fields of view are <35 cm DSV.

This study addressed only static field inhomogeneities. It does not address temporal changes in B₀ associated with eddy currents and magnet drift. Temporal fluctuations in B₀ can cause signal degradation, image artifacts, and quantitative errors.^66,67 Thermal fluctuations from high duty cycles can affect B₀ field homogeneity and stability if thermal management is inadequate.⁶⁸

5. CONCLUSIONS

The priority for MRI systems used in RT is delineating lesion from OARs. Therefore, MRI quality recommendations for field homogeneity were designed to limit spatial distortion, blurring, and major artifacts from poor fat saturation and null bands (MR-IGRT).

Field homogeneity specifications for MRI simulators use VRMS while MR-IGRT systems currently use pk-pk or FWHM. The relationship between these measurement units is variable so it is not possible to compare such systems without direct measurements. VRMS and SD measurements of ΔB tend to be more stable and sensitive to field inhomogeneities in RT applications than pk-pk and FWHM.

Based on the specifications and recommendations for field homogeneity, some MRI systems are not well suited to meet the strict demands of RT particularly for the large imaging volumes used in body MRI. Because of the effects of field inhomogeneities, the usable field of view may be smaller than the largest DSV included in the MRI specification for field homogeneity.

Finally, new AAPM recommendations for MRI in RT are needed given that the existing recommendations were designed for diagnostic MRI and are inconsistent with current MRI manufacturer specifications (e.g., AAPM Report No. 100).

Supplementary Material

Supplementary material

Data S1. Supplemental materials including manufacturer magnetic field homogeneity specifications, detailed field homogeneity measurements and statistical analysis results, measurement repeatabilities, and estimated fields of view.