Human Prostate MRI at Ultra-high-performance Gradient: A Feasibility Study

Abstract

Purpose:

To demonstrate the technical feasibility and the value of ultra-high-performance gradient in imaging the prostate in a 3T MRI system.

Methods:

In this local institutional review board-approved study, prostate MRI was performed on four healthy men. Each subject was scanned in a prototype 3T MRI system with a 42-cm inner diameter gradient coil that achieves a maximum gradient amplitude of 200 mT/m and slew rate of 500 T/m/s. PI-RADS V2.1 compliant axial T₂-weighted anatomical imaging and single-shot echo planar diffusion weighted imaging (DWI) at standard gradient of 70 mT/m and 150 T/m/s were obtained, followed by DWI at maximum performance, i.e., 200 mT/m and 500 T/m/s. In comparison to state-of-the-art clinical whole-body MRI systems, the high slew rate improved echo spacing from 1020 μs to 596 μs and, together with a high gradient amplitude for diffusion encoding, the echo time was reduced from 55 ms to 36 ms.

Results:

In all four subjects (waist circumference = 81–91 cm, age = 45–65 years), no peripheral nerve stimulation sensation was reported during DWI. Reduced image distortion in the posterior peripheral zone prostate gland and higher signal intensity (e.g., in the surrounding muscle of high-gradient DWI) were noted.

Conclusion:

Human prostate MRI at simultaneously high gradient amplitude of 200 mT/m and slew rate of 500 T/m/s is feasible, demonstrating that improved gradient performance can address image distortion and T₂-decay induced signal-to-noise ratio issues for in vivo prostate imaging.

INTRODUCTION

Prostate cancer is the second most common cancer in men¹. Multiparametric MRI², including diffusion MRI, plays an essential role in the identification of clinically significant prostate cancer (Gleason score ≥ 3+4). However, standard clinical diffusion MRI with single-shot echo planar imaging readout cannot achieve necessarily high phase-encoding bandwidth and short echo time (TE). The low phase-encoding bandwidth results in image distortion in regions with an inhomogeneous magnetic field due to proximity to rectum, colon, and associated gas—specifically, the posterior peripheral zone of the prostate gland where majority of cancer lesions are located. The prolonged TE results in significant T₂ decay-induced poor signal characteristics and low signal-to-noise ratio (SNR), especially for prostate cancer lesions which have relatively short T2 relaxation rates. Image quality and clinical performance of diffusion MRI are directly influenced by the gradient performance of clinical whole-body MRI systems, where the maximum gradient amplitude is 33-80 mT/m and slew rate is 150–200 T/m/s. Ultra-high-performance gradients with simultaneously high gradient amplitude and slew rate could be very advantageous for prostate MRI.

Ultra-high-performance gradient systems for human MRI at 3T and 7T^3,4 have been recently introduced for advanced brain imaging, which can simultaneously achieve a maximum gradient amplitude of 200 mT/m and slew rate of 500–900 T/m/s. These ultra-high-performance MRI systems are equipped with gradient coils that have a small inner diameter, measuring 42–44 cm, compared to the 65–85 cm available in whole-body MRI systems. The inner diameter of the bore is further reduced to 37–39 cm after mounting dedicated birdcage radiofrequency (RF) transmit/receive coils. The patient bore diameter steps up to 54 cm at the shoulders and 74.4 cm at the arms and torso, at brain landmark position³, where most people can fit in the bore for brain imaging. Such a high-performance gradient MRI system would significantly advance the characterization of human brain microstructure and understanding the brain functional connectivity.

According to the demographic distribution of men in the United States, approximately 15% of men over 40 years have a waist circumference less than 92 cm⁵ and thus could fit in a bore with an inner diameter of 37 cm. Therefore, such a system, even with bore-size limitation, provides a unique opportunity to investigate ultra-high-performance MRI for prostate imaging. In this manuscript, we report on ultra-high-performance gradient MRI of the prostate gland in human subjects using small diameter gradient coils. Our work points to the potential value in having an ultra-high-performance gradient for prostate imaging. However, we do not propose, given the current coil constraints, that this approach be considered for clinical adoption at this time.

The purpose of this study was to evaluate the technical feasibility of prostate MRI in an ultra-high-performance human 3T MRI system that can simultaneously achieve high gradient amplitude and slew rate.

METHODS

Subjects

All human MRI scans were in accordance with local institutional review board-approved protocols. All subjects provided written informed consent. Healthy men with waist circumference below 92 cm and age older than 40 years were recruited.

Ultra-high-gradient MRI setup and protocol

The setup of the MRI system for prostate imaging is shown in Figure 1. MRI experiments were performed on a specialized 3T MRI system (SIGNA MR750, GE Healthcare, WI, USA). The system is equipped with the 42-cm diameter MAGNUS (Microstructure Anatomy Gradient for Neuroimaging with Ultrafast Scanning)³ gradient coil, operating at a maximum gradient amplitude of 200 mT/m and a maximum slew rate of 500 T/m/s with 1 MVA gradient driver for each gradient axis, and a dedicated birdcage RF transmit/receive coil. The inner diameter of this system is 37 cm, so it is therefore compatible with waist circumferences below approximately 92 cm. In this study, signals were received by two 16-channel phased-array RF coils for 3T MRI (GEM Flex, NeoCoil, LLC, WI, USA) placed anterior and posterior to the pelvis. Each subject was carefully positioned such that the prostate was located close to the isocenter of the gradient system.

Figure 1.

(A). Setup of prostate imaging in the 3T MRI system equipped with the MAGNUS gradient coil with a 42-cm inner diameter, capable of operating at maximum gradients of 200 mT/m and 500 T/m/s. The dedicated birdcage RF transmit/receive coil reduces the bore inner diameter to 37 cm. (B). The human body is less exposed to the dedicated birdcage RF transmit/receive coil, compared to the coil in the clinical whole-body MRI systems. (C). Two 16-channel phased-array RF receive coils for signal receiving at 3T MRI were placed anterior and posterior to the pelvis.

Protection of Human Safety

The short length of the dedicated birdcage RF transmit/receive coil for brain imaging (Figure 1) results in a smaller body mass exposed to the RF field, which has been demonstrated to reduce overall specific absorption rate (SAR)⁶. In this study, we preliminarily evaluated SAR safety by performing electromagnetic simulations (Sim4Life, Zurich MedTech, Zurich, Switzerland). More specifically, simulations were performed using a validated 16-rung high-pass RF coil model with an inner diameter of 37 cm and length of 40 cm in a flanged RF shield⁶. A human body model (named Glenn, IT’IS Foundation, Zurich, Switzerland) was used to predict electromagnetic field distributions in the prostate exposed mass region. The finite difference time domain⁷ voxelized mesh size for the coil components and human body models were set to a maximum volume of 2×2×2 mm³. An absorbing boundary condition was used with the simulation time set to automatically terminate once a steady-state convergence factor of −30 dB was reached. A fixed prostate landmark position was used for the human body model in the RF coil. Electromagnetic transmit magnetic field (B₁⁺) and SAR distributions from the simulations were evaluated for the prostate exposed mass region. Safety assessment was made by comparing to the upper limits following the IEC 60601-2-33 3^rd edition⁸.

Rapid gradient switching can cause unwanted action potential to fire in nerves which can lead to uncomfortable nerve stimulation⁹. Therefore, the gradient slew rate is constrained by the peripheral nerve stimulation (PNS) threshold¹⁰ for human safety. The MAGNUS gradient coil has been validated to allow a five times higher peripheral nerve stimulation (PNS) threshold for brain imaging^3,11,12, when compared to whole-body gradient coils. The use of higher slew rates for human prostate imaging in the MAGNUS MRI system without PNS sensation was tested in this study.

Imaging of Healthy Subjects

In each imaging session, non-conductive paddings were put in between two thighs to avoid skin touch-introduced circuit in the lower body and the consequent RF heating/burning. The male pelvic region was tightly sandwiched by the two RF receive coils and the body (exterior thigh)-to-bore contact was avoided by the non-conductive padding built for the RF receive coils. After set up in the bore, all subjects were provided with a squeeze-bulb alarm and were instructed to use it when experiencing any PNS during the scan. Each subject was first imaged with clinical PI-RADS V2.1 compliant axial T₂-weighted anatomical imaging and axial diffusion weighted imaging (DWI), constrained by the clinical whole-body MRI system limit with maximum gradients of 70 mT/m and 150 T/m/s and the PNS threshold of whole-body gradient coils^3,11,12. Thus, the first DWI was acquired with a TE of 55 ms, two b-values of 50 s/mm² and 1000 s/mm², echo spacing of 1020 μs, and an in-plane resolution of 1.5×1.5 mm². A second set of DWI fully utilized the MAGNUS system limit with maximum gradients of 200 mT/m and 500 T/m/s, and a five times higher PNS threshold^3,11,12. This enabled a TE of 36 ms and an echo spacing of 596 μs. Reduced field-of-view¹³, which is used as the standard technique in clinical prostate DWI for distortion mitigation, was applied for all DWI scans. Detailed imaging parameters are listed in Table 1. The echo planar readout gradients in the two DWI acquisitions were designed with the equal constraint of the readout gradient plateau to total sampling time, yielding the echo spacings in Table 1.

Table 1.

Prostate T2-weighted anatomical imaging and diffusion weighted imaging parameters.

	Diffusion weighted imaging		T2-weighted anatomical imaging
	200 mT/m 500 T/m/s	70 mT/m 150 T/m/s	Two-dimensional fast spin echo
Echo spacing ŝ)	596	1020	-
Phase-encoding bandwidth (Hz)	28.0	16.3	-
Total echo train length (ms)	27.4	46.9	-
Partial Fourier factor	0.77	0.77	-
Echo time (ms)	36	55	122
Scan time (mins)	03:52	03:52	02:40
Sampling bandwidth (kHz)	±250		±62.5
Repetition time (ms)	4000		2500–5000
Field of view (mm)	180 × 90		200 × 200
Image matrix	120 × 60		380 × 240
Slice thickness (mm)	3		3 or 4
Voxel size (mm3)	1.5 × 1.5 × 3.0		0.5 × 0.8 × 3 (or 4)
Number of slices	16		16
2 b-value (sec/mm2)	50 (3 repeated) 1000 (16 repeated)		-
Number of encoding directions	3		-
Averages	-		3

Data Analysis

For each subject and each diffusion acquisition, apparent diffusion coefficient (ADC) maps were calculated using signals at the two b-values of 50 s/mm² and 1000 s/mm², following the standard processing for clinical diffusion-weighted images. Notably, the effects of gradient nonlinearity were accounted for in the estimation of the ADC¹⁴. Denoising was not performed.

SNR of diffusion-weighted images at the lower acquired b-value (i.e., b=50 s/mm²) was compared between MAGNUS DWI and whole-body DWI. The standard deviation of signals in the nine b=50 s/mm² images was calculated for each voxel. SNR of each image voxel was obtained by dividing the averaged signals of the nine b=50 s/mm² (3 diffusion encoding directions X 3 repeated scans per direction) to the standard deviation. Three regions-of-interest (ROIs) were delineated in each DWI dataset for each subject, including the peripheral zone of prostate, the transition zone of prostate, and adjacent bilateral muscles. Mann-Whitney U-test (not assuming normal distribution) was then performed to compare the SNR in corresponding ROIs between the MAGNUS DWI and whole-body DWI. A p-value below 0.05 was deemed statistically significant.

RESULTS

All SAR values in extracted regions were scaled to a B₁⁺ of 2 μT, which was spatially averaged over a 2 mm thick axial slice where the prostate could be observed. A value of 2 μT was chosen to represent the typical flip angles encountered in human prostate scanning on the MAGNUS MRI system. Figure 2 shows the simulated map of 10 gram local SAR normalized to 2 μT B₁⁺ field in a coronal and axial orientation. At the prostate landmark, the exposed mass of the human body model was 31.1 kg (total body mass = 61.1 kg), with a B₁⁺-normalized average and peak local SAR values of 1.17 W/kg and 6.29 W/kg, respectively. The simulated average SAR was compared with respect to the IEC 6-minute time-averaged SAR limit⁸. The example slice in Figure 2 contains the location for the peak local SAR in the glute muscle.

Figure 2.

The coronal and axial views of the simulated local SAR (10 gram), normalized to 2 μΤ B₁⁺ field, at the prostate landmark. The peak SAR was 6.29 W/kg at the gluteal muscle, as indicated in the axial view (red line) of the specific absorption rate map.

Reported SAR numbers for prostate imaging on the MAGNUS system at 2 μT were compared to partial body SAR value limits in a volume transmit coil in IEC 60601-2-33 3^rd edition, Table 201.105. In the case of normal operating mode, our average SAR of 1.17 W/kg (at 2 μT B₁⁺) was well below the IEC 6-minute average SAR limit of 5.93 W/kg (= 10 W/kg minus 8 W/kg times exposed mass (kg) / total mass (kg) = 10-8x31.1/61.1). Simulation results presented here indicate that patient scanning on MAGNUS at a prostate landmark in the body did not exceed the IEC 60601-2-33 3rd edition SAR safety limits.

Four healthy subjects were recruited. Demographic characteristics of the subjects were: waist circumference of 81–91 cm, hip circumference of 90–104 cm, weight of 66–84 kg, and age of 45–65 years.

The top and bottom slices containing the prostate gland of each subject along the superior-inferior direction were at −86.1 mm and −58.1 mm, 1.5 mm and 28.5 mm, −29.7 mm and −2.7 mm, and −29.5 mm and 9.5 mm, respectively. None of the four subjects reported any PNS sensation during DWI acquisitions.

Geometric distortion in diffusion-weighted images acquired with ultra-high-performance gradient, especially in regions near the prostate-rectum boundaries, was reduced in three subjects (Figure 3). One subject had no visible distortion in any diffusion-weighted image. Higher signal intensity was observed at a shorter TE, as indicated by the preservation of signal in the muscle on either side of the prostate (Figure 4). The image contrast between the peripheral zone and the transition zone was lower as a result of shorter TE. ADC was robustly estimated in the prostate, bladder, and muscles from the MAGNUS DWI acquisitions with the shorter TE. In comparison, more numbers of voxels with physically implausible ADC estimations, which were masked out as dark dots in the bladder and muscle, were observed in the DWI acquisition with the longer TE.

Figure 3.

A shorter echo spacing yields reduced image distortion of echo planar imaging (e.g., b = 50 s/mm²), especially in prostate-rectum boundaries pointed to by the yellow arrows. The red contour of the prostate gland was delineated in the axial T₂ anatomical image. TR: repetition time.

Figure 4.

Higher signal intensity in DWI and more robust ADC estimation at 200 mT/m and 500 T/m/s were achieved due to the shortening of TE, compared to that at 70 mT/m and 150 T/m/s in clinical whole-body maximum gradient performance. The two DWI acquisitions were from the same subject in the same imaging session. The different maximum gradient performances were implemented via software configuration. TR: repetition time. NEX: number of repeated scans.

The SNR map of diffusion-weighted images at b=50 s/mm² in a representative slice, the histogram of SNR in different ROIs, and the comparison of SNR between MAGNUS DWI at b=50 s/mm² and whole-body DWI at b=50 s/mm² are presented in Supplementary Materials. In muscles of all the subjects, SNR of MAGNUS DWI was significant higher compared to SNR of whole-body DWI (Supporting Table 1 and Supporting Figures 1 and 2). In the transition zone of the prostate, SNR in MAGNUS DWI was statistically tested no different in three subjects and lower in one subject, compared to SNR in whole-body DWI. In the peripheral zone of the prostate, SNR in MAGNUS DWI was statistically tested no different in two subjects, higher in one subject, and lower in one subject.

DISCUSSION

Our study demonstrates, for the first time, the feasibility of performing prostate MRI at simultaneously high gradient amplitude of 200 mT/m and slew rate of 500 T/m/s in the MAGNUS MRI system. Initial electromagnetic simulations indicated that human scanning at a prostate landmark in MAGNUS, where a dedicated birdcage RF transmit/receive coil was used, did not exceed the IEC 60601-2-33 3rd edition SAR limits. In this preliminary study, four healthy subjects with a waist circumference less than 92 cm were comfortably able to fit in the scanner. None of the four subjects reported PNS sensations during echo planar imaging readout that utilized a gradient slew rate of 500 T/m/s, indicating that a high PNS threshold can be achieved for prostate imaging in the MAGNUS MRI system. With a short echo spacing, the image distortion was markedly reduced, especially in the posterior peripheral zone where the presence of rectal gas and/or stool can cause marked field inhomogeneity. By contrast, ultra-high-performance whole-body gradient MRI system, e.g., the Connectome MRI system¹⁵, could employ higher gradient amplitude (300 mT/m) but no improvement in slew rate (200 T/m/s), which cannot achieve a short echo spacing or reduce image distortion.

The shortened TE in DWI will result in 15-25% higher signal intensity, assuming a T₂ relaxation rate of 85–140 ms in healthy prostate gland and prostate cancer¹⁶. The SNR benefit due to the shortened TE was primarily appreciated in the tissues with short T2, e.g., in the muscles (~40 ms). However, due to the shorter total readout time, SNR in the MAGNUS diffusion-weighted images of the healthy prostate was not significantly different in three subjects and was ~10% lower in one subject compared to SNR in the whole-body DWI. This emphasized the tradeoff between reduced distortion due to the shorter echo spacing versus the resulting shorter echo planar readout-induced SNR penalty at fixed k-space acquisition lines, as detailed in Supporting Figure S3. The SNR benefit due to the shortened TE will be more appreciated in prostate cancer due to its shorter T₂, and DWI with high b-values due to more TE shortening (Supporting Figures S3E-G).

The ultra-high-performance MRI platform provides a unique opportunity to develop and validate advanced diffusion encoding techniques to characterize tissues at the meso- and micro-scales in humans^17-19. Emerging time-dependent diffusion MRI^20,21, multi-dimensional diffusion MRI^22-24, and hybrid diffusion-relaxation MRI^25,26 have shown great promise to non-invasively characterize the stroma, epithelium, and lumen compartments of each voxel in the prostate gland. These specific microscopic features have shown to better identify prostate cancer and assess tumor aggressiveness^21,26. The advanced diffusion MRI techniques will benefit from the ultra-high-performance gradients and will lead to translation of ex vivo and pre-clinical research findings in human prostate studies.

Furthermore, our studies will guide optimization of ultra-high-performance MRI systems with larger bore size to increase the dissemination of high-quality prostate MRI in conjunction with the use of high gradient amplitude, high slew rate, and an elevated PNS threshold. For example, a 50-cm inner diameter gradient coil constrained with high PNS threshold may operate at 220–250 mT/m and 560–640 T/m/s when equipped with highly efficient 3–4 MVA gradient drivers, as detailed in Appendix A1. Upon future successful implementation and integration of this system, ~80% of men over 40 years old in the United States will be able to fit in it.

This study has several limitations. First, the SNR of prostate MRI in MAGNUS was degraded due to the limited choice of RF receive coils, compared to that in clinical whole-body MRI. Two 16-channel phased-array RF receive coils, normally for extremity MRI, were used in the experimental settings. The coils could not provide a comparable SNR efficiency compared to the two 48-channel coils used in the clinical whole-body MRI systems. Light-weight flexible RF receive coils with more channels^27,28 may be used to further improve the SNR for prostate MRI. We also note that the SNR calculation was based on the magnitude signal and the ADC was calculated by a linear fitting. This straightforward approach may be biased by the rectified noise floor at low SNR values^29,30. We further note that the repetition time was kept the same for both the MAGNUS DWI and whole-body DWI in this study. Future studies may boost SNR in MAGNUS DWI by using more image averaging and a shorter repetition time without lengthening scan time. Furthermore, using b=50 s/mm² images to evaluate image distortion is a limitation. Evaluation of image distortions at high b-values would be more comprehensive for comparisons of DW image quality. Another limitation of this study is that the small number of subjects limit the assessment of precise PNS thresholds for prostate imaging in the MAGNUS gradient coil. Comprehensive evaluation of human PNS for prostate imaging in the MAGNUS gradient coil will be performed in the future. Also, it is important to note that these subjects represent only approximately 15% of men over 40 years old in the United States⁵—thus, these small diameter gradient coils may not be applicable for widespread adoption at this time. However, the significance of this work is the feasibility and potential value in having ultra-high-performance gradient for in vivo prostate imaging.

In conclusion, human in vivo prostate MRI at maximum gradients of 200 mT/m and 500 T/m/s and at a high PNS threshold, is feasible in the ultra-high-performance MRI system. Future studies with a large group of subjects will be performed to establish the precise PNS threshold for prostate imaging in the MAGNUS gradient coil. Patient studies will be performed to assess the clinical benefit of advanced diffusion MRI enabled by the ultra-high-performance gradient.

APPENDIX 1. TECHNIQUE FEASIBLITY OF ULTRA-HIGH-PERFORMANCE PROSTATE MRI SYSTEM

First, a high-power gradient driver is a necessity for ultra-high-performance whole-body MRI system. In recent years, high-power gradient drivers delivering 3 MVA and above have been developed^31,32. To avoid the need for upgrades in electrical power and cooling infrastructure, silicon carbide (SiC) metal oxide semiconductor field effect transistors (MOSFETs) have emerged as a cutting-edge solution that enables high-power gradients at significantly lower power loss^32-34. Therefore, highly power-efficient high-power gradient drivers delivering 3–4 MVA power^32-34 will be available for building ultra-high-performance whole-body MRI systems, providing a pathway for wider adoption across hospitals without extensive and expensive facility electrical service and cooling modifications.

Lee and Bernstein³⁵ derived and validated that the product of maximum gradient amplitude and slew rate, ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$, scales to 1) the inverse of the fifth power of the gradient coil diameter (d⁻⁵), and 2) the product of maximum current and voltage delivered to the gradient coil, i.e., the gradient driver power. At a fixed gradient driver power, ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ of a 50-cm diameter gradient coil is 0.418 (= 50⁻⁵/42⁻⁵) times ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ of the 42-cm diameter MAGNUS gradient coil. With any given gradient coil, ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ with 3–4 MVA power input is 3–4 times the ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ with 1 MVA power input. Therefore, ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ of a 50-cm diameter gradient coil operating at 3 (or 4) MVA can achieve 1.255 (= 0.418 x 3) or 1.673 (= 0.418 x 4) times ${\mathrm{G}}_{\max }\mathsf{x}{\text{SR}}_{\max }$ of the current MAGNUS system, i.e., 42-cm diameter coil with 1 MVA power input. Note that we only get the product of ${\mathrm{G}}_{\max }$ and ${\text{SR}}_{\max }$ in this calculation. The balance between ${\mathrm{G}}_{\max }$ and ${\text{SR}}_{\max }$ can be adjusted by changing the number of turns.

Assuming evenly scaling ${\mathrm{G}}_{\max }$ and ${\text{SR}}_{\max }$ from the values of current MAGNUS operating at 1 MVA (i.e., 200 mT/m and 500 T/m/s), we can achieve:

• 50-cm diameter gradient and 3 MVA gradient driver: ${\mathrm{G}}_{\max }=200\mathsf{x}\sqrt{1.255}=224\mathsf{mT}∕\mathsf{m}$, ${\text{SR}}_{\max }=500\mathsf{x}\sqrt{1.255}=560\mathsf{T}∕\mathsf{m}∕\mathsf{s}$

• 50-cm diameter gradient and 4 MVA gradient driver: ${\mathrm{G}}_{\max }=200\mathsf{x}\sqrt{1.673}=259\mathsf{mT}∕\mathsf{m}$, ${\text{SR}}_{\max }=500\mathsf{x}\sqrt{1.673}=647\mathsf{T}∕\mathsf{m}∕\mathsf{s}$

Therefore, the system or hardware performance capable of operating at 220–250 mT/m and 560–640 T/m/s can be achieved.

The PNS is determined by both the electromagnetic field and the distribution of nerves. PNS-informed design of gradient coils^36-38 has emerged to achieve high PNS threshold, compared to whole-body gradient coils. With electromagnetic and neurodynamic simulations, it is feasible to design gradient coils achieving high PNS at prostate iso-center position.