Magnetization transfer contrast in fat-suppressed steady-state three-dimensional MR images

qMTNet: Accelerated quantitative magnetization transfer imaging with artificial neural networks

PURPOSE

To develop a set of artificial neural networks, collectively termed qMTNet, to accelerate data acquisition and fitting for quantitative magnetization transfer (qMT) imaging.

METHODS

Conventional and interslice qMT data were acquired with two flip angles at six offset frequencies from seven subjects for developing the networks and from four young and four older subjects for testing the generalizability. Two subnetworks, qMTNet-acq and qMTNet-fit, were developed and trained to accelerate data acquisition and fitting, respectively. qMTNet-2 is the sequential application of qMTNet-acq and qMTNet-fit to produce qMT parameters (exchange rate, pool fraction) from undersampled qMT data (two offset frequencies rather than six). qMTNet-1 is one single integrated network having the same functionality as qMTNet-2. qMTNet-fit was compared with a Gaussian kernel-based fitting. qMT parameters generated by the networks were compared with those from ground truth fitted with a dictionary-driven approach.

RESULTS

The proposed networks achieved high peak signal-to-noise ratio (>30) and structural similarity index (>97) in reference to the ground truth. qMTNet-fit produced qMT parameters in concordance with the ground truth with better performance than the Gaussian kernel-based fitting. qMTNet-2 and qMTNet-1 could accelerate data acquisition at threefold and accelerate fitting at 5800- and 4218-fold, respectively. qMTNet-1 showed slightly better performance than qMTNet-2, whereas qMTNet-2 was more flexible for applications.

CONCLUSION

The proposed single (qMTNet-1) and two joint neural networks (qMTNet-2) can accelerate qMT workflow for both data acquisition and fitting significantly. qMTNet has the potential to accelerate qMT imaging for clinical applications, which warrants further investigation.

Simultaneous fat saturation and magnetization transfer contrast imaging with steady-state incoherent sequences

PURPOSE

This work combines an n-dimensional fat sat(uration) radiofrequency (RF) pulse with steady-state incoherent (SSI) pulse sequences, e.g., spoiled gradient-echo sequence, to simultaneously produce B0 insensitive fat suppression and magnetization transfer (MT) contrast. This pulse is then referred to as "fat sat and MT contrast pulse."

THEORY

We discuss the features of the fat sat and MT contrast pulse and the MT sensitivities of the SSI sequences when combining with fat sat. Moreover, we also introduce an adapted RF spoiling scheme for SSI sequences with fat sat.

METHODS

Simulations and phantom experiments were conducted to demonstrate the adapted RF spoiling. Fat suppression and MT effects are shown in 3T phantom experiments and in vivo experiments, including brain imaging, cartilage imaging, and angiography.

RESULTS

To ensure that the sequence reaches steady state, the adapted RF spoiling is required for fat sat SSI sequences. Fat sat and MT contrast pulse works robustly with field inhomogeneity and also produces MT contrasts.

CONCLUSION

SSI sequences with fat sat and MT contrast pulse and adapted RF spoiling can robustly produce fat suppressed and MT contrast images in the presence of field inhomogeneity.

A case of fat necrosis after breast quadrantectomy in which preoperative diagnosis was enabled by MRI with fat suppression technique

A 63-year-old woman was found to have a left breast mass after quadrantectomy and radiation for bilateral breast cancer on postoperative cyclic examination. Intramammary recurrence could not be excluded by physical examination, mammography, or ultrasound examination. MR imaging with fat suppression technique revealed an oil-containing lesion, indicating fat necrosis. It was confirmed histologically that the mass-forming lesion included no cancer tissue. MR imaging with fat suppression technique appears to be a promising method for identification of postoperative mass lesions of the breast.

Magnetization transfer imaging of bone marrow with and without fat suppression

RATIONALE AND OBJECTIVES

To assess magnetization transfer (MT) ratios of bone marrow at magnetic resonance (MR) imaging performed with and without fat suppression.

MATERIALS AND METHODS

The authors performed MR imaging in 30 regions of normal bone marrow in 10 subjects by using four types of gradient-echo sequences: a combination of MT and fat-suppression techniques, only the fat-suppression technique, only the MT technique, and without MT or fat-suppression techniques. MT ratios of marrow obtained with and without the fat-suppression technique were quantitatively compared.

RESULTS

The average MT ratio of marrow with fat suppression was significantly higher than that without fat suppression (P < .01).

CONCLUSION

Because bone marrow includes both water and fat, the MT ratio of marrow was underestimated when MT imaging was performed without fat suppression. A fat-suppression technique should be used.

Understanding pulsed magnetization transfer

Using a two-pool exchange model of magnetization transfer (MT), numeric simulations were developed to predict the time dependence of longitudinal magnetization in both semisolid and liquid pools for arbitrary pulsed radiofrequency (RF) irradiation. Whereas RF excitation of the liquid pool was modeled using the time-dependent Bloch equations, RF saturation of the semisolid pool was described by a time-dependent rate proportional to both the absorption lineshape of the semisolid pool and the square of the RF pulse amplitude. Simulations show good agreement with experimental results for a 4% agar gel aqueous system in which the two-pool kinetics have been well studied previously. These simulations provide a method for interpreting pulsed MT effects, are easily extended to biologic tissues, and provide a basis for optimizing clinical imaging applications that exploit MT contrast.

Improved visualization of breast lesions with gadolinium-enhanced magnetization transfer MR imaging

A pulse sequence with magnetization transfer as the main contrast mechanism (MT-FLASH) was developed for improved imaging of breast lesions that requires neither fat suppression nor postprocessing. After optimization of the sequence in phantom and volunteer studies, a clinical pilot study with 14 patients was performed. In carcinomas the relative signal increase after Gd-DTPA administration was on average 34% in MT-FLASH images compared with 169% in conventional T1-weighted (T1W) three-dimensional FLASH images. In MT-FLASH images, all lesions demonstrated a signal intensity higher than that of fat; in T1W images, all lesions have a lower signal intensity. The average postcontrast carcinoma-to-fat contrast-to-noise ratios were +11.6 and -14.2, respectively. The conspicuity of 12 of 13 carcinomas was improved in postcontrast MT-FLASH imaging enables excellent visualization of Gd-DTPA-enhancing breast lesions.

Pulsed magnetization transfer contrast for MR imaging with application to breast

The relative populations and transverse relaxation times of the solid-like hydrogen pool (PB and T2B) and the magnetization transfer (MT) rates between the solid-like and liquid-like hydrogen pools (kappa) have been determined for three different agar gel concentrations (2%, 4%, and 8% by weight) as well as excised fibroglandular breast tissue specimens. PB was determined to be .003(.001), .01(.002), .02(.01), and .06(.01); T2B was determined to be 13.0(.2), 14.0(.1), 14.5(.1) and 15.2(1.3) microseconds; and kappa was determined to be 0.78(.01), 1.15(.02), 2.00(.02), and 3.55(1.5) sec-1 for the 2%, 4%, and 8% agar gels and the fibroglandular tissue, respectively. The image signal intensities of a pulsed MTC-prepared gradient-echo imaging technique are predicted using these MT parameters and are shown to agree well with experimental data obtained from a clinical MR imaging system. This technique is shown to suppress signal intensity of fibroglandular breast tissue by 40%-50% without exceeding SAR limits (< or = 8 W/kg) and is helpful for visualizing lesions and silicone implants.

An interleaved sequence for accurate and reproducible clinical measurement of magnetization transfer ratio

We demonstrate an interleaved dual spin echo-based sequence for quantitative measurement of Magnetisation Transfer Ratio (MTR) in a clinical environment that overcomes the problems of patient motion between scans faced by noninterleaved methods. The sequence also provides proton density and T2-weighted images, allowing direct comparison among the three contrast regimes. Phantom studies and in vivo measurements on normal controls show the sequence to be robust in normal use. The values of MTR calculated from the sequence are shown to be precise and reproducible enough to allow regional variations to be identified within and between white matter and other brain tissues.

Pulsed magnetization transfer contrast MRI by a sequence with water selective excitation

OBJECTIVE

A water selective SE imaging sequence was developed providing suitable properties for the assessment of magnetization transfer (MT) effects in tissues with considerable amounts of fat.

MATERIALS AND METHODS

The sequence with water selective excitation and slice selective refocusing combines the following features: The RF exposure on the macromolecular protons is relatively low for single slice imaging without MT prepulses, since no additional pulses for fat saturation are necessary. Water selection by frequency selective excitation diminishes faults in the subtraction of images recorded with and without MT prepulses (which might arise from movements).

RESULTS

High differences in the signal amplitudes from hyaline cartilage and muscle tissue were obtained comparing images recorded with irradiation of the series of prepulses for MT and those lacking MT prepulses. Utilizations of the described water selective approach for the assessment of MT effects in lesions of cartilage and bone are demonstrated. MT saturation was also examined in muscles with fatty degeneration of patients suffering from progressive muscular dystrophy.

CONCLUSION

The described technique allows determination of MT effects with good precision in a single slice, especially in regions with dominating fat signals.

Magnetization transfer in hemopoietic bone marrow examined by localized proton spectroscopy

The sensitivity of hemopoietic bone marrow to magnetization transfer is analyzed in 15 healthy volunteers and seven patients with different hematological disorders (inflammation, plasmacytoma, hemopoietic reconstitution after bone marrow transplantation). To obtain sufficient signal-to-noise ratio, a 90 degrees - 180 degrees - 180 degrees double spin echo (PRESS) single voxel spectroscopic method was combined with pulsed magnetization transfer. Several spectra were recorded from each volume element inside the vertebral marrow, alternately with and without prepulses for magnetization transfer. Water signals from marrow with increased content of extracellular water due to inflammation or edema revealed less magnetization transfer effects than marrow with increased intracellular water content due to high cellularity. The preliminary results show magnetization transfer to be a promising tool for the clinically important characterization of the water composition in red bone marrow.

Magnetization transfer by simple sequences of rectangular pulses

On-resonant radio frequency (RF) sequences composed of a train of short rectangular pulses of the same kind were optimized in order to obtain selective saturation of protons with short transverse relaxation times for magnetization transfer purposes. It is demonstrated that the sequences regarded allow a good adaptation to different requirements for magnetization transfer examinations on whole-body imagers. The sequences presented here provide relatively strong saturation of protons with very short transverse relaxation times T2 approximately less than 50 microseconds, whereas signals from protons with long T2 to be recorded are hardly influenced in a broad frequency range. The sequences are especially advantageous for applications in pulse files with limited numbers of support points.

Analysis of on- and off-resonance magnetization transfer techniques

Three methods of performing magnetization transfer (MT) MR imaging are analyzed: (a) off-resonance continuous wave, (b) off-resonance shaped pulses, and (c) on-resonance binomial pulses. With two-pool Bloch-model simulations, signal levels from "MT active" spin systems were calculated, with reference to direct saturation of "MT inactive" systems, allowing calculation of contrast due to MT. Simulations demonstrate several trends with variation of excitation amplitude and offset frequency for the off-resonance methods and with variation of excitation amplitude and pulse shape "order" for binomial pulses. The simulations show that nominally optimized versions of each of these approaches provide essentially equivalent contrast at a given level of applied MT power, contrary to previous claims. Experiments with an MT-inactive phantom, with a whole-body system, show results with off-resonance pulses to be in good agreement with simulations, whereas binomial-pulse experiments show anomalously large direct saturation.

Off-resonance spin locking for MR imaging

Off-resonance spin locking is investigated as a low power method for achieving low field spin-lattice relaxation contrast using high field clinical MR imaging systems (e.g., 1.5 tesla). Spin-lattice relaxation times and equilibrium magnetizations in the off-resonance rotating frame (T1 rho(off), beta) were measured for tissue-mimicking phantom materials as a function of the ratio of the amplitude to the resonance offset of the spin-locking pulse (f1/delta). The phantom materials consisted of vegetable oil to simulate fat and two different gels containing 2% and 4% agar to simulate nonfatty tissues with different macromolecular compositions. These measurements were used to verify a signal strength equation for a multislice off-resonance spin-locking technique implemented on a clinical MR imaging system operating at 1.5 tesla. Although the oil showed little change in image contrast with increasing f1/delta, the two gels demonstrated a strong variation which provided improved discrimination compared to T1-weighted imaging. Off-resonance spin locking is suggested as a method for improving delineation of breast lesions and a preliminary clinical example from a patient volunteer is presented.

Estimation of Bloch model MT spin system parameters from Z-spectral data

Previous studies have described magnetization transfer (MT) Z-spectra in terms of a two-pool Bloch model, with six spin-system parameters KA, F, T1A, T1B, T2A, and T2B. By simulation, we show that a process including nonlinear constrained optimization can be used to accurately and uniquely estimate spin-system parameters from MT Z-spectra prepared by continuous wave (CW) RF irradiation. Experiments producing Z-spectra by pulsed RF irradiation give substantially different magnetization values, relative to MT acquisitions obtained by CW RF irradiation, at small offset frequencies, with a consequence that only T2B can be uniquely determined. However, several equalities and bounds involving four of the other parameters (KA, F, T1A, and T1B) are derived, which are applicable to pulsed data. These relationships allow calculation of "free pool" magnetization corresponding to complete saturation of the restricted pool, without requiring that this complete saturation be experimentally achieved. MT experimental data from pulsed RF irradiation on boiled egg albumin, obtained using a clinical whole-body MRI system, are analyzed using an optimization algorithm and the derived expressions.