Multiple spin-echo magnetic resonance imaging

Improving Accuracy and Reproducibility of Cartilage T2 Mapping in the OAI Dataset Through Extended Phase Graph Modeling

BACKGROUND

The Osteoarthritis Initiative (OAI) collected extensive imaging data, including Multi-Echo Spin-Echo (MESE) sequences for measuring knee cartilage T2 relaxation times. Mono-exponential models are used in the OAI for T2 fitting, which neglects stimulated echoes and B1 inhomogeneities. Extended Phase Graph (EPG) modeling addresses these limitations but has not been applied to the OAI dataset.

PURPOSE

To assess how different fitting methods, including EPG-based and exponential-based approaches, affect the accuracy and reproducibility of cartilage T2 in the OAI dataset.

STUDY TYPE

Retrospective.

POPULATION

From OAI dataset, 50 subjects, stratified by osteoarthritis (OA) severity using Kellgren-Lawrence grades (KLG), and 50 subjects without OA diagnosis during OAI duration were selected (each group: 25 females, mean ages ~61 years).

FIELD STRENGTH/SEQUENCE

3-T, two-dimensional (2D) MESE sequence.

ASSESSMENT

Femoral and tibial cartilages were segmented from DESS images, subdivided into seven sub-regions, and co-registered to MESE. T2 maps were obtained using three EPG-based methods (nonlinear least squares, dictionary matching, and deep learning) and three mono-exponential approaches (linear least squares, nonlinear least squares, and noise-corrected exponential). Average T2 values within sub-regions were obtained. Pair-wise agreement among fitting methods was evaluated using the stratified subjects, while reproducibility using healthy subjects. Each method's T2 accuracy and repeatability varying signal-to-noise ratio (SNR) were assessed with simulations.

STATISTICAL TESTS

Bland-Altman analysis, Lin's concordance coefficient, and coefficient of variation assessed agreement, repeatability, and reproducibility. Statistical significance was set at P-value <0.05.

RESULTS

EPG-based methods demonstrated superior T2 accuracy (mean absolute error below 0.5 msec at SNR > 100) compared to mono-exponential methods (error > 7 msec). EPG-based approaches had better reproducibility, with limits of agreement 1.5-5 msec narrower than exponential-based methods. T2 values from EPG methods were systematically 10-17 msec lower than those from mono-exponential fitting.

DATA CONCLUSION

EPG modeling improved agreement and reproducibility of cartilage T2 mapping in subjects from the OAI dataset.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 1.

Personalized synthetic MR imaging with deep learning enhancements

PURPOSE

Personalized synthetic MRI (syn-MRI) uses MR images of an individual subject acquired at a few design parameters (echo time, repetition time, flip angle) to obtain underlying parametric ( ρ , T 1 , T 2 ) $$ \left(\rho, {\mathrm{T}}_1,{\mathrm{T}}_2\right) $$ maps, from where MR images of that individual at other design parameter settings are synthesized. However, classical methods that use least-squares (LS) or maximum likelihood estimators (MLE) are unsatisfactory at higher noise levels because the underlying inverse problem is ill-posed. This article provides a pipeline to enhance the synthesis of such images in three-dimensional (3D) using a deep learning (DL) neural network architecture for spatial regularization in a personalized setting where having more than a few training images is impractical.

METHODS

Our DL enhancements employ a Deep Image Prior (DIP) with a U-net type denoising architecture that includes situations with minimal training data, such as personalized syn-MRI. We provide a general workflow for syn-MRI from three or more training images. Our workflow, called DIPsyn-MRI, uses DIP to enhance training images, then obtains parametric images using LS or MLE before synthesizing images at desired design parameter settings. DIPsyn-MRI is implemented in our publicly available Python package DeepSynMRI available at: https://github.com/StatPal/DeepSynMRI.

RESULTS

We demonstrate feasibility and improved performance of DIPsyn-MRI on 3D datasets acquired using the Brainweb interface for spin-echo and FLASH imaging sequences, at different noise levels. Our DL enhancements improve syn-MRI in the presence of different intensity nonuniformity levels of the magnetic field, for all but very low noise levels.

CONCLUSION

This article provides recipes and software to realistically facilitate DL-enhanced personalized syn-MRI.

3D MRI: Technical Considerations and Practical Integration

One of the main advantages of three-dimensional (3D) magnetic resonance imaging (MRI) is the possibility of isotropic voxels and reconstructed planar cuts through the volumetric data set in any orientation with multiplanar reformation software through real-time evaluation. For example, reformats by the radiologist during reporting allows exploitation of the full potential of isotropic 3D volumetric acquisition or through standardized retrospective reformats of thicker predefined slices of an isotropic volumetric data set by technologists. The main challenges for integrating 3D fast spin echo (FSE) and turbo spin-echo (TSE) MRI in clinical practice are a long acquisition time and some artifacts, whereas for integrating 3D gradient-recalled echo protocols, the main challenges are lower signal-to-noise ratios (SNRs) and the inability to produce intermediate, and T2-weighted contrast. The implementation of bidirectional parallel imaging acquisition and random undersampling acceleration strategies of 3D TSE pulse sequences substantially shortens the examination time with only minor SNR reductions. This article provides an overview of general technical considerations of 3D FSE and TSE sequences in musculoskeletal MRI. It also describes how these sequences achieve efficient data acquisition and reviews the main advantages and challenges for their introduction to clinical practice.

An information theory model for optimizing quantitative magnetic resonance imaging acquisitions

Acquisition parameter selection is currently performed empirically for many quantitative MRI (qMRI) acquisitions. Tuning parameters for different scan times, tissues, and resolutions requires some amount of trial and error. There is an opportunity to quantitatively optimize these acquisition parameters in order to minimize variability of quantitative maps and post-processing techniques such as synthetic image generation. The objective of this work is to introduce and evaluate a quantitative method for selecting parameters that minimize image variability. An information theory framework was developed for this purpose and applied to a 3D-quantification using an interleaved Look-Locker acquisition sequence with T2 preparation pulse (3D-QALAS) signal model for qMRI. In this framework, mutual information is used to measure the information gained by a measurement as a function of acquisition parameters, quantifying the information content of potential acquisitions and allowing informed parameter selection. The information theory framework was tested on artificial data generated from a representative mathematical phantom, measurements acquired on a qMRI multiparametric imaging standard phantom, and in vivo measurements in a human brain. The phantom measurements showed that higher mutual information calculated by the model correlated with smaller coefficient of variation in the reconstructed parametric maps, and in vivo measurements demonstrated that information-based calibration of acquisition parameters resulted in a decrease in parametric map variability consistent with model predictions.

B1 inhomogeneity correction of RARE MRI with transceive surface radiofrequency probes

PURPOSE

The use of surface radiofrequency (RF) coils is common practice to boost sensitivity in (pre)clinical MRI. The number of transceive surface RF coils is rapidly growing due to the surge in cryogenically cooled RF technology and ultrahigh-field MRI. Consequently, there is an increasing need for effective correction of the excitation field ( B 1 + ) inhomogeneity inherent in these coils. Retrospective B₁ correction permits quantitative MRI, but this usually requires a pulse sequence-specific analytical signal intensity (SI) equation. Such an equation is not available for fast spin-echo (Rapid Acquisition with Relaxation Enhancement, RARE) MRI. Here we present, test, and validate retrospective B₁ correction methods for RARE.

METHODS

We implemented the commonly used sensitivity correction and developed an empirical model-based method and a hybrid combination of both. Tests and validations were performed with a cryogenically cooled RF probe and a single-loop RF coil. Accuracy of SI quantification and T₁ contrast were evaluated after correction.

RESULTS

The three described correction methods achieved dramatic improvements in B₁ homogeneity and significantly improved SI quantification and T₁ contrast, with mean SI errors reduced from >40% to >10% following correction in all cases. Upon correction, images of phantoms and mouse heads demonstrated homogeneity comparable to that of images acquired with a volume resonator. This was quantified by SI profile, SI ratio (error < 10%), and percentage of integral uniformity (PIU > 80% in vivo and ex vivo compared to PIU > 87% with the reference RF coil).

CONCLUSION

This work demonstrates the efficacy of three B₁ correction methods tailored for transceive surface RF probes and RARE MRI. The corrected images are suitable for quantification and show comparable results between the three methods, opening the way for T₁ measurements and X-nuclei quantification using surface transceiver RF coils. This approach is applicable to other MR techniques for which no analytical SI exists.

Accelerating the estimation of 3D spatially resolved T2 distributions

Obtaining quantitative, 3D spatially-resolved T₂ distributions (T₂ maps) from magnetic resonance data is of importance in both medical and porous media applications. Due to the long acquisition time, there is considerable interest in accelerating the experiments by applying undersampling schemes during the acquisition and developing reconstruction techniques for obtaining the 3D T₂ maps from the undersampled data. A multi-echo spin echo pulse sequence is used in this work to acquire the undersampled data according to two different sampling patterns: a conventional coherent sampling pattern where the same set of lines in k-space is sampled for all equally-spaced echoes in the echo train, and a proposed incoherent sampling pattern where an independent set of k-space lines is sampled for each echo. The conventional reconstruction technique of total variation regularization is compared to the more recent techniques of nuclear norm regularization and Nuclear Total Generalized Variation (NTGV) regularization. It is shown that best reconstructions are obtained when the data acquired using an incoherent sampling scheme are processed using NTGV regularization. Using an incoherent sampling pattern and NTGV regularization as the reconstruction technique, quantitative results are obtained at sampling percentages as low as 3.1% of k-space, corresponding to a 32-fold decrease in the acquisition time, compared to a fully sampled dataset.

Quantitative in vivo T2 mapping using fast spin echo techniques - A linear correction procedure

A method is presented for correcting the effects of stimulated and indirect echoes on quantitative T2 mapping data acquired with multiple spin echo techniques, such as turbo spin echo. In contrast to similar correction techniques proposed in the literature, the method does not require a priori knowledge of the radio frequency (RF) pulse profiles. In a first step, for the T2 mapping protocol under investigation, signal decay curves S(TE) are simulated for a range of different RF pulse profiles. The actual signal decay S(TE) is then measured on a phantom with known T2, so the approximate RF pulse profiles can be derived via comparison with the simulated decay curves. In a second step, with the RF pulses obtained from step one, signal decay curves S(TE) are simulated for different T2 values and fitted mono-exponentially, thus allowing to deduce the relationship between true T2 and the apparent T2 (T2app) values. Results show that this relationship is approximately linear, allowing for a direct correction of T2app maps. If the amplitude of the transmitted RF field (B1) does not exceed the nominal value by more than 10%, it is shown that a B1-independent correction of T2app maps yields sufficiently accurate results for T2. A B1-dependent version is also presented. The method is tested in vitro on a phantom with different T2 values and in vivo on healthy subjects.

Free-breathing inner-volume black-blood imaging of the human heart using two-dimensionally selective local excitation at 3 T

Black-blood fast spin-echo imaging is a powerful technique for the evaluation of cardiac anatomy. To avoid fold-over artifacts, using a sufficiently large field of view in phase-encoding direction is mandatory. The related oversampling affects scanning time and respiratory chest motion artifacts are commonly observed. The excitation of a volume that exclusively includes the heart without its surrounding structures may help to improve scan efficiency and minimize motion artifacts. Therefore, and by building on previously reported inner-volume approach, the combination of a black-blood fast spin-echo sequence with a two-dimensionally selective radiofrequency pulse is proposed for selective "local excitation" small field of view imaging of the heart. This local excitation technique has been developed, implemented, and tested in phantoms and in vivo. With this method, small field of view imaging of a user-specified region in the human thorax is feasible, scanning becomes more time efficient, motion artifacts can be minimized, and additional flexibility in the choice of imaging parameters can be exploited.

Implementation of multi-echo-based correlated spectroscopic imaging and pilot findings in human brain and calf muscle

PURPOSE

To implement a spatially encoded correlated spectroscopic imaging (COSI) sequence on 3 Tesla (T) MRI/MR spectroscopy scanners incorporating four echoes to collect four phase-encoded acquisitions per repetition time (TR), and to evaluate the performance and reliability of this four-dimensional (4D) multi-echo COSI (ME-COSI) sequence in brain and calf muscle.

MATERIALS AND METHODS

Typical scan parameters for the 4D datasets were as follows: repetition time = 1500 ms, 2000 Hz bandwidth, 8 × 8 spatial encoding, one average, 64 Δt(1) increments and the scan duration was 25 min. The performance and test-retest reliability of ME-COSI were evaluated with phantoms and in the occipitoparietal brain tissues and calf of six healthy volunteers (mean age = 32 years old).

RESULTS

Regional differences in concentrations of lipids, creatine (Cr), choline (Ch), and carnosine (Car) were observed between spectra from voxels located in tibial marrow, tibialis anterior, and soleus muscle. Diagonal and cross-peak resonances were identified from several brain metabolites including N-acetyl aspartate (NAA), Ch, Cr, lactate (Lac), aspartate (Asp), glutathione (GSH), and glutamine\glutamate (Glx). Coefficients of variation (CV) in metabolite ratios across repeated measurements were <15% for diagonal and <25% for cross-peaks observed in vivo.

CONCLUSION

The ME-COSI sequence reliably acquired spatially resolved 2D Correlated Spectroscopy (COSY) spectra demonstrating the feasibility of differentiating spatial variation of metabolites in different tissues. Multi-echo acquisition shortens scan duration to clinically feasible times.

Two-gradient-echo, two-dimensional, Fourier transform multisection imaging: comparison with spin-echo imaging

A multisection, two-dimensional, Fourier transform, double-gradient-echo magnetic resonance imaging sequence with partial flip-angle excitation and section doubling by radiofrequency encoding approximately doubles the signal-to-noise ratio obtainable from a conventional spin-echo sequence at low field strength, while maintaining essentially equivalent contrast.

Comparison of multi echo and contrast-enhanced MR scans: image contrast and delineation of intracranial tumors

In 56 patients with intracranial tumor T2-weighted multi echo (ME) [TR: 1600 ms, TE: 30 to 240 ms (8 echoes)] and T1-weighted Gd-DTPA-enhanced images (TR: 400 ms, TE: 30 ms or 35 ms) were compared with respect to tumor contrast and tumor delineation. In subgroups of 20 and 14 patients, respectively, tumor/brain- and tumor/edema-contrast were assessed on both a subjective and a quantitative basis (contrast-to-noise ratio, CNR). CNR's for tumor/brain-contrast were significantly higher on Gd-DTPA enhanced images than on ME scans. No significant differences between techniques were found for CNR's of tumor/edema-contrast. ME images yielded pathological findings in 54 of the 56 patients but visually assessed tumor delineation was achieved in only 36 cases. Fifteen out of 25 intraaxial tumors and 5 out of 31 extraaxial tumors could not be sufficiently delineated on ME images. In contrast, Gd-DTPA-enhanced images provided differentiation of tumor and surrounding tissues in 55 of 56 studies. In conclusion, ME images have proved highly sensitive in detecting intracranial lesions. However, brain tumors, especially intraaxial tumors could often not be differentiated from adjacent tissues on ME images. After Gd-DTPA, T1-weighted images provided more reliable tumor detection and delineation.)

Optimum acquisition times of two spin echoes for MR image synthesis

Spin-echo images can be synthesized at arbitrary values of echo time TEs if two images are acquired at the same repetition time and two different echo times TE1 and TE2. Depending on the value of TEs, the noise in the synthetic images can either be greater or less than the acquisition noise. This note shows that if the time between the acquired echoes TE2 - TE1 is equal to T2, the noise level in the synthetic images is no larger than the acquisition noise for TEs greater than or equal to TE1. This is the lowest possible noise bound for two-echo acquisition. Also, the noise bound for images synthesized with O less than or equal to TEs less than or equal to TE1 is minimized by making TE1 as short as possible.

Magnetic resonance imaging of the neoplasms involving the head

Magnetic resonance (MR) imaging is a new method for producing high quality tomographic images with improved soft-tissue definition. MR imaging is conceptually quite different from conventional imaging modalities. A number of variations in MR technique may be chosen to provide better resolution or improved specificity. Some of the current and future roles of MR in the evaluation of both cranial and extracranial neoplasms involving the head are discussed.