Extraction of artefactual MRS patterns from a large database using non-negative matrix factorization.

Despite the success of automated pattern recognition methods in problems of human brain tumor diagnostic classification, limited attention has been paid to the issue of automated data quality assessment in the field of MRS for neuro-oncology. Beyond some early attempts to address this issue, the current standard in practice is MRS quality control through human (expert-based) assessment. One aspect of automatic quality control is the problem of detecting artefacts in MRS data. Artefacts, whose variety has already been reviewed in some detail and some of which may even escape human quality control, have a negative influence in pattern recognition methods attempting to assist tumor characterization. The automatic detection of MRS artefacts should be beneficial for radiology as it guarantees more reliable tumor characterizations, as well as the development of more robust pattern recognition-based tumor classifiers and more trustable MRS data processing and analysis pipelines. Feature extraction methods have previously been used to help distinguishing between good and bad quality spectra to apply subsequent supervised pattern recognition techniques. In this study, we apply feature extraction differently and use a variant of a method for blind source separation, namely Convex Non-Negative Matrix Factorization, to unveil MRS signal sources in a completely unsupervised way. We hypothesize that, while most sources will correspond to the different tumor patterns, some of them will reflect signal artefacts. The experimental work reported in this paper, analyzing a combined short and long echo time 1 H-MRS database of more than 2000 spectra acquired at 1.5T and corresponding to different tumor types and other anomalous masses, provides a first proof of concept that points to the possible validity of this approach.

Arterial spin labeling MR image denoising and reconstruction using unsupervised deep learning.

Arterial spin labeling (ASL) imaging is a powerful magnetic resonance imaging technique that allows to quantitatively measure blood perfusion non-invasively, which has great potential for assessing tissue viability in various clinical settings. However, the clinical applications of ASL are currently limited by its low signal-to-noise ratio (SNR), limited spatial resolution, and long imaging time. In this work, we propose an unsupervised deep learning-based image denoising and reconstruction framework to improve the SNR and accelerate the imaging speed of high resolution ASL imaging. The unique feature of the proposed framework is that it does not require any prior training pairs but only the subject's own anatomical prior, such as T1-weighted images, as network input. The neural network was trained from scratch in the denoising or reconstruction process, with noisy images or sparely sampled k-space data as training labels. Performance of the proposed method was evaluated using in vivo experiment data obtained from 3 healthy subjects on a 3T MR scanner, using ASL images acquired with 44-min acquisition time as the ground truth. Both qualitative and quantitative analyses demonstrate the superior performance of the proposed txtc framework over the reference methods. In summary, our proposed unsupervised deep learning-based denoising and reconstruction framework can improve the image quality and accelerate the imaging speed of ASL imaging.

Spectral binning as an approach to post-acquisition processing of high resolution FIE-MS metabolome fingerprinting data.

INTRODUCTION: Flow infusion electrospray high resolution mass spectrometry (FIE-HRMS) fingerprinting produces complex, high dimensional data sets which require specialist in-silico software tools to process the data prior to analysis. OBJECTIVES: Present spectral binning as a pragmatic approach to post-acquisition procession of FIE-HRMS metabolome fingerprinting data. METHODS: A spectral binning approach was developed that included the elimination of single scan m/z events, the binning of spectra and the averaging of spectra across the infusion profile. The modal accurate m/z was then extracted for each bin. This approach was assessed using four different biological matrices and a mix of 31 known chemical standards analysed by FIE-HRMS using an Exactive Orbitrap. Bin purity and centrality metrics were developed to objectively assess the distribution and position of accurate m/z within an individual bin respectively. RESULTS: The optimal spectral binning width was found to be 0.01 amu. 80.8% of the extracted accurate m/z matched to predicted ionisation products of the chemical standards mix were found to have an error of below 3 ppm. The open-source R package binneR was developed as a user friendly implementation of the approach. This was able to process 100 data files using 4 Central Processing Units (CPU) workers in only 55 seconds with a maximum memory usage of 1.36 GB. CONCLUSION: Spectral binning is a fast and robust method for the post-acquisition processing of FIE-HRMS data. The open-source R package binneR allows users to efficiently process data from FIE-HRMS experiments with the resources available on a standard desktop computer.

Supervised segmentation framework for evaluation of diffusion tensor imaging indices in skeletal muscle.

Diffusion tensor imaging (DTI) is becoming a relevant diagnostic tool to understand muscle disease and map muscle recovery processes following physical activity or after injury. Segmenting all the individual leg muscles, necessary for quantification, is still a time-consuming manual process. The purpose of this study was to evaluate the impact of a supervised semi-automatic segmentation pipeline on the quantification of DTI indices in individual upper leg muscles. Longitudinally acquired MRI datasets (baseline, post-marathon and follow-up) of the upper legs of 11 subjects were used in this study. MR datasets consisted of a DTI and Dixon acquisition. Semi-automatic segmentations for the upper leg muscles were performed using a transversal propagation approach developed by Ogier et al on the out-of-phase Dixon images at baseline. These segmentations were longitudinally propagated for the post-marathon and follow-up time points. Manual segmentations were performed on the water image of the Dixon for each of the time points. Dice similarity coefficients (DSCs) were calculated to compare the manual and semi-automatic segmentations. Bland-Altman and regression analyses were performed, to evaluate the impact of the two segmentation methods on mean diffusivity (MD), fractional anisotropy (FA) and the third eigenvalue (λ3 ). The average DSC for all analyzed muscles over all time points was 0.92 ± 0.01, ranging between 0.48 and 0.99. Bland-Altman analysis showed that the 95% limits of agreement for MD, FA and λ3 ranged between 0.5% and 3.0% for the transversal propagation and between 0.7% and 3.0% for the longitudinal propagations. Similarly, regression analysis showed good correlation for MD, FA and λ3 (r = 0.99, p < 60; 0.0001). In conclusion, the supervised semi-automatic segmentation framework successfully quantified DTI indices in the upper-leg muscles compared with manual segmentation while only requiring manual input of 30% of the slices, resulting in a threefold reduction in segmentation time.

Feasibility of precise and reliable glucose quantification in human whole blood samples by 1 tesla benchtop NMR.

The standard procedure for blood glucose measurements is enzymatic testing. This method is cheap, but requires small samples of open blood with direct contact to the test medium. In principle, NMR provides non-contact analysis of body fluids, but high-field spectrometers are expensive and cannot be easily utilized under clinical conditions. Low-field NMR systems with permanent magnets are becoming increasingly smaller and more affordable. The studies presented here aim at exploring the capabilities of low-field NMR for measuring glucose concentrations in whole blood. For this purpose, a modern 1 T benchtop NMR spectrometer was used. Challenges arise from broad spectral lines, the glucose peak locations close to the water signal, low SNR and the interference with signals from other blood components. Whole blood as a sample comprises even more boundary conditions: crucial for reliable results are avoiding the separation of plasma and cells by gravitation and reliable reference values. First, the accuracy of glucose levels measured by NMR was tested using aqueous glucose solutions and commercially available bovine plasma. Then, 117 blood samples from oral glucose tolerance testing were measured with minimal preparation by simple pulse-acquire NMR experiments. The analysis itself is the key to achieve high precision, so several approaches were investigated: peak integration, orthogonal projection to latent structure analysis and support vector machine regression. Correlations between results from the NMR spectra and the routine laboratory automated analyzer revealed an RMSE of 7.90 mg/dL for the best model. 91.5% of the model output lies within the limits of the German Medical Association guidelines, which require the glucose measurement to be within 11% of the reference method. It is concluded that spectral quantification of glucose in whole blood samples by high-quality NMR spectrometers operating at 1 T is feasible with sufficient accuracy.

Analysis of metabolic abnormalities in high-grade glioma using MRSI and convex NMF.

Clinical use of MRSI is limited by the level of experience required to properly translate MRSI examinations into relevant clinical information. To solve this, several methods have been proposed to automatically recognize a predefined set of reference metabolic patterns. Given the variety of metabolic patterns seen in glioma patients, the decision on the optimal number of patterns that need to be used to describe the data is not trivial. In this paper, we propose a novel framework to (1) separate healthy from abnormal metabolic patterns and (2) retrieve an optimal number of reference patterns describing the most important types of abnormality. Using 41 MRSI examinations (1.5 T, PRESS, TE 135 ms) from 22 glioma patients, four different patterns describing different types of abnormality were detected: edema, healthy without Glx, active tumor and necrosis. The identified patterns were then evaluated on 17 MRSI examinations from nine different glioma patients. The results were compared against BraTumIA, an automatic segmentation method trained to identify different tumor compartments on structural MRI data. Finally, the ability to predict future contrast enhancement using the proposed approach was also evaluated.

Early regional cuprizone-induced demyelination in a rat model revealed with MRI.

The cuprizone model of demyelination is well established in the mouse as a tool for the study of the mechanisms of both demyelination and remyelination. It is often desirable, however, to have a larger model, such as the rat, especially for imaging-based studies, yet initial work has failed to show demyelination in cuprizone-fed rats. Several recent studies have demonstrated demyelination in the rat, but only in the corpus callosum. In this study, we acquired high-resolution, three-dimensional images of the whole brain every 2 weeks, using a T1 -weighted magnetization-prepared rapid acquisition gradient echo imaging sequence, optimized for myelin contrast, in order to assess myelination across the entire rat brain over a period of 8 weeks on a 1% cuprizone diet. We observed a consistent pattern of demyelination, beginning in the cerebellum by 4 weeks and involving more rostral regions of the brain by 8 weeks on the cuprizone diet, with validation using Luxol fast blue histology. This imaging technique permits the effects of cuprizone-induced demyelination to be followed longitudinally in a single animal, over the entire brain. In turn, this may facilitate the establishment of the cuprizone model of demyelination in the rat.

Separation of collagen-bound and porous bone water transverse relaxation in mice: proposal of a multi-step approach.

The separation and quantification of collagen-bound water (CBW) and pore water (PW) components of the cortical bone signal are important because of their different contribution to bone mechanical properties. Ultrashort TE (UTE) imaging can be used to exploit the transverse relaxation from CBW and PW, allowing their quantification. We tested, for the first time, the feasibility of UTE measurements in mice for the separation and quantification of the transverse relaxation of CBW and PW in vivo using three different approaches for T2 * determination. UTE sequences were acquired at 4.7 T in six mice with 10 different TEs (50-5000 µs). The transverse relaxation time T2 * of CBW (T2 *cbw ) and PW (T2 *pw ) and the CBW fraction (bwf) were computed using a mono-exponential (i), a standard bi-exponential (ii) and a new multi-step bi-exponential (iii) approach. Regions of interest were drawn at multiple levels of the femur and vertebral body cortical bone for each mouse. The sum of the normalized squared residuals (Res) and the homogeneity of variance were tested to compare the different methods. In the femur, approach (i) yielded mean T2 * ± standard deviation (SD) of 657 ± 234 µs. With approach (ii), T2 *cbw , T2 *pw and bwf were 464 ± 153 µs, 15 777 ± 10 864 µs and 57.6 ± 9.9%, respectively. For approach (iii), T2 *cbw , T2 *pw and bwf were 387 ± 108 µs, 7534 ± 2765 µs and 42.5 ± 6.2%, respectively. Similar values were obtained from vertebral bodies. Res with approach (ii) was lower than with the two other approaches (p < 0.007), but T2 *pw and bwf variance was lower with approach (iii) than with approach (ii) (p < 0.048). We demonstrated that the separation and quantification of cortical bone water components with UTE sequences is feasible in vivo in mouse models. The direct bi-exponential approach exhibited the best approximation to the measured signal curve with the lowest residuals; however, the newly proposed multi-step algorithm resulted in substantially lower variability of the computed parameters. Copyright © 2016 John Wiley & Sons, Ltd.

Characterization of trabecular bone density with ultra-short echo-time MRI at 1.5, 3.0 and 7.0 T--comparison with micro-computed tomography.

The goal of this study was to test the potential of ultra-short echo-time (UTE) MRI at 1.5, 3.0 and 7.0 T for depiction of trabecular bone structure (of the wrist bones), to evaluate whether T2* relaxation times of bone water and parametric maps of T2* of trabecular bone could be obtained at all three field strengths, and to compare the T2* relaxation times with structural parameters obtained from micro-computed tomography (micro-CT) as a reference standard. Ex vivo carpal bones of six wrists were excised en bloc and underwent MRI at 1.5, 3.0 and 7.0 T in a whole-body MR imager using the head coil. A three-dimensional radial fat-suppressed UTE sequence was applied with subsequent acquisitions, with six different echo times TE of 150, 300, 600, 1200, 3500 and 7000 µs. The T2* relaxation time and pixel-wise computed T2* parametric maps were compared with a micro-computed-tomography reference standard providing trabecular bone structural parameters including porosity (defined as the bone-free fraction within a region of interest), trabecular thickness, trabecular separation, trabecular number and fractal dimension (Dk). T2* relaxation curves and parametric maps could be computed from datasets acquired at all field strengths. Mean T2* relaxation times of trabecular bone were 4580 ± 1040 µs at 1.5 T, 2420 ± 560 µs at 3.0 T and 1220 ± 300 µs at 7.0 T, when averaged over all carpal bones. A positive correlation of T2* with trabecular bone porosity and trabecular separation, and a negative correlation of T2* relaxation time with trabecular thickness, trabecular number and fractal dimension, was detected (p < 0.01 for all field strengths and micro-CT parameters). We conclude that UTE MRI may be useful to characterize the structure of trabecular bone, comparable to micro-CT.

Prostate cancer detection and diagnosis: the role of MR and its comparison with other diagnostic modalities--a radiologist's perspective.

It is now universally recognized that many prostate cancers are over-diagnosed and over-treated. The European Randomized Study of Screening for Prostate Cancer from 2009 evidenced that, to save one man from death from prostate cancer, over 1400 men need to be screened, and 48 need to undergo treatment. The detection of prostate cancer is traditionally based on digital rectal examination (DRE) and the measurement of serum prostate-specific antigen (PSA), followed by ultrasound-guided biopsy. The primary role of imaging for the detection and diagnosis of prostate cancer has been transrectal ultrasound (TRUS) guidance during biopsy. Traditionally, MRI has been used primarily for the staging of disease in men with biopsy-proven cancer. It has a well-established role in the detection of T3 disease, planning of radiation therapy, especially three-dimensional conformal or intensity-modulated external beam radiation therapy, and planning and guiding of interstitial seed implant or brachytherapy. New advances have now established that prostate MRI can accurately characterize focal lesions within the gland, an ability that has led to new opportunities for improved cancer detection and guidance for biopsy. Two new approaches to prostate biopsy are under investigation. Both use pre-biopsy MRI to define potential targets for sampling, and the biopsy is performed either with direct real-time MR guidance (in-bore) or MR fusion/registration with TRUS images (out-of-bore). In-bore and out-of-bore MRI-guided prostate biopsies have the advantage of using the MR target definition for the accurate localization and sampling of targets or suspicious lesions. The out-of-bore method uses combined MRI/TRUS with fusion software that provides target localization and increases the sampling accuracy of TRUS-guided biopsies by integrating prostate MRI information with TRUS. Newer parameters for each imaging modality, such as sonoelastography or shear wave elastography, contrast-enhanced ultrasound and MRI elastography, show promise to further enrich datasets.

A portable single-sided magnet system for remote NMR measurements of pulmonary function.

In this work, we report initial results from a light-weight, low field magnetic resonance device designed to make relative pulmonary density measurements at the bedside. The development of this device necessarily involves special considerations for the magnet, RF and data acquisition schemes as well as a careful analysis of what is needed to provide useful information in the ICU. A homogeneous field region is created remotely from the surface of the magnet such that when the magnet is placed against the chest, an NMR signal is measured from a small volume in the lung. In order to achieve portability, one must trade off field strength and therefore spatial resolution. We report initial measurements from a ping-pong ball size region in the lung as a function of lung volume. As expected, we measured decreased signal at larger lung volumes since lung density decreases with increasing lung volume. Using a CPMG sequence with ΔTE=3.5 ms and a 20 echo train, a signal to noise ratio ~1100 was obtained from an 8.8mT planar magnet after signal averaging for 43 s. This is the first demonstration of NMR measurements made on a human lung with a light-weight planar NMR device. We argue that very low spatial resolution measurements of different lobar lung regions will provide useful diagnostic information for clinicians treating Acute Respiratory Distress Syndrome as clinicians want to avoid ventilator pressures that cause either lung over distension (too much pressure) or lung collapse (too little pressure).

Quantitative comparison of post-processing methods for reduction of frequency modulation sidebands in non-water suppression 1H MRS.

Non-water suppression MRS (NWS MRS) has several advantages. First, the unsuppressed water signal can be used as internal calibration for metabolite quantification and as a reliable frequency/phase reference for retrospective motion correction. Second, it avoids the potential artifacts caused by incomplete water suppression (WS) and extra radiofrequency deposition from WS pulses. However, the frequency modulation (FM) sidebands originating from a large water signal will distort the spectrum. Among the methods proposed to solve the problems caused by FM sidebands, post-acquisition processing methods are superior in flexibility for general use compared with experimental methods. In this study, we propose two algorithms based on advanced matrix decomposition to remove the FM sidebands. These methods, the simultaneous diagonalization (QZ) algorithm and its subsequent variant, the simultaneously generalized Schur decomposition (SGSD) algorithm, were numerically evaluated using computer simulations. In addition, we quantitatively compared the performance of these methods and the modulus method in an in vitro experiment and in vivo NWS MRS against conventional WS data. Our results show that the proposed SGSD algorithm can reduce the FM sidebands to achieve superior estimation of concentration on three major metabolites. This method can be applied directly to spectra pre-acquired under various experimental conditions without modifying the acquisition sequences.

Frequency-modulated bSSFP for phase-sensitive separation of water and fat.

Our study proposes the use of a frequency-modulated acquisition which suppresses banding artefacts in combination with a phase-sensitive water-fat separation algorithm. The performance of the phase-sensitive separation for standard bSSFP, complex sum combination thereof, and frequency-modulated bSSFP were compared in in vivo measurements of the upper and lower legs at 1.5 and 3 T. It is shown, that the standard acquisition suffered from banding artefacts and major swaps between tissues. The dual-acquisition bSSFP could alleviate banding artefacts and only minor swaps occurred, but it comes at the expense of a doubled acquisition. In the frequency-modulated acquisitions all banding artefacts and the associated phase jumps were eliminated and no swaps between tissues occurred. It therefore provides a means to robustly separate water and fat, in one single radial bSSFP scan, using the phase-sensitive approach, even in the presence of high field inhomogeneities.

Repeated Activation of a CS-US-Contingency Memory Results in Sustained Conditioned Responding.

Individuals seem to differ in conditionability, i.e., the ease by which the contingent presentation of two stimuli will lead to a conditioned response. In contemporary learning theory, individual differences in the etiology and maintenance of anxiety disorders are, among others, explained by individual differences in temperamental variables (Mineka and Zinbarg, 2006). One such individual difference variable is how people process a learning experience when the conditioning stimuli are no longer present. Repeatedly thinking about the conditioning experience, as in worry or rumination, might prolong the initial (fear) reactions and as such, might leave certain individuals more vulnerable to developing an anxiety disorder. However, in human conditioning research, relatively little attention has been devoted to the processing of a memory trace after its initial acquisition, despite its potential influences on subsequent performance. Post-acquisition processing can be induced by mental reiteration of a conditioned stimulus-unconditioned stimulus (CS-US)-contingency. Using a human conditioned suppression paradigm, we investigated the effect of repeated activations of a CS-US-contingency memory on the level of conditioned responding at a later test. Results of three experiments showed more sustained responding to a "rehearsed" CS+ as compared to a "non-rehearsed" CS+. Moreover, the second experiment showed no effect of rehearsal when only the CS was rehearsed instead of the CS-US-contingency. The third experiment demonstrated that mental CS-US-rehearsal has the same effect regardless of whether it was cued by the CS and a verbal reference to the US or by a neutral signal, making the rehearsal "purely mental." In sum, it was demonstrated that post-acquisition activation of a CS-US-contingency memory can impact conditioned responding, underlining the importance of post-acquisition processes in conditioning. This might indicate that individuals who are more prone to mentally rehearse information condition more easily.