Single-voxel Spectroscopy Data Research Articles

Generating Synthetic MR Spectroscopic Imaging Data with Generative Adversarial Networks to Train Machine Learning Models.

To develop a new method to generate synthetic MR spectroscopic imaging (MRSI) data for training machine learning models. This study targeted routine MRI examination protocols with single voxel spectroscopy (SVS). A novel model derived from pix2pix generative adversarial networks was proposed to generate synthetic MRSI data using MRI and SVS data as inputs. T1- and T2-weighted, SVS, and reference MRSI data were acquired from healthy brains with clinically available sequences. The proposed model was trained to generate synthetic MRSI data. Quantitative evaluation involved the calculation of the mean squared error (MSE) against the reference and metabolite ratio value. The effect of the location of and the number of the SVS data on the quality of the synthetic MRSI data was investigated using the MSE. The synthetic MRSI data generated from the proposed model were visually closer to the reference. The 95% confidence interval (CI) of the metabolite ratio value of synthetic MRSI data overlapped with the reference for seven of eight metabolite ratios. The MSEs tended to be lower in the same location than in different locations. The MSEs among groups of numbers of SVS data were not significantly different. A new method was developed to generate MRSI data by integrating MRI and SVS data. Our method can potentially increase the volume of MRSI data training for other machine learning models by adding SVS acquisition to routine MRI examinations.

A method for cranial target delineation in radiotherapy treatment planning aided by single-voxel magnetic resonance spectroscopy: evaluation using a custom-designed gel-based phantom and simulations.

Magnetic resonance spectroscopy (MRS) has been useful in radiotherapy treatment planning (RTP) especially in tumor delineation. Routinely, 2D/3D MRSI data are used for this application. However, not all centers have access to 2D/3D MRSI. The objective of this study was to introduce a method of using single-voxel spectroscopy (SVS) data in target delineation and assess its reliability. A gel-based phantom containing Creatine (Cr), N-acetyl-l-aspartic-acid (NAA), and Choline (Cho) was designed and built. The metabolite ratios simulate the normal and tumoral part of the brain. The jMRUI software (v. 6.0) was used to simulate a 1.5 T GE MRI scanner. The metabolite spectra provided by different time of echos (TE)s of the Point-RESolved Spectroscopy pulse-sequence (PRESS), different data-points, and post-processings were quantized by jMRUI. PseudoMRSI maps of Cho/Cr, NAA/Cr, and Cho + Cr/NAA were created. A conformity index (CI) was used to determine which metabolite-ratio isolines are more appropriate for tumor delineation. The simulation accuracy was verified. There were no differences > 4% between the measured and simulated spectra in peak regions. The pseudoMRSI map of Cho + Cr/NAA smoothly followed the complicated geometry of the tumor inside the gel-based phantom. The results showed that the single-voxel spectra produced by the PRESS pulse sequence with the TE of 144 ms, 512 data-points, and minimum post-processings of water suppression, eddy current correction, and baseline correction can be used for target delineation. This study suggests that SVS data can be used to aid target delineation by using a mathematical approach. This can enable a wider use of MR-derived information in radiotherapy. To the best of our knowledge, until now, 2D or 3D MRSI data provided from 3T MRI scanners have been used for MRS-based radiotherapy treatment planning. However, there are a lot of centers that are equipped to 1.5 T MRI scanners and some of them just equipped to SVS. This study introduces a mathematical approach to help these centers to take the benefits of MRS-based treatment planning.

Macromolecule mapping of the brain using ultrashort-TE acquisition and reference-based metabolite removal.

To develop a practical method for mapping macromolecule distribution in the brain using ultrashort-TE MRSI data. An FID-based chemical shift imaging acquisition without metabolite-nulling pulses was used to acquire ultrashort-TE MRSI data that capture the macromolecule signals with high signal-to-noise-ratio (SNR) efficiency. To remove the metabolite signals from the ultrashort-TE data, single voxel spectroscopy data were obtained to determine a set of high-quality metabolite reference spectra. These spectra were then incorporated into a generalized series (GS) model to represent general metabolite spatiospectral distributions. A time-segmented algorithm was developed to back-extrapolate the GS model-based metabolite distribution from truncated FIDs and remove it from the MRSI data. Numerical simulations and in vivo experiments have been performed to evaluate the proposed method. Simulation results demonstrate accurate metabolite signal extrapolation by the proposed method given a high-quality reference. For in vivo experiments, the proposed method is able to produce spatiospectral distributions of macromolecules in the brain with high SNR from data acquired in about 10 minutes. We further demonstrate that the high-dimensional macromolecule spatiospectral distribution resides in a low-dimensional subspace. This finding provides a new opportunity to use subspace models for quantification and accelerated macromolecule mapping. Robustness of the proposed method is also demonstrated using multiple data sets from the same and different subjects. The proposed method is able to obtain macromolecule distributions in the brain from ultrashort-TE acquisitions. It can also be used for acquiring training data to determine a low-dimensional subspace to represent the macromolecule signals for subspace-based MRSI. Magn Reson Med 79:2460-2469, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

A Novel Semi-Supervised Methodology for Extracting Tumor Type-Specific MRS Sources in Human Brain Data

BackgroundThe clinical investigation of human brain tumors often starts with a non-invasive imaging study, providing information about the tumor extent and location, but little insight into the biochemistry of the analyzed tissue. Magnetic Resonance Spectroscopy can complement imaging by supplying a metabolic fingerprint of the tissue. This study analyzes single-voxel magnetic resonance spectra, which represent signal information in the frequency domain. Given that a single voxel may contain a heterogeneous mix of tissues, signal source identification is a relevant challenge for the problem of tumor type classification from the spectroscopic signal.Methodology/Principal FindingsNon-negative matrix factorization techniques have recently shown their potential for the identification of meaningful sources from brain tissue spectroscopy data. In this study, we use a convex variant of these methods that is capable of handling negatively-valued data and generating sources that can be interpreted as tumor class prototypes. A novel approach to convex non-negative matrix factorization is proposed, in which prior knowledge about class information is utilized in model optimization. Class-specific information is integrated into this semi-supervised process by setting the metric of a latent variable space where the matrix factorization is carried out. The reported experimental study comprises 196 cases from different tumor types drawn from two international, multi-center databases. The results indicate that the proposed approach outperforms a purely unsupervised process by achieving near perfect correlation of the extracted sources with the mean spectra of the tumor types. It also improves tissue type classification.Conclusions/SignificanceWe show that source extraction by unsupervised matrix factorization benefits from the integration of the available class information, so operating in a semi-supervised learning manner, for discriminative source identification and brain tumor labeling from single-voxel spectroscopy data. We are confident that the proposed methodology has wider applicability for biomedical signal processing.

Effects of fat on MR‐measured metabolite signal strengths: implications for in vivo MRS studies of the human brain

Recent MRS studies have indicated that a higher body mass index (BMI) is associated with lower brain metabolite levels. Generally, individuals with higher BMIs have more body fat deposits than individuals with normal BMIs. This single-voxel spectroscopy (SVS) study investigated possible effects of fat on MR-measured metabolite signal areas, which may at least partly explain the observed associations of BMI with MR-measured brain metabolite levels in vivo. SVS data were acquired at 4 T from a phantom containing N-acetylaspartate, glutamate and creatine, as well as from three healthy male adults. Back fat obtained from pig was used to assess the effects of fat on metabolite signals. With the same voxel size and placement, the phantom was first scanned without fat (baseline), and then with 0.7-cm- and 1.4-cm-thick fat layers placed on it. Each participant was also scanned first without fat and then with two 0.7-cm fat layers, one placed beneath the occiput and the other on the forehead. Two spectra were acquired per participant from the anterior cingulate and the parieto-occipital cortices. The metabolite resonance and corresponding water peak areas were then fitted and metabolite to water signal ratios were used for analyses. In both phantom and in vivo experiments, the metabolite-to-water ratios decreased in the presence of fat relative to baseline metabolite-to-water ratios. The reduced metabolite signals in the presence of fat reported here are reminiscent of the negative correlations observed between BMI and MR-measured metabolite levels. These apparent physical effects of fat have potentially far-reaching consequences for the accuracy of MR measurements of brain metabolite levels and their interpretation, particularly when large fat stores exist around the skull, such as in individuals with higher BMI.

Abstract P2-06-16: Total Choline Measurement in Human Breast Using High-Speed MR Spectroscopic Imaging at 3T

Abstract PURPOSE: We developed a novel, quantitative and high-speed MR spectroscopic imaging (MRSI) method to map total Choline (tCho), a sensitive biomarker of breast tumor status, as an adjunct to enhance the limited specificity of routine dynamic-contrast enhanced (DCE) MRI. Quantitative tCho maps measured in 7 minutes were compared with tCho obtained with conventional Single Voxel Spectroscopy (SVS). METHOD AND MATERIALS: Measurements on a total of 18 healthy female subjects (mean age: 25.6±5) were performed using 3T MR scanners (Siemens Trio, Erlangen, Germany) located at the two partner sites equipped with 4-channel breast coil (Siemens, Erlangen, Germany) or 8-channel breast coil (Sentinelle Medical, Toronto, Canada). 2D MRSI data of an entire oblique slice were collected using Proton-Echo-Planar-Spectroscopic-Imaging (PEPSI) [1] with MEGA lipid suppression. Acquisition parameters were: TR/TE=1500ms/125ms, matrix size=32x32, voxel size=2x2x2mm3 (8cc), number of signal averages= 16 with weighted k-space sampling and total acquisition time=7 minutes. PRESS SVS data were acquired with 8 cc voxel size using identical TR/TE and acquisition time. The absolute metabolite concentration was calculated in reference to tissue water (millimoles of tCho per kilogram of solute) using LCModel (s-provencher.com) fitting to estimate the Choline peak baseline and subsequent spectral integration using a Cramer-Rao lower bound threshold of 25%. RESULTS: tCho was detected in 7 of the 15 subjects (47%) in both SVS and PEPSI data. In the PEPSI data sets, tCho was detected in multiple voxels (Fig. 1). Subjects in which tCho was detected exhibited narrower water line width and smaller lipid content than subjects in which tCho was not detectable (2-tailed t-test, P<0.01). The absolute tCho concentrations corrected for relaxation effects in these 7 subjects using SVS and PEPSI was 0.43±0.34 mmol/kg and 0.51±0.19 mmol/kg, respectively. In comparison with SVS data (21.8±8.6Hz), PEPSI spectra demonstrated larger water line width (33.9±12.6Hz) and displayed greater lipid contamination from adipose tissue areas and larger baseline distortion due to the spatial point spread function. CONCLUSION: Despite less favorable shimming and lipid suppression conditions compared to SVS, it is feasible to quantitatively map tCho in healthy breast tissue using high-speed MRSI, with concentration values that are consistent with those from SVS. Studies in breast cancer patients are in progress to assess the feasibility of breast cancer diagnosis and treatment monitoring with MRSI. Results will be reported at the Symposium. The long-term goals are to utilize high-speed MRSI as an early predictor of treatment failure in women undergoing systemic therapy (i.e. chemotherapy, endocrine therapy) for breast cancer and to develop an improved screening protocol for high risk patients. Fig. 1: PEPSI slice localization (left) and spectral array (right) with superimposed LCModel fit and integrated tCho peak Ref: (1) Posse et al. Magn. Reson. Med. 2007;58(2):236-244. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P2-06-16.