Search

MRS thermometry has been utilized to measure temperature changes in the brain, which may aid in the diagnosis of brain trauma and tumours. However, the temperature calibration of the technique has been shown to be sensitive to non-temperature-based factors, which may provide unique information on the tissue microenvironment if the mechanisms can be further understood. The focus of this study was to investigate the effects of varied protein content on the calibration of MRS thermometry at 3 T, which has not been thoroughly explored in the literature. The effects of ionic concentration and magnetic field strength were also considered. Temperature reference materials were controlled by water circulation and freezing organic fixed-point compounds (diphenyl ether and ethylene carbonate) stable to within 0.2 °C. The temperature was measured throughout the scan time with a fluoro-optic probe, with an uncertainty of 0.16 °C. The probe was calibrated at the National Physical Laboratory (NPL) with traceability to the International Temperature Scale 1990 (ITS-90). MRS thermometry measures were based on single-voxel spectroscopy chemical shift differences between water and N-acetylaspartate (NAA), Δ(H20-NAA), using a Philips Achieva 3 T scanner. Six different phantom solutions with varying protein or ionic concentration, simulating potential tissue differences, were investigated within a temperature range of 21–42 °C. Results were compared with a similar study performed at 1.5 T to observe the effect of field strengths. Temperature calibration curves were plotted to convert Δ(H20-NAA) to apparent temperature. The apparent temperature changed by −0.2 °C/% of bovine serum albumin (BSA) and a trend of 0.5 °C/50 mM ionic concentration was observed. Differences in the calibration coefficients for the 10% BSA solution were seen in this study at 3 T compared with a study at 1.5 T. MRS thermometry may be utilized to measure temperature and the tissue microenvironment, which could provide unique unexplored information for brain abnormalities and other pathologies.

Brain tumours are the most common solid cancers in children in the UK and are the most common cause of cancer deaths in this age group. Despite current advances in MRI, non-invasive diagnosis of paediatric brain tumours has yet to find its way into routine clinical practice. Radiomics, the high-throughput extraction and analysis of quantitative image features (e.g. texture), offers potential solutions for tumour characterization and decision support. In the search for diagnostic oncological markers, the primary aim of this work was to study the application of MRI texture analysis (TA) for the classification of paediatric brain tumours. A multicentre study was carried out, within a supervised classification framework, on clinical MR images, and a support vector machine (SVM) was trained with 3D textural attributes obtained from conventional MRI. To determine the cross-centre transferability of TA, an assessment of how SVM performs on unseen datasets was carried out through rigorous pairwise testing. The study also investigated the nature of features that are most likely to train classifiers that can generalize well with the data. Finally, the issue of class imbalance, which arises due to some tumour types being more common than others, was explored. For each of the tests carried out through pairwise testing, the optimal area under the receiver operating characteristic curve ranged between 76% and 86%, suggesting that the model was able to capture transferable tumour information. Feature selection results suggest that similar aspects of tumour texture are enhanced by MR images obtained at different hospitals. Our results also suggest that the availability of equally represented classes has enabled SVM to better characterize the data points. The findings of the study presented here support the use of 3D TA on conventional MR images to aid diagnostic classification of paediatric brain tumours.

Independent component analysis (ICA) can automatically extract individual metabolite, macromolecular and lipid (MMLip) components from a series of in vivo MR spectra. The traditional feature extraction (FE)-based ICA approach is limited, in that a large sample size is required and a combination of metabolite and MMLip components can appear in the same independent component. The alternative ICA approach, based on blind source separation (BSS), is weak when dealing with overlapping peaks. Combining the advantages of both BSS and FE methods may lead to better results. Thus, we propose an ICA approach involving a hybrid of the BSS and FE techniques for the automated decomposition of a series of MR spectra. Experiments were performed on synthesised and patient in vivo childhood brain tumour MR spectra datasets. The hybrid ICA method showed an improvement in the decomposition ability compared with BSS-ICA or FE-ICA, with an increased correlation between the independent components and simulated metabolite and MMLip signals. Furthermore, we were able to automatically extract metabolites from the patient MR spectra dataset that were not in commonly used basis sets (e.g. guanidinoacetate).

A number of algorithms designed to determine metabolite concentrations from in vivo (1)H MRS require a collection of single metabolite spectra, known as a basis set, which can be obtained experimentally or by simulation. It has been assumed that basis sets can be used interchangeably, but no systematic study has investigated the effects of small variations in basis functions on the metabolite values obtained. The aim of this study was to compare the results of simulated with experimental basis sets when used to fit short-TE (1)H MRS data of variable quality at 1.5 T. Two hundred and twelve paediatric brain tumour spectra were included in the analysis, and each was analysed twice with LCModel™ using a simulated and experimental basis set. To determine the influence of data quality on quantification, each spectrum was assessed and 152 were classified as being of 'good' quality. Bland-Altman statistics were used to measure the agreement between the two basis sets for all available spectra and only 'good'-quality spectra. Monte-Carlo simulations were performed to investigate the influence of minor shifts in metabolite frequencies on metabolite concentration estimates. All metabolites showed good agreement between the two basis sets, and the average metabolite limits of agreement were approximately ±3.84 mM for all available data and ±0.99 mM for good-quality data. Errors obtained from the Monte-Carlo analysis were found to be more accurate than the Cramer-Rao lower bounds (CRLB) for 12 of 15 metabolites when metabolite frequency shifting was considered. For the majority of purposes, a level of agreement of ±0.99 mM between simulated and experimental basis sets is sufficiently small for them to be used interchangeably. Multiple analyses using slightly modified basis sets may be useful in estimating fitting errors, which are not predicted by CRLBs.

Independent component analysis (ICA) has the potential of determining automatically the metabolite signals which make up MR spectra. However, the realiability with which this is accomplished and the optimal approach for investigating in vivo MRS have not been determined. Furthermore, the properties of ICA in brain tumour MRS with respect to dataset size and data quality have not been systematically explored. The two common techniques for applying ICA, blind source separation (BSS) and feature extraction (FE) were examined in this study using simulated data and the findings confirmed on patient data. Short echo time (TE 30 ms), low and high field (1.5 and 3 T) in vivo brain tumour MR spectra of childhood astrocytoma, ependymoma and medulloblastoma were generated by using a quantum mechanical simulator with ten metabolite and lipid components. Patient data (TE 30 ms, 1.5 T) were acquired from children with brain tumours. ICA of simulated data shows that individual metabolite components can be extracted from a set of MRS data. The BSS method generates independent components with a closer correlation to the original metabolite and lipid components than the FE method when the number of spectra in the dataset is small. The experiments also show that stable results are achieved with 300 MRS at an SNR equal to 10. The FE method is relatively insensitive to different ranges of full width at half maximum (FWHM) (from 0 to 3 Hz), whereas the BSS method degrades on increasing the range of FWHM. The peak frequency variations do not affect the results within the range of ±0.08 ppm for the FE method, and ±0.05 ppm for the BSS method. When the methods were applied to the patient dataset, results consistent with the synthesized experiments were obtained.

(1)H MRS has great potential for the clinical investigation of childhood brain tumours, but the low incidence in, and difficulties of performing trials on, children have hampered progress in this area. Most studies have used a long-TE, thus limiting the metabolite information obtained, and multivariate analysis has been largely unexplored. Thirty-five children with untreated cerebellar tumours (18 medulloblastomas, 12 pilocytic astrocytomas and five ependymomas) were investigated using a single-voxel short-TE PRESS sequence on a 1.5 T scanner. Spectra were analysed using LCModel to yield metabolite profiles, and key metabolite assignments were verified by comparison with high-resolution magic-angle-spinning NMR of representative tumour biopsy samples. In addition to univariate metabolite comparisons, the use of multivariate classifiers was investigated. Principal component analysis was used for dimension reduction, and linear discriminant analysis was used for variable selection and classification. A bootstrap cross-validation method suitable for estimating the true performance of classifiers in small datasets was used. The discriminant function coefficients were stable and showed that medulloblastomas were characterised by high taurine, phosphocholine and glutamate and low glutamine, astrocytomas were distinguished by low creatine and high N-acetylaspartate, and ependymomas were differentiated by high myo-inositol and glycerophosphocholine. The same metabolite features were seen in NMR spectra of ex vivo samples. Successful classification was achieved for glial-cell (astrocytoma + ependymoma) versus non-glial-cell (medulloblastoma) tumours, with a bootstrap 0.632 + error, e(B.632+), of 5.3%. For astrocytoma vs medulloblastoma and astrocytoma vs medulloblastoma vs ependymoma classification, the e(B.632+) was 6.9% and 7.1%, respectively. The study showed that (1)H MRS detects key differences in the metabolite profiles for the main types of childhood cerebellar tumours and that discriminant analysis of metabolite profiles is a promising tool for classification. The findings warrant confirmation by larger multi-centre studies.

MRS thermometry has been utilized to measure temperature changes in the brain, which may aid in the diagnosis of brain trauma and tumours. However, the temperature calibration of the technique has been shown to be sensitive to non-temperature-based factors, which may provide unique information on the tissue microenvironment if the mechanisms can be further understood. The focus of this study was to investigate the effects of varied protein content on the calibration of MRS thermometry at 3 T, which has not been thoroughly explored in the literature. The effects of ionic concentration and magnetic field strength were also considered. Temperature reference materials were controlled by water circulation and freezing organic fixed-point compounds (diphenyl ether and ethylene carbonate) stable to within 0.2 °C. The temperature was measured throughout the scan time with a fluoro-optic probe, with an uncertainty of 0.16 °C. The probe was calibrated at the National Physical Laboratory (NPL) with traceability to the International Temperature Scale 1990 (ITS-90). MRS thermometry measures were based on single-voxel spectroscopy chemical shift differences between water and N-acetylaspartate (NAA), Δ(H20-NAA), using a Philips Achieva 3 T scanner. Six different phantom solutions with varying protein or ionic concentration, simulating potential tissue differences, were investigated within a temperature range of 21-42 °C. Results were compared with a similar study performed at 1.5 T to observe the effect of field strengths. Temperature calibration curves were plotted to convert Δ(H20-NAA) to apparent temperature. The apparent temperature changed by -0.2 °C/% of bovine serum albumin (BSA) and a trend of 0.5 °C/50 mM ionic concentration was observed. Differences in the calibration coefficients for the 10% BSA solution were seen in this study at 3 T compared with a study at 1.5 T. MRS thermometry may be utilized to measure temperature and the tissue microenvironment, which could provide unique unexplored information for brain abnormalities and other pathologies.

1H MRS thermometry has been investigated for brain trauma and hypothermia monitoring applications but has not been explored in brain tumours. The proton resonance frequency (PRF) of water is dependent on temperature but is also influenced by microenvironment factors, such as fast proton exchange with macromolecules, ionic concentration and magnetic susceptibility. 1H MRS has been utilized for brain tumour diagnostic and prognostic purposes in children; however, the water PRF measure may provide complementary information to further improve characterization. Water PRF values were investigated from a repository of MRS data acquired from childhood brain tumours and children with apparently normal brains. The cohort consisted of histologically proven glioma (22), medulloblastoma (19) and control groups (28, MRS in both the basal ganglia and parietal white matter regions). All data were acquired at 1.5 T using a short TE (30 ms) single voxel spectroscopy (PRESS) protocol. Water PRF values were calculated using methyl creatine and total choline. Spectral peak amplitude weighted averaging was used to improve the accuracy of the measurements. Mean PRF values were significantly larger for medulloblastoma compared with glioma, with a difference in the means of 0.0147 ppm (p < 0.05), while the mean PRF for glioma was significantly lower than for the healthy cohort, with a difference in the means of 0.0061 ppm (p < 0.05). This would suggest the apparent temperature of the glioma group was ∼1.5 °C higher than the medulloblastomas and ∼0.7 °C higher than a healthy brain. However, the PRF shift may not reflect a change in temperature, given that alterations in protein content, microstructure and ionic concentration contribute to PRF shifts. Measurement of these effects could also be used as a supplementary biomarker, and further investigation is required. This study has shown that the water PRF value has the potential to be used for characterizing childhood brain tumours, which has not been reported previously. © 2014 The Authors. NMR in Biomedicine published by John Wiley & Sons, Ltd.

Brain tumours are the most common solid tumours in children, representing 20% of all cancers. The most frequent posterior fossa tumours are medulloblastomas, pilocytic astrocytomas and ependymomas. Texture analysis (TA) of MR images can be used to support the diagnosis of these tumours by providing additional quantitative information. MaZda software was used to perform TA on T1- and T2-weighted images of children with pilocytic astrocytomas, medulloblastomas and ependymomas of the posterior fossa, who had MRI at Birmingham Children's Hospital prior to treatment. The region of interest was selected on three slices per patient in Image J, using thresholding and manual outlining. TA produced 279 features, which were reduced using principal component analysis (PCA). The principal components (PCs) explaining 95% of the variance were used in a linear discriminant analysis (LDA) and a probabilistic neural network (PNN) to classify the cases, using DTREG statistics software. PCA of texture features from both T1- and T2-weighted images yielded 13 PCs to explain >95% of the variance. The PNN classifier for T1-weighted images achieved 100% accuracy on training the data and 90% on leave-one-out cross-validation (LOOCV); for T2-weighted images, the accuracy was 100% on training the data and 93.3% on LOOCV. A PNN classifier with T1 and T2 PCs achieved 100% accuracy on training the data and 85.8% on LOOCV. LDA classification accuracies were noticeably poorer. The features found to hold the highest discriminating potential were all co-occurrence matrix derived, where adjacent pixels had highly correlated intensities. This study shows that TA can be performed on standard T1- and T2-weighted images of childhood posterior fossa tumours using readily available software to provide high diagnostic accuracy. Discriminatory features do not correspond to those used in the clinical interpretation of the images and therefore provide novel tumour information. Copyright © 2014 John Wiley & Sons, Ltd.

Independent component analysis (ICA) can automatically extract individual metabolite, macromolecular and lipid (MMLip) components from a series of in vivo MR spectra. The traditional feature extraction (FE)-based ICA approach is limited, in that a large sample size is required and a combination of metabolite and MMLip components can appear in the same independent component. The alternative ICA approach, based on blind source separation (BSS), is weak when dealing with overlapping peaks. Combining the advantages of both BSS and FE methods may lead to better results. Thus, we propose an ICA approach involving a hybrid of the BSS and FE techniques for the automated decomposition of a series of MR spectra. Experiments were performed on synthesised and patient in vivo childhood brain tumour MR spectra datasets. The hybrid ICA method showed an improvement in the decomposition ability compared with BSS-ICA or FE-ICA, with an increased correlation between the independent components and simulated metabolite and MMLip signals. Furthermore, we were able to automatically extract metabolites from the patient MR spectra dataset that were not in commonly used basis sets (e.g. guanidinoacetate).

Management of brain tumours in children would benefit from improved non-invasive diagnosis, characterisation and prognostic biomarkers. Metabolite profiles derived from in-vivo MRS have been shown to provide such information. Studies indicate that using optimum a priori information on metabolite contents in the construction of linear combination (LC) models of MR spectra leads to improved metabolite profile estimation. Glycine (Gly) is usually neglected in such models due to strong overlap with myo-inositol (mI) and a low concentration in normal brain. However, biological studies indicate that Gly is abundant in high-grade brain tumours. This study aimed to investigate the quantitation of Gly in paediatric brain tumours using MRS analysed by LCModel™, and its potential as a non-invasive biomarker of malignancy. Single-voxel MRS was performed using PRESS (TR 1500 ms, TE 30 ms/135 ms) on a 1.5 T scanner. Forty-seven cases (18 high grade (HG), 17 low grade (LG), 12 ungraded) were retrospectively selected if both short-TE and long-TE MRS (n = 33) or short-TE MRS and high-resolution magic-angle spinning (HRMAS) of matched surgical samples (n = 15) were available. The inclusion of Gly in LCModel™ analyses led to significantly reduced fit residues for both short-TE and long-TE MRS (p

Independent component analysis (ICA) has the potential of determining automatically the metabolite signals which make up MR spectra. However, the reliability with which this is accomplished and the optimal approach for investigating in vivo MRS have not been determined. Furthermore, the properties of ICA in brain tumour MRS with respect to dataset size and data quality have not been systematically explored. The two common techniques for applying ICA, blind source separation (BSS) and feature extraction (FE) were examined in this study using simulated data and the findings confirmed on patient data. Short echo time (TE 30 ms), low and high field (1.5 and 3 T) in vivo brain tumour MR spectra of childhood astrocytoma, ependymoma and medulloblastoma were generated by using a quantum mechanical simulator with ten metabolite and lipid components. Patient data (TE 30 ms, 1.5 T) were acquired from children with brain tumours. ICA of simulated data shows that individual metabolite components can be extracted from a set of MRS data. The BSS method generates independent components with a closer correlation to the original metabolite and lipid components than the FE method when the number of spectra in the dataset is small. The experiments also show that stable results are achieved with 300 MRS at an SNR equal to 10. The FE method is relatively insensitive to different ranges of full width at half maximum (FWHM) (from 0 to 3 Hz), whereas the BSS method degrades on increasing the range of FWHM. The peak frequency variations do not affect the results within the range of +/-0.08 ppm for the FE method, and +/-0.05 ppm for the BSS method. When the methods were applied to the patient dataset, results consistent with the synthesized experiments were obtained.

A number of algorithms designed to determine metabolite concentrations from in vivo (1)H MRS require a collection of single metabolite spectra, known as a basis set, which can be obtained experimentally or by simulation. It has been assumed that basis sets can be used interchangeably, but no systematic study has investigated the effects of small variations in basis functions on the metabolite values obtained. The aim of this study was to compare the results of simulated with experimental basis sets when used to fit short-TE (1)H MRS data of variable quality at 1.5 T. Two hundred and twelve paediatric brain tumour spectra were included in the analysis, and each was analysed twice with LCModel™ using a simulated and experimental basis set. To determine the influence of data quality on quantification, each spectrum was assessed and 152 were classified as being of 'good' quality. Bland-Altman statistics were used to measure the agreement between the two basis sets for all available spectra and only 'good'-quality spectra. Monte-Carlo simulations were performed to investigate the influence of minor shifts in metabolite frequencies on metabolite concentration estimates. All metabolites showed good agreement between the two basis sets, and the average metabolite limits of agreement were approximately ±3.84 mM for all available data and ±0.99 mM for good-quality data. Errors obtained from the Monte-Carlo analysis were found to be more accurate than the Cramer-Rao lower bounds (CRLB) for 12 of 15 metabolites when metabolite frequency shifting was considered. For the majority of purposes, a level of agreement of ±0.99 mM between simulated and experimental basis sets is sufficiently small for them to be used interchangeably. Multiple analyses using slightly modified basis sets may be useful in estimating fitting errors, which are not predicted by CRLBs.