A hybrid method of application of independent component analysis to in vivo 1H MR spectra of childhood 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).     

Linear discriminant analysis of brain tumour (1)H MR spectra: a comparison of classification using whole spectra versus metabolite quantification

(1)H MRS is an attractive choice for non-invasively diagnosing brain tumours. Many studies have been performed to create an objective decision support system, but there is not yet a consensus as to the best techniques of MRS acquisition or data processing to be used for optimum classification. In this study, we investigate whether LCModel analysis of short-TE (30 ms), single-voxel tumour spectra provide a better input for classification than the use of the original spectra. A total of 145 histologically diagnosed brain tumour spectra were acquired [14 astrocytoma grade II (AS2), 15 astrocytoma grade III (AS3), 42 glioblastoma (GBM), 41 metastases (MET) and 33 meningioma (MNG)], and linear discriminant analyses (LDA) were performed on the LCModel analysis of the spectra and the original spectra. The results consistently suggest improvement in classification when the LCModel concentrations are used. LDA of AS2, MNG and high-grade tumours (HG, comprising GBM and MET) correctly classified 94% using the LCModel dataset compared with 93% using the spectral dataset. The inclusion of AS3 reduced the accuracy to 82% and 78% for LCModel analysis and the original spectra, respectively, and further separating HG into GBM and MET gave 70% compared with 60%. Generally MNG spectra have profiles that are visually distinct from those of the other tumour types, but the classification accuracy was typically about 80%, with MNG with substantial lipid/macromolecule signals being classified as HG. Omission of the lipid/macromolecule concentrations in the LCModel dataset provided an improvement in classification of MNG (91% compared with 76%). In conclusion, there appears to be an advantage to performing pattern recognition on the quantitative analysis of tumour spectra rather than using the whole spectra. However, the results suggest that a two-step LDA process may help in classifying the five tumour groups to provide optimum classification of MNG with high lipid/macromolecule contributions which maybe misclassified as HG.     

Brain GABA editing by localized in vivo (1)H magnetic resonance spectroscopy

Editing of GABA by (1)H MRS in a specific brain area is a unique tool for in vivo non-invasive investigation of neurotransmission disorders. Selective GABA detection is achieved using sequences based on double quantum coherence (DQC). Our pulse sequence makes accurate measurements without artefacts due to spatial localization. The sequence was tested on a phantom solution. The effect of vigabatrin, a specific inhibitor of GABA transaminase, was measured in rat brain and GABA detection was performed in vivo in monkey brain using this procedure. Rats were split into two groups. In the control group, the rats had access to water and, in the other group (vigabatrin, VGB, rats), animals were allowed free access to drinking water containing vigabatrin. After 3 weeks of treatment, rats were anesthetized for in vivo NMR spectroscopy investigation. At the end of the experiment, brains were quickly removed, freeze-clamped and extracted with 4% perchloric acid. One part of the acid extract was used for GABA concentrations assessment by ion exchange chromatography with ninhydrin detection. The second was used for high-resolution NMR analysis. By chromatography measurements, the GABA concentration was 1.23+/-0.06 micromol/g for controls, while for vigabatrin-treated rats the GABA concentration was 4.89+/-1.60 micromol/g. The NMR in vivo results were closely correlated with the NMR ex vivo (r=0.99, p<0.01) and chromatography results (r=0.98, p<0.01). The correlation between ex vivo results and chromatography results was also high (r=0.99, p<0.001). This pulse sequence performed GABA editing from a 376 microl voxel located on the right basal ganglia area in a non-human primate brain. This in vivo GABA editing scheme can thus be proposed for accurate measurement of brain GABA concentrations.     

An investigation of human brain tumour lipids by high-resolution magic angle spinning 1H MRS and histological analysis

NMR-visible lipid signals detected in vivo by 1H MRS are associated with tumour aggression and believed to arise from cytoplasmic lipid droplets. High-resolution magic angle spinning (HRMAS) 1H MRS and Nile Red staining were performed on human brain tumour biopsy specimens to investigate how NMR-visible lipid signals relate to viable cells and levels of necrosis across different grades of glioma. Presaturation spectra were acquired from 24 adult human astrocytoma biopsy samples of grades II (8), III (2) and IV (14) using HRMAS 1H MRS and quantified using LCModel to determine lipid concentrations. Each biopsy sample was then refrozen, cryostat sectioned, and stained with Nile Red, to determine the number of lipid droplets and droplet size distribution, and with Haematoxylin and Eosin, to determine cell density and percentage necrosis. A strong correlation (R=0.92, P<0.0001) was found between the number of Nile Red-stained droplets and the approximately 1.3 ppm lipid proton concentration by 1H MRS. Droplet sizes ranged from 1 to 10 microm in diameter, and the size distribution was constant independent of tumour grade. In the non-necrotic biopsy samples, the number of lipid droplets correlated with cell density, whereas in the necrotic samples, there were greater numbers of droplets that showed a positive correlation with percentage necrosis. The correlation between 1H MRS lipid signals and number of Nile Red-stained droplets, and the presence of lipid droplets in the non-necrotic biopsy specimens provide good evidence that the in vivo NMR-visible lipid signals are cytoplasmic in origin and that formation of lipid droplets precedes necrosis.     

An assessment of the effects of sample ischaemia and spinning time on the metabolic profile of brain tumour biopsy specimens as determined by high-resolution magic angle spinning (1)H NMR

High-resolution magic angle spinning (HRMAS) (1)H NMR of biopsy tissue provides a biochemical profile that has potential diagnostic and prognostic value, and can aid interpretation of the lower-resolution (1)H-NMR spectra obtained in vivo. However, the biochemical profile obtained may be affected by experimental factors such as a period of ischaemia before snap-freezing of the biopsy tissue for subsequent analysis and the mechanical stress of the spinning procedure of HRMAS itself. We have used normal rat brain cortex as a 'gold standard', either funnel-frozen or deliberately allowed to become ischaemic for set periods of time before snap-freezing, to quantitatively investigate these two effects. In addition, we have compared biochemical changes that occur in normal rat brain during HRMAS (spun continuously at 5 kHz for 4 h at 4 degrees C as could be required for a two-dimensional acquisition) with those that occur in biopsy samples from low-grade and high-grade adult human astrocytomas, during the same HRMAS procedure. Significant changes due to delayed initial sample freezing were noted in metabolites associated with glycolysis (alanine, glucose and lactate), as expected. However, for the funnel-frozen rat tissue at 4 degrees C, there were even more significant changes, which appear to be the result of extended spinning at 5 kHz. In particular, the 18% total creatine increase observed is unlikely to be de novo synthesis of creatine. More likely, the asymptotic exponential increase in creatine suggests an exponential release of an NMR-invisible bound creatine store as a result of tissue damage from mechanical stress of sample spinning. Overall, it appears that tissue ischaemia during biopsy excision and delays in snap-freezing may have less significant effects on metabolite profile than the prolonged spinning times required for two-dimensional HRMAS, and this must be accounted for when results are being interpreted.     

A quantitative comparison of metabolite signals as detected by in vivo MRS with ex vivo 1H HR‐MAS for childhood brain tumours

¹H MRS provides a powerful method for investigating tumour metabolism by allowing the measurement of metabolites in vivo. Recently, the technique of ¹H high‐resolution magic angle spinning (HR‐MAS) has been shown to produce high‐quality data, allowing the accurate measurement of many metabolites present in unprocessed biopsy tissue. The purpose of this study was to evaluate the agreement between the techniques of in vivo MRS and ex vivo HR‐MAS for investigating childhood brain tumours. Short‐TE (30 ms), single‐voxel, in vivo MRS was performed on 16 paediatric patients with brain tumours at 1.5 T. A frozen biopsy sample was available for each patient. HR‐MAS was performed on the biopsy samples, and metabolite quantities were determined from the MRS and HR‐MAS data using the LCModel? and TARQUIN algorithms, respectively. Linear regression was performed on the metabolite quantities to asses the agreement between MRS and HR‐MAS. Eight of the 12 metabolite quantities were found to correlate significantly (P < 0.05). The four worst correlating metabolites were aspartate, scyllo‐inositol, glycerophosphocholine and N‐acetylaspartate, and, except for glycerophosphocholine, this error was reflected in their higher Cramer–Rao lower bounds (CRLBs), suggesting that low signal‐to‐noise was the greatest source of error for these metabolites. Glycerophosphocholine had a lower CRLB implying that interference with phosphocholine and choline was the most significant source of error. The generally good agreement observed between the two techniques suggests that both MRS and HR‐MAS can be used to reliably estimate metabolite quantities in brain tumour tissue and that tumour heterogeneity and metabolite degradation do not have an important effect on the HR‐MAS metabolite profile for the tumours investigated. HR‐MAS can be used to improve the analysis and understanding of MRS data. Copyright © 2009 John Wiley & Sons, Ltd.     

A comparative study of automatic techniques for ocular artifact reduction in spontaneous EEG signals based on clinical target variables: a simulation case

Eye movement artifacts represent a critical issue for quantitative electroencephalography (EEG) analysis and a number of mathematical approaches have been proposed to reduce their contribution in EEG recordings. The aim of this paper was to objectively and quantitatively evaluate the performance of ocular filtering methods with respect to spectral target variables widely used in clinical and functional EEG studies. In particular the following methods were applied: regression analysis and some blind source separation (BSS) techniques based on second-order statistics (PCA, AMUSE and SOBI) and on higher-order statistics (JADE, INFOMAX and FASTICA). Considering blind source decomposition methods, a completely automatic procedure of BSS based on logical rules related to spectral and topographical information was proposed in order to identify the components related to ocular interference. The automatic procedure was applied in different montages of simulated EEG and electrooculography (EOG) recordings: a full montage with 19 EEG and 2 EOG channels, a reduced one with only 6 EEG leads and a third one where EOG channels were not available. Time and frequency results in all of them indicated that AMUSE and SOBI algorithms preserved and recovered more brain activity than the other methods mainly at anterior regions. In the case of full montage: (i) errors were lower than 5% for all spectral variables at anterior sites; and (ii) the highest improvement in the signal-to-artifact (SAR) ratio was obtained up to 40dB at these anterior sites. Finally, we concluded that second-order BSS-based algorithms (AMUSE and SOBI) provided an effective technique for eye movement removal even when EOG recordings were not available or when data length was short.     

A comparative study of short‐ and long‐TE 1H MRS at 3 T for in vivo detection of 2‐hydroxyglutarate in brain tumors

2‐Hydroxyglutarate (2HG) is produced in gliomas with mutations of isocitrate dehydrogenase (IDH) 1 and 2. The ¹H resonances of the J‐coupled spins of 2HG are extensively overlapped with signals from other metabolites. Here, we report a comparative study at 3 T of the utility of the point‐resolved spectroscopy sequence with a standard short TE (35 ms) and a long TE (97 ms), which had been theoretically designed for the detection of the 2HG 2.25‐ppm resonance. The performance of the methods is evaluated using data from phantoms, seven healthy volunteers and 22 subjects with IDH‐mutated gliomas. The results indicate that TE = 97 ms provides higher detectability of 2HG than TE = 35 ms, and that this improved capability is gained when data are analyzed with basis spectra that include the effects of the volume localizing radiofrequency and gradient pulses. Copyright © 2013 John Wiley & Sons, Ltd.     

Multiple clusters of A(H1N1)pdm09 virus circulating in severe cases of influenza during the 2010–2011 season: A phylogenetic and molecular analysis of the neuraminidase gene

The molecular characterization of circulating influenza A viruses is crucial to detect mutations potentially involved in increased virulence, drug resistance and immune escape. A molecular and phylogenetic analysis of A(H1N1)pdm09 neuraminidase (NA) gene sequences from different patient categories defined according to the severity of influenza infection were analyzed. A total of 126 influenza A(H1N1)pdm09 positive samples from patients with severe infections in comparison with those with moderate and mild infections was performed in Lombardy (Northern Italy, nearly 10 million inhabitants) during the 2010–2011 season. NA sequences included in this study segregated into five distinct clusters. Nineteen amino acid substitutions were detected exclusively in NA sequences of viruses identified in patients with severe or moderate influenza infection. Three of them (F74S, S79P, E287K) were observed in virus strains with the 222G/N hemagglutinin mutation. None of NA sequences under study had mutations related to the resistance to the NA inhibitors. Four out of 126 (3.2%) NA sequences from patients with severe infection lost a N‐linked glycosylation site due to the change from N to K at residue 386. Two additional N‐linked glycosylation sites in the NA stalk region (residues 42 and 44) were found in 12 (9.5%) NA sequences. Sporadic NA mutations were detected in NA viral sequences from critically ill patients, and no variants with reduced sensitivity to NA inhibitors were observed either in treated or untreated patients. J. Med. Virol. 85: 944–952, 2013. © 2013 Wiley Periodicals, Inc.     

Elucidating the role of 5-HT(1A) and 5-HT(7) receptors on 8-OH-DPAT-induced behavioral recovery after experimental traumatic brain injury

8-OH-DPAT is a 5-HT(1A/7) receptor agonist that enhances behavioral recovery after traumatic brain injury (TBI). This study is a first attempt to decipher whether the benefits induced by 8-OH-DPAT after TBI are mediated by 5-HT(1A) or 5-HT(7) receptors. A single i.p. injection of 8-OH-DPAT (0.5 mg/kg) alone or co-administered with either the 5-HT(1A) or 5-HT(7) receptor antagonists WAY 100635 (0.5 mg/kg) or SB 269970 HCl (2.0 mg/kg), respectively, or vehicle control (1.0 mL/kg) was given 15 min after cortical impact or sham injury. Function was assessed by established motor and cognitive tests. No difference in motor performance was observed among the TBI groups. Spatial acquisition was enhanced, relative to vehicle controls, by 8-OH-DPAT alone and when co-administered with WAY 100635, but not when combined with SB 269970 HCl. These data imply that 5-HT(1A) receptor antagonism does not abate the 8-OH-DPAT-induced cognitive benefits, but 5-HT(7) receptor antagonism does, which suggests that the 8-OH-DPAT-induced benefits in this single administration paradigm may be mediated more by 5-HT(7) versus 5-HT(1A) receptors. Evaluation of a specific 5-HT(7) receptor agonist will further elucidate the contribution of 5-HT(1A) and 5-HT(7) receptors on behavioral recovery conferred by acute 8-OH-DPAT treatment after TBI.