In vivo proton MR spectroscopy of human gliomas: definition of metabolic coordinates for multi-dimensional classification

SPectroscOpic prediction of bRain Tumours (SPORT): study protocol of a prospective imaging trial

Background

The revised 2016 WHO-Classification of CNS-tumours now integrates molecular information of glial brain tumours for accurate diagnosis as well as for the development of targeted therapies. In this prospective study, our aim is to investigate the predictive value of MR-spectroscopy in order to establish a solid preoperative molecular stratification algorithm of these tumours. We will process a 1H MR-spectroscopy sequence within a radiomics analytics pipeline.

Methods

Patients treated at our institution with WHO-Grade II, III and IV gliomas will receive preoperative anatomical (T2- and T1-weighted imaging with and without contrast enhancement) and proton MR spectroscopy (MRS) by using chemical shift imaging (MRS) (5 × 5 × 15 mm³ voxel size). Tumour regions will be segmented and co-registered to corresponding spectroscopic voxels. Raw signals will be processed by a deep-learning approach for identifying patterns in metabolic data that provides information with respect to the histological diagnosis as well patient characteristics obtained and genomic data such as target sequencing and transcriptional data.

Discussion

By imaging the metabolic profile of a glioma using a customized chemical shift 1H MR spectroscopy sequence and by processing the metabolic profiles with a machine learning tool we intend to non-invasively uncover the genetic signature of gliomas. This work-up will support surgical and oncological decisions to improve personalized tumour treatment.

Trial registration

This study was initially registered under another name and was later retrospectively registered under the current name at the German Clinical Trials Register (DRKS) under DRKS00019855.

The integrative metabolomic-transcriptomic landscape of glioblastome multiforme

The purpose of this study was to map the landscape of metabolic-transcriptional alterations in glioblastoma multiforme. Omic-datasets were acquired by metabolic profiling (1D-NMR spectroscopy n=33 Patient) and transcriptomic profiling (n=48 Patients). Both datasets were analyzed by integrative network modeling. The computed model concluded in four different metabolic-transcriptomic signatures containing: oligodendrocytic differentiation, cell-cycle functions, immune response and hypoxia. These clusters were found being distinguished by individual metabolism and distinct transcriptional programs. The study highlighted the association between metabolism and hallmarks of oncogenic signaling such as cell-cycle alterations, immune escape mechanism and other cancer pathway alterations. In conclusion, this study showed the strong influence of metabolic alterations in the wide scope of oncogenic transcriptional alterations.

Magnetic resonance spectroscopy in intracranial tumours of glial origin

BACKGROUND AND PURPOSE

To determine in vivo magnetic resonance spectroscopy (MRS) characteristics of intracranial glial tumours and to assess MRS reliability in glioma grading and discrimination between different histopathological types of tumours.

MATERIAL AND METHODS

Analysis of spectra of 26 patients with glioblastomas, 6 with fibrillary astrocytomas, 4 with anaplastic astrocytomas, 2 with pilocytic astrocytoma, 3 with oligodendrogliomas, 3 with anaplastic oligodendrogliomas and 17 control spectra taken from healthy hemispheres.

RESULTS

All tumours' metabolite ratios, except for Cho/Cr in fibrillary astrocytomas (p = 0.06), were statistically significantly different from the control. The tumours showed decreased Naa and Cr contents and a high Cho signal. The Lac-Lip signal was high in grade III astrocytomas and glioblastomas. Reports that Cho/Cr ratio increases with glioma's grade whereas Naa/Cr decreases were not confirmed. Anaplastic astrocytomas compared to grade II astrocytomas had a statistically significantly greater mI/Cr ratio (p = 0.02). In pilocytic astrocytomas the Naa/Cr value (2.58 ± 0.39) was greater, whilst the Cho/Naa ratio was lower (2.14 ± 0.64) than in the other astrocytomas. The specific feature of oligodendrogliomas was the presence of glutamate/glutamine peak Glx. However, this peak was absent in two out of three anaplastic oligodendrogliomas. Characteristically, the latter tumours had a high Lac-Lip signal.

CONCLUSIONS

MRS in vivo cannot be used as a reliable method for glioma grading. The method is useful in discrimination between WHO grade I and WHO grade II astrocytomas as well as oligodendrogliomas from other gliomas.

Improving the classification of brain tumors in mice with perturbation enhanced (PE)-MRSI

Classifiers based on statistical pattern recognition analysis of MRSI data are becoming important tools for the non-invasive diagnosis of human brain tumors. Here we investigate the potential interest of perturbation-enhanced MRSI (PE-MRSI), in this case acute hyperglycemia, for improving the discrimination between mouse brain MRS patterns of glioblastoma multiforme (GBM), oligodendroglioma (ODG), and non-tumor brain parenchyma (NT). Six GBM-bearing mice and three ODG-bearing mice were scanned at 7 Tesla by PRESS-MRSI with 12 and 136 ms echo-time, during euglycemia (Eug) and also during induced acute hyperglycemia (Hyp), generating altogether four datasets per animal (echo time + glycemic condition): 12Eug, 136Eug, 12Hyp, and 136Hyp. For classifier development all spectral vectors (spv) selected from the MRSI matrix were unit length normalized (UL2) and used either as a training set (76 GBM spv, four mice; 70 ODG spv, two mice; 54 NT spv) or as an independent testing set (61 GBM spv, two mice; 31 ODG, one mouse; 23 NT spv). All Fisher's LDA classifiers obtained were evaluated as far as their descriptive performance-correctly classified cases of the training set (bootstrapping)-and predictive accuracy-balanced error rate of independent testing set classification. MRSI-based classifiers at 12Hyp were consistently more efficient in separating GBM, ODG, and NT regions, with overall accuracies always >80% and up to 95-96%; remaining classifiers were within the 48-85% range. This was also confirmed by user-independent selection of training and testing sets, using leave-one-out (LOO). This highlights the potential interest of perturbation-enhanced MRSI protocols for improving the non-invasive characterization of preclinical brain tumors.

Clinical pitfalls related to short and long echo times in cerebral MR spectroscopy

MR-spectroscopy (MRS) is a multiparameter diagnostic tool and modification of each parameter results in spectrum morphology changes. In particular, changing the echo time (TE) represents a useful tool to highlight different diagnostic elements, but also has significant impact on the spectrum morphology. Diagnostic errors can result if the role of TE is not properly considered. This article reviews the four most common TE-related pitfalls of MRS interpretation. Clinical practical methods to avoid such pitfalls are also suggested.

Role of proton magnetic resonance spectroscopy in differentiating oligodendrogliomas from astrocytomas

BACKGROUND AND PURPOSE

Preoperative differentiation of astrocytomas from oligodendrogliomas is clinically important, as oligodendrogliomas are more sensitive to chemotherapy. The purpose of this study was to assess the role of proton magnetic resonance spectroscopy in distinguishing astrocytomas from oligodendrogliomas.

METHODS

Forty-six patients [astrocytomas (n= 17) and oligodendrogliomas (n= 29)] underwent magnetic resonance imaging and multi voxel proton magnetic resonance spectroscopic imaging before treatment. Peak areas for N-acetylaspartate (NAA), creatine (Cr), choline (Cho), myo-inositol (mI), glutamate/glutamine (Glx), and lipids + lactate (Lip+Lac) were analyzed from voxels that exhibited hyperintensity on fluid-attenuated inversion recovery images and were normalized to Cr from each voxel. The average metabolite/Cr ratios from these voxels were then compared between astrocytomas and oligodendrogliomas. Receiver-operating curve analyses were used as measures of differentiation accuracy of metabolite ratios. A threshold value for a metabolite ratio was estimated by maximizing the sum of sensitivity and specificity.

RESULTS

A significant difference in mI/Cr was observed between astrocytomas and oligodendrogliomas (.50 +/- .18 vs. 0.66 +/- 0.20, P < .05). Using a threshold value of .56 for mI/Cr ratio, it was possible to differentiate oligodendrogliomas from astrocytomas with a sensitivity of 72.4% and specificity of 76.4%.

CONCLUSION

These results suggest that mI/Cr might aid in distinguishing oligodendrogliomas from astrocytomas.

Short echo time MR spectroscopy of brain tumors: grading of cerebral gliomas by correlation analysis of normalized spectral amplitudes

PURPOSE

To process single voxel spectra of low- and high-grade gliomas. To propose correlation analysis of the scatter plots of normalized spectral amplitudes as a pattern recognition tool for the classification (grading) of brain tumors. To propose a spectrum processing approach that improves the differentiation of proton spectra with dominating macromolecule and lipid peaks.

MATERIALS AND METHODS

LCModel was used to process spectra. Mean metabolite concentrations and mean normalized spectra were obtained for normal white matter and for gliomas. The mean spectra of macromolecules and lipids (ML) in the range 1.4-0.9 ppm, and mean difference spectra (DS) without ML and lactate were computed. Correlation analysis of the scatter plot of the patient and mean normalized spectral amplitudes and dispersion of the scatter plot points were used for classification and grading of tumors.

RESULTS

It was found advantageous to perform the classifications using DS spectra. The shape of ML spectrum and concentration of tCr seem to be a good markers for glioma grade.

CONCLUSION

Combining a qualitative comparison of the patient and mean DS spectra of the tumors using correlation analysis of normalized spectra amplitudes with a quantitative comparison of metabolite concentrations is a powerful tool in studying brain lesions.

Molecular classification of brain tumor biopsies using solid-state magic angle spinning proton magnetic resonance spectroscopy and robust classifiers

Brain tumors are one of the leading causes of death in adults with cancer; however, molecular classification of these tumors with in vivo magnetic resonance spectroscopy (MRS) is limited because of the small number of metabolites detected. In vitro MRS provides highly informative biomarker profiles at higher fields, but also consumes the sample so that it is unavailable for subsequent analysis. In contrast, ex vivo high-resolution magic angle spinning (HRMAS) MRS conserves the sample but requires large samples and can pose technical challenges for producing accurate data, depending on the sample testing temperature. We developed a novel approach that combines a two-dimensional (2D), solid-state, HRMAS proton (1H) NMR method, TOBSY (total through-bond spectroscopy), which maximizes the advantages of HRMAS and a robust classification strategy. We used approximately 2 mg of tissue at -8 degrees C from each of 55 brain biopsies, and reliably detected 16 different biologically relevant molecular species. We compared two classification strategies, the support vector machine (SVM) classifier and a feed-forward neural network using the Levenberg-Marquardt back-propagation algorithm. We used the minimum redundancy/maximum relevance (MRMR) method as a powerful feature-selection scheme along with the SVM classifier. We suggest that molecular characterization of brain tumors based on highly informative 2D MRS should enable us to type and prognose even inoperable patients with high accuracy in vivo.

Identification and characterisation of childhood cerebellar tumours by in vivo proton MRS

(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.

Metabolic fingerprinting as a diagnostic tool

Within the framework of systems biology, functional analyses at all 'omic levels have seen an intense level of activity during the first decade of the twenty-first century. These include genomics, transcriptomics, proteomics, metabolomics and lipidomics. It could be said that metabolomics offers some unique advantages over the other 'omics disciplines and one of the core approaches of metabolomics for disease diagnostics is metabolic fingerprinting. This review provides an overview of the main metabolic fingerprinting approaches used for disease diagnostics and includes: infrared and Raman spectroscopy, Nuclear magnetic resonance (NMR) spectroscopy, followed by an introduction to a wide range of novel mass spectrometry-based methods, which are currently under intense investigation and developmental activity in laboratories worldwide. It is hoped that this review will act as a springboard for researchers and clinicians across a wide range of disciplines in this exciting era of multidisciplinary and novel approaches to disease diagnostics.

Characterization of Glioblastoma by Contrast-Enhanced Flair Sequences

The tissues placed on the edge of a glioblastoma's necrotic cavities are more vascularized than other pseudocystic central nervous system (CNS) tumours, both benign and malignant. The post-contrast enhancement is greater in Fluid-Attenuated Inversion-Recovery (FLAIR) images than in Spin Echo T1-weighted (SE T1w) sequences above all in the CNS tissues with a low concentration of gadolinium. The purpose of this study was to distinguish pseudocystic glioblastomas from other cystic CNS tumors by comparing post-contrast pseudocystic rim enhancement in FLAIR and SE T1-w magnetic resonance (MR) images. We investigated 32 extensive sets of MR images relating to histologically diagnosed pseudocystic CNS tumors; 14/32 were glioblastoma. Fast Spin Echo (FSE) T2-weighted and Proton Density, SE T1w and FSE FLAIR sequences were acquired in all the studies. After contrast media administration SE T1w and FLAIR sequences were acquired. In post-contrast T1w SE and T2w FLAIR acquisitions, pseudocyst rim enhancement was evaluated assigning scores: 4 = rim enhancement completely surrounds perimeter; 3 = rim enhancement in ≥50% of perimeter; 2 ? rim in < 50% of perimeter; 1 = rim enhancement absent. Mean scores were calculated and the results were compared with statistical methods (Student's t test) for glioblastomas and all other tumors. Moreover differences between FLAIR and SE scores was assessed in each patient. If the difference was 0 glioblastoma was assumed, if the difference was ≥ 1 another tumor was assumed; the sensitivity and specificity of this diagnosis compared to the histological diagnosis were assessed. Mean Tl-weighted SE scores did not differ in glioblastomas and other tumors. FLAIR scores in glioblastomas were less than half those of other tumors (p < 0.005). Glioblastoma diagnosis based on score difference identified 13 true positives (glioblastomas), 16 true negatives (non glioblastomas), two false positives and two false negatives. The sensitivity for glioblastoma was 86.7% and the specificity was 94.1%. Comparison of post-contrast rim enhancement in T1w SE and FLAIR sequences distinguishes glioblastomas from other pseudocystic CNS tumors, assisting the differential diagnosis of glioblastomas, that in many cases are not distinguishable from metastases even with advanced MR techniques.

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.

Arterial spin-labeling and MR spectroscopy in the differentiation of gliomas

BACKGROUND AND PURPOSE

Noninvasive grading of gliomas remains a challenge despite its important role in the prognosis and management of patients with intracranial neoplasms. In this study, we evaluated the ability of cerebral blood flow (CBF)-guided voxel-by-voxel analysis of multivoxel proton MR spectroscopic imaging ((1)H-MRSI) to differentiate low-grade from high-grade gliomas.

MATERIALS AND METHODS

A total of 35 patients with primary gliomas (22 high grade and 13 low grade) underwent continuous arterial spin-labeling perfusion-weighted imaging (PWI) and (1)H-MRSI. Different regions of the gliomas were categorized as "hypoperfused," "isoperfused," and "hyperperfused" on the basis of the average CBF obtained from contralateral healthy white matter. (1)H-MRSI indices were computed from these regions and compared between low- and high-grade gliomas. Using a similar approach, we applied a subgroup analysis to differentiate low- from high-grade oligodendrogliomas because they show different physiologic and genetic characteristics.

RESULTS

Cho(glioma (G)/white matter (WM)), Glx(G/WM), and Lip+Lac(G)/Cr(WM) were significantly higher in the "hyperperfused" regions of high-grade gliomas compared with low-grade gliomas. Cho(G/WM) and Lip+Lac(G)/Cr(WM) were also significantly higher in the "hyperperfused" regions of high-grade oligodendrogliomas. However, metabolite ratios from the "hypoperfused" or "isoperfused" regions did not exhibit any significant differences between high-grade and low-grade gliomas.

CONCLUSION

The results suggest that (1)H-MRSI indices from the "hyperperfused" regions of gliomas, on the basis of PWI, may be helpful in distinguishing high-grade from low-grade gliomas including oligodendrogliomas.

Dissecting drug and vehicle metabolic effects in rats by a metabonomic approach

A combined application of high resolution (1)H NMR spectroscopy and multivariate statistical techniques focused on establishing a consistent statistical approach to metabonomic studies was tested. The data reduction, which is preliminary to the application of multivariate analysis to NMR spectra, was carried out by means of two complementary methods: pure Pattern Recognition (PR) and Assigned Signal Analysis (ASA). The simultaneous use of both approaches allowed us to obtain additional information in the analysis of metabonomic data, compared to the use of PR alone. This additional information consists in the possibility of a biochemical interpretation of the effects induced by treatment with xenobiotics, such as drugs or drug vehicles, on the metabolic networks of the systems under investigation. This approach allowed us to ascertain that a single-dose treatment with ST1959 vehicled by Sesame oil affects the production of hepatic glucose associated to an increment of the amino acid ketogenic process.

A metabolomics perspective of human brain tumours

During the past decade or so, a wealth of information about metabolites in various human brain tumour preparations (cultured cells, tissue specimens, tumours in vivo) has been accumulated by global profiling tools. Such holistic approaches to cellular biochemistry have been termed metabolomics. Inherent and specific metabolic profiles of major brain tumour cell types, as determined by proton nuclear magnetic resonance spectroscopy ((1)H MRS), have also been used to define metabolite phenotypes in tumours in vivo. This minireview examines the recent advances in the field of human brain tumour metabolomics research, including advances in MRS and mass spectrometry technologies, and data analysis.

Application of proton NMR spectroscopy in the study of lipid metabolites in a rat hepatocarcinogenesis model

Liver cancer is one of the most common cancers worldwide. Altered lipid metabolism in the liver is a key feature of developing liver nodules and tumors. Methods of analysis vary from the most sophisticated chromatography to the in vivo nuclear magnetic resonance (NMR) spectroscopy. In this study, we present a systematic method for the identification and quantitation of signature signals from lipid metabolites using 1D NMR proton spectroscopy. We assessed lipid metabolites in an epigenetic rat hepatocarcinogenesis model induced by treatment with a choline-deficient diet (CDAA, choline-deficient l-amino acid defined) over a period of 1 year, from the formation of steatosis, to the development of nodules and adenomas. A comparable choline-sufficient (CSAA) diet was used for the controls. The resonances of the methylene protons of the glycerol backbone in phospholipids were used to quantify the total concentration of such compounds. CDAA rat livers were found to have significantly higher levels of phospholipids, when compared to CSAA, throughout the entire carcinogenesis period. The tri-methyl protons of choline compounds serves to quantify total choline, and the vinyl and bis-allyl proton resonances can be used to not only quantify fatty acid concentrations but also to probe the number of double bonds in a fatty acid moiety. Early stages of carcinogenesis indicate a lower degree of double bonds in fatty acyl containing compounds in CDAA rat livers, when compared to CSAA. The results of this study are in agreement with those previously published in the literature on other rat hepatocarcinogenesis models.

Minute glioma identified by 3.0 Tesla magnetic resonance spectroscopy--case report

A 33-year-old man presented with a minute tumor incidentally detected by magnetic resonance (MR) imaging screening. 1.5 Tesla MR spectroscopy indicated normal brain tissue whereas 3.0 Tesla MR spectroscopy indicated neoplasm. The tumor was completely resected. The histological diagnosis was fibrillary astrocytoma. Minute glioma, measuring less than 15 mm in diameter on MR imaging, can be completely resected, resulting in a good prognosis. 3.0 Tesla MR spectroscopy can establish the diagnosis in the early stage of glioma.

Proton magnetic resonance spectroscopic evaluation of brain tumor metabolism

Magnetic resonance imaging (MRI) is the neuroimaging method of choice for the noninvasive monitoring of patients with brain tumors due to the enormous amount of information it yields regarding the morphologic features of the lesion and surrounding parenchyma. Over the past decade, proton magnetic resonance spectroscopy (1H-MRS), which uses the same technology as MRI and can be performed during a routine clinical imaging examination, has been used to glean information about the metabolic status of the brain. Accurate interpretation of 1H-MRS data from individual patients requires an understanding of the various techniques for acquiring the data, the physiologic basis of the metabolic signatures obtained from different types of tumors, and the specificity of the technique. This review covers the basic physics of 1H-MRS, the spectral and physiological characteristics of the metabolites that are typically measured in various types of brain tumors, and the clinical utility of 1H-MRS with respect to diagnosis, therapeutic planning, and the assessment of response to treatment.

Classification of brain tumours using short echo time 1H MR spectra

The purpose was to objectively compare the application of several techniques and the use of several input features for brain tumour classification using Magnetic Resonance Spectroscopy (MRS). Short echo time 1H MRS signals from patients with glioblastomas (n = 87), meningiomas (n = 57), metastases (n = 39), and astrocytomas grade II (n = 22) were provided by six centres in the European Union funded INTERPRET project. Linear discriminant analysis, least squares support vector machines (LS-SVM) with a linear kernel and LS-SVM with radial basis function kernel were applied and evaluated over 100 stratified random splittings of the dataset into training and test sets. The area under the receiver operating characteristic curve (AUC) was used to measure the performance of binary classifiers, while the percentage of correct classifications was used to evaluate the multiclass classifiers. The influence of several factors on the classification performance has been tested: L2- vs. water normalization, magnitude vs. real spectra and baseline correction. The effect of input feature reduction was also investigated by using only the selected frequency regions containing the most discriminatory information, and peak integrated values. Using L2-normalized complete spectra the automated binary classifiers reached a mean test AUC of more than 0.95, except for glioblastomas vs. metastases. Similar results were obtained for all classification techniques and input features except for water normalized spectra, where classification performance was lower. This indicates that data acquisition and processing can be simplified for classification purposes, excluding the need for separate water signal acquisition, baseline correction or phasing.

1H-MRSI of radiation effects in normal-appearing white matter: dose-dependence and impact on automated spectral classification

PURPOSE

To identify radiation-induced changes in healthy white-matter spectra in the first six months following radiotherapy, and assess the impact of these changes on an automated algorithm for detecting spectral abnormalities.

MATERIALS AND METHODS

1H-MRSI was performed on 10 patients with grade IV gliomas who were to undergo radiation therapy. Choline (Cho), creatine (Cr), and N-acetylaspartate (NAA) ratios were studied as a function of dose and time. The impact of these spectral changes on a spectral analysis algorithm was evaluated.

RESULTS

The Cho/NAA ratios rose to values of 0.66 +/- 0.15, 0.75 +/- 0.21, and 0.73 +/- 0.15 two months after therapy, compared to immediate post-therapy values of 0.56 +/- 0.15, 0.60 +/- 0.16, and 0.61 +/- 0.15 for the < 25, 25-50, and > 50 Gy dose groups, respectively. These maxima were followed by a dose-dependent recovery. A similar trend was found in the Cho/Cr ratio. The automated spectral analysis system incorporated the changing Cho/NAA ratio into a global redefinition of healthy tissue, but did not account for dose-dependent spatial variations in Cho/NAA ratios.

CONCLUSION

Radiation significantly alters the spectra of healthy tissues in the first six months after radiotherapy. This suggests that the radiation dose distribution should be considered during analysis of post-therapy spectra.

A Chemometric Approach for Brain Tumor Classification Using Magnetic Resonance Imaging and Spectroscopy

A new classification approach was developed to improve the noninvasive diagnosis of brain tumors. Within this approach, information is extracted from magnetic resonance imaging and spectroscopy data, from which the relative location and distribution of selected tumor classes in feature space can be calculated. This relative location and distribution is used to select the best information extraction procedure, to identify overlapping tumor classes, and to calculate probabilities of class membership. These probabilities are very important, since they provide information about the reliability of classification and might provide information about the heterogeneity of the tissue. Classification boundaries were calculated by setting thresholds for each investigated tumor class, which enabled the classification of new objects. Results on histopathologically determined tumors are excellent, demonstrated by spatial maps showing a high probability for the correctly identified tumor class and, moreover, low probabilities for other tumor classes.

Gliomatosis cerebri: a review of 22 cases

OBJECTIVE

Gliomatosis cerebri is an enigmatic diffuse brain neoplasm whose prognosis is grim. We reviewed data for patients with gliomatosis who were treated at the University of California, San Francisco, during a 10-year period. Our focus was on presentation, radiological and pathological features, and outcomes.

METHODS

We reviewed hospital and clinic records and magnetic resonance imaging scans for 22 patients with gliomatosis. The diagnosis was based on magnetic resonance imaging findings and tissue confirmation for all patients. Seven patients also underwent magnetic resonance spectroscopy. Eleven patients were male (50%), and the median age at presentation was 49 years (range, 7-79 yr).

RESULTS

Kaplan-Meier analysis demonstrated median lengths of survival as follows: no treatment, 1 month (n = 4); radiotherapy alone, 28 months (95% confidence interval, 5-51 mo; n = 13); radiotherapy followed by chemotherapy, two patients, alive at 28 and 104 months; radiotherapy and chemotherapy simultaneously, three patients, one alive at 18 months and the others dead at 7 and 9 months. There was no significant difference between radiotherapy alone and radiotherapy combined with chemotherapy (P = 0.69). Karnofsky Performance Scale scores of >/=70 and grade were both significantly related to length of survival in univariate analyses (P < 0.05); these correlations were confirmed in the multivariate analysis, although the small numbers of patients and deaths precluded reliable interpretation.

CONCLUSION

Although the small number of patients in our study and its retrospective nature preclude definitive conclusions regarding the utility of treatment, our findings suggest that biopsies are useful not only for diagnosis but also for prediction of the length of survival.

Differential MRI diagnosis between brain abscesses and necrotic or cystic brain tumors using the apparent diffusion coefficient and normalized diffusion-weighted images

Magnetic Resonance Diffusion-Weighted Imaging (DWI) has been reported to be helpful for the differential diagnosis between abscesses and cystic/necrotic brain tumors. However the number of patients is still limited, and the sensitivity and specificity of the method remain to be confirmed. The primary purpose of this study was to investigate a larger sample of patients, all investigated under the same experimental conditions, in order to obtain statistically significant data. Moreover, there is no consensus about the appropriate values of b required to use to make an accurate diagnosis from DWI. The secondary purpose of this study was to determine the discriminating threshold b values for raw diffusion-weighted images and for normalized diffusion-weighted images. On the basis of 14 abscesses, 10 high-grade gliomas and 2 metastases, we show that the calculation of accurate Apparent Diffusion Coefficient (ADC) values gives a specificity rate of 100%. Without ADC calculation, we show that image normalization is required to make an accurate differential diagnosis, and we highlight the ability of DWI to discriminate between brain abscesses and cystic/necrotic brain tumors using normalized signal intensity at lower b values (503 s/mm(2)) than usual.

Could assessment of glioma methylene lipid resonance by in vivo (1)H-MRS be of clinical value?

The potential clinical role of in vivo (1)H-MRS ((1)H-magnetic resonance spectroscopy) lipid methylene resonance measurements of human glioma has been assessed. 20 patients, 14 with low grade and 6 with high grade gliomas have been investigated using single voxel (1)H-MRS. Three of the low grade group had undergone transformation by clinical and imaging criteria. Short echo time (TE=20 ms, TR=2500 ms) single voxel Stimulated Echo Acquisition (STEAM) spectra with (acquisitions=64) and without (acquisitions=4) water suppression were acquired. Additionally, T(1) weighted (T(1)W) water spectra (TE=20 ms, TR=888 ms) were acquired pre- and post-injection of Gd-DTPA (0.2 mmol x kg(-1)). The T(1)W water spectra were used to determine the water proton enhancement occurring within the spectroscopic voxel. The enhancement expressed as a percentage was compared with the lipid methylene peak. All the high grade tumours had significantly higher levels of lipid than low grade tumours (p=0.002). Low grade tumours had significantly less water proton enhancement than transformers (p=0.04) and high grade tumours (p=0.001). The lipid methylene signal correlated strongly with the voxel water enhancement (r(2)=0.74, p<0.0001). The data support the view that the spectroscopically detected lipid methylene signal may be a useful criterion in grading glioma. The correlation of the lipid methylene signal with blood-brain barrier breakdown suggests that detection of a previously absent (1)H-MRS lipid methylene signal in low grade tumours might be an early indicator of transformation.

Preoperative proton-MR spectroscopy of gliomas--correlation with quantitative nuclear morphology in surgical specimen

A comparison between data from proton-MR spectroscopy (1HMRS) and quantitative histomorphology of tumor cell nuclei in gliomas has not been reported up to now. Therefore, the question must be answered, if there are any significant correlations between histomorphology of gliomas and quantitative data from 1HMRS concerning tissue metabolites. Surgical glioma specimen (glioblastomas, astrocytomas, oligodendrogliomas) from 46 patients with tumor grades II-IV according to WHO have been evaluated by means of a digital image analysis system using Ki-67-immunostained paraffin sections. Nuclear density, Ki-67-proliferation index, nuclear area and shape variables (roundness factor, Fourier-amplitudes) have been determined from 200 randomly selected tumor cell nuclei in each tumor specimen. These data have been correlated with preoperative data from 1HMRS. A positive correlation between Fourier-amplitudes, choline peak and lipide peak was observed, as well as a negative correlation between these variables and the nuclear roundness factor. This result indicates higher choline and lipide peaks with increasing irregularity of nuclear outlines. Proliferation index Ki-67 was positively correlated with the lipide peak, nuclear density showed a positive correlation with the choline peak. Glioblastomas (n = 29) showed an additional positive correlation between mean nuclear size and total creatine. Anaplastic gliomas (n = 12) showed a positive correlation between lactate peak and the standard deviation of the nuclear roundness factor. Further multivariate analyses have shown, that for the present collective of 46 cases, histometric variables have a higher significance than spectroscopic data for the differentiation of the different tumor grades. These results verify a significant correlation between preoperative data from 1HMRS and histomorphology of tumor cell nuclei in gliomas, supporting the biological significance of both histomorphometry and 1HMRS for the evaluation of these tumors.

Metabolic profiles of human brain tumors using quantitative in vivo 1H magnetic resonance spectroscopy

Proton spectroscopy can noninvasively provide useful information on brain tumor type and grade. Short- (30 ms) and long- (136 ms) echo time (TE) (1)H spectra were acquired from normal white matter (NWM), meningiomas, grade II astrocytomas, anaplastic astrocytomas, glioblastomas, and metastases. Very low myo-Inositol ([mI]) and creatine ([Cr]) were characteristic of meningiomas, and high [mI] characteristic of grade II astrocytomas. Tumor choline ([Cho]) was greater than NWM and increased with grade for grade II and anaplastic astrocytomas, but was highly variable for glioblastomas. Higher [Cho] and [Cr] correlated with low lipid and lactate (P < 0.05), indicating a dilution of metabolite concentrations due to necrosis in high-grade tumors. Metabolite peak area ratios showed no correlation with lipids and mI/Cho (at TE = 30 ms), and Cr/Cho (at TE = 136 ms) best correlated with tumor grade. The quantified lipid, macromolecule, and lactate levels increased with grade of tumor, consistent with progression from hypoxia to necrosis. Quantification of lipids and macromolecules at short TE provided a good marker for tumor grade, and a scatter plot of the sum of alanine, lactate, and delta 1.3 lipid signals vs. mI/Cho provided a simple way to separate most tumors by type and grade.

Automated classification of short echo time in in vivo 1H brain tumor spectra: a multicenter study

Automated pattern recognition techniques are needed to help radiologists categorize MRS data of brain tumors according to histological type and grade. A major question is whether a computer program "trained" on spectra from one hospital will be able to classify those from another, particularly if the acquisition protocol is different. A subset of 144 histopathologically validated brain tumor spectra in the INTERPRET database, obtained from three of the collaborating centers, was grouped into meningiomas, low-grade astrocytomas, and "aggressive tumors" (glioblastomas and metastases). Spectra from two centers formed the training set (94 spectra) while the third acted as the test set (50 spectra). Linear discriminant analysis successfully classified 48/50 in the test set; the remaining two were atypical cases. When the training and test sets were combined, 133 of the 144 spectra were correctly classified using the leave-one-out procedure. These spectra had been obtained using different sequences (STEAM and PRESS), different echo times (20, 30, 31, and 32 ms), different repetition times (1600 and 2000 ms), and different manufacturers' instruments (GE and Philips). Pattern recognition algorithms are less sensitive to acquisition parameters than had been expected.

Parameterized evaluation of macromolecules and lipids in proton MR spectroscopy of brain diseases

Short echo time (TE) proton MR spectra of the brain include signals of several metabolites as well as macromolecules. In various pathologies, such as brain tumors and multiple sclerosis (MS), the presence of mobile lipids or pathologically altered macromolecules may provide useful additional diagnostic information. A reliable quantitation of these resonances, however, is often not possible due to the lack of adequate prior knowledge. Furthermore, even if advanced fitting procedures are used, a reliable evaluation of metabolites in the presence of pathological lipids or macromolecules often fails if the latter are omitted in the spectral evaluation. In this study, a method is presented for the simultaneous evaluation of all visible components, including metabolites, lipids, and macromolecules, by the use of the fitting procedure LCModel. A standard basis set of brain metabolites was extended by inclusion of parameterized components for macromolecules and lipids that were derived from metabolite-nulled in vivo spectra of normal brain and high-grade gliomas, respectively. The improved spectral quantitation is demonstrated in glial brain tumors and MS lesions as well as in normal brain. It is pointed out that both macromolecules and lipids must be included to provide a proper spectral evaluation.

In vivo 3-T MR spectroscopy in the distinction of recurrent glioma versus radiation effects: initial experience

PURPOSE

To determine if 3-T magnetic resonance (MR) spectroscopy allows accurate distinction of recurrent tumor from radiation effects in patients with gliomas of grade II or higher.

MATERIALS AND METHODS

This blinded prospective study included 14 patients who underwent in vivo 3-T MR spectroscopy prior to stereotactic biopsy. All patients received a previous diagnosis of glioma (grade II or higher) and high-dose radiation therapy (>54 Gy). Prior to MR spectroscopy, conventional MR imaging was performed at 1.5 T to identify a gadolinium-enhanced region within the irradiated volume. Diagnosis was assigned by means of histopathologic analysis of the biopsy samples.

RESULTS

Sixteen of 17 biopsy locations could be classified as predominantly tumor or predominantly radiation effect on the basis of the ratio of choline at the biopsy site to normal creatine level by using a value greater than 1.3 as the criterion for tumor. The remaining case, classified as recurrent tumor on the basis of MR spectroscopy results, was diagnosed as predominantly radiation effect on the basis of histopathologic findings. Disease in this patient progressed to biopsy-proven recurrence within 3 months. Overall, the ratio of choline at the biopsy site to normal creatine level was significantly elevated (unpaired two-tailed Student t test, P <.002) in those biopsy samples composed predominantly of tumor (n = 9) compared with those containing predominantly radiation effects (n = 8). The ratio was not significantly different between the two histopathologic groups.

CONCLUSION

In vivo 3-T MR spectroscopy has sufficient spatial resolution and chemical specificity to allow distinction of recurrent tumor from radiation effects in patients with treated gliomas.

Characterization of untreated gliomas by magnetic resonance spectroscopic imaging

Although there are trends in the morphologic, metabolic, hemodynamic, and structural properties of untreated gliomas that are reflected in MR measurements, there is considerable heterogeneity both within and between lesions of the same histologic grade. The spatial extent of the abnormality in ADC and RA images is similar to the T2 lesion, but there is no obvious difference in intensity between grades. The rCBV is significantly increased in the enhancing volume of grade 4 lesions but is similar or reduced in intensity for most grade 3 lesions. There are clear differences between the enhancing volumes and the regions with increased Cho that may be highly significant for planning focal therapy. The location and intensity of the Lac/Lip peaks are consistent with those representing regions of necrosis for grade 4 lesions. The fact that small Lac/Lip peaks can also be seen in grade 2 and grade 3 lesions suggests that their presence may be indicative of regions that are likely to progress to a higher grade. If this were the case, it would be valuable for directing biopsies. The correlations between rCBV, Cho, and ADC suggest that cellularity, membrane turnover, and vascularity are linked in grade 4 lesions. It is not clear whether there is any relationship between these parameters regions in grade 2 or grade 3 gliomas. While further work is required to optimize the methodology associated with these MR parameters, it seems likely that combining the information from such measurements may be valuable for predicting outcome and tailoring therapy to individual patients.

An efficient chemical shift imaging scheme for magnetic resonance-guided neurosurgery

An efficient magnetic resonance spectroscopic imaging (MRSI) or chemical shift imaging (CSI) technique based on multiple spin echoes (MSE) has been implemented, validated, and used in both phantom and in vivo MR-guided neurosurgical applications. The key concept of the method is to employ MSE to significantly speed up the data collection rate for mapping hydrogen-containing metabolites. Using an echo train length (ETL) of three per excitation to simultaneously fill three consecutive k-space areas, the total scan time for a spectroscopic image matrix size of 32 x 32 has been shortened to approximately 11 minutes. An interecho spacing time of 273 msec was used to null the phase anomalies of lactate double peaks due to the J-coupling. This allowed a sufficient long data sampling time to achieve 4 Hz spectral resolution. Performing CSI intraopertively during an MR-guided neurosurgical procedure was shown to be feasible at 1.5 T. More importantly, it was shown that more relevant information can be obtained regarding neurochemistry about a targeted lesion, in addition to conventional MR morphological imaging noninvasively. In 25 MR-guided neurosurgical cases, the alleviated choline signal has been found to be consistent with the existence of rapid tumor cell proliferation in the corresponding area. The actual neurobiopsy guided by the spectroscopic imaging method demonstrated that it could provide valuable information in specifying the optimal site in a biopsy procedure, especially in the case involving a nonenhancing tumor. The multiecho scheme has made the CSI technique efficient enough to be routinely used in MR-guided surgical procedures at 1.5 T and also allows the possibility of taking full advantage of MRI capability.

An automated technique for the quantitative assessment of 3D-MRSI data from patients with glioma

Although proton magnetic resonance spectroscopic imaging (1H-MRSI) has been shown to be effective for localizing tumor in patients with gliomas, it is not a routinely used clinical tool. This is due, in part, to the lack of a standardized, objective method for analyzing spectra. We present an automated technique for a) selecting a population of voxels from each patient that have the spectral features of normal brain regions, and b) using the selected voxels as internal controls for quantifying the probability of abnormality at each voxel location. The technique was demonstrated on a phantom, 14 normal volunteers, and 30 patients with histologically proven tumor. In addition, we demonstrated the usefulness of the method for monitoring patients in serial studies from two glioma patients with progressive disease.

Combined proton MR spectroscopy and dynamic contrast enhanced MR imaging of human intracranial tumours in vivo

A study was undertaken to determine if the vascular characteristics measured by dynamic contrast-enhanced magnetic resonance imaging (primarily permeability surface area product and extracellular-extravascular tissue volume fraction) would be beneficial in explaining the inter-lesion metabolic heterogeneity displayed by human intracranial tumours. Magnetic resonance spectroscopy was carried out using a single-voxel STEAM sequence and dynamic imaging was carried out using a combination of pre-contrast proton density-weighted FSPGR images (to remove the influence of native tissue T1), bolus injection of Gd-DTPA and subsequent T1-weighted FSPGR dynamic imaging. A two-compartment pharmacokinetic model was employed to determine vascular characteristics. Results obtained from 12 meningiomas suggest a possible correlation between the level of lipids/macromolecules and permeability surface area product, although the confounding issue of extra-voxel contamination arising from lipids in the scalp and skull marrow cannot be ruled out in the more superficial lesions. Results obtained from 11 gliomas (four low and seven high grade) demonstrate that permeability surface area product is not specific for the range of vascular characteristics and metabolite profiles observed in gliomas and is therefore unable to explain metabolic heterogeneity in these lesions.

MR spectroscopy: a powerful tool for investigating brain function and neurological diseases

Magnetic resonance spectroscopy (MRS) has attracted much attention in recent years and has become an important tool to study in vivo particular biochemical aspects of brain disorders. Since the proton is the most sensitive stable nucleus for MRS, and since almost all metabolites contain hydrogen atoms, investigation by in vivo 1H MRS provides chemical information on tissue metabolites, thus enabling a non-invasive assessment of changes in brain metabolism underlying several brain diseases. In this review a brief description of the basic principles of MRS is given. Moreover, we provide some explanations on the techniques and technical problems related to the use of 1H MRS in vivo including water suppression, localization, editing, quantitation and interpretation of 1H spectra. Finally, we discuss the more recent advancement in three major areas of neurological diseases: brain tumors, multiple sclerosis, and inborn errors of metabolism.

Quantification of plasma lipids and apolipoproteins by use of proton NMR spectroscopy, multivariate and neural network analysis

New approaches for quantification of human blood plasma lipids and apolipoproteins are presented. One method is based on multivariate analysis of proton nuclear magnetic resonance spectra of human blood plasma. Although similar approaches have been developed previously, this is the first time principal component analysis (PCA) and partial least squares regression (PLS) have been applied to this particular task. Further, a large proportion of the subjects in this study were cancer patients undergoing treatment, which introduced a new dimension to the quantification of lipoprotein distributions. Calibration models for prediction of lipids and apolipoproteins were constructed by use of PLS, and blind samples were used to test the predictive ability. Comparison of the predicted vs observed data obtained by standard clinical chemical procedures gave good agreement; the correlation coefficient for total plasma triglyceride was 0.99, for total plasma cholesterol 0.98, for LDL cholesterol 0. 97, and for HDL cholesterol 0.88. These results are comparable with those obtained with other methods. The quantitative analysis of 14 components (including total cholesterol and total triglyceride) of human blood plasma was also undertaken using various neural network (NN) analyses of selected portions of the spectra. Conventional fully connected backpropagation neural network topologies were capable of providing excellent predictions for the majority of the variables, confirming and reinforcing literature related to this approach. However HDL triglycerides were poorly predicted, while intermediate-quality results were obtained for the LDL cholesterol, plasma apoA1 and LDL apoB variables. In these instances, applying significantly different neural network algorithms involving either general regression or polynomial neural networks in combination with genetic adaptive components for parameter optimisation made improved predictions.

Magnetic resonance spectroscopy of brain tumors

Magnetic resonance spectroscopy provides metabolic information about brain tumors beyond what can be obtained from anatomic images. In contrast to other metabolism-based imaging techniques such as single photon emission computed tomography and positron-emission tomography, magnetic resonance spectroscopy yields multiparametric data, does not require radio-labeled tracers or ionizing radiation, and can be performed in conjunction with other magnetic resonance imaging studies. Magnetic resonance spectral patterns have been shown to be distinct for different tumor types and grades. Response to radiation therapy is also reflected by magnetic resonance spectral patterns. Although there are quantitative issues still to be addressed, correlation of in vivo spectral patterns with ex vivo spectral patterns obtained from actual biopsy samples indicates that magnetic resonance spectroscopy is a fundamentally valid tool for monitoring disease progression and therapeutic response in patients with brain tumors.

Distinction between normal and renal cell carcinoma kidney cortical biopsy samples using pattern recognition of (1)H magic angle spinning (MAS) NMR spectra

The technique of magic angle spinning (MAS) high resolution (1)H NMR spectroscopy applied to intact tissues provides excellent peak resolution and thus much biochemical information. The use of computer-based pattern recognition techniques to classify human renal cortex tissue samples as normal or tumour based on their (1)H MAS NMR spectra has been investigated. In this preliminary study of 22 paired control and tumour samples, exploratory data analysis using principal components based on NMR spectral intensities showed clear separation of the two classes. Furthermore, using the supervised method of linear discriminant analysis, based on individual data point intensities or on integrated spectral regions, it was possible to distinguish between the normal and tumour kidney cortex tissue with 100% accuracy, including a single example of a metastatic tumour from a primary lung carcinoma. A tumour sample from the collecting duct of the kidney showed a different NMR spectral profile, and pattern recognition indicated that this sample did not classify with the cortical tumours.

In vivo proton magnetic resonance spectroscopy of central neurocytomas

OBJECTIVE

The authors report on the metabolic features of central neurocytomas observed during in vivo single-voxel proton magnetic resonance spectroscopy.

METHODS

Volume-selective single-voxel proton magnetic resonance spectroscopy was performed with a 1.5-T unit using a point-resolved spectroscopy sequence (TR/TE = 2000 ms/135 and 270 ms) to obtain spectra of a single 8-cc voxel. The subjects were five patients in the Department of Neurosurgery of Seoul National University Hospital whose central neurocytomas had been diagnosed histologically. The peak intensities of compounds containing choline (Cho), N-acetylaspartate, creatine/phosphocreatine, and lactate were analyzed.

RESULTS

The ratios of Cho to creatine/phosphocreatine and Cho to N-acetylaspartate were significantly higher than ratios in normal brains. A lactate signal was present, and an unidentified signal was also observed at 3.55 ppm, which might have been produced by inositol or glycine.

CONCLUSION

A combination of the signal at 3.55 ppm and a prominent Cho peak seems to be a characteristic feature of central neurocytomas. Volume-selective single-voxel proton magnetic resonance spectroscopy could provide additional information to aid in diagnosing this condition.

Pediatric low-grade gliomas: prognosis with proton magnetic resonance spectroscopic imaging

OBJECTIVE

Our aim was to assess the correlation between the low-grade glioma (LGG) metabolic profile and tumor progression. Using in vivo proton magnetic resonance spectroscopic imaging, we specifically asked whether and which metabolic features are associated with tumor regrowth or recurrence.

METHODS

Eleven pediatric patients with histologically proven partially resected (<20% resection) midline LGG were treated and followed up for a period of 2 years. All patients underwent proton magnetic resonance spectroscopic imaging studies before any management was determined. Tumor progression was defined as radiological evidence of mass enlargement (>25%) during the follow-up period. Proton magnetic resonance spectroscopic imaging was performed using a PRESS-CSI sequence on a General Electric 1.5-tesla scanner (General Electric Medical System, Waukesha, WI). The signal intensities of N-acetylaspartate, choline (CHO), and creatine from the tumor and the normal brain were used to calculate normalized metabolite intensities and metabolite ratios.

RESULTS

Tumors that progressed during a 2-year period displayed higher normalized CHO than those that remained stable (Mann-Whitney test, P < 0.03). The majority (five of six) of the rapidly growing LGG showed values of normalized CHO of at least 1, whereas the nonprogressors had a normalized CHO value of less than 1.

CONCLUSION

In association with pediatric LGG, high normalized CHO values seem to herald the potential for rapid tumor growth. These observations may be valuable for defining subsets of patients with LGG who may benefit from early therapeutic interventions.

In vivo differentiation of astrocytic brain tumors and isolated demyelinating lesions of the type seen in multiple sclerosis using 1H magnetic resonance spectroscopic imaging

We used computer pattern recognition of proton magnetic resonance spectroscopic image data to differentiate between brain tumors and large, isolated, demyelinating lesions of the type seen in multiple sclerosis. Leave-one-out linear discriminant analyses correctly classified resonance profiles from five acute demyelinating lesions, 20 low-grade astrocytomas, 22 anoplastic astrocytomas, and 24 glioblastomas. Classification of nonacute lesions will require further development, as the metabolic profiles of demyelinating lesions evolve over time.

Pattern recognition analysis of 1H NMR spectra from perchloric acid extracts of human brain tumor biopsies

Pattern recognition techniques (factor analysis and neural networks) were used to investigate and classify human brain tumors based on the 1H NMR spectra of chemically extracted biopsies (n = 118). After removing information from lactate (because of variable ischemia times), unsupervised learning suggested that the spectra separated naturally into two groups: meningiomas and other tumors. Principal component analysis reduced the dimensionality of the data. A back-propagation neural network using the first 30 principal components gave 85% correct classification of meningiomas and nonmeningiomas. Simplification by vector rotation gave vectors that could be assigned to various metabolites, making it possible to use or to reject their information for neural network classification. Using scores calculated from the four rotated vectors due to creatine and glutamine gave the best classification into meningiomas and nonmeningiomas (89% correct). Classification of gliomas (n = 47) gave 62% correct within one grade. Only inositol showed a significant correlation with glioma grade.

From magnetic resonance spectroscopy to classification of tumors. A review of pattern recognition methods

This article reviews the wealth of different pattern recognition methods that have been used for magnetic resonance spectroscopy (MRS) based tumor classification. The methods have in common that the entire MR spectra is used to develop linear and non-linear classifiers. The following issues are addressed: (i) pre-processing, such as normalization and digitization, (ii) extraction of relevant spectral features by multivariate methods, such as principal component analysis, linear discriminant analysis (LDA), and optimal discriminant vector, and (iii) classification by LDA, cluster analysis and artificial neural networks. Different approaches are compared and discussed in view of practical and theoretical considerations.

Towards a method for automated classification of 1H MRS spectra from brain tumours

Recent studies have shown that MRS can substantially improve the non-invasive categorization of human brain tumours. However, in order for MRS to be used routinely by clinicians, it will be necessary to develop reliable automated classification methods that can be fully validated. This paper is in two parts: the first part reviews the progress that has been made towards this goal, together with the problems that are involved in the design of automated methods to process and classify the spectra. The second part describes the development of a simple prototype system for classifying 1H single voxel spectra, obtained at an echo time (TE) of 135 ms, of the four most common types of brain tumour (meningioma (MM), astrocytic (AST), oligodendroglioma (OD) and metastasis (ME)) and cysts. This system was developed in two stages: firstly, an initial database of spectra was used to develop a prototype classifier, based on a linear discriminant analysis (LDA) of selected data points. Secondly, this classifier was tested on an independent test set of 15 newly acquired spectra, and the system was refined on the basis of these results. The system correctly classified all the non-astrocytic tumours. However, the results for the the astrocytic group were poorer (between 55 and 100%, depending on the binary comparison). Approximately 50% of high grade astrocytoma (glioblastoma) spectra in our data base showed very little lipid signal, which may account for the poorer results for this class. Consequently, for the refined system, the astrocytomas were subdivided into two subgroups for comparison against other tumour classes: those with high lipid content and those without.

Using pattern analysis of in vivo proton MRSI data to improve the diagnosis and surgical management of patients with brain tumors

We have used pattern analysis of proton magnetic resonance spectroscopic imaging (1H MRSI) data in a variety of situations related to the clinical management of patients with brain tumors and other cerebral space-occupying lesions (SOLs). Here, we review how 'leave-one-out' linear discriminant analyses (LDAs) of in vivo 1H MRSI spectral patterns have enabled us to quickly, accurately, and non-invasively: (1) discriminate amongst tissue arising from the five most common types of supratentorial tumors found in adults, and (2) use the metabolic heterogeneity of cerebral SOLs to predict certain pathological characteristics that are useful in guiding stereotaxic biopsy and selective tumor resection. These findings suggest that pattern analysis of 1H MRSI data can significantly improve the diagnostic specificity and surgical management of patients with certain cerebral SOLs.

Assignment of glial brain tumors in humans by in vivo 1H-magnetic resonance spectroscopy and multidimensional metabolic classification

This study presents a simple approach for the noninvasive assignment of glial brain tumors according to malignancy by single-voxel proton magnetic resonance spectroscopy at short echo times (TE < or = 50 milliseconds). Based on peak area ratios, a five-dimensional data set was obtained for each investigated subject. This vector was then projected along metabolic coordinates in a two-dimensional metabolic space. These coordinates had been determined in a previous study (Hagberg G et al., 1995, Magn Reson Med 34: 242-252). Tumor assignment was done without any knowledge of histology by comparing the location of the new cases to the features of the previous study. All 11 investigated glioblastomas multiforme, as well as 4 of 5 astrocytomas grade II, could easily be assigned to the groups of high- and low-grade tumors, respectively. Classification was more difficult in the case of a cystic astrocytoma grade II and one astrocytoma grade III. Two spectra measured in normal-appearing matter of glioblastoma patients were not classified as healthy. Using single-voxel proton magnetic resonance spectroscopy at short echo times with the knowledge of a base study, a straightforward, fast, and noninvasive differential diagnosis of glial brain tumors is possible.

Functional characterization of brain tumors: an overview of the potential clinical value

Early detection and characterization are still challenging issues in the diagnostic approach to brain tumors. Among functional imaging techniques, a clinical role for positron emission tomography studies with [18F]-fluorodeoxyglucose and for single photon emission computed tomography studies with [201Tl]-thallium-chloride has emerged. The clinical role of magnetic resonance spectroscopy is still being defined, whereas functional magnetic resonance imaging seems able to provide useful data for presurgical localization of critical cortical areas. Integration of morphostructural information provided by computed tomography and magnetic resonance imaging, with functional characterization and cyto-histologic evaluation of biologic markers, may assist in answering the open diagnostic questions concerning brain tumors.