Incremental Gaussian Discriminant Analysis based on Graybill and Deal weighted combination of estimators for brain tumour diagnosis.

In the last decade, machine learning (ML) techniques have been used for developing classifiers for automatic brain tumour diagnosis. However, the development of these ML models rely on a unique training set and learning stops once this set has been processed. Training these classifiers requires a representative amount of data, but the gathering, preprocess, and validation of samples is expensive and time-consuming. Therefore, for a classical, non-incremental approach to ML, it is necessary to wait long enough to collect all the required data. In contrast, an incremental learning approach may allow us to build an initial classifier with a smaller number of samples and update it incrementally when new data are collected. In this study, an incremental learning algorithm for Gaussian Discriminant Analysis (iGDA) based on the Graybill and Deal weighted combination of estimators is introduced. Each time a new set of data becomes available, a new estimation is carried out and a combination with a previous estimation is performed. iGDA does not require access to the previously used data and is able to include new classes that were not in the original analysis, thus allowing the customization of the models to the distribution of data at a particular clinical center. An evaluation using five benchmark databases has been used to evaluate the behaviour of the iGDA algorithm in terms of stability-plasticity, class inclusion and order effect. Finally, the iGDA algorithm has been applied to automatic brain tumour classification with magnetic resonance spectroscopy, and compared with two state-of-the-art incremental algorithms. The empirical results obtained show the ability of the algorithm to learn in an incremental fashion, improving the performance of the models when new information is available, and converging in the course of time. Furthermore, the algorithm shows a negligible instance and concept order effect, avoiding the bias that such effects could introduce.

Compatibility between 3T 1H SV-MRS data and automatic brain tumour diagnosis support systems based on databases of 1.5T 1H SV-MRS spectra.

OBJECT: This study demonstrates that 3T SV-MRS data can be used with the currently available automatic brain tumour diagnostic classifiers which were trained on databases of 1.5T spectra. This will allow the existing large databases of 1.5T MRS data to be used for diagnostic classification of 3T spectra, and perhaps also the combination of 1.5T and 3T databases. MATERIALS AND METHODS: Brain tumour classifiers trained with 154 1.5T spectra to discriminate among high grade malignant tumours and common grade II glial tumours were evaluated with a subsequently-acquired set of 155 1.5T and 37 3T spectra. A similarity study between spectra and main brain tumour metabolite ratios for both field strengths (1.5T and 3T) was also performed. RESULTS: Our results showed that classifiers trained with 1.5T samples had similar accuracy for both test datasets (0.87 ± 0.03 for 1.5T and 0.88 ± 0.03 for 3.0T). Moreover, non-significant differences were observed with most metabolite ratios and spectral patterns. CONCLUSION: These results encourage the use of existing classifiers based on 1.5T datasets for diagnosis with 3T (1)H SV-MRS. The large 1.5T databases compiled throughout many years and the prediction models based on 1.5T acquisitions can therefore continue to be used with data from the new 3T instruments.

Multiproject-multicenter evaluation of automatic brain tumor classification by magnetic resonance spectroscopy.

JUSTIFICATION: Automatic brain tumor classification by MRS has been under development for more than a decade. Nonetheless, to our knowledge, there are no published evaluations of predictive models with unseen cases that are subsequently acquired in different centers. The multicenter eTUMOUR project (2004-2009), which builds upon previous expertise from the INTERPRET project (2000-2002) has allowed such an evaluation to take place. MATERIALS AND METHODS: A total of 253 pairwise classifiers for glioblastoma, meningioma, metastasis, and low-grade glial diagnosis were inferred based on 211 SV short TE INTERPRET MR spectra obtained at 1.5 T (PRESS or STEAM, 20-32 ms) and automatically pre-processed. Afterwards, the classifiers were tested with 97 spectra, which were subsequently compiled during eTUMOUR. RESULTS: In our results based on subsequently acquired spectra, accuracies of around 90% were achieved for most of the pairwise discrimination problems. The exception was for the glioblastoma versus metastasis discrimination, which was below 78%. A more clear definition of metastases may be obtained by other approaches, such as MRSI + MRI. CONCLUSIONS: The prediction of the tumor type of in-vivo MRS is possible using classifiers developed from previously acquired data, in different hospitals with different instrumentation under the same acquisition protocols. This methodology may find application for assisting in the diagnosis of new brain tumor cases and for the quality control of multicenter MRS databases.

Liver triglyceride content in HIV-1-infected patients on combination antiretroviral therapy studied with 1H-MR spectroscopy.

OBJECTIVE: To carry out an exploratory evaluation of liver triglyceride content in HIV-1-infected patients receiving highly active antiretroviral therapy (HAART) using proton magnetic resonance spectroscopy and to study how both the treatment itself and the biochemical and physiological variables in which the treatment causes alterations are related to liver fat content. METHODS: Intracellular hepatic triglyceride content was determined in 29 HIV-1-infected patients on their first HAART regime by means of localized water-unsuppressed single voxel proton spectra. Other measurements were body mass index, waist-to-hip ratio, lipodystrophy assessment and a detailed blood biochemical analysis. The relationship between intracellular hepatic triglycerides and relevant descriptive, treatment and biochemical variables was studied by correlation and regression analysis. RESULTS: Intrahepatic triglycerides were detected in 58.6% of the patients and 13.8% showed a triglyceride content compatible with liver steatosis. Many variables (body mass index, waist-to-hip ratio, cumulative exposure to PIs, lactate, insulin, insulin resistance measured by the homeostasis model assessment method [HOMA-R index], pH, total triglycerides, high density lipoprotein cholesterol and very low density lipoprotein [VLDL] cholesterol) correlated individually with the amount of triglycerides. Stepwise multiple regression analysis showed that the combination of insulin or HOMA-R index and VLDL cholesterol accounted for up to 50.2% of the triglyceride liver variance. A positive relationship was found between the concomitant presence of the metabolic syndrome components (insulin resistance, dyslipidaemia and central obesity) and intrahepatic triglyceride content. CONCLUSIONS: The study showed that intrahepatic triglyceride deposit appears to be a frequent feature of HIV-1-infected patients receiving HAART. A coherent multifactorial combination of biochemical and physiological factors associated with the deposit suggested that cumulative exposure to PIs might be a possible trigger event.

Comparison between neuroimaging classifications and histopathological diagnoses using an international multicenter brain tumor magnetic resonance imaging database.

OBJECT: The aim of this study was to estimate the accuracy of routine magnetic resonance (MR) imaging studies in the classification of brain tumors in terms of both cell type and grade of malignancy. METHODS: The authors retrospectively assessed the correlation between neuroimaging classifications and histopathological diagnoses by using multicenter database records from 393 patients with brain tumors. An ontology was devised to establish diagnostic agreement. Each tumor category was compared with the corresponding histopathological diagnoses by dichotomization. Sensitivity, specificity, positive and negative predictive values (PPVs and NPVs, respectively), and the Wilson 95% confidence intervals (CI) for each were calculated. In routine reporting of MR imaging examinations, tumor types and grades were classified with a high specificity (85.2-100%); sensitivity varied, depending on the tumor type and grade, alone or in combination. The recognition of broad diagnostic categories (neuroepithelial or meningeal lesions) was highly sensitive, whereas when both detailed type and grade were considered, sensitivity diverged, being highest in low-grade meningioma (sensitivity 100%, 95% CI 96.2-100.0%) and lowest in high-grade meningioma (sensitivity 0.0%, 95% CI 0.0-65.8%) and low-grade oligodendroglioma (sensitivity 15%, 95% CI 5.2-36.0%). In neuroepithelial tumors, sensitivity was inversely related to the precision in reporting of grade and cellular origin; "glioma" was a frequent neuroimaging classification associated with higher sensitivity in the corresponding category. The PPVs varied among categories, in general being greater than their prevalence in this dataset. The NPV was high in all categories (69.8-100%). CONCLUSIONS: The PPVs and NPVs provided in this study may be used as estimates of posttest probabilities of diagnostic accuracy using MR imaging. This study targets the need for noninvasively increasing sensitivity in categorizing most brain tumor types while retaining high specificity, especially in the differentiation of high- and low-grade glial tumor classes.

Age-related metabolic changes in the upper brainstem tegmentum by MR spectroscopy.

Several neurodegenerative disorders have a profound metabolic and structural impact on the brainstem. MR spectroscopy provides metabolic information non-invasively and has the potential to characterize the changes associated with normal aging and differentiate them from neurodegenerative alterations. The present work was aimed at studying the upper brainstem tegmentum at the midbrain and pontine levels in 57 adult normal volunteers, aged 23-79 years, with long-echo time proton MR spectroscopy to evaluate possible regional differences and the effect of age. Higher ratios of N-acetyl aspartate (NAA)/total creatine (Cr) and choline-containing compounds (Cho)/Cr were observed in the pons compared to the midbrain, resulting from higher net NAA and Cho content. In the midbrain, there was a linear decline of NAA and Cho with age in subjects over 50, most probably related to neuronal tissue loss. In the pons, such an aging effect was not observed, with subjects over 50 showing higher Cr and Cho than the under-50 subjects. Our findings provided evidence of regional differences and suggest different effects of age on the two studied brainstem segments, hitherto undescribed.

Strategies for annotation and curation of translational databases: the eTUMOUR project.

The eTUMOUR (eT) multi-centre project gathered in vivo and ex vivo magnetic resonance (MR) data, as well as transcriptomic and clinical information from brain tumour patients, with the purpose of improving the diagnostic and prognostic evaluation of future patients. In order to carry this out, among other work, a database--the eTDB--was developed. In addition to complex permission rules and software and management quality control (QC), it was necessary to develop anonymization, processing and data visualization tools for the data uploaded. It was also necessary to develop sophisticated curation strategies that involved on one hand, dedicated fields for QC-generated meta-data and specialized queries and global permissions for senior curators and on the other, to establish a set of metrics to quantify its contents. The indispensable dataset (ID), completeness and pairedness indices were set. The database contains 1317 cases created as a result of the eT project and 304 from a previous project, INTERPRET. The number of cases fulfilling the ID was 656. Completeness and pairedness were heterogeneous, depending on the data type involved.

On the relevance of automatically selected single-voxel MRS and multimodal MRI and MRSI features for brain tumour differentiation.

In order to evaluate the relevance of magnetic resonance (MR) features selected by automatic feature selection techniques to build classifiers for differential diagnosis and tissue segmentation two data sets containing MR spectroscopy data from patients with brain tumours were investigated. The automatically selected features were evaluated using literature and clinical experience. It was observed that a significant part of the automatically selected features correspond to what is known from the literature and clinical experience. We conclude that automatic feature selection is a useful tool to obtain relevant and possibly interesting features, but evaluation of the obtained features remains necessary.

Work-related pain in extrinsic finger extensor musculature of instrumentalists is associated with intracellular pH compartmentation during exercise.

BACKGROUND: Although non-specific pain in the upper limb muscles of workers engaged in mild repetitive tasks is a common occupational health problem, much is unknown about the associated structural and biochemical changes. In this study, we compared the muscle energy metabolism of the extrinsic finger extensor musculature in instrumentalists suffering from work-related pain with that of healthy control instrumentalists using non-invasive phosphorus magnetic resonance spectroscopy ((31)P-MRS). We hypothesize that the affected muscles will show alterations related with an impaired energy metabolism. METHODOLOGY/PRINCIPAL FINDINGS: We studied 19 volunteer instrumentalists (11 subjects with work-related pain affecting the extrinsic finger extensor musculature and 8 healthy controls). We used (31)P-MRS to find deviations from the expected metabolic response to exercise in phosphocreatine (PCr), inorganic phosphate (Pi), Pi/PCr ratio and intracellular pH kinetics. We observed a reduced finger extensor exercise tolerance in instrumentalists with myalgia, an intracellular pH compartmentation in the form of neutral and acid compartments, as detected by Pi peak splitting in (31)P-MRS spectra, predominantly in myalgic muscles, and a strong association of this pattern with the condition. CONCLUSIONS/SIGNIFICANCE: Work-related pain in the finger extrinsic extensor muscles is associated with intracellular pH compartmentation during exercise, non-invasively detectable by (31)P-MRS and consistent with the simultaneous energy production by oxidative metabolism and glycolysis. We speculate that a deficit in energy production by oxidative pathways may exist in the affected muscles. Two possible explanations for this would be the partial and/or local reduction of blood supply and the reduction of the muscle oxidative capacity itself.

Development of a predictor for human brain tumors based on gene expression values obtained from two types of microarray technologies.

Development of molecular diagnostics that can reliably differentiate amongst different subtypes of brain tumors is an important unmet clinical need in postgenomics medicine and clinical oncology. A simple linear formula derived from gene expression values of four genes (GFAP, PTPRZ1, GPM6B, and PRELP) measured from cDNA microarrays (n = 35) have distinguished glioblastoma and meningioma cases in a previous study. We herein extend this work further and report that the above predictor formula showed its robustness when applied to Affymetrix microarray data acquired prospectively in our laboratory (n = 80) as well as publicly available data (n = 98). Importantly, GFAP and GPM6B were both retained as being significant in the predictive model upon using the Affymetrix data obtained in our laboratory, whereas the other two predictor genes were SFRP2 and SLC6A2. These results collectively indicate the importance of the expression values of GFAP and GPM6B genes sampled from the two types of microarray technologies tested. The high prediction accuracy obtained in these instances demonstrates the robustness of the predictors across microarray platforms used. This result would require further validation with a larger population of meningioma and glioblastoma cases. At any rate, this study paves the way for further application of gene signatures to more stringent biopsy discrimination challenges.

The effect of combining two echo times in automatic brain tumor classification by MRS.

(1)H MRS is becoming an accurate, non-invasive technique for initial examination of brain masses. We investigated if the combination of single-voxel (1)H MRS at 1.5 T at two different (TEs), short TE (PRESS or STEAM, 20-32 ms) and long TE (PRESS, 135-136 ms), improves the classification of brain tumors over using only one echo TE. A clinically validated dataset of 50 low-grade meningiomas, 105 aggressive tumors (glioblastoma and metastasis), and 30 low-grade glial tumors (astrocytomas grade II, oligodendrogliomas and oligoastrocytomas) was used to fit predictive models based on the combination of features from short-TEs and long-TE spectra. A new approach that combines the two consecutively was used to produce a single data vector from which relevant features of the two TE spectra could be extracted by means of three algorithms: stepwise, reliefF, and principal components analysis. Least squares support vector machines and linear discriminant analysis were applied to fit the pairwise and multiclass classifiers, respectively. Significant differences in performance were found when short-TE, long-TE or both spectra combined were used as input. In our dataset, to discriminate meningiomas, the combination of the two TE acquisitions produced optimal performance. To discriminate aggressive tumors from low-grade glial tumours, the use of short-TE acquisition alone was preferable. The classifier development strategy used here lends itself to automated learning and test performance processes, which may be of use for future web-based multicentric classifier development studies.

Automated quality control protocol for MR spectra of brain tumors.

eTUMOUR (http://www.etumour.net/) is acquiring a large database of brain tumor (1)H MR spectra to develop automated pattern recognition methods and decision support system (DSS) for tumor diagnosis. Development of accurate pattern-recognition algorithms requires spectra undistorted by artifacts, low signal-to-noise, or broad lines. eTUMOUR currently uses panels of expert spectroscopists to subjectively grade spectra as being acceptable or unacceptable. Automated quality control (QC) would be more satisfactory for several reasons: 1) to provide a reproducible objective classification of spectrum quality; 2) for use within the future DSS to prevent misdiagnosis due to poor spectrum quality; 3) to rapidly process the very large datasets of 1H spectra being accrued. An automated QC method using independent component analysis for feature extraction with a least-squares support vector machine classifier is presented. Separate training (n=144) and test sets (n=98) of single-voxel spectra from brain tumors and other lesions were acquired at multiple clinical centers with short and long echo times. Pairs of expert spectroscopists classified the test set an average of 85% the same. The automated QC classification agreed with an expert for 87% of test spectra, on average, suggesting the method classifies spectrum quality as accurately as expert spectroscopists.

MRS quality assessment in a multicentre study on MRS-based classification of brain tumours.

This paper reports on quality assessment of MRS in the European Union-funded multicentre project INTERPRET (International Network for Pattern Recognition of Tumours Using Magnetic Resonance; http://azizu.uab.es/INTERPRET), which has developed brain tumour classification software using in vivo proton MR spectra. The quality assessment consisted of both MR system quality assurance (SQA) and quality control (QC) of spectral data acquired from patients and healthy volunteers. The system performance of the MR spectrometers at all participating centres was checked bimonthly by a short measurement protocol using a specially designed INTERPRET phantom. In addition, a more extended SQA protocol was performed yearly and after each hardware or software upgrade. To compare the system performance for in vivo measurements, each centre acquired MR spectra from the brain of five healthy volunteers. All MR systems fulfilled generally accepted minimal system performance for brain MRS during the entire data acquisition period. The QC procedure of the MR spectra in the database comprised automatic determination of the signal-to-noise ratio (SNR) in a water-suppressed spectrum and of the line width of the water resonance (water band width, WBW) in the corresponding non-suppressed spectrum. Values of SNR > 10 and WBW < 8 Hz at 1.5 T were determined empirically as conservative threshold levels required for spectra to be of acceptable quality. These thresholds only hold for SNR and WBW values using the definitions and data processing described in this article. A final QC check consisted of visual inspection of each clinically validated water-suppressed metabolite spectrum by two, or, in the case of disagreement, three, experienced MR spectroscopists, to detect artefacts such as large baseline distortions, exceptionally broadened metabolite peaks, insufficient removal of the water line, large phase errors, and signals originating from outside the voxel. In the end, 10% of 889 spectra with completed spectroscopic judgement were discarded.

Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra.

A computer-based decision support system to assist radiologists in diagnosing and grading brain tumours has been developed by the multi-centre INTERPRET project. Spectra from a database of 1H single-voxel spectra of different types of brain tumours, acquired in vivo from 334 patients at four different centres, are clustered according to their pathology, using automated pattern recognition techniques and the results are presented as a two-dimensional scatterplot using an intuitive graphical user interface (GUI). Formal quality control procedures were performed to standardize the performance of the instruments and check each spectrum, and teams of expert neuroradiologists, neurosurgeons, neurologists and neuropathologists clinically validated each case. The prototype decision support system (DSS) successfully classified 89% of the cases in an independent test set of 91 cases of the most frequent tumour types (meningiomas, low-grade gliomas and high-grade malignant tumours--glioblastomas and metastases). It also helps to resolve diagnostic difficulty in borderline cases. When the prototype was tested by radiologists and other clinicians it was favourably received. Results of the preliminary clinical analysis of the added value of using the DSS for brain tumour diagnosis with MRS showed a small but significant improvement over MRI used alone. In the comparison of individual pathologies, PNETs were significantly better diagnosed with the DSS than with MRI alone.

Taurine detection by proton magnetic resonance spectroscopy in medulloblastoma: contribution to noninvasive differential diagnosis with cerebellar astrocytoma.

OBJECTIVE: We sought to evaluate whether taurine detection in short-echo (20 ms) proton magnetic resonance spectroscopy contributes to the noninvasive differential diagnosis between medulloblastoma and cerebellar astrocytoma in children and young adults. These two types of tumor have very different prognoses and may be difficult to differentiate by neuroradiological or clinical means. METHODS: Single-voxel proton magnetic resonance spectra of tumors were acquired at 1.5 T in 14 patients with biopsy-proven primary cerebellar tumors (six medulloblastomas, seven astrocytomas, and one mixed astroependymoma) using short-echo time (20 ms) and long-echo time (135 ms). For taurine assignment, qualitative analysis was performed on short-echo time spectra and results were compared in vitro with spectra of model solutions. Perchloric acid extracts of postsurgical tumor biopsies were performed in two medulloblastoma cases. RESULTS: Taurine detection was demonstrated in all patients with medulloblastoma and in none of those with astrocytoma. We were unable to ascertain any relationship between taurine and metastatic spread within the medulloblastoma group. CONCLUSION: Medulloblastomas characteristically seem to show taurine detectable in vivo by short-echo proton magnetic resonance spectroscopy, which may help to discriminate medulloblastoma from cerebellar astrocytoma.