Classification of ¹H MR spectra of human brain neoplasms: The influence of preprocessing and computerized consensus diagnosis on classification accuracy

Abstract

We study how classification accuracy can be improved when both different data preprocessing methods and computerized consensus diagnosis (CCD) are applied to 1H magnetic resonance (MR) spectra of astrocytomas, meningiomas, and epileptic brain tissue. The MR spectra (360 MHz, 37 degrees C) of tissue specimens (biopsies) from subjects with meningiomas (95; 26 cases), astrocytomas (74; 26 cases), and epilepsy (37; 8 cases) were preprocessed by several methods. Each data set was partitioned into training and validation sets. Robust classification was carried out via linear discriminant analysis (LDA), artificial neural nets (NN), and CCD, and the results were compared with histopathological diagnosis of the MR specimens. Normalization of the relevant spectral regions affects classification accuracy significantly. The spectra-based average three-class classification accuracies of LDA and NN increased from 81.7% (unnormalized data sets) to 89.9% (normalized). CCD increased the classification accuracy of the normalized sets to an average of 91.8%. CCD invariably decreases the fraction of unclassifiable spectra. The same trends prevail, with improved results, for case-based classification. Preprocessing the 1H MR spectra is essential for accurate and reliable classification of astrocytomas, meningiomas, and nontumorous epileptic brain tissue. CCD improves classification accuracy, with an attendant decrease in the fraction of unclassifiable spectra or cases.

Combination of feature-reduced MR spectroscopic and MR imaging data for improved brain tumor classification

“…The -errors are between 8 and 25%. There is virtually no difference between nonnormalized (method 1-4) and normalized methods (5)(6)(7)(8). When information from the images is also used during classification, the -error ranges between 6 and 12% and the -error between 4 and 8%.…”

“…The interpretation of this data is difficult and subjective and therefore classification based on pattern recognition is under development. [5][6][7][8][9][10] At the moment, most MR-related strategies to obtain information about brain tumors use either MRI or MRSI. A combination of both data sources has not been extensively investigated yet.…”

“…Several methods to perform data reduction and feature selection have been described in the literature. 6,[11][12][13] The last step in automated tumor classification, involves the actual classifier. In our case the classifier was based on the Mahalanobis distance.…”

Fast quantification of proton magnetic resonance spectroscopic imaging with artificial neural networks

“…[12]), were used to train the RBFNN. [12] where the Lorentzian peak is defined as, [13] and the Gaussian peak is defined as,…”

“…Such a goal could possibly be realized using methods based on artificial neural networks (ANN). The ability of ANN for classification of spectra for various pathologies has been demonstrated [6][7][8][9][10][11][12][13][14]. However, relatively little attention has been paid to metabolite quantification using ANN [15][16].…”