Effect of feature extraction for brain tumor classification based on short echo time ¹H MR spectra

Abstract

This study examines the effect of feature extraction methods prior to automated pattern recognition based on magnetic resonance spectroscopy (MRS) for brain tumor diagnosis. Since individual inspection of spectra is time-consuming and requires specific spectroscopic expertise, the introduction of clinical decision support systems (DSSs) is expected to strongly promote the clinical use of MRS. This study focuses on the feature extraction step in the preprocessing protocol of MRS when using a DSS. On two independent data sets, encompassing single-voxel and multi-voxel data, it is observed that the use of the full spectra together with a kernel-based technique, handling high dimensional data, or using an automated pattern recognition method based on independent component analysis or Relief-F achieves accurate performances. In addition, these approaches have low cost and are easy to automate. When sophisticated quantification methods are used in a DSS, user interaction should be minimized. The computationally intensive quantification techniques do not tend to increase the performance in these circumstances. The results suggest to simplify the feature reduction step in the preprocessing protocol when using a DSS purely for classification purposes. This can greatly speed up the execution of classifiers and DSSs and may accelerate their introduction into clinical practice.Magn Reson Med 60:288-298, 2008.

Accurate classification of childhood brain tumours by in vivo 1H MRS – A multi-centre study

“…Classifiers were designed and evaluated using features from Short-TE and Long-TE alone and a combination of both TEs, ShortTE+Long-TE. Our results were compared with those in previous studies [27][28][29][30].Based on the results of previous studies [15,20,26,29 …”

“…PI automatically estimates with proportionality to the concentration of 11 main metabolites for Short-TE and 8 metabolites for Long-TE MRS [15,26]. Details of these estimations are given in the Supplementary…”

“…Based on the results of previous studies [15,20,26 [31]. Details about the evaluation methodology are described in the Supplementary Material.…”

Clinical Decision Support Systems for Brain Tumour Diagnosis: Classification and Evaluation Approaches

“…Analogously to the PR methods selected for classification, PCA has been widely accepted in the literature and is a simple, standard and straigh multivariate technique that obtains reasonably good results in MRS feature extraction [83,84].…”

“…They reported that PR methods perform at least as well as the ones based on manual quantitation obtaining 5%-10% higher accuracy when automatic techniques were applied. Classifiers for in vivo Short-TE spectra and Further investigation with LS-SVM and feature extraction methods for MRS was carried out by Luts et al [84] during a subsequent EC project called eTUMOUR. Luts et al studied the effect of automating the feature extraction step in the preprocessing protocol of Short-TE spectra.…”