Classification of brain tumours using short echo time 1 H MR spectra A. Devos a, * , L. Lukas a , J.A.K. Suykens a , L. Vanhamme a , A.R. Tate b , F.A. Howe c , C. Majo ´s d , A. Moreno-Torres e , M. van der Graaf f , C. Aru ´s g , S. Van Huffel a a SCD-SISTA, Department of Electrical Engineering, Katholieke Universiteit Leuven, Kasteelpark Arenberg 10, 3001 Heverlee (Leuven), Belgium b Department of Paediatric Epidemiology and Biostatistics, Institute of Child Health, London WC1N 1EH, UK c CRUK Biomedical Magnetic Resonance Research Group, Department of Biochemistry and Immunology, St. GeorgeÕs Hospital Medical School, Cranmer Terrace, London SW17 0RE, UK d Institut de Diagno ` stic per la Imatge (IDI), CSU de Bellvitge, Autovia de Castelldefels km 2.7, LÕHospitalet de Llobregat, 08907 Barcelona, Spain e Centre Diagno ` stic Pedralbes, Unitat Esplugues, C/ Josep Anselm Clave ´, 100, 08950 Esplugues de Llobregat, Spain f Department of Radiology, University Medical Center Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands g Departament de Bioquı ´mica i Biologia Molecular, Unitat de Cie `ncies, Edifici Cs, Universitat Autono ` ma de Barcelona, 08193 Cerdanyola del Valle `s, Spain Received 1 December 2003; revised 26 May 2004 Available online 20 July 2004 Abstract The purpose was to objectively compare the application of several techniques and the use of several input features for brain tu- mour classification using Magnetic Resonance Spectroscopy (MRS). Short echo time 1 H MRS signals from patients with glioblas- tomas (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 per- formance of binary classifiers, while the percentage of correct classifications was used to evaluate the multiclass classifiers. The in- fluence 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 con- taining the most discriminatory information, and peak integrated values. Using L2-normalized complete spectra the automated bi- nary 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. Ó 2004 Elsevier Inc. All rights reserved. Keywords: Brain tumour classification; Short echo time MRS; Linear discriminant analysis; Least squares support vector machines 1. Introduction In vivo magnetic resonance spectroscopy (MRS) is a noninvasive technique which provides chemical informa- tion of metabolites present in living tissue and can be used to help characterize human brain tumours [1–3]. A histopathological analysis of a biopsy is the present gold standard for diagnosis of an abnormal brain mass suspected of being a brain tumour. A biopsy is not with- out risk of morbidity and mortality and cannot be car- ried out in all instances (e.g., brain stem tumours, paediatric tumours). Additionally, there are inherent inaccuracies in the gold standard [4] which can lead to www.elsevier.com/locate/jmr Journal of Magnetic Resonance 170 (2004) 164–175 1090-7807/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.jmr.2004.06.010 * Corresponding author. Fax: +32-16-321970. E-mail address: adevos@esat.kuleuven.ac.be (A. Devos).