European Journal of Radiology xxx (2005) xxx–xxx Independent component analysis to proton spectroscopic imaging data of human brain tumours J. Pulkkinen a , A.-M. H¨ akkinen b , N. Lundbom c , A. Paetau d , R.A. Kauppinen e , Y. Hiltunen f,* a Department of Biomedical NMR, A.I. Virtanen Institute, University of Kuopio, Finland b Department of Oncology, Helsinki University Central Hospital, Finland c Department of Radiology, Helsinki University Central Hospital, Finland d Department of Pathology, Helsinki University Central Hospital, Finland e School of Biological Sciences, University of Manchester, England, UK f Department of Environmental Sciences, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland Received 2 March 2005; received in revised form 5 March 2005; accepted 8 March 2005 Abstract In proton magnetic resonance spectroscopic imaging ( 1 H MRSI), the recorded spectra are often linear combinations of spectra from different cell and tissue types within the voxel. This produces problems for data analysis and interpretation. A sophisticated approach is proposed here to handle the complexity of tissue heterogeneity in MRSI data. The independent component analysis (ICA) method was applied without prior knowledge to decompose the proton spectral components that relate to the heterogeneous cell populations with different proliferation and metabolism that are present in gliomas. The ability to classify brain tumours based on IC decomposite spectra was studied by grouping the components with histopathology. To this end, 10 controls and 34 patients with primary brain tumours were studied. The results indicate that ICA may reveal useful information from metabolic profiling for clinical purposes using long echo time MRSI of gliomas. © 2005 Elsevier Ireland Ltd. All rights reserved. Keywords: Independent component; Brain tumour; Magnetic resonance spectroscopic imaging 1. Introduction Proton magnetic resonance spectroscopy ( 1 H MRS) and spectroscopic imaging (MRSI) provide information that is independent of and additive to MRI. MRSI complements non-invasive diagnosis and treatment monitoring of human brain disorders [1–3]. However, the spectroscopic methods have not been widely used in clinical practice because of time-consuming data acquisition and lack of standardized spectral analysing methods. Special expertise is required for collection, analysis and interpretation of spectral data. Data acquisition and readout methods have improved, and mul- tislice MRSI methods now yield large anatomical coverage [4,5]. Also, several automated methods have been introduced * Corresponding author. Mobile: +358 40 7320355; fax: +358 17 162373. E-mail address: yrjo.hiltunen@uku.fi (Y. Hiltunen). to MRSI data analysis [6–9], i.e. artificial neural networks (ANN) [10–14] that can efficiently analyse biomedical MRS data. Automated ANN analysis using standardized protocols are expected to benefit clinical MRSI examinations [15]. Biological systems are inherently complex and produce spectral sets that are often linear combinations of spectra from different tissue types that comprise of several distinct cell populations within a voxel. This is particularly true in brain tumours with inherent histological and metabolic heterogeneity [16]. In vitro tumour biopsy studies have shown metabolite patterns that are indicative of histological type of malignancy [17,18]. For diagnostic purposes, it would be useful to decompose the spectral components and determine their concentrations in the composite spectrum. However, in vivo both the spectral compo- nents and their respective metabolite concentrations are 0720-048X/$ – see front matter © 2005 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2005.03.018