TOWARDS A METHOD FOR AUTOMATED CLASSIFICATION OF H-1 MRS SPECTRA FROMBRAIN-TUMORS

Recent studies have shown that MRS can substantially improve the non-i nvasive categorization of human brain tumours. However, in order for M RS to be used routinely by clinicians, it will be necessary to develop reliable automated classification methods that can be fully validated . This paper is in two parts: the first part reviews the progress that has been made towards this goal, together with the problems that are involved in the design of automated methods to process and classify th e spectra. The second part describes the development of a simple proto type system for classifying H-1 single voxel spectra, obtained at an e cho time (TE) of 135 ms, of the four most common types of brain tumour (meningioma (MM), astrocytic (AST), oligodendroglioma (OD) and metast asis (ME)) and cysts. This system was developed in two stages: firstly , an initial database of spectra was used to develop a prototype class ifier, based on a linear discriminant analysis (LDA) of selected data points. Secondly, this classifier was tested on an independent test se t of 15 newly acquired spectra, and the system was refined on the basi s of these results. The system correctly classified all the non-astroc ytic tumours. However, the results for the the astrocytic group were p oorer (between 55 and 100%, depending on the binary comparison). Appro ximately 50% of high grade astrocytoma (glioblastoma) spectra in our d ata base showed very little lipid signal, which may account for the po orer results for this class. Consequently, far the refined system, the astrocytomas were subdivided into two subgroups for comparison agains t other tumour classes: those with high lipid content and those withou t. (C) 1998 John Wiley & Sons, Ltd.