The INTERPRET Decision-Support System version 3.0 for evaluation of Magnetic Resonance Spectroscopy data from human brain tumours and other abnormal brain masses

Table 1 DSS 3.0 Datasets (number of cases) and classifiers they served to train.

				*Superclasses used for training the classifiers*
*Dataset*	*Spectra available*	*Classification*		*Meningioma*	*Aggressive*	*Low-grade glial*	*Subtotal*	*Cases from other classes*	*Total*
		*problem*	*Method(s)*
INTERPRET	Short TE (20-32 ms)	MCTT	LDA	58	124	35	217	87	304
	Long TE (135-144 ms)	MCTT	LDA	55	109	31	195	71	266
	Short + Long TE	MCTT	LDA	55	109	31	195	71	266
				*Pseudotumoural*	*Tumoural*	*Normal* *brain*	*Subtotal*	*Cases from other classes*	*Total*
IDI-Bellvitge	Short TE (30 ms)	T vs.PS	LDA	19	46	5	70	0	70
	Long TE (135 ms)	T vs.PS	LDA	19	46	5	70	0	70
	Short + Long TE	T vs.PS	LDA and Ratios[14]	19	46	5	70	0	70

Specifications of the two main datasets included in the system, the INTERPRET and the IDI-Bellvitge dataset. Each dataset has short and long TE spectra and both short and long TE spectra concatenated. Different classification problems have been analysed with these datasets. Furthermore, in the IDI-Bellvitge dataset, the same classification problem has been solved in two different ways, either by an LDA classification or by a peak height ratio-based classifier [14]. The INTERPRET dataset contains cases used for training the classifiers as well as from other less common types of tumours. Note that for INTERPRET the number of cases available at short and long TE is different. MCTT: Most common tumour types; T vs.PS: Tumour vs. pseudotumoural disease. See [4, 13] for further details on superclass definition.

ISSN: 1471-2105