Multiparametric imaging of tumor boundaries using linear discriminant analysis
Benito-Vicente M, Lpez-Larrubia P, Barrios L, Garca-Martn M, Cerdn S, Sanchez-Garca P
Instituto de Investigaciones Biomdicas Alberto Sols CSIC/UAM
We provide a new method for tumor boundary evaluation based on the combination of T2, ADC and % MT (Magnetization Transfer) maps rather than on the use of the individual parameter maps. Rats bearing C6 gliomas were subjected to T2, ADC and % MT imaging providing the corresponding maps. Regions containing proliferative tumor, edema and contralateral healthy brain were selected and analyzed individually pixel by pixel, or combined using linear discriminant analysis (LDA). LDA improved the classifications of individual parameters allowing the identification of tumor tissue in 98.9 %, edema in 93.5% and healthy brain in 91.9% of the cases.