Meeting Banner
Abstract #0981

Age-Filtered MRS Classifier to Overcome the Differences in Childhood and Adulthood Brain Tumours

Javier Vicente1, Juan Miguel Garca-Gmez1, Salvador Tortajada1, Elies Fuster-Garcia1, Antoni Capdevila2, Andrew Charles Peet3, Bernardo Celda4, Monserrat Robles1

1IBIME-ITACA, Universidad Politcnica de Valencia, Valencia, Spain; 2Hospital Sant Joan de Du, Barcelona, Spain; 3Birmingham Children's Hospital NHS Foundation Trust, University of Birmingham, Birmingham, UK; 4Departamento de Qumica Fsica, Universidad de Valencia, Valencia, Spain

Several studies confirm that the nature of child Brain Tumours (BT) may be totally different from adults. We have developed classifiers for adulthood and childhood BT and compared performances with independent test sets of children and adult patients using 489 (93 children, 396 adults) SV 1H-MRS at 1.5T histopathologically diagnosed brain tumor cases. Performance dramatically lowered when children classifiers were tested with an adult test set and vice-versa. A filter based on the normal probability density function of the training datasets age can successfully overcome these differences and obtain a classifier that globally behaves as predicted by the training performance.