Juan M Garca-Gmez1, Irene Epifanio2, Margarida Juli-Sap3,4, Daniel Monlen3,5, Javier Vicente1, Salvador Tortajada1, Elies Fuster1, Angel Moreno-Torres6, Andrew Peet7,8, Franklyn Howe9, Bernardo Celda3,10, Carles Ars3,4, Montserrat Robles1
1ITACA-IBIME, Universidad Politcnica de Valencia, Valencia, Spain; 2Departament de Matemtiques, Universitat Jaume I, Valencia, Spain; 3CIBER de Bioingeniera, Biomateriales y Nanomedicina, Spain; 4Departament de Bioqumica i Biologia Molecular, Universitat Autnoma de Barcelona, Cerdanyola del Valls, Spain; 5Fundacin de Investigacin del Hospital Clnico Universitario de Valencia, Valencia, Spain; 6Research Department, Centre Diagnstic Pedralbes, Barcelona, Spain; 7University of Birmingham, Birmingham, UK; 8Birmingham Childrens Hospital, Birmingham, UK; 9St Georges Hospital Medical School, London, UK; 10Departamento de Qumica-Fsica, Universitat de Valncia, Valencia, Spain
When designing a Clinical Decision Support System for Brain Tumors based on MRS, it would be of interest to deal with any prospective case. Besides, due to the possible acquisition artifacts, baseline differences, or molecular artifacts in MRS-SV, the in-vivo MRS pattern may be heterogeneous within each diagnostic class. We present a possibilistic classifier evaluated on the largest multicenter database of MRS of brain tumors available to us based on FDA and subpattern analysis. It overperformed the classical approaches. The detected in-vivo MR spectral pattern subtypes could be useful for the interpretation of the natural heterogeneity of the diagnoses.