Francesc
Xavier Aymerich1,2, Julio Alonso1, Manuel Comabella3,
Miquel E. Cabaas4, Mar Tintor3, Xavier Montalban3,
Alex Rovira1
1Unitat
RM Vall Hebron (IDI), Hospital Vall Hebron, Barcelona, Spain; 2Enginyeria
de Sistemes, Automtica i Informtica
Industrial, Universitat Politcnica de Catalunya, Barcelona, Spain; 3Unitat Neuroimmunologia Clnica, CEM Cat, Hospital
Vall Hebron, Barcelona, Spain; 4Servei Ressonncia Magntica
Nuclear, Universitat Autnoma de Barcelona, Cerdanyola del Valls, Barcelona,
Spain
The
purpose was the design of a fuzzy classifier to differentiate among primary
progressive multiple sclerosis (MS), relapsing remitting MS and non-MS
conditions by 1H-NMR spectroscopy of cerebrospinal fluid. The design
considered the fusion of classifiers based on fuzzy decision trees. We
considered three different datasets (aliphatic region, aromatic region and
the aggregation of both regions). We evaluated for each dataset the
classifier performance by means of two classification quality indexes
(correctness and robustness). Results showed mean classification correctness
and robustness in the intervals [0.92,1] and [0.34,0.50] respectively. The
aggregation of aliphatic and aromatic regions provided the best results.