Classification of Biomedical Spectra Using Fuzzy Interquartile Encoding and Stochastic Feature Selection

Accurate classification of biomedical spectra is often difficult due to the large number of features, which tends to have a confounding effect. We present a strategy where the original spectral feature space is transformed using a fuzzy set theoretic method, which analyzes the features' interquartile ranges, coupled with a stochastic feature selection mechanism, which identifies highly discriminatory feature subsets. We demonstrate the effectiveness of this strategy using biofluid data acquired from a magnetic resonance spectrometer