International Journal of Engineering and Technology

This paper proposes a new method for classification of Capillary Electrophoretic Registers (CER) retrieved from cerebrospinal fluid sample taken from meningitis contaminated rats. The proposed approach applies several signal processing tools such as, wavelet analysis (WA), dynamic programming, principal component analysis (PCA) and support vector machines (SVM), for data pre-processing, feature extraction and CER classification. Furthermore, an algorithm is developed that detects zones in the CER where local energy variations between study groups (meningitis group and control group) are observed. This algorithm help us to identify the effects that Kliebsella Pneumonie (KP) bacteria produce in certain substances (aminoacids) that are part of the cerebrospinal fluid samples. It is shown that Meningitis disease can be effectively detected, analyzing the CER with the proposed methods. Futhermore, we show that exploiting the information related to the local energy variation improves the classification correctness rate up to 97.3%. This classification performance is obtained using least square SVM (LS-SVM) as classification tools and the parameterized CER representation proposed in this paper.