An approach to the elimination of inter-individual variability in tumor detection

Abstract

The inter-individual variability among biological samples is different from the random error produced in experimental examination because that reveals the characteristic of samples themselves, and often works on the results of analysis and diagnosis in biomedicine research. In this paper, the methodology of signal processing and the technique of wavelet analysis were introduced to decrease the inter-individual variability of samples in tumor detection. The 16 antibody biomarkers for tumor detection were determined by means of the BioPlex system based on 199 plasma samples, and were regarded as 16 signal channels. Then, as a noise signal, the inter-individual variability of samples was reduced by wavelet transforms, which was evaluated by a rank sun hypothesis test, receiver operating characteristic curve analysis and classification tree models. After inter-individual variability was removed using the wavelet transform, the tumor detection algorithms produced results that had more accuracy and greater reliability. Our study provided a novel approach to the pretreatment of biomedical data.