Identification of novel serum biomarkers of hepatocellular carcinoma (HCC) is needed for early-stage disease detection and to improve patients' survival. The aim of this study was to evaluate the potential of serum Fourier transform infrared (FTIR) spectroscopy for differentiating sera from cirrhotic patients with and without HCC. Serum samples were collected from 2 sets of patients: cirrhotic patients with HCC (n = 39) and without HCC (n = 40). The FTIR spectra (10 per sample) were acquired in the transmission mode, and data homogeneity was tested by cluster analysis to exclude outliers. After data preprocessing by extended multiplicative signal correction and principal component analysis, the Support Vector Machine (SVM) method was applied using a leave-one-out cross-validation algorithm to classify the spectra into 2 classes of cirrhotic patients with and without HCC. When SVM was applied to all spectra (n = 790), the sensitivity and the specificity for the diagnosis of HCC were, respectively, 82.02% and 82.5%. When applied to the subset of spectra excluding the outliers (n = 739), SVM classification led to a sensitivity and specificity of 87.18% and 85%, respectively. Using median spectra for each patient instead of all replicates, the sensitivity and specificity were 84.62% and 82.50%, respectively. The overall accuracy rate was 82%-86%. In conclusion, this study suggests that FTIR spectroscopy combined with advanced methods of pattern analysis shows potential for differentiating sera from cirrhotic patients with and without HCC.
Keywords: AFP; AFSSAPS; Agence Française de Sécurité Sanitaire des Produits de Santé; EMSC; Extended Multiplicative Signal Correction; FTIR; Fourier transform infrared; HCA; HCC; Hierarchical Cluster Analysis; LOOCV; PC; PCA; SVM; alpha-fetoprotein; hepatocellular carcinoma; leave-one-out cross-validation; principal component; principal component analysis; support vector machine.
Copyright © 2013 Mosby, Inc. All rights reserved.