Detection of basal cell carcinoma by automatic classification of Confocal Raman spectra

Abstract

Raman spectroscopy has strong potential for providing noninvasive dermatological diagnosis of skin cancer. In this study, we investigated various classification methods with confocal Raman spectra for the detection of basal cell carcinoma (BCC), which is one of the most common skin cancer. The methods include maximum a posteriori (MAP) probability, probabilistic neural networks (PNN), k-nearest neighbor (KNN), multilayer perceptron networks (MLP), and support vector machine (SVM). The classification framework consists of preprocessing of Raman spectra, feature extraction, and classification. In the preprocessing step, a simple half Harming method is adopted to obtain robust features. Classification results involving 216 spectra gave about 97% true classification rate in case of MLP and SVM, which is an evident proof of the effectiveness of confocal Raman spectra for BCC detection. In addition to it, spectral regions, which are important for classification, are examined by sensitivity analysis.