An Enhancement on the Detection of Skin Cancer Based on Band Approach of Confocal Raman Spectra

Abstract

In this study, we investigated skin cancer classification methods based on confocal Raman spectroscopy using maximum a posterior probability, fuzzy algorithm and support vector machine (SVM). The preprocessing steps were consisted of data normalization with minmax method and dimension reduction with principal components analysis. To enhance the classification performance, we divided Raman spectra into three significant protein bands, i.e., Amide I mode composed of two bands, normal and BCC tissue and Amide III mode. All the features were extracted independently. Classification results involving 216 spectra showed 97.2 % true classification in case of SVM, which is an evident proof of the effectiveness of three band approach for skin cancer detection.