Empirical mode decomposition of blood flow data for melanoma classification

Blood flow signals can characterize a notable feature correlating to melanoma, the most serious type of skin cancer. In this study, blood flow signals are decomposed into their instituent components in terms of intrinsic mode functions (IMFs) using the empirical mode decomposition (EMD). In particular, only the first two IMF components of blood flow signals ae examined. The blood flow data analyzed were recorded from four different lesions (melanoma, psoriasis, benign atypical nevi, and benign typical nevi) using laser Doppler flowmetry (LDF). The root mean square (RMS) of IMF components of blood flow signals is used a quantitative feature for melanoma classification. From the computational results, it is shown that there are significant differences between the RMS values of IMF components of blood flow signals recorded from the melanoma lesion and those recorded from other lesions. Furthermore, the computational results suggest that overall the first IMF components of blood flow signals provide a better performance on melanoma classification.