Support Vector Machine Classification of Non-melanoma Skin Lesions based on Fluorescence Lifetime Imaging Microscopy.

Early diagnosis of malignant skin lesions is critical for prompt treatment and a clinical prognosis of skin cancers. However, it is difficult to precisely evaluate the development stage of nonmelanoma skin cancers because they are derived from the same tissues as a result of the uncontrolled growth of abnormal squamous keratinocytes in the epidermis layer of the skin. In the present study, we developed a linear-kernel support vector machine (LSVM) model to distinguish basal cell carcinoma (BCC) from actinic keratosis (AK) and Bowen's disease (BD). The input parameters of the LSVM model consist of appropriate lifetime components and entropy values, which were extracted from two-photon fluorescence lifetime imaging of hematoxylin and eosin (H&E)-stained biopsy sections. Different features used as inputs for SVM training were compared and evaluated. In constructing the SVM models, features obtained from the lifetime (tau(2)) of the second component were found to be significantly more predictive than the average fluorescence lifetime (tau(m)) in terms of diagnostic accuracy, sensitivity, and specificity. The above findings were confirmed on the basis of the receiver operating characteristic (ROC) curves of diagnostic models. Shannon entropy was added to the SVM models as an independent feature to further improve the diagnostic accuracy. Therefore, fluorescence lifetime analysis and entropy calculations can provide highly informative features for the accurate detection of skin neoplasm disorders. In summary, fluorescence lifetime imaging microscopy (FLIM) combined with the SVM classification exhibited great potential for developing an effective computer-aided diagnostic criterion and accurate cancer detection in dermatology.