Extracting morphological high-level intuitive features (HLIF) for enhancing skin lesion classification

Feature extraction of skin lesions is necessary to provide automated tools for the detection of skin cancer. High-level intuitive features (HLIF) that measure border irregularity of skin lesion images obtained with standard cameras are presented. Existing feature sets have defined many low-level unintuitive features. Incorporating HLIFs into a set of low-level features gives more semantic meaning to the feature set, and allows the system to provide intuitive rationale for the classification decision. Promising experimental results show that adding a small set of HLIFs to the large state-of-the-art low-level skin lesion feature set increases sensitivity, specificity, and accuracy, while decreasing the cross-validation error.