New approaches for computer-assisted skin cancer diagnosis

With the wide-spread availability of advanced digital cameras dermoscopy has become nowadays a standard technique to help the doctors taking diagnosis for many types of skin lesions. Further, computer-assisted techniques and image processing methods can be used for image filtering and for feature extraction and pattern recognition in the selected images. Apart from standard approaches based on geometrical features and color/pattern analysis we propose to enhance the computer-aided diagnostic tools by adding non-standard image decompositions and applying classification techniques based on statistical learning and model ensembling. Ensembles of classifiers based on the extended feature set show improved performance figures suggesting that the proposed approach could be used as powerful tool assisting medical diagnosis.