Pigmented network structure detection using semi-smart adaptive filters

This paper demonstrates a method for detecting pigment based dermatoscopic structure called pigment network. This structure is used in dermatoscopy as one of the criteria in clinical evaluation of pigmented skin lesions and can indicate if a lesion is of malignant nature. For detection process we have developed an adaptive filter, inspired by Swarm Intelligence (SI) optimization algorithms. The introduced filtering method is applied in a non-linear manner, to processed dermatoscopic image of a skin lesion. The non-linear approach derives from SI algorithms, and allows selective image filtering. In the beginning of filtration process, the filters (agents) are randomly applied to sections of the image, where each of them adapts its output based on the neighborhood surrounding it. Agents share its information with other agents that are located in immediate vicinity. This is a new approach to the problem of dermatoscopic structure detection, and it is highly flexible, as it can be applied to images without the need of previous pre-processing stage. This feature is highly desirable, mainly due to the fact that in most cases of computer aided diagnostic, input images need to be pre-processed (e.g.: brightness normalization, histogram equation, contrast enhancement, color normalization) and results of this can introduce unwanted artifacts, so this step need to be verified by human. Results of applying the introduced method can be used as one of the differential structures criteria for calculating the Total Dermatoscopy Score (TDS) of the ABCD rule.