Application of Hidden Markov Models to Melanoma Diagnosis

摘要

In this paper we present a clinical decision support system for melanoma diagnosis.Unlike other systems based on diagnosis obtained just from one image, in this work it is em-ployed an image set, that represents the evolution of damaged tissues, taken in different in-stances of time (for example once a month). Therefore, the system analyses the image sequenceextracting the affected area and using the gradient orientations histogram of each area to com-pose a description which allows achieving a decision about the input. Hidden Markov Modelsare proposed as classify method, obtaining classification rates of 77%.