COMPARATIVE STUDY FOR MELANOMA SEGMENTATION IN SKIN LESION IMAGES

Resumo

Melanoma is the leading cause of fatalities among skin can-cers and the discovery of the pathology in the early stagesis essential to increase the chances of cure. Computationalmethods through medical imaging are being developed tofacilitate the detection of melanoma. To interpret informa-tion in these images eciently, it is necessary to isolate theaected region. In our research, a comparison was made be-tween segmentation techniques, rstly a method based onthe Otsu algorithm, secondly the K-means clustering algo-rithm and nally,the U-net deep learning was developed.The tests performed on the PH2 images base had promisingresults, especially U-net.