Classification of Skin Lesions using CNN
Palavras-chave:
Convolutional Neural Networks, Skin Lesion Classification, Image ClassificationResumo
Various computational methodologies can be found in the specialized literature, with different
applications, among which the classification of skin lesions in dermoscopic images stands out in this work.
Although the initial analysis of skin lesions was performed using a set of visual rules known as the ABCDErule
(Asymmetry, Borders, Lesion Color, Diameter, and Evolution), the performance of this visual analysis is
influenced by factors such as lighting variation during image capture, the presence of artifacts that cause noise,
and the specialist's eye strain during image analysis. A mistaken initial analysis can delay the development of an
adequate treatment plan, affecting the effectiveness of this treatment. In the task of computational recognition of
elements in an image, the Convolutional Neural Network (CNN)stands out. In this context, this work presents the
results of the application of CNN ResNet for the identification of melanomas. To carry out this work, TensorFlow
and a database with 9144 images were used. The results were promising, reaching approximately 75% accuracy.
