Ponente
Descripción
The incidence of skin cancer cases increases considerably every year. Cuba is not exempt from this growth being this type of cancer the most reported since 2020. Within these malignant skin tumors there are different classifications such as Melanoma and Non Melanoma skin cancer, the latter includes Basal Cell Carcinoma and Spinocellular Carcinoma. For its early detection a non-invasive technique called dermatoscopy is used to visualize deep cutaneous structures not detectable with the naked eye. In this work we study and compare essentially two ways for classification. An ensemble of neural networks is proposed to classify a dermatoscopic image into four classes: melanoma, basal cell carcinoma, squamous cell carcinoma and others. In addition, the "7-point list" procedure used by dermatologists to classify using feature vectors is followed. We experimented with different fusion methods that achieved a small increase in the number of identified cases. This classification is integrated into a web application that supports the diagnosis and follow-up of this type of lesions from the processing and analysis of dermatoscopic images.
