Análise e classificação de lesões de pele por meio de estratégias de aprendizado ativo

• Ano de defesa: 2022
• Autor(a) principal: Batista, Lucas Garcia
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Cornelio Procopio; Brasil; Programa de Pós-Graduação em Informática; UTFPR
• Palavras-chave em Português: Aprendizado do computador; Câncer - Diagnóstico; Processamento de imagens; Machine learning; Cancer - Diagnosis; Image processing; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; Ciência da Computação
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/30266

According to the National Cancer Institute, among all malignant tumors, non-melanoma skin cancer and melanoma are the most frequent in Brazil. Despite having a lower incidence, compared to the non-melanoma type, the melanoma type has accelerated growth and greater lethality. Early diagnosis helps in the treatment and cure of patients. Several studies have been carried out in recent years in the area of computer vision to assist in the early diagnosis of skin cancer. Deep learning approaches, using convolutional neural network architectures through transfer of learning, have been widely adopted. Despite presenting significant results, a large amount of annotated training data is needed. However, obtaining labeled data by a specialist is not feasible in the case of large data sets. Active learning approaches make it possible to select a small set of more informative data for classifier training. Therefore, this research project explores active learning approaches, including the proposal and validation of selection strategies of more informative samples for the classification of skin lesions.