Uma Abordagem Totalmente Automática para Diagnóstico de COVID-19 em Imagens de TC: Integrando Segmentação de Pulmões, Fine-Tuning e Visualização em Grad-CAM

• Ano de defesa: 2025
• Autor(a) principal: Santos, Matheus Araújo dos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Transferência de Aprendizado; COVID-19; Tomografia Computadorizada (TC); Aprendizado Profundo; Segmentação Profunda; Ajuste Fino; Fine-Tuning; Transfer Learning; Computed Tomography (CT); Deep Learning; Deep Segmentation; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://repositorio.ufc.br/handle/riufc/80552

The pandemic caused by the COVID-19 virus has highlighted the need for efficient and automated medical diagnostic tools to assist healthcare professionals in the fast and accurate identification of the disease. This study introduces a fully automatic approach for classifying COVID-19 in thoracic computed tomography (CT) images from the SARS-CoV-2 CT-scan dataset, employing deep learning, automatic lung segmentation, and fine-tuning models to the specific problem. In the first experiment, transfer learning experiments were performed, which revealed that the deep MobileNet model was the most effective architecture for extracting features from these exams. After choosing the model, Detectron2 was integrated for lung segmentation. With the segmented exams, the previously selected MobileNet model was fine-tuned. As a result, the classification performance improved significantly, which was confirmed through a visual analysis of the model's interpretability using Grad-CAM. As a result, the proposed fully automatic method achieved superior evaluation metrics compared to other methods in the literature, reaching values of 99.60% for accuracy, precision, and F1-Score, along with 99.59% for recall and 99.20% for the Matthews Correlation Coefficient (MCC). This demonstrated that the network focused on the lung regions with segmented images and the refined model, improving the model's diagnostic understanding.