Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa. DISCENTE: JUSTINO DUARTE SANTOS DATA: 31/03/2022 HORA: 14:00 LOCAL: Sala Virtual do Google Meet TÍTULO: Glomerulosclerosis detection with pre-trained CNNs ensemble PALAVRAS-CHAVES: Transfer Learning, Kidney Disease, Computer-Aided Diagnosis, Image Analysis PÁGINAS: 22 GRANDE ÁREA: Ciências Exatas e da Terra ÁREA: Ciência da Computação RESUMO: Glomerulosclerosis is common kidney disease. It characterizes the advanced stages of most forms of primary kidney disease. Its accurate diagnosis relies on histological analysis of renal cortex biopsy. It is paramount to guide the appropriate treatment and minimize the chances of the disease progressing to chronic stages. This article presents an ensemble approach composed of five convolutional neural networks (CNNs) - VGG-19, Inception-V3, ResNet-50, DenseNet-201, and EfficientNet-B2 - to detect glomerulosclerosis in glomerulus images. We fine-tuned the CNNs and evaluated several configurations for the fully connected layers. In total, 25 different models were analyzed. These CNNs, individually, demonstrated effectiveness in the task; however, we verified that the union of these five well-known CNNs improves the hit rates and decreases the standard deviations of current techniques. The experiments were carried out in a dataset composed of 1,028 images, on which we apply data-augmentation techniques in the training set. The proposed CNNs ensemble achieved an accuracy of 99.0\% and kappa of 98.0\%, considered excellent. MEMBROS DA BANCA: Presidente - 641.754.563-68 - RODRIGO DE MELO SOUZA VERAS Externo à Instituição - LUCAS FERRARI DE OLIVEIRA - UFPR Externo à Instituição - PEDRO PEDROSA REBOUCAS FILHO - IFCE