Deep-learning-based membranous nephropathy classification and Monte-Carlo dropout uncertainty estimation

Abstract

Membranous Nephropathy (MN) is one of the most common glomerular diseases that cause adult nephrotic syndrome. To assist pathologists on MN classification, we evaluated three deep-learning-based architectures, namely, ResNet-18, DenseNet and Wide-ResNet. In addition, to accomplish more reliable results, we applied Monte-Carlo Dropout for uncertainty estimation. We achieved average F1-Scores above 92% for all models, with Wide-ResNet obtaining the highest average F1-Score (93.2%). For uncertainty estimation on Wide-ResNet, the uncertainty scores showed high relation with incorrect classifications, proving that these uncertainty estimates can support pathologists on the analysis of model predictions.

Aprendizado Profundo na Classificação de Lesões Crescentes Glomerulares: modelos e condições

“…al. [9] fazem uso de três arquiteturas baseadas em deep learning para a classificac ¸ão de nefropatia membranosa, em um dataset com imagens de glomérulos normais, de glomérulos com nefropatia membranosa, com hipercelularidade e com esclerose, e obtiveram resultados de F1-score acima de 92% para os três modelos.…”

“…al. [9] foram além de tarefa de distinguir imagens normais de imagens de uma certa lesão, e consideraram as classes de membranosa, outras lesões, e normais. Somente Uchino et.…”