Justinas Besusparis

This paper introduces GlomExtractor, a tool for extracting glomerular patches from kidney biopsy images, and shows how processing strategies affect machine learning performance.

This study uses shape analysis to detect and prioritize errors in deep learning-based glomerular segmentation, enabling more efficient correction by pathologists.

The paper investigates the feasibility of unsupervised clustering for classifying globally sclerosed versus non-globally sclerosed glomeruli using pre-trained CNN models, achieving over 95% accuracy in most datasets.