Toward automated detection of microbleeds with anatomical scale localization using deep learning

Jun-Ho Kim, Young Noh, Haejoon Lee, Seul Lee, Woo-Ram Kim, Koung Mi Kang, Eung-Yeop Kim, Mohammed A. Al-masni, Donghyun Kim

Abstract: Highlights•A novel framework for microbleed detection and anatomical localization in 3D images.•Our hard sample prototype learning (HSPL) makes a loss focusing on hard samples.•HSPL boosts the performance of detection model.•Our single-stage detector outperforms the state-of-the-art two-stage detector.•Anatomical localization task reduces the false positives and increases a precision.