A single-stage detector of cerebral microbleeds using 3D feature fused region proposal network (FFRP-Net)

Jun-Ho Kim, Mohammed A. Al-masni, Haejoon Lee, Yoon-Seok Choi, Dong-Hyun Kim

Abstract: Cerebral Microbleeds (CMBs) are chronic deposits of small blood products in the brain tissues, which have explicit relation to cerebrovascular diseases including cognitive decline, intracerebral hemorrhage, cerebral infarction. However, manual detection of the CMBs is a time-consuming and error-prone process. In this paper, we propose an efficient single-stage deep learning framework for automatic detection of the CMBs. The framework consists of a 3D U-Net and a region proposal network employing a feature fusing method (FFRP-Net) for detecting small objects. This model utilizes Susceptibility-Weighted Imaging (SWI) and phase images as 3D input to efficiently capture 3D contextual information. The performance of the proposed FFRP-Net records a sensitivity of 94.66% and an average number of false positives per subject (FPavg) of 8.82.