AUTOMATIC DETECTION OF CEREBRAL MICROBLEEDS VIA DEEP LEARNING BASED 3D FEATURE REPRESENTATION

Clinical identification and rating of the cerebral microbleeds (CMBs) are important in vascular diseases and dementia diagnosis. However, manual labeling is time-consuming with low reproducibility. In this paper, we present an automatic method via deep learning based 3D feature representation, which solves this detection problem with three steps: candidates localization with high sensitivity, feature representation, and precise classification for reducing false positives. Different from previous methods by exploiting low-level features, e.g., shape features and intensity values, we utilize the deep learning based high-level feature representation. Experimental results validate the efficacy of our approach, which outperforms other methods by a large margin with a high sensitivity while significantly reducing false positives per subject