Automatic perivascular spaces (PVS) segmentation from magnetic resonance images enables whole brain morphometric assessment of glia-lymphatic network. While conventional deep learning segmentations are shown to produce reliable results, they heavily rely on the presence of ground truth for training. In this work we introduce an unsupervised learning method for PVS segmentation, by combining a rule-based image processing approach with a deep learning algorithm. The experiment results showed that the proposed method increased segmentation accuracy by effectively increasing the true positive rate compared to the rule-based method and decreasing false positive rate compared to the deep learning method.
This abstract and the presentation materials are available to members only; a login is required.