A method for automatic venous vessel segmentation is presented that uses a Random Forest classifier supplied with a number of appearance and shape features computed separately from magnitude images, phase images and QSMs of a multi-echo T2*-weighted GE scan. The importance of each feature, and thus each echo, is investigated. The approach was tested on whole-brain 7T scans of four subjects, two of which were manually annotated, and was effective in segmenting both internal and surface veins.
This abstract and the presentation materials are available to members only; a login is required.