Common carotid artery lumen automatic segmentation from cine fast spin echo magnetic resonance imaging

Abstract

Atherosclerosis is one of the main causes of stroke and is responsible for millions of deaths every year. Magnetic resonance (MR) is a common way of assessing carotid artery atherosclerosis. Cine fast spin echo (FSE) imaging is a new MR method that can now obtain dynamic image data of the carotid artery across the cardiac cycle. This work introduces a post-processing technique that segments the common carotid artery (CCA) wall-blood boundary across the cardiac cycle without human interaction. To the best of our knowledge, the proposed method is the first automatic technique proposed for segmenting cardiac cycle-resolved cine FSE images. The technique overcomes some inherent limitations of dynamic FSE images compared to static images (e.g., lower spatial resolution). It combines a priori knowledge about the size and shape of the CCA, with the max-tree data structure, random forest classifier and tie-zone watershed transform from identified internal and external markers to segment the vessel lumen. Segmentation performance was assessed using 3-fold cross validation with 15 cine FSE data sets in the test set per fold, each sequence consisting of 16 temporal bins over the cardiac cycle. The automatic segmentation was compared against manually segmented images. Our technique achieved an average Dice coefficient, sensitivity and false positive rate of 0.926 ± 0.005 (mean ± standard deviation), 0.909 ± 0.011 and 0.056 ± 0.003, respectively, compared to the majority voting consensus of manual segmentation from three experts.