Automatic lumen size measurement in carotid atherosclerosis with phase sensitive magnetic resonance angiography(MRA) using self-trained radial basis function kernel support vector machine

A self-trained algorithm based on Ostu’s method and a radial basis function (RBF) kernel support vector machine (SVM) model was developed for automatic lumen detection and quantification for the negative polarity map of SNAP magnetic resonance angiography(MRA). Based on an analysis of 15 arteries with carotid stenosis, the proposed automatic lumen segmentation algorithm demonstrated good agreement with manual lumen segmentation of SNAP MRA (intraclass correlation coefficient (ICC=0.95). The automated method also had good agreement with manual segmentation of CE-MRA (ICC = 0.90), which was comparable to the agreement between manually segmented SNAP MRA and CE-MRA (ICC = 0.93).