Arterial tortuosity measurement system for examining correlations with vascular disease

High arterial tortuosity, or twistedness, is a sign of many vascular diseases. Some ocular diseases are clinically diagnosed in part by assessment of increased tortuosity of ocular blood vessels. Increased arterial tortuosity is seen in other vascular diseases but is not commonly used for clinical diagnosis. This study develops the use of existing magnetic resonance angiography (MRA) image data to study arterial tortuosity in a range of arteries of hypertensive and intracranial aneurysm patients. The accuracy of several centerline extraction algorithms based on Dijkstra's algorithm was measured in numeric phantoms. The stability of the algorithms was measured in brain arteries. A centerline extraction algorithm was selected based on its accuracy. A centerline tortuosity metric was developed using a curve of tortuosity scores. This tortuosity metric was tested on phantoms and compared to observer-based tortuosity rankings on a test data set. The tortuosity metric was then used to measure and compare with negative controls the tortuosity of brain arteries from intracranial aneurysm and hypertension patients. A Dijkstra based centerline extraction algorithm employing a distance-from-edge weighted center of mass (DFE-COM) cost function of the segmented arteries was selected based on generating 15/16 anatomically correct centerlines in a looping artery iv compared to 15/16 for the center of mass (COM) cost function and 7/16 for the inverse modified distance from edge cost function. The DFE-COM cost function had a lower root mean square error in a lopsided phantom (0.413) than the COM cost function (0.879). The tortuosity metric successfully ordered electronic phantoms of arteries by tortuosity. The tortuosity metric detected an increase in arterial tortuosity in hypertensive patients in 13/13 (10/13 significant at α = 0.05). The metric detected increased tortuosity in a subset of the aneurysm patients with Loeys-Dietz syndrome (LDS) in 7/7 (three significant at α = 0.001). The tortuosity measurement combination of the centerline algorithm and the distance factor metric tortuosity curve was able to detect increases in arterial tortuosity in hypertensives and LDS patients. Therefore the methods validated here can be used to study arterial tortuosity in other hypertensive population samples and in genetic subsets related to LDS.