Vascular centerline extraction in 3D MR angiograms to optimize acquisition plane for blood flow measurement by phase contrast MRI

Cerebral blood flow (CBF) is an important parameter in the diagnosis and choice of therapeutic strategy of the cerebro-vascular disorders. Reliable quantification of the CBF requires a preliminary determination of an imaging plane close to orthogonal cross-sections of the vessel of interest. This can be calculated for each vessel separately, but a simultaneous determination for the four principal arteries irrigating the brain would improve the acquisition process. This article presents two algorithms devised for this purpose. The first one determines the centerline of each of the four vessels of interest. The centerline extraction is done by an iterative prediction-estimation tracking technique based on a multi-scale analysis of image moments and on a shape model close to “snakes”. The second algorithm seeks the plane that minimizes a criterion of orthogonality with respect to these centerlines. The method was evaluated on 3D images of 35 patients. All but one centerline, out of 140 arteries, were correctly extracted. For each patient the algorithm found a plane with an orthogonality error not exceeding 12° per artery. The operating time (interaction + computation) was within 1 min by patient on a standard PC.