Recursive tracking of vascular tree axes in 3D medical images

Abstract

Object This article describes a method for automated extraction of branching structures in three dimensional (3D) medical images. Materials and methods The algorithm recursively tracks branches and detects bifurcations by analyzing the binary connected components on the surface of a sphere that moves along the vessels. Local segmentation within the sphere is performed using a clustering algorithm based on both geometric and intensity information. It minimizes a combination of the intra-class intensity variances and of the inertia moment of the "vessel" class, which emphasizes the cylindrical structures. The algorithm was applied to 16 MRA and 12 CTA 3D images of different anatomic regions. Its capability of extracting all the branches and avoiding spurious detections was evaluated by comparing the number of extracted branches with the number of branches found by visual inspection of the datasets. Its reproducibility and sensitivity to parameter variation were also assessed. Results With a fixed parameter setting, 68 out of 286 perceptible branches were missed or partly extracted and 11 spurious branches were obtained. Increasing the weight of the geometric criterion helped in tracking

Angiographic Image Analysis

“…It can also be performed based on a local optimisation of 3D models [102,112], which may also lead to the determination of the vessel axis orientation. More classically, the next tracking point may be determined according to the best fit of a sphere modelling the vessel into the image [12,47,69].…”

Geodesic voting for the automatic extraction of tree structures. Methods and applications

“…While there are some studies dedicated to the segmentation of microglia structures, there are many studies dedicated to the extraction of vas-cular or airway trees. For a review of such methods see [3,4,5,6,7,8].…”

Probabilistic branching node detection using hybrid local features

“…In [11], Adaboost [13] is applied on features for classification of lung bronchovascular anatomy. Tracking-based approaches have also been applied for vessel detection [14,10]. A thorough survey on vessel detection is given in [15].…”