Guillaume Pizaine

Vessel geometry modeling and segmentation using convolution surfaces and an implicit medial axis

In the context of vessel tree structures segmentation with implicit deformable models, we propose to exploit convolution surfaces to introduce a novel variational formulation, robust to bifurcations, tangential vessels and aneurysms. Vessels are represented by an implicit function resulting from the convolution of the centerlines of the vessels, modeled as a second implicit function, with localized kernels of continuously-varying scales. The advantages of this coupled representation are twofold. First, it allows for a joint determination of the vessels centerlines and radii, with a single model relevant for segmentation and visualization tasks. Second, it allows us to define a new shape constraint on the implicit function representing the centerlines, to enforce the tubular shape of the segmented objects. The algorithm has been evaluated on the segmentation of the portal veins in 20 CT-scans of the liver from the 3D-IRCADb-01 database, achieving an average recovery of 73% of the trees with fast computational times.

Fast solver for some computational imaging problems: A regularized weighted least-squares approach

In this paper we propose to solve a range of computational imaging problems under a unified perspective of a regularized weighted least-squares (RWLS) framework. These problems include data smoothing and completion, edge-preserving filtering, gradient-vector flow estimation, and image registration. Although originally very different, they are special cases of the RWLS model using different data weightings and regularization penalties. Numerically, we propose a preconditioned conjugate gradient scheme which is particularly efficient in solving RWLS problems. We provide a detailed analysis of the system conditioning justifying our choice of the preconditioner that improves the convergence. This numerical solver, which is simple, scalable and parallelizable, is found to outperform most of the existing schemes for these imaging problems in terms of convergence rate.

Accuracy of a Cone-Beam CT Virtual Parenchymal Perfusion Algorithm for Liver Cancer Targeting during Intra-arterial Therapy

PURPOSE To evaluate accuracy of virtual parenchymal perfusion (VPP) algorithm developed for targeting liver cancer during intra-arterial therapy (IAT) using cone-beam CT guidance. MATERIALS AND METHODS VPP was retrospectively applied to 15 patients who underwent IAT for liver cancer. Virtual territory (VT) was estimated after positioning a virtual injection point on nonselective dual-phase (DP) cone-beam CT images acquired during hepatic arteriography at the same position chosen for selective treatment. Targeted territory (TT) was used as the gold standard and was defined by parenchymal phase enhancement of selective DP cone-beam CT performed before treatment start. Qualitative evaluation of anatomic conformity between VT and TT was performed using a 3-rank scale (poor, acceptable, excellent) by 3 double-blinded readers. VT and TT were also quantitatively compared using spatial overlap-based (Dice similarity coefficient [DSC], sensitivity, and positive predictive value), distance-based (mean surface distance [MSD]), and volume-based (absolute volume error and correlation between pairwise volumes) metrics. Interreader agreement was evaluated for the 2 evaluation methods. RESULTS Eighteen DP cone-beam CT scans were performed. Qualitative evaluation showed excellent overlap between VT and TT in 88.9%-94.4%, depending on the readers. DSC was 0.78 ± 0.1, sensitivity was 80%, positive predictive value was 83%, and MSD was 5.1 mm ± 2.4. Absolute volume error was 15%, and R2 Pearson correlation factor was 0.99. Interreader agreement was good for both qualitative and quantitative evaluations. CONCLUSIONS VPP algorithm is accurate and reliable in identification of liver arterial territories during IAT using cone-beam CT guidance.

Implicit medial representation for vessel segmentation

In the context of mathematical modeling of complex vessel tree structures with deformable models, we present a novel level set formulation to evolve both the vessel surface and its centerline. The implicit function is computed as the convolution of a geometric primitive, representing the centerline, with localized kernels of continuously-varying scales allowing accurate estimation of the vessel width. The centerline itself is derived as the characteristic function of an underlying signed medialness function, to enforce a tubular shape for the segmented object, and evolves under shape and medialness constraints. Given a set of initial medial loci and radii, this representation first allows for simultaneous recovery of the vessels centerlines and radii, thus enabling surface reconstruction. Secondly, due to the topological adaptivity of the level set segmentation setting, it can handle tree-like structures and bifurcations without additional junction detection schemes nor user inputs. We discuss the shape parameters involved, their tuning and their influence on the control of the segmented shapes, and we present some segmentation results on synthetic images, 2D angiographies, 3D rotational angiographies and 3D-CT scans.

Segmentation-free and multiscale-free extraction of medial information using Gradient Vector Flow — Application to vascular structures

Gradient Vector Flow has become a popular method to recover medial information in medical imaging, in particular for vessels centerline extraction. This renewed interest has been motivated by its ability to process gray-scale images without prior segmentation. However, another interesting property lies in the diffusion process used to solve the underlying variational problem. We propose a method to recover scale information in the context of vascular structures extraction, relying on analytical properties of the Gradient Vector Flow only, with no multiscale analysis. Through simple one-dimensional considerations, we demonstrate the ability of our approach to estimate the radii of the vessels with an error of 10% only in the presence of noise and less than 3% without noise. Our approach is evaluated on convolved bar-like templates and is illustrated on 2D X-ray angiographic images.

Simultaneous image segmentation and medial structure estimation: application to 2D and 3D vessel tree extraction

We propose a variational approach which combines automatic segmentation and medial structure extraction in a single computationally efficient algorithm. In this paper, we apply our approach to the analysis of vessels in 2D X-ray angiography and 3D X-ray rotational angiography of the brain. Other variational methods proposed in the literature encode the medial structure of vessel trees as a skeleton with associated vessel radii. In contrast, our method provides a dense smooth level set map which sign provides the segmentation. The ridges of this map define the segmented regions skeleton. The differential structure of the smooth map (in particular the Hessian) allows the discrimination between tubular and other structures. In 3D, both circular and non-circular tubular cross-sections and tubular branching can be handled conveniently. This algorithm allows accurate segmentation of complex vessel structures. It also provides key tools for extracting anatomically labeled vessel tree graphs and for dealing with challenging issues like kissing vessel discrimination and separation of entangled 3D vessel trees.