Roberto Ardon

Automatic Detection and Segmentation of Kidneys in 3D CT Images Using Random Forests

Kidney segmentation in 3D CT images allows extracting useful information for nephrologists. For practical use in clinical routine, such an algorithm should be fast, automatic and robust to contrast-agent enhancement and fields of view. By combining and refining state-of-the-art techniques (random forests and template deformation), we demonstrate the possibility of building an algorithm that meets these requirements. Kidneys are localized with random forests following a coarse-to-fine strategy. Their initial positions detected with global contextual information are refined with a cascade of local regression forests. A classification forest is then used to obtain a probabilistic segmentation of both kidneys. The final segmentation is performed with an implicit template deformation algorithm driven by these kidney probability maps. Our method has been validated on a highly heterogeneous database of 233 CT scans from 89 patients. 80% of the kidneys were accurately detected and segmented (Dice coefficient > 0.90) in a few seconds per volume.

Fuzzy region competition: a convex two-phase segmentation framework

Philips Medical Systems Research Paris,51 rue Carnot, B.P. 301, F-92156 SURESNES Cedex, FRANCE{benoit.mory, roberto.ardon}@philips.comAbstract. We describe a novel framework for two-phase image segmen-tation, namely the Fuzzy Region Competition. The functional involvedin several existing models related to the idea of Region Competition isextended by the introduction of a fuzzy membership function. The newproblem is convex and the set of its global solutions turns out to be stableunder thresholding, operation that also provides solutions to the corre-sponding classical formulations. The advantages are then shown in thepiecewise-constant case. Finally, motivated by medical applications suchas angiography, we derive a fast algorithm for segmenting images into twonon-overlapping smooth regions. Compared to existing piecewise-smoothapproaches, this last model has the unique advantage of featuring closed-form solutions for the approximation functions in each region based onnormalized convolutions. Results are shown on synthetic 2D images andreal 3D volumes.

Quantitative and Volumetric European Association for the Study of the Liver and Response Evaluation Criteria in Solid Tumors Measurements: Feasibility of a Semiautomated Software Method to Assess Tumor Response after Transcatheter Arterial Chemoembolization

PURPOSE To show that hepatic tumor volume and enhancement pattern measurements can be obtained in a time-efficient and reproducible manner on a voxel-by-voxel basis to provide a true three-dimensional (3D) volumetric assessment. MATERIALS AND METHODS Magnetic resonance (MR) imaging data obtained from 20 patients recruited for a single-institution prospective study were retrospectively evaluated. All patients had a diagnosis of hepatocellular carcinoma (HCC) and underwent drug-eluting beads (DEB) transcatheter arterial chemoembolization for the first time. All patients had undergone contrast-enhanced MR imaging before and after DEB transcatheter arterial chemoembolization; poor image quality excluded 3 patients, resulting in a final count of 17 patients. Volumetric RECIST (vRECIST) and quantitative EASL (qEASL) were measured, and segmentation and processing times were recorded. RESULTS There were 34 scans analyzed. The time for semiautomatic segmentation was 65 seconds±33 (range, 40-200 seconds). vRECIST and qEASL of each tumor were computed<1 minute for each. CONCLUSIONS Semiautomatic quantitative tumor enhancement (qEASL) and volume (vRECIST) assessment is feasible in a workflow-efficient time frame. Clinical correlation is necessary, but vRECIST and qEASL could become part of the assessment of intraarterial therapy for interventional radiologists.

Fast Constrained Surface Extraction by Minimal Paths

In this paper we consider a new approach for single object segmentation in 3D images. Our method improves the classical geodesic active surface model. It greatly simplifies the model initialization and naturally avoids local minima by incorporating user extra information into the segmentation process. The initialization procedure is reduced to introducing 3D curves into the image. These curves are supposed to belong to the surface to extract and thus, also constitute user given information. Hence, our model finds a surface that has these curves as boundary conditions and that minimizes the integral of a potential function that corresponds to the image features. Our goal is achieved by using globally minimal paths. We approximate the surface to extract by a discrete network of paths. Furthermore, an interpolation method is used to build a mesh or an implicit representation based on the information retrieved from the network of paths. Our paper describes a fast construction obtained by exploiting the Fast Marching algorithm and a fast analytical interpolation method. Moreover, a Level set method can be used to refine the segmentation when higher accuracy is required. The algorithm has been successfully applied to 3D medical images and synthetic images.

Variational Segmentation using Fuzzy Region Competition and Local Non-Parametric Probability Density Functions

We describe a novel variational segmentation algorithm designed to split an image in two regions based on their intensity distributions. A functional is proposed to integrate the unknown probability density functions of both regions within the optimization process. The method simultaneously performs segmentation and non-parametric density estimation. It does not make any assumption on the underlying distributions, hence it is flexible and can be applied to a wide range of applications. Although a boundary evolution scheme may be used to minimize the functional, we choose to consider an alternative formulation with a membership function. The latter has the advantage of being convex in each variable, so that the minimization is faster and less sensitive to initial conditions. Finally, to improve the accuracy and the robustness to low-frequency artifacts, we present an extension for the more general case of local space-varying probability densities. The approach readily extends to vectorial images and 3D volumes, and we show several results on synthetic and photographic images, as well as on 3D medical data.

Semiautomatic volumetric tumor segmentation for hepatocellular carcinoma: comparison between C-arm cone beam computed tomography and MRI.

RATIONALE AND OBJECTIVES To evaluate the precision and reproducibility of a semiautomatic tumor segmentation software in measuring tumor volume of hepatocellular carcinoma (HCC) before the first transarterial chemo-embolization (TACE) on contrast-enhancement magnetic resonance imaging (CE-MRI) and intraprocedural dual-phase C-arm cone beam computed tomography (DP-CBCT) images. MATERIALS AND METHODS Nineteen HCCs were targeted in 19 patients (one per patient) who underwent baseline diagnostic CE-MRI and an intraprocedural DP-CBCT. The images were obtained from CE-MRI (arterial phase of an intravenous contrast medium injection) and DP-CBCT (delayed phase of an intra-arterial contrast medium injection) before the actual embolization. Three readers measured tumor volumes using a semiautomatic three-dimensional volumetric segmentation software that used a region-growing method employing non-Euclidean radial basis functions. Segmentation time and spatial position were recorded. The tumor volume measurements between image sets were compared using linear regression and Students t-test, and evaluated with intraclass-correlation analysis (ICC). The inter-rater Dice similarity coefficient (DSC) assessed the segmentation spatial localization. RESULTS All 19 HCCs were analyzed. On CE-MRI and DP-CBCT examinations, respectively, 1) the mean segmented tumor volumes were 87 ± 8 cm(3) (2-873) and 92 ± 10 cm(3) (1-954), with no statistical difference of segmented volumes by readers of each tumor between the two imaging modalities and the mean time required for segmentation was 66 ± 45 seconds (21-173) and 85 ± 34 seconds (17-214) (P = .19); 2) the ICCs were 0.99 and 0.974, showing a strong correlation among readers; and 3) the inter-rater DSCs showed a good to excellent inter-user agreement on the spatial localization of the tumor segmentation (0.70 ± 0.07 and 0.74 ± 0.05, P = .07). CONCLUSION This study shows a strong correlation, a high precision, and excellent reproducibility of semiautomatic tumor segmentation software in measuring tumor volume on CE-MRI and DP-CBCT images. The use of the segmentation software on DP-CBCT and CE-MRI can be a valuable and highly accurate tool to measure the volume of hepatic tumors.

Non-Euclidean image-adaptive Radial Basis Functions for 3D interactive segmentation

In the context of variational image segmentation, we propose a new finite-dimensional implicit surface representation. The key idea is to span a subset of implicit functions with linear combinations of spatially-localized kernels that follow image features. This is achieved by replacing the Euclidean distance in conventional Radial Basis Functions with non-Euclidean, image-dependent distances. For the minimization of an objective region-based criterion, this representation yields more accurate results with fewer control points than its Euclidean counterpart. If the user positions these control points, the non-Euclidean distance enables to further specify our localized kernels for a target object in the image. Moreover, an intuitive control of the result of the segmentation is obtained by casting inside/outside labels as linear inequality constraints. Finally, we discuss several algorithmic aspects needed for a responsive interactive workflow. We have applied this framework to 3D medical imaging and built a real-time prototype with which the segmentation of whole organs is only a few clicks away.

Real-Time 3d image segmentation by user-constrained template deformation

We describe an algorithm for 3D interactive image segmentation by non-rigid implicit template deformation, with two main original features. First, our formulation incorporates user input as inside/outside labeled points to drive the deformation and improve both robustness and accuracy. This yields inequality constraints, solved using an Augmented Lagrangian approach. Secondly, a fast implementation of non-rigid template-to-image registration enables interactions with a real-time visual feedback. We validated this generic technique on 21 Contrast-Enhanced Ultrasound images of kidneys and obtained accurate segmentation results (Dice > 0.93) in less than 3 clicks in average.

Kidney detection and real-time segmentation in 3D contrast-enhanced ultrasound images

In this paper, we present an automatic method to segment the kidney in 3D contrast-enhanced ultrasound (CEUS) images. This modality has lately benefited of an increasing interest for diagnosis and intervention planning, as it allows to visualize blood flow in real-time harmlessly for the patient. Our method is composed of two steps: first, the kidney is automatically localized by a novel robust ellipsoid detector; then, segmentation is obtained through the deformation of this ellipsoid with a model-based approach. To cope with low image quality and strong organ variability induced by pathologies, the algorithm allows the user to refine the result by real-time interactions. Our method has been validated on a representative clinical database.

A Generic, Robust and Fully-Automatic Workflow for 3D CT Liver Segmentation

Liver segmentation in 3D CT images is a fundamental step for surgery planning and follow-up. Robustness, automation and speed are required to fulfill this task efficiently. We propose a fully-automatic workflow for liver segmentation built on state-of-the-art algorithmic components to meet these requirements. The liver is first localized using regression forests. A liver probability map is computed, followed by a global-to-local segmentation strategy using a template deformation framework. We evaluate our method on the SLIVER07 reference database and confirm its state-of-the-art results on a large, varied database of 268 CT volumes. This extensive validation demonstrates the robustness of our approach to variable fields of view, liver contrast, shape and pathologies. Our framework is an attractive tradeoff between robustness, accuracy mean distance to ground truth of 1.7mm and computational speedi¾?46s. We also emphasize the genericity and relative simplicity of our framework, which requires very limited liver-specific tuning.