David Lesage

A review of 3D vessel lumen segmentation techniques: Models, features and extraction schemes

Vascular diseases are among the most important public health problems in developed countries. Given the size and complexity of modern angiographic acquisitions, segmentation is a key step toward the accurate visualization, diagnosis and quantification of vascular pathologies. Despite the tremendous amount of past and on-going dedicated research, vascular segmentation remains a challenging task. In this paper, we review state-of-the-art literature on vascular segmentation, with a particular focus on 3D contrast-enhanced imaging modalities (MRA and CTA). We structure our analysis along three axes: models, features and extraction schemes. We first detail model-based assumptions on the vessel appearance and geometry which can embedded in a segmentation approach. We then review the image features that can be extracted to evaluate these models. Finally, we discuss how existing extraction schemes combine model and feature information to perform the segmentation task. Each component (model, feature and extraction scheme) plays a crucial role toward the efficient, robust and accurate segmentation of vessels of interest. Along each axis of study, we discuss the theoretical and practical properties of recent approaches and highlight the most advanced and promising ones.

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.

Radiologic-Pathologic Analysis of Contrast-enhanced and Diffusion-weighted MR Imaging in Patients with HCC after TACE: Diagnostic Accuracy of 3D Quantitative Image Analysis

PURPOSE To evaluate the diagnostic performance of three-dimensional ( 3D three-dimensional ) quantitative enhancement-based and diffusion-weighted volumetric magnetic resonance (MR) imaging assessment of hepatocellular carcinoma ( HCC hepatocellular carcinoma ) lesions in determining the extent of pathologic tumor necrosis after transarterial chemoembolization ( TACE transarterial chemoembolization ). MATERIALS AND METHODS This institutional review board-approved retrospective study included 17 patients with HCC hepatocellular carcinoma who underwent TACE transarterial chemoembolization before surgery. Semiautomatic 3D three-dimensional volumetric segmentation of target lesions was performed at the last MR examination before orthotopic liver transplantation or surgical resection. The amount of necrotic tumor tissue on contrast material-enhanced arterial phase MR images and the amount of diffusion-restricted tumor tissue on apparent diffusion coefficient ( ADC apparent diffusion coefficient ) maps were expressed as a percentage of the total tumor volume. Visual assessment of the extent of tumor necrosis and tumor response according to European Association for the Study of the Liver ( EASL European Association for the Study of the Liver ) criteria was performed. Pathologic tumor necrosis was quantified by using slide-by-slide segmentation. Correlation analysis was performed to evaluate the predictive values of the radiologic techniques. RESULTS At histopathologic examination, the mean percentage of tumor necrosis was 70% (range, 10%-100%). Both 3D three-dimensional quantitative techniques demonstrated a strong correlation with tumor necrosis at pathologic examination (R(2) = 0.9657 and R(2) = 0.9662 for quantitative EASL European Association for the Study of the Liver and quantitative ADC apparent diffusion coefficient , respectively) and a strong intermethod agreement (R(2) = 0.9585). Both methods showed a significantly lower discrepancy with pathologically measured necrosis (residual standard error [ RSE residual standard error ] = 6.38 and 6.33 for quantitative EASL European Association for the Study of the Liver and quantitative ADC apparent diffusion coefficient , respectively), when compared with non- 3D three-dimensional techniques ( RSE residual standard error = 12.18 for visual assessment). CONCLUSION This radiologic-pathologic correlation study demonstrates the diagnostic accuracy of 3D three-dimensional quantitative MR imaging techniques in identifying pathologically measured tumor necrosis in HCC hepatocellular carcinoma lesions treated with TACE transarterial chemoembolization .

Medial-based Bayesian tracking for vascular segmentation: Application to coronary arteries in 3D CT angiography

We propose a new Bayesian, stochastic tracking algorithm for the segmentation of blood vessels from 3D medical image data. Inspired by the recent developments in particle filtering, it relies on a constrained, medial-based geometric model and on an original sampling scheme for the selection of tracking hypotheses. A key property of this new sampling scheme is the ability to take into account a distribution of hypotheses broader than similar methods such as classical particle filters, while remaining computationally efficient. The proposed method was applied to the challenging and medically critical task of coronary artery segmentation from 3D cardiac computed tomography (CT) images. Prior knowledge, injected in the process, was learned from a manually segmented database of 19 cases. Qualitative and quantitative evaluation is presented on clinical data, including pathologies and local anomalies.

Three-dimensional Evaluation of Lipiodol Retention in HCC after Chemoembolization: A Quantitative Comparison between CBCT and MDCT

RATIONALE AND OBJECTIVES To evaluate the capability of cone-beam computed tomography (CBCT) acquired immediately after transcatheter arterial chemoembolization (TACE) in determining lipiodol retention quantitatively and volumetrically when compared to 1-day postprocedure unenhanced multidetector computed tomography (MDCT). MATERIALS AND METHODS From June to December 2012, 15 patients met the inclusion criteria of unresectable hepatocellular carcinoma (HCC) that was treated with conventional TACE (cTACE) and had intraprocedural CBCT and 1-day post-TACE MDCT. Four patients were excluded because the lipiodol was diffuse throughout the entire liver or lipiodol deposition was not clear on both CBCT and MDCT. Eleven patients with a total of 31 target lesions were included in the analysis. A quantitative three-dimensional software was used to assess complete, localized, and diffuse lipiodol deposition. Tumor volume, lipiodol volume in the tumor, percent lipiodol retention, and lipiodol enhancement in Hounsfield units (HU) were calculated and compared between CBCT and MDCT using two-tailed Students t test and Bland-Altman plots. RESULTS The mean value of tumor volume, lipiodol-deposited regions, calculated average percent lipiodol retention, and HU value of CBCT were not significantly different from those of MDCT (tumor volume: 9.37 ± 11.35 cm(3) vs 9.34 ± 11.44 cm(3), P = .991; lipiodol volume: 7.84 ± 9.34 cm(3) vs 7.84 ± 9.60 cm(3), P = .998; lipiodol retention: 89.3% ± 14.7% vs. 90.2% ± 14.9%, P = .811; HU value: 307.7 ± 160.1 HU vs. 257.2 ± 120.0 HU, P = .139). Bland-Altman plots showed only minimal difference and high agreement when comparing CBCT to MDCT. CONCLUSIONS CBCT has a similar capability, intraprocedurally, to assess lipiodol deposition in three dimensions for patients with HCC treated with cTACE when compared to MDCT.

Design and study of flux-based features for 3D vascular tracking

In this paper, we present and study two local features for the tracking of vascular structures on 3D angiograms. The first one, Flux, measures the inward gradient flux through circular cross-sections. The second one, MFlux, introduces a non-linear penalization of asymmetric flux contributions to reduce false positive responses.

Multi-organ localization with cascaded global-to-local regression and shape prior

We propose a method for fast, accurate and robust localization of several organs in medical images. We generalize the global-to-local cascade of regression random forest to multiple organs. A first regressor encodes the global relationships between organs, learning simultaneously all organs parameters. Then subsequent regressors refine the localization of each organ locally and independently for improved accuracy. By combining the regression vote distribution and the organ shape prior (through probabilistic atlas representation) we compute confidence maps that are organ-dedicated probability maps. They are used within the cascade itself, to better select the test voxels for the second set of regressors, and to provide richer information than the classical bounding boxes result thanks to the shape prior. We propose an extensive study of the different learning and testing parameters, showing both their robustness to reasonable perturbations and their influence on the final algorithm accuracy. Finally we demonstrate the robustness and accuracy of our approach by evaluating the localization of six abdominal organs (liver, two kidneys, spleen, gallbladder and stomach) on a large and diverse database of 130 CT volumes. Moreover, the comparison of our results with two existing methods shows significant improvements brought by our approach and our deep understanding and optimization of the parameters.

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.

Multi-organ Localization Combining Global-to-Local Regression and Confidence Maps

We propose a method for fast, accurate and robust localization of several organs in medical images. We generalize global-to-local cascades of regression forests [1] to multiple organs. A first regressor encodes global relationships between organs. Subsequent regressors refine the localization of each organ locally and independently for improved accuracy. We introduce confidence maps, which incorporate information about both the regression vote distribution and the organ shape through probabilistic atlases. They are used within the cascade itself, to better select the test voxels for the second set of regressors, and to provide richer information than the classical bounding boxes thanks to the shape prior. We demonstrate the robustness and accuracy of our approach through a quantitative evaluation on a large database of 130 CT volumes.

Multiple template deformation application to abdominal organ segmentation

We propose a fast, automatic and versatile framework for the segmentation of multiple anatomical structures from 2D and 3D images. We extend the work of [1] on implicit template deformation to multiple targets. Our variational formulation optimizes the non-rigid transformation of a set of templates according to image-driven forces. It embeds non-overlapping constraints ensuring a consistent segmentation result. We demonstrate the potential of our approach on the segmentation of abdominal organs (liver, kidneys, spleen and gallbladder) with an evaluation on CT volumes (50 for training and 50 for testing). Our method reaches state-of-the-art accuracy, ranging from 2mm (liver and kidneys) to 8mm (gallbladder).