Chantal Revol-Muller

Automated 3D region growing algorithm based on an assessment function

A new region growing algorithm is proposed for the automated segmentation of three-dimensional images. No initial parameters such as the homogeneity threshold or the seeds location have to be adjusted. The principle of our method is to build a region growing sequence by increasing the maximal homogeneity threshold from a very small value to a large one. On each segmented region, a 3D parameter that has been validated on a test image assesses the segmentation quality. This set of values called assessment function is used to determine of the optimal homogeneity criterion. Our algorithm was tested on 3D MR images for the segmentation of trabecular bone samples in order to quantify osteoporosis. A comparison to automated and manual thresholding showed that our algorithm performs better. Its main advantages are to eliminate isolated points due to the noise and to preserve connectivity of the bone structure.

Influence of age and sex on aortic distensibility assessed by MRI in healthy subjects

Magnetic resonance imaging (MRI) is particularly well adapted to the evaluation of aortic distensibility. The calculation of this parameter, based on the change in vessel cross-sectional area per unit change in blood pressure, requires precise delineation of the aortic wall on a series of cine-MR images. Firstly, the study consisted in validating a new automatic method to assess aortic elasticity. Secondly, aortic distensibility was studied for the ascending and descending thoracic aortas in 26 healthy subjects. Two homogeneous groups were available to evaluate the influence of sex and age (with an age limit value of 35 years). The automatic postprocessing method proved to be robust and reliable enough to automatically determine aortic distensibility, even on artefacted images. In the 26 healthy volunteers, a marked decrease in distensibility appears with age, although this decrease is only significant for the ascending aorta (8.97+/-2.69 10(-3) mmHg(-1) vs. 5.97+/-2.02 10(-3) mmHg(-1)). Women have a higher aortic distensibility than men but only significantly at the level of the descending aorta (7.20+/-1.61 10(-3) mmHg(-1) vs. 5.05+/-2.40 10(-3) mmHg(-1)). Through our automatic contouring method, the aortic distensibility from routine cine-MRI has been studied on a healthy subject population providing reference values of aortic stiffness. The aortic distensibility calculation shows that age and sex are causes of aortic stiffness variations in healthy subjects.

Automated 3D region growing algorithm governed by an evaluation function

A new region growing algorithm is proposed for the automated segmentation of three-dimensional images. No initial parameters such as the homogeneity threshold or the seeds location have to be adjusted. The principle of the authors method is to build a region growing sequence in increasing the maximal homogeneity threshold from a very small value to large one. On each segmented region, a 3D parameter which has been validated on a test image, evaluates the segmentation quality. This set of values called evaluation function is used to the determination of the best segmentation. The authors algorithm was tested on 3D MR images for the segmentation of trabecular bone samples in order to quantify osteoporosis. A comparison to automated and manual thresholding showed that the authors algorithm performs better. Its main advantages are to eliminate isolated points due to the noise and to preserve connectivity of the bone structure.

3D Robust Adaptive Region Growing for segmenting [18F] fluoride ion PET images

We propose a new robust adaptive region growing method (RoAd RG) based on two local parameters: the local mean value of the intensity function and the local mean value of the norm of the intensity gradient. This approach enables a better spread of the region growing inside the region of interest while avoiding the merge of outlier pixels. We tested our method on a synthesized noisy image, and demonstrated that RoAd RG gives better result than non adaptive or not fully adaptive methods. We applied positively our method to 3D [18F]fluoride ion PET images for segmenting bone structures, and showed its superiority compared to a non adaptive method.

Vesselness-guided variational segmentation of cellular networks from 3D micro-CT

Advances in imaging techniques lead to nondestructive 3D visualization of biological tissue at a sub-cellular scale. As a consequence, new demands emerge to segment complex structures. For instance, synchrotron radiation micro-CT, makes it possible to image the lacunar-canalicular porosity in bone tissue. This structure contains a dense network of slender channels interconnecting the cells. Their size (~300-600 nanometers in diameter) is at the limit of the acquisition system resolution (280 nm) making their detection difficult. In this work is proposed a variational region growing segmentation method adapted for cellular networks. To control the evolution of the segmentation through tubular structures a vesselness map is introduced in the expression of the functional to minimize. The method is tested on synthetic images and applied to experimental data.

Shape Prior Integrated in an Automated 3D Region Growing Method

We propose a new automated region growing method integrating shape prior (RGISP). The aim of this work is to improve region growing segmentation by taking into account a reference model. Our algorithm is assessed on a synthesized image and compared with two other methods in order to point up the contribution of shape prior. It was also applied to segment in-vivo mu-CT images of mouse kidneys in the framework of small animal imaging. RGISP gives promising results and appears to be well adapted to satisfy small animal imaging constraints.

Hybrid Approach for Multiparametric Mean Shift Filtering

In ultrasound imaging, robustness of the diagnosis can be improved by using many images of parameters. In this paper we propose a hybrid approach for improving multi-parametric mean shift filtering (MPMS). Multi-parametric filtering is really attractive since it works conjointly in the spatial-range domain, taking into account the spatial location of the data as well as the range values of many parameters. Hybrid MPMS is an. iterative method that combines two mean shift procedures called nonblurring and blurring. Our method was positively tested on a set of simulated ultrasound data. The results show the superiority of hybrid MPMS compared to the simple MPMS filtering.

Automated seeds location for whole body NaF PET segmentation

18F-labeled NaF, also called [/sup 18/F] fluoride ion, positron emission tomography (NaF PET) is a specific imaging modality of bone activity and allows bone tumor detection. In this paper, we propose a fast-automated method to locate anatomical structures by special planes and seeds in whole body NaF PET images. This step is crucial in registration and segmentation processes such as region growing. Our method proceeds in two steps: first, it delineates anatomical objects such as bladder, head, spine and legs with two horizontal planes (bottom and top); secondly, it determines labeled seeds in these objects. Processes are based on the analysis of the slice energy (SE) and the computation of the center of gravity (CG) of all transverse slices. This method was applied to initialize seeds in order to segment the whole skeleton. This segmentation is necessary to detect bone tumors and quantify bone uptake. Results are fast and quite satisfying. Robustness of our method was positively tested on a set of eight NaF PET images.

Shape prior criterion based on Tchebichef moments in variational region growing

Region growing has become a popular method for 3D segmentation. Starting from a seed, this approach allows one to extract a region by merging all its neighbors and comparing the extracted region to a reference. Here, we present an alternative approach to constrain the evolution of the region growing method in respect to a fixed reference shape. This approach is based on a shape description by the Tchebichef moments. The evolution equation minimizes a function that represents the distance between the evolving region and the reference shape. Experimental results show the ability of our method to drive the segmentation towards a desired shape in 2D and 3D data. Finally, we apply our shape prior conjointly to a bimodal segmentation functional, showing its benefits on segmentation results.

3D region growing integrating adaptive shape prior

We propose an automated region growing integrating adaptive shape prior in order to segment biomedical images. In our work, the segmentation method is improved by taking into account a shape reference model by non-linear way. Thus, the proposed method is driven by statistical data computed from the evolving region and by a priori shape information given by the model. An improvement of the method is proposed by adapting automatically the degree of integration of shape prior for each pixel of the image. The proposed method was applied for segmenting 3D micro-CT image of mouse skull in the framework of small animal imaging. The method gives promising results and appears to be well adapted to the context.