B. Macq

Interactive delineation of brain sulci and their merging into functional PET images

A set of software tools has been developed to assist the neuro(physio)logist in the analysis of a series of cerebral images from single subjects by fusing sulci manually delineated on MRI brain surface into PET functional data. The procedure requires coregistered datasets and involves 4 steps: (i) segmentation of anatomical MRI data in order to extract the brain surface, (ii) generate the brain surface views by parallel ray casting, (iii) manual delineation of the relevant sulci from the surface views and (iv) fusion of the landmarks into any coregistered dataset from the same subject. The brain surface is segmented automatically from 3D MRI data using a new Directional Watershed Transform algorithm. From the segmented brain surface, 8 orthogonal surface views are calculated as visual support for interactive stereo definition of the major brain sulci. Each sulcus is built as a 3D trace line using a few vertices which are manually defined on one or several surface views. This technique allows one to follow the brain surface curvature rather independently of the number and the position of the vertices. The sulci are saved in an individual file for further use. The brain surface viewer is linked (via the 3D cursor position) to an independent volume viewer containing a coregistered (anatomical or functional) volume. Sulcal landmarks are finally projected onto this volume allowing further volume of interest definition. The use of the tool set is illustrated by a single subject brain activation study after /sup 15/O water injection.

Morphological shape and region description

Abstract A new algorithm for multiscale description of binary digital regions is given. A region is represented by a growing sequence of subsets approximating it; each approximation is obtained from the previous one by addition of the opening by a structuring element chosen in a finite family. The structuring elements vary in size and shape, and are chosen at each step according to two criteria: largest structuring element size and greatest increment in the size of the approximation. The algorithm satisfies meaningful algebraic properties and converges to a final result in a bounded number of steps. It also allows an economical representation of shape in terms of structuring elements, and so it can be applied in binary image coding.

Three-dimensional image quality measurement for the benchmarking of 3D watermarking schemes

Three-dimensional image quality assessment causes new challenges for a wide set of applications and particularly for emerging 3-D watermarking schemes. First, new metrics have to be drawn for the distortion measurement from an original 3-D surface to its deformed version: this metric is necessary to address distortions that are acceptable and to which a 3-D watermarking algorithm should resist. In this paper, we focus on distortion energy evaluation extending works on distortion minimization for planar and spherical parameterization. Secondly, a key perceptual assessment of 3-D geometrical transforms is their impact on the various 2-D views that can be extracted from the object. As a matter of fact, most of the applications (games, avatars, ...) are targeting users owning 2-D screens. In this paper we restrict our study to 3-D shape distortion analysis, assuming standard lighting conditions and we do not address the textures distortion issues. We analyze how to automatically select relevant pairs of 2D projections which needs an initial registration between both shapes to compare. We use a mutual information criterion to assess the distortion for each projection pair and eventually derive a global score by weighting the contributions of each view.

Multimodal nonrigid registration using a stochastic gradient approximation

We present a new fast implementation of a nonrigid registration algorithm, based on a finite element elastic deformation model using the mutual information metric with a linear elastic regularization constraint. The algorithm was parallelized for symmetric multiprocessor architectures. A simultaneous perturbation stochastic approximation (SPSA) optimization scheme was used to maximize the objective function. This algorithm was applied to capture nonrigid deformations of pre-procedural to post-procedural images of tumor radio-frequency (RF) ablation in the liver and nonrigid deformations of intraoperative to preoperative prostate images.

Quality Assessment of Non-Rigid Registration Methods for Atlas-Based Segmentation in Head-Neck Radiotherapy

In this paper we compare three non-rigid registration methods for atlas-based segmentation: B-splines, morphons and a combination of morphons and demons. To assess the quality of each method, we use a data set of four patients, containing for each patient the computed tomography (CT) image and a manual segmentation of the organs at risk performed by an expert of the head and neck anatomy. Non-rigid registration algorithms have been used to match the patient and atlas images. Each deformation field, resulting from the non-rigid deformation, have been applied on the masks corresponding to segmented regions in the atlas. The atlas based segmented masks have been compared to manual segmentations performed by the expert. The results show that the combined method (morphons + demons) achieves the best performances on this dataset resulting in an average improvement of 6% with respect to morphons and 18% with respect to B-spline.

Feature-based watermarking of 3D objects: toward robustness against remeshing and desynchronization

In this paper, we propose a blind watermarking scheme based on automatic feature points detection. The irregular sampling of 3D shapes is a challenging issue for extending well-known signal processing tools. 3D shape watermarking schemes have to resist to common resampling operations used for example in some compression applications. We propose an automatic selection of intrinsic feature points that are robust against surface remeshing. They are detected as multi-scale robust degeneracies of the shape curvature tensor field. The impact of the sampling on the curvature estimation is studied. These points are then used as seeds in the partition of the shape into fast approximated geodesic triangles. Each of them is then remeshed with a regular connectivity and watermarked in the mesh spectral domain. The watermark perturbations computed on the remeshed triangles are the projected on the original points of the 3D object. We discuss the robustness of the feature points and of the overall scheme under various watermarking attacks.

A versatile knowledge-based clinical imaging annotation system for breast cancer screening

Medical information is evolving towards more complex multimedia data representation, as new imaging modalities are made available by sophisticated devices. Features such as segmented lesions can now be extracted through analysis techniques and need to be integrated into clinical patient data. The management of structured information extracted from multimedia has been addressed in knowledge based annotation systems providing methods to attach interpretative semantics to multimedia content. Building on these methods, we develop a new clinical imaging annotation system for computer aided breast cancer screening. The proposed system aims at more consistent, efficient and standardised data mark-up of digital and digitalised radiology images. The objective is to provide detailed characterisation of abnormalities as an aid in the diagnostic task through integrated annotation management. The system combines imaging analysis results and radiologist diagnostic information about suspicious findings by mapping well-established visual and low-level descriptors into pathology specific profiles. The versatile characterisation allows differentiating annotation descriptors for different types of findings. Our approach of semi-automatic integrated annotations supports increased quality assurance in screening practice. This is achieved through detailed and objective patient imaging information while providing user-friendly means for their manipulation that is oriented to relieving the radiologists workload.

Accurate segmentation of 3-D magnetic resonance images of the head using a directional watershed transform

A new algorithm for automatic and accurate segmentation of 3-D magnetic resonance images (MRI) is presented. It improves the classical watershed transformation (WST) whose results are inaccurate when applied on 3-D noisy data. It uses directional information and is less sensitive to noise since it does not require the gradient image. Initial results are presented.

Interactive breast cancer segmentation based on relevance feedback: from user-centered design to evaluation

Computer systems play an important role in medical imaging industry since radiologists depend on it for visualization, interpretation, communication and archiving. In particular, computer-aided diagnosis (CAD) systems help in lesion detection tasks. This paper presents the design and the development of an interactive segmentation tool for breast cancer screening and diagnosis. The tool conception is based upon a user-centered approach in order to ensure that the application is of real benefit to radiologists. The analysis of user expectations, workflow and decision-making practices give rise to the need for an interactive reporting system based on the BIRADS, that would not only include the numerical features extracted from the segmentation of the findings in a structured manner, but also support human relevance feedback as well. This way, the numerical results from segmentation can be either validated by end-users or enhanced thanks to domain-experts subjective interpretation. Such a domain-expert centered system requires the segmentation to be sufficiently accurate and locally adapted, and the features to be carefully selected in order to best suit users knowledge and to be of use in enhancing segmentation. Improving segmentation accuracy with relevance feedback and providing radiologists with a user-friendly interface to support image analysis are the contributions of this work. The preliminary result is first the tool conception, and second the improvement of the segmentation precision.

RoI-based multiresolution compression of heart MR images

In this paper, we present an image compression scheme based on the automatic segmentation of regions of interest (RoI) and a lossy wavelet compression algorithm adapted to this segmentation. Quasi-lossless compression is applied to the RoI while lossy compression is allowed outside the RoI, preserving at best the visual quality of the decoded image within a defined RoI. In fact, for diagnostic accuracy purposes, quasi- lossless compression is often mandatory, while high compression ratio can only be achieved by lossy compression methods. The proposed technique is applied to heart MR images where the RoI is the entire heart. First, an unsupervised segmentation of the heart is performed in the original MR images, and the RoI is modeled by an ellipse fitted by means of a genetic algorithm. This model is defined by only 5 parameters, providing an efficient representation of the surrounding shape of the RoI. Compression is then applied using a wavelet-based multiresolution scheme. The quantization factor applied to the wavelet coefficients is adapted to the region and the subband, leading to the quasi-lossless compression in the RoI and lossy compression outside this RoI. The quantization difference between inside and outside RoI is also optimized for the desired compression ratio. Finally, the compressed bitstream is transmitted to the decoder together with the parameters of the RoI, allowing the reconstruction of the RoI surrounding shape in the decoder. Results of this RoI- based compression scheme are presented and further compared with the JPEG standard.