Biomedical subjects

Petia Radeva

Publications and source records attributed to Petia Radeva.

6 recordsLinked to original sources

PubMed2006

Statistical strategy for anisotropic adventitia modelling in IVUS.

Vessel plaque assessment by analysis of intravascular ultrasound sequences is a useful tool for cardiac disease diagnosis and intervention. Manual detection of luminal (inner) and media-adventitia (external) vessel borders is the main activity of physicians in the process of lumen narrowing (plaque) quantification. Difficult definition of vessel border descriptors, as well as, shades, artifacts, and blurred signal response due to ultrasound physical properties trouble automated adventitia segmentation. In order to efficiently approach such a complex problem, we propose blending advanced anisotropic filtering operators and statistical classification techniques into a vessel border modelling strategy. Our systematic statistical analysis shows that the reported adventitia detection achieves an accuracy in the range of interobserver variability regardless of plaque nature, vessel geometry, and incomplete vessel borders.

Algorithms↗

PubMed2006

Discriminant ECOC: a heuristic method for application dependent design of error correcting output codes.

We present a heuristic method for learning error correcting output codes matrices based on a hierarchical partition of the class space that maximizes a discriminative criterion. To achieve this goal, the optimal codeword separation is sacrificed in favor of a maximum class discrimination in the partitions. The creation of the hierarchical partition set is performed using a binary tree. As a result, a compact matrix with high discrimination power is obtained. Our method is validated using the UCI database and applied to a real problem, the classification of traffic sign images.

Algorithms↗

PubMed2005

Medical image retrieval based on plaque appearance and image registration.

The increasing amount of medical images produced and stored daily in hospitals needs a datrabase management system that organizes them in a meaningful way, without the necessity of time-consuming textual annotations for each image. One of the basic ways to organize medical images in taxonomies consists of clustering them depending of plaque appearance (for example, intravascular ultrasound images). Although lately, there has been a lot of research in the field of Content-Based Image Retrieval systems, mostly these systems are designed for dealing a wide range of images but not medical images. Medical image retrieval by content is still an emerging field, and few works are presented in spite of the obvious applications and the complexity of the images demanding research studies. In this chapter, we overview the work on medical image retrieval and present a general framework of medical image retrieval based on plaque appearance. We stress on two basic features of medical image retrieval based on plaque appearance: plaque medical images contain complex information requiring not only local and global descriptors but also context determined by image features and their spatial relations. Additionally, given that most objects in medical images usually have high intra- and inter-patient shape variance, retrieval based on plaque should be invariant to a family of transformations predetermined by the application domain. To illustrate the medical image retrieval based on plaque appearance, we consider a specific image modality: intravascular ultrasound images and present extensive results on the retrieval performance.

Humans↗

PubMed2005

On the assessment of texture feature descriptors in intravascular ultrasound images: a boosting approach to a feasible plaque classification.

Intravascular ultrasound images represent a unique tool to guide interventional coronary procedures, this technique allows to supervise the cross-sectional locations of the vessel morphology and to provide quantitative and qualitative information about the causes and severity of coronary diseases. At the moment, the automatic extraction of this kind of information is performed without taking into account the basic signal principles that guide the process of image generation. In this chapter, we overview the main physical principles and factors that affect the IVUS generation; we propose a simple physics-based approach for IVUS image simulation that is defined as a discrete representation of the tissue by individual scatterrers elements with given spatial distribution and backscattering differential cross sections. In order to generate the physical model that allows to construct synthetic IVUS images, we analyze the process of pulse emission, transmission and reception of the ultrasound signal as well as its interaction with the different tissues scatterrers of the simulated artery. In order to obtain the 3D synthetic image sequences we involve the dynamic behavior of the heart/arteries and the catheter movement in the image generation model. Having an image formation model allows to study rhe physics parameters that participate during the image generation and to achieve a better understanding and robust interpreting of IVUS image structures. Moreover, this model allows to comprehend, simulate and solve several limitations of IVUS sequences, to extract important image parameters to be taken into account when developing robust image processing algorithms as well as to construct wide synthetic image sequence databases in order to validate different image processing techniques.

Algorithms↗

PubMed2004

A regularized curvature flow designed for a selective shape restoration.

Among all filtering techniques, those based exclusively on image level sets (geometric flows) have proven to be the less sensitive to the nature of noise and the most contrast preserving. A common feature to existent curvature flows is that they penalize high curvature, regardless of the curve regularity. This constitutes a major drawback since curvature extreme values are standard descriptors of the contour geometry. We argue that an operator designed with shape recovery purposes should include a term penalizing irregularity in the curvature rather than its magnitude. To this purpose, we present a novel geometric flow that includes a function that measures the degree of local irregularity present in the curve. A main advantage is that it achieves non-trivial steady states representing a smooth model of level curves in a noisy image. Performance of our approach is compared to classical filtering techniques in terms of quality in the restored image/shape and asymptotic behavior. We empirically prove that our approach is the technique that achieves the best compromise between image quality and evolution stabilization.

Algorithms↗

PubMed2002

Predictive (un)distortion model and 3-D reconstruction by biplane snakes.

This paper is concerned with the three-dimensional (3-D) reconstruction of coronary vessel centerlines and with how distortion of X-ray angiographic images affects it. Angiographies suffer from pincushion and other geometrical distortions, caused by the peripheral concavity of the image intensifier (II) and the nonlinearity of electronic acquisition devices. In routine clinical practice, where a field-of-view (FOV) of 17-23 cm is commonly used for the acquisition of coronary vessels, this distortion introduces a positional error of up to 7 pixels for an image matrix size of 512 x 512 and an FOV of 17 cm. This error increases with the size of the FOV. Geometrical distortions have a significant effect on the validity of the 3-D reconstruction of vessels from these images. We show how this effect can be reduced by integrating a predictive model of (un)distortion into the biplane snakes formulation for 3-D reconstruction. First, we prove that the distortion can be accurately modeled using a polynomial for each view. Also, we show that the estimated polynomial is independent of focal length, but not of changes in anatomical angles, as the II is influenced by the earth's magnetic field. Thus, we decompose the polynomial into two components: the steady and the orientation-dependent component. We determine the optimal polynomial degree for each component, which is empirically determined to be five for the steady component and three for the orientation-dependent component. This fact simplifies the prediction of the orientation-dependent polynomial, since the number of polynomial coefficients to be predicted is lower. The integration of this model into the biplane snakes formulation enables us to avoid image unwarping, which deteriorates image quality and therefore complicates vessel centerline feature extraction. Moreover, we improve the biplane snake behavior when dealing with wavy vessels, by means of using generalized gradient vector flow. Our experiments show that the proposed methods in this paper decrease up to 88% the reconstruction error obtained when geometrical distortion effects are ignored. Tests on imaged phantoms and real cardiac images are presented as well.

Angiography, Digital Subtraction↗