Catálogo de publicaciones - libros

Recently, several works have approached the HIV-1 protease specificity problem by applying a number of classifier creation and combination methods, from the field of machine learning. In this work we propose a hierarchical classifier (HC) architecture. Moreover, we show that radial basis function-support vector machines may obtain a lower error rate than linear support vector machines, if a step of feature selection and a step of feature transformation is performed. The error rate decreases from 9.1% using linear support vector machines to 6.85% using the new hierarchical classifier.

We present an algorithm to extract a medial representation of proteins in volume images. The representation (MGR) takes into account the internal grey-level distribution of the protein and can be extracted without first segmenting the image into object and background. We show how MGR can facilitate the analysis of the structure of the proteins and thereby also classification. Results are shown on two types of protein images.

A recent trend in digital mammography are CAD systems, which are computerized tools designed to help radiologists. Most of these systems are used for the automatic detection of abnormalities. However, recent studies have shown that their sensitivity is significantly decreased as the density of the breast is increased. In addition, the suitability of abnormality segmentation approaches tends to depend on breast tissue density. In this paper we propose a new approach to the classification of mammographic images according to the breast parenchymal density. Our classification is based on gross segmentation and the underlying texture contained within the breast tissue. Robustness and classification performance are evaluated on a set of digitized mammograms, applying different classifiers and leave-one-out for training. Results demonstrate the feasibility of estimating breast density using computer vision techniques.

Ultrasound B-mode images of thyroid gland were previously analyzed to distinguish normal tissue from inflamed tissue due to Hashimoto’s Lymphocytic Thyroiditis. This is a two-class recognition problem. Sensitivity and specificity of 100% was reported using Bayesian classifier with selected texture features. These results were obtained on 99 subjects at a fixed setting of one specific sonograph, for a given manual thyroid gland segmentation and sonographic scan orientation (longitudinal, transversal). To evaluate the reproducibility of the method, sensitivity analysis is the topic of this paper. A general method for determining feature sensitivity to variables influencing the scanning process is proposed. Jensen Shannon distances between modified and unmodified inter- and intra-class feature probability distributions capture the changes induced by the variables. Among selected features, the least sensitive one is found. The proposed sensitivity evaluation method can be used in other problems with complex and non-linear dependencies on variables that cannot be controlled.

This paper presents a segmentation of cardiac valve structure method in ultrasound sequence. Prior knowledge on certain complex object is a powerful guidance in image segmentation. We represent region and shape prior of the cardiac valve in a form of speed field and incorporate it into image segmentation process within level set framework. Region prior constrains the zero level set evolving in certain region and shape prior pulls the curve to the ideal contour. Experiments on a large quantity of 3D valve sequences show that the algorithm improves accuracy of segmentation and reduces the manual intervention.

In this work we propose a new method to estimate the scale hyperparameter for transmission tomography in Nuclear Medicine image reconstruction problems. Within the Bayesian paradigm, Evidence Analysis and circulant preconditioners are used to obtain the scale hyperparameter. For the prior distribution, we use Generalized Gaussian Markov Random Fields (GGMRF), a nonquadratic function that preserves the edges in the reconstructed image. The experimental results indicate that the proposed method produces satisfactory reconstructions.

We propose an automatic segmentation and registration method for matching lung surfaces of temporal CT scans. Our method consists of three steps. First, an automatic segmentation is used for accurately identifying lung surfaces. Second, initial registration using an optimal cube is performed for correcting the gross translational mismatch. Third, the initial alignment is step by step refined by the iterative surface registration. For the fast and robust convergence of the distance measure to the optimal value, a 3D distance map is generated by the narrow band distance propagation. Experimental results show that our segmentation and registration method extracts accurate lung surfaces and aligns them much faster than conventional ones using a distance measure.

Previous works on breast tissue identification and abnormalities detection notice that the feature extraction process is affected if the region processed is not well focused. Thereby, it is important to split the mammogram into interesting regions to achieve optimal breast parenchyma measurements, breast registration or to put into focus a technique when we search for abnormalities. In this paper, we review most of the relevant work that has been presented from 80’s to nowadays. Secondly, an automated technique for segmenting a digital mammogram into breast region and background, with pectoral muscle suppression is presented.

The present work seeks to develop a computational tool to suggest the malignancy or benignity of Solitary Lung Nodules by means of analyzing texture measures obtained from computerized tomography images.Two methods are proposed, that analyze the nodules’ texture by means of the Spatial Gray Level Dependence Method and a geostatistical function denominated semivariogram. A sample with 36 nodules, 29 benign and 7 malignant, was analyzed and the preliminary results of these methods are very promising in characterizing lung nodules. The obtained results suggested that the proposed methods have great potential in the discrimination and classification of Solitary Lung Nodules.

An anisotropic method of 3D reconstruction method for time sequence echocardiographic images is proposed in this paper. First, a Bayesian model based on MAP-MRF is described to reconstruct 3D volume, and extended to deal with the images acquired by rotation scanning method. Second, the spatial and temporal nature of ultrasound images is taken into account for the selection of parameter of energy function, which makes this statistical model anisotropic. Hence not only can this method reconstruct 3D ultrasound images, but also remove the speckle noise anisotropically. Finally, we illustrate the experiments of our method on the synthetic and medical images and compare with the isotropic reconstruction method.