Hyxia Villegas

Implementation of a fast coral detector using a supervised machine learning and Gabor Wavelet feature descriptors

The task of reef restoration is very challenging for volunteer SCUBA divers, if it has to be carried out at deep sea, 200 meters, and low temperatures. This kind of task can be properly performed by an Autonomous Underwater Vehicle (AUV); able to detect the location of reef areas and approach them. The aim of this study is the development of a vision system for coral detections based on supervised machine learning. In order to achieve this, we use a bank of Gabor Wavelet filters to extract texture feature descriptors, we use learning classifiers, from OpenCV library, to discriminate coral from non-coral reef. We compare: running time, accuracy, specificity and sensitivity of nine different learning classifiers. We select Decision Trees algorithm because it shows the fastest and the most accurate performance. For the evaluation of this system, we use a database of 621 images (developed for this purpose), that represents the coral reef located in Belize: 110 for training the classifiers and 511 for testing the coral detector.

Segmentation, reconstruction, modeling and 3D visualization of the ventricles in echocardiographics images

This paper addresses the processing and reconstruction of the heart from echocardiographic images. This work is divided into two processing chains: reconstruction and visualization, each one of them is divided in several modules denominated as: 3D acquisition, 3D filtering, 3D segmentation, 3D reconstruction, 3D modeling and 3D visualization. The second processing chain uses an a priori model of the organ. It is adjusted to the data issued from the segmentation of the 4D (3D+t) volume captured from the patient. This is a new processing method because it is possible to capture a series of useful parameters (such as the ejection fraction, position, displacement and torsion of the ventricle) and at the same time it allows a 3D visualization of the virtual model. Two different models have been used in this work: a) a model of front propagation that allows segmentation of the internal wall of the ventricle with enough precision; b) a model with deformable superquadrics which obtains some useful geometric parameters to describe the shape and the dynamics of the ventricles. When we observe the differences between the virtual model and the resulting images of the 3D reconstruction of the heart, the physician has more information for the diagnosis, so that cardio-vascular illnesses will be detected with more precision.

Medical decision-making and collaborative reasoning

An overview is presented of different constraints in conventional medical decision-making for improving collaborative reasoning. When designing and implementing an effective assisted diagnoses system it is imperative to make a complete study of reasoning processes used by physicians every day. We are particularly interested in studying the process of conventional decision-making when, at the first stage, a physician faces a diagnosis decision and, later, when this decision is made by several physicians, that is, when collaborative medical reasoning is involved. The design of an architecture that supports the requirements of medical practices in this framework is presented.

A Software Tool for Reading DICOM Directory Files

DICOMDIR directory files are useful in medical software applications because they allow organized access to images and information sets that come from radiological studies that are stored in conformance with the digital imaging and communication in medicine (DICOM) standard. During the medical application software development, specialized programming libraries are commonly used in order to solve the requirements of computation and scientific visualization. However, these libraries do not provide suitable tools for reading DICOMDIR files, making necessary the implementation of a flexible tool for reading these files, which can be also easily integrated into applications under development. To solve this problem, this work introduces an object-oriented design and an open-source implementation for such reading tool. It produces an output data tree containing the information of the DICOM images and their related radiological studies, which can be browsed easily in a structured way through navigation interfaces coupled to it.

Collaborative medical reasoning in telemedicine

During medical practice, doctors use different aspects of reasoning for diagnosis and treatment. There is a particular process when several doctors are involved in taking a medical decision. In this paper we present an overview of different constraints in the process of conventional decision making and collaborative medical reasoning. General notions for an assisted decision cooperation system are introduced, bearing in mind general medical practices; without neglecting the physicians role.

Lector De Historias Clínicas Electrónicas codificadas en el estándar Health Level 7 / Clinical Document Architecture para su Aplicación en Servicios de Telemedicina

El objetivo principal de este trabajo es desarrollar un software libre que permita leer historias clinicas electronicas basadas en el estandar Health Level 7 / Clinical Document Architecture (HL7/CDA), desde el Web. El software recibe un archivo XML, de una historia clinica, codificado en HL7/CDA, lo lee y organiza su informacion en una estructura de datos almacenada en la memoria del computador, evitando asi lecturas innecesarias del documento HL7/CDA. El intercambio de datos entre el navegador de paginas Web y el servidor Web se realiza utilizando Javascript y XML asincronos (AJAX). El desarrollo del software se realizo siguiendo la metodologia Programacion Extrema (XP) y Programacion Orientada a Objetos en el lenguaje PHP, a traves del editor multiproposito Eclipse. Las pruebas del Lector fueron realizadas en Linux Ubuntu y en Windows XP, utilizando Apache 2 y PHP 5. Los resultados obtenidos muestran que el software funciona correctamente en los dos sistemas operativos utilizados en este estudio. El software desarrollado facilita la gestion de la informacion de las historias clinicas electronicas codificadas en HL7/CDA luego de su recepcion y permite mantener la integridad de la informacion clinica, por tal motivo esta herramienta representa una contribucion al intercambio de informacion clinica para ser utilizada en servicios de telemedicina.

Estimación Espacial y Temporal del Movimiento Cardíaco 3D, Utilizando Algoritmos de Flujo Óptico

This paper proposes a new technique for 3D motion estimation of the left ventricle from a sequence of a heartbeat. Accurate motion estimation of the movement of cardiac walls has been shown to be very important for studying the cardiovascular illnesses. This technique is based on a processing chain from the acquisition to the 3D segmentation of the left ventricle area obtained from each image during the cardiac cycle. With the purpose of estimating the movement of the Left Ventricle we calculate the optical flow starting from a sequence of images using the method proposed by Horn and Schunck [1].

Implementación de una Interfaz Cerebro Computador Basada en la Interpretación del Electroencefalograma para el Control a Distancia de Dispositivos Electrónicos

This paper describes the design and implementation of a prototype off line brain computer interface (BCI) used to simulate the operation of a remote control operated only by means of brain activity. The system is based on the spectral analysis of brain activity during simple mental tasks performing. Seven independent modules were developed and implemented in the form of ActiveX controls. Every ActiveX control makes one processing stage required and when they are combined allows building the interface between the brain and the machine. The classification of the mental tasks associated patterns was done by a multilayer perceptron artificial neural network. The system implemented, can simulate the remote control of standard device like a TV or Radio by the brain.

Estudio Comparativo de Varios Métodos de Segmentación Dinámica 3D para la extracción de Estructuras del Cerebro en Volúmenes de MRI

The segmentation of cerebral structures in medical images has numerous clinical applications. It can provide significant aids for the diagnosis of some pathology. In this work, we propose a comparative study of the new segmentation method based on the theory of deformable models along with the front propagation method, which it can make an semiautomatic segmentation of some cerebral structures. This new methodology for 3D information extraction is based on a processing chain that possesses the following modules: 1) 3D Filtering: the purpose is to preserve the contours of the structures and to smooth the homogeneous areas; several filters were tested and finally an anisotropic diffusion filter was used. 2) 3D Segmentation: this module compare three different methods, and finally proposes a segmentation method based on the front propagation method that allows the making of the reconstruction of the internal walls of the anatomical structures of the brain. 3) The new contribution of this work consists on the 3D visualization of the segmented model. This segmentation procedure is faster than the manual segmentation of images, with the advantage that it allows to use the same patient as anatomical reference, which has more precision than a generic atlas.

Segmentation of the Left Ventricle in 3D Echocardiographic Images and Simultaneous Visualization Using the 3D Texture Method

The new echocardiographic equipment produces volumes of 3D images. However, the processing and the information extraction of these volumetric data is troubled by complexity due to the high quantity of speckle noise, low resolution, and the loss of regions at the echocardiographic images. Our work presents a new methodology to extract the 3D information. It is based on a processing chain composed by the following modules: 1) Filtering: to preserve the contours of the heart structures and to smooth the homogeneous areas; several filters were tested and finally an anisotropic diffusion filter was selected. 2) Segmentation: this module proposes a segmentation technique based on a 3D level set method, which allows making the precision for the reconstruction of the internal wall of the left ventricle (LV). 3) Visualization: the new contribution from this work consists on the simultaneous visualization of the segmented model with the previous method, and the 3D visualization of the echocardiographic volume using 3D textures. The analysis of the Mean Square Error (MSE) between the calculated model and the real LV volume is smaller than 5% when we evaluated the segmentation method. In conclusion, the simultaneous 3D visualization will allow us to navigate inside of the 3D volume of data, which would not otherwise be possible when we visualized the data volume without the 3D information extraction, as proposed with this method. Finally, the segmentation obtains the model of the LV, and it is thus possible to observe the ischemic heart diseases using this method.