MEDICAL TECHNOLOGY

Author: MARTÍNEZ CORTÉS JUAN PABLO.
Year: 2004.
University: ZARAGOZA.
Place of defense: CENTRO POLITÉCNICO SUPERIOR.
Place of preparation: INSTITUTO DE INVESTIGACIÓN EN INGENIERÍA DE ARAGÓN (I3A).

Summary: This dissertation examines the analysis methodologies Alternancias Wave T (AOT) in the electrocardiogram signal. This phenomenon has been revealed in recent years as one of the non-invasive clinical indices for most promising stratify the risk of malignant arrhythmias and sudden cardiac death. First an effort is made to systematize approaches posed by the various authors. It performs a thorough review of existing methods, suggesting a unified framework for the analysis of AOT. In light of this scheme widespread discusses various methods, ranked in terms of how to calculate point estimates and statistical decision. It shows that some methods are equivalent despite departing from very different approaches. In this thesis is used tools theories statistics detection and estimation to derive and propose new analysis techniques AOT. It is proposed a methodological approach based on models. In particular, derive the maximum likelihood estimator (GSS), the ratio test verosimilitudes widespread (TCVG) and detectors asintóticamente equivalents Wald and Rao for different models considered. In them, alternating signal is characterized according to a linear model with deterministic unknown parameters. Considered models gaussian noise, as well as the family of distributions gaussianas widespread. On the other hand, depending on the degree of stationarity and a priori knowledge of the noise defined three types of models: stationary and non-stationary adaptive. With the use of these models is intended to carry out a more realistic characterization of the problem as the actual noise present in the ECG is leptocúrtico and clearly not static. Benefits have been derived from theoretical methods (either accurate, either asymptotic) and its sensitivity to mismatches between the model and analyze the signal. The validation of methods is completed before assessing its performance with AOT simulated signals under realistic conditions. This will use two ECG signal generators, whose common feature is that the noise present in the simulated signals is real, either from a database containing records of physiological noise, or their own ECG signals on which simulates alternation. It found that the detectors AOT they get the best benefits are TCVG Laplacian adaptive, along with detectors based on non-stationary models (both gaussianos as laplacianos). The results also show that any of the three above methods clearly exceeds the benefits of the most widely used methods in the literature. As clinical application of the proposed methodologies, presents a study to analyze ECG signals recorded during an intervention by percutaneous transluminal coronary angioplasty (PTCA). The induced ischemia controlled during the speech offers an opportunity to characterize the phenomenon of AOT produced in the first few minutes of acute myocardial ischemia. Using a methodology based on the model Laplacian adaptive explores its prevalence and magnitude depending on the location of the occlusion. It is also characterized the evolution of the rotation during periods of occlusion and reperfusion, it is found that the spatial distribution of shunts depends on the occluded artery and looks at the differences in their timing inside the compound ST-T.

Author: ROCON DE LIMA EDUARDO.
Year: 2005.
University: POLITÉCNICA DE MADRID.
Place of defense: E.T.S. ING. INDUSTRIALES.
Place of preparation: E.T.S. ING. INDUSTRIALES.

Summary: This thesis has been developed under the project European DRIFTS. The project DRIFTS began in April 2002 with the main objective of validating the concept of active suppression of pathological tremor through outpatient brace. To achieve this objective has been defined a work methodology based on a detailed overview of the state of art of gracious bibliography of tremor. In this review criticizes discusses the shortcomings in the removal ortésica Quake and lays the foundation for the future development work presented in this thesis memory. The work of this thesis attempts to fill these gaps. The main contributions of this work are presented in summary form in the following points: 1 - Presentation of a descriptive overview of the state of art of the tremor in their different areas. This forecast detected shortcomings in the area of the Abolition ortésica Quake and defined work goals. 2-Development of a system for measuring ambulatory activity shaken in the senior member humans. This system has several advantages over the systems used until the publication of this thesis. 3-Introduction of the Empirical Mode Decomposition as a new tool for the study of movement shaky. This technique solves several limitations of current techniques for analysis of time series tremor. There are indications of the ability of this technique to detect automatically otrostipos of involuntary movements. This technique also identifies new features of the shaky movement that can be used for the diagnosis of pathology which originates. 4-depth study of the behavior of the shaky movement in the joints of senior member humans. 5-Developing an algorithm capable of separating the shaky movement of voluntary movement. This algorithm has the following features for application in ambulatory devices for removal of tremor: A-generates an estimate of voluntary movement. B-generates an estimate of movement shaky. C - Has a delay in the estimation of a sample period (1 ms) and a time of convergence of 2 seconds or less. D-It is capable of estimating the magnitude and frequency of tremor for applications that require removal of tremor real time. E-Holds low computational cost. 6, - Development of a robotic exoskeletons for the monitoring and control of the tremor in patients. This robotic exoskeletons is capable of applying dynamic forces between two joints of the human upper member and thus, control independently of the same three movements: flexión-extensión of the wrist, pronación-supinación forearm and flexión-extensión elbow. Its characteristics greatly expanded its scope, and may be used in robotics rehabilitation and in industrial applications or teleoperación. 7-Study, definition and validation of control strategies for the elimination of tremor through exoesqueletos robotic. Have been proposed and validated two strategies of control. One is a completely novel approach in the elimination of tremor. 8-Validation of the technical feasibility of the concept of eliminating tremor through exoesqueletos robotic. 9-Defining constraints, requirements and milestones in the field of robotics for ambulatory robotic devices for the removal of excess tremor status exoesqueletos laboratory. 10-Transfer technology part of the knowledge generated in the context of the thesis to the firm 8 closed. 2 f4 This transfer has created a device that helps the digital inclusion of people who suffer from tremor.

Author: LAMATA DE LA ORDEN PABLO.
Year: 2005.
University: POLITÉCNICA DE MADRID.
Place of defense: ESCUELA TÉCNICA SUPERIOR DE INGENIEROS DE TELECOMUNICACIÓN.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIEROS DE TELECOMUNICACIÓN.

Summary: The problem faced is that of training in laparoscopic surgery in an effective and efficient manner by using virtual reality simulators (RSV). Specifically focuses on the design of the simulator, in identifying the characteristics that make it a useful teaching tool. The approach taken is done from a perspective of Biomedical Engineering, seeking to draw a bridge between the needs of clinical training and technology to achieve that RSV optimal solution. To address this problem are plotted three methodological approaches: 1 - The development of a conceptual framework for the analysis, design and validation of models. 2 - The study of sensory ability of surgeons to clarify the requirements for simulation. 3 - The design of a training simulator for laparoscopic surgery to improve the performance of existing ones. It is proposed to design the simulator RSV as a training tool that is built through teaching resources of technologies RV: 1-fidelity resources, employees to emulate reality. 2-resources training, new forms of interaction and guide the user. 3-Resource Assessment, used to measure the performance of the user and give constructive regulation. This conceptual framework is nothing but a taxonomy specification simulator RV. Its application in the study of various commercial products has revealed how more advanced simulators RSV have a realism similar to that offered simulators physical bodies ex-vivo (59% and 62% respectively). It also proposes a methodology to achieve what would be an optimal design based on an analysis of the value of the various resources simulator RV. His greatest difficulty is the lack of a metric for the effectiveness of training with a power of discrimination sufficient. On the other hand, focuses on the analysis and modeling of the interaction forces in laparoscopy. This is carried out: 1-A characterization of the perception of the forces of traction. 2 - The acquisition of these forces vivo and the biomechanical properties of the ex-vivo tissue. 3, - The development of a model for these forces to a simulator of RSV. The results have revealed the existence of a "tactile memory" in expert surgeons and not a "visual touch" as other researchers have suggested, as well as surgeons have been able to distinguish between four tissues. This information comes mostly touch, indicating the desirability of a filling up of forces in a simulator if he wants to be realistic. Analysis of the results has led to propose a model of the perceived consistency of the tissue, which is based on two main parameters of touch, indicating the desirability of a filling up of forces in a simulator if he wants to be realistic. Analysis of the results has led to propose a model of the perceived consistency of the tissue, which is based on two major parameters: the type of tissue (rigidity) and the "degree of fixation." Finally it is proposed design of two teaching modules to train laparoscopic RV, one of "basic skills" and one "Nissen fundoplications had." The need for training in the procedure Nissen has been defined through a Hierarchical Task Analysis, which has led to the selection of three critical steps for his or require specific motor skills. The designs proposed teaching are the result of several meetings of "validating content" with experts in surgery laparasocópica and their training. The major contributions to scientific issue addressed are: 1 - The conceptual framework, a tool that wants to help focus research efforts and development of simulations RV. 2-A simple model of the interaction forces in the draft tissue. 3-A methodology for analyzing sensory interaction in the field laparoscopic. 4 - The hypothetical clarification of the concept of "visual touch", which has proved to be an ability to "momoria touch." 5 - The design of two teaching simulators for training laparoscopic RV, one of "basic skills" and one "Nissen." It was proposed last several lines of work ahead, as the definition of what has been called the Surgical Signs of Circulation, or the strengthening of the capacity of educational simulators making greater use of resources for training and evaluation. It also provides a brief overview of how future research and technological development will define the future of the surgical arena.

Author: VIDAURRE ARBIZU MARÍA CARMEN.
Year: 2005.
University: PÚBLICA DE NAVARRA.
Place of defense: ESCUELA TÉCNICA SUPERIOR DE INGENIEROS INDUSTRIALES Y DE TELECOMUNICACIÓN.
Place of preparation: E.T.S.I.I. Y DE TELECOMUNICACIÓN, UNIVERSIDAD PUBLICA DE NAVARRA.

Summary: A brain computer interface (BCI) allows people with problems related to their environment interaction using electrophysiological signals from the brain. The interface used in this project is an EET system "cue-based" two classes whose strategy is the imagination of the movement of the right and left hands. Two channels are used for the extraction of bipolar components, which were parameters adaptive auto-regresivos (AAR) and estimators logarithmic power band (BP), which was subsequently classified by various types of discriminant analysis. It chose the paradigm of the basket for experiments. The performance of the subjects described in the error rate and mutual information, the result of the analysis of evidence independent of the data under study. The results were analyzed using different statistical tests (parametric and non - parametric). We studied different types of discriminant analysis for application to BCIs using the components AAR and BP. The study was carried out using exhaustive search algorithms and genetic data from three different subjects. The results conclude that the linear discriminant analysis (LDA) is the best classifier for ill-defined problems and that all classifiers based on discriminant function analysis in a manner very similar problems in well-defined, so that they can avoid regularization. Given the above findings, it was decided to explore algorithms to transform classifiers based on linear and quadratic discriminant analysis in automatically adapatativos in real time. The adaptation of the QDA is based on the motto of the inverse matrix. It analyzes data previously recorded 6 subjects to find stable estimators of the information matrix and parameters for the optimal classifier. These parameters are the adjustment coefficient UC, the length of the window for adaptation, N and the initial time in beginning such adaptation, Tini. It has also developed an algorithm to estimate Tini adaptively. But UC and N remained constant and were found by optimization through a comprehensive search because of the low number of iterations. After experiments were conducted online with 12 subjects with no previous experience in BCIs. Six of them received feedback parameters AAR and six other estimators with BP. The results show that the system is stable and that the subjects were able to control in a very short time. In addition, no difference was found in the performance of genetic parameters AAR over the estimators BP. It also developed an adaptive version of LDA-based filtering technique Kalma. This approach has two advantages over classifier QDA adaptive. The first is that the adjustment coefficient is estimated at UC line through the profit equation Kalma. The second is that the evidence of all kinds can be used to update the equations of LDA. Using this classifier, UC and Tini were adaptive and N remained constant. To find appropriate parameters for the classifier, a comprehensive search data previously recorded 6 subjects. There were tested with different types of components and found that the chain of parameters estimators AAR and BP was the best solution. Later experiments were conducted in line with 6 subjects. The results showed again that the system is stable and subjects could reduce its error rate by approximately 12% between the first woe the third session pilot. The experimental results show that all systems networking estimators BP and AAR parameters is significantly better than using either of these two components separately. There were no significant differences between the two algorithms used 8 two for 615 estimate the initial time for adaptation, nor in the results of classifying data using components AAR and BP. The analysis of experiments with classifiers not automatically adaptive, conducted with 6 subjects with no previous experience in BCIs, shows the best performance of the system automatically adaptive. According to our knowledge, this paper describes the first classifiers automatically adaptive to BCIs published. Experiments conducted with 18 subjects, showed that all systems are stable, that the subjects were able to reach controlled with the use and classifiers automatically adaptive allow better performance than non-adaptive automatic. The extension of these systems to multiclase and / or without prior knowledge of the class (systems "non cue-based"), will be part of the work in the future.

Author: Pascau González-Garzón Javier.
Year: 2006.
University: POLITÉCNICA DE MADRID.
Place of defense: E.T.S. de Ingenieros de Telecomunicación.
Place of preparation: Escuela Técnica Superior de Ingenieros de Telecomunicación.

Summary: This deepens Doctoral Thesis in the methods of registration studies multimodal (anatomical and functional) of laboratory animals using techniques based on information theory. It performs an analysis of the influence of parameters of the algorithm registration, as well as a systematic validation of the results. To evaluate the functioning of the recording methods based on information theory, presents a series of studies pattern occurring thanks to a set of external markers. These studies estimated the accuracy of the registration is achieved through markers, validating the design and configuration of these. Assesses the convergence and accuracy of the algorithms registration, and their dependence on various factors (desalineamiento initial parameters of the algorithm), by means of an experimental design that allows a systematic validation. The results allow to quantify the influence of factors in the outcome. Given that the methods described in this Doctoral Thesis is the main constraint desalineamiento initial maximum they are able to solve, it is proposed a method based on trunk roads and the centers of mass of images to prealinear studies. Using this method is achieved by improving the performance of the automatic algorithms, getting a percentage of successes for the changes evaluated close to 100%. Finally, all the algorithms developed and used in this Doctoral Thesis are integrated within an interface acquisition, reconstruction, analysis and visualization called MMWKS, specifically designed for studies of laboratory animals. In this way, users of such devices can integrate information from different modalities with ease.