Zugl.: Karlsruhe, Universiẗat, Diss., 1996.

Dans ce travail, nous avons développé une méthode de couplage d'un processus de diagnostic à base de modèle avec un processus de prognostic issu de l'expérience au moyen d'une architecture OSA-CBM (Open System Architecture for Condition Based Maintenance). Pour cela, le modèle 2M7P de la cellule photovoltaïque est optimisé par la mise en œuvre d'une méthode métaheuristique d'extraction des sept paramètres, FodPSO. Cette optimisation permet de mieux reproduire la performance de la cellule surtout en conditions réelles de fonctionnement. L'approche du diagnostic est traitée par la quantification du résidu différentiel au moyen de la distance euclidienne en vue de caractériser la signature électrique des cinq défauts ciblés. La connaissance de la signature des défauts permet la détection puis la discrimination. La partie pronostic évolue depuis une première approche issue des modèles de dégradation de Pan et Monroe d'une part et de Vazquez et al. d'autre part vers une approche par analyse de fiabilité basée sur l'expérience. Le couplage proposé dans ce travail, consiste en un échange d'information entre les deux processus de diagnostic et de pronostic pour l'aide à la décision selon une architecture où les décisions sont faites grâce à un test de seuil sur l'indicateur choisi, le Corrected Performance Ration. Les informations sont restituées à l'utilisateur sur la dernière couche de l'architecture. L'ensemble des approches ont été validées avec les données expérimentales issues de différentes centrales solaires (UPJV, Catania Univ…). / In this work, we have developed a method of coupling a model-based diagnostic process with a prognostic process derived from the experience using an OSA-CBM (Open System Architecture for Condition Based Maintenance) architecture. For that, the 2M7P model of the photovoltaic cell is optimized by the implementation of a metaheuristic method of extraction of seven parameters, FodPSO. This optimization makes it possible to better reproduce cell's performance, especially under actual operating conditions. The diagnostic approach is treated by quantifying the differential residual using Euclidean distance in order to characterize electrical signatures of five targeted faults. The knowledge of faults signatures allows detection and discrimination. The prognostic part evolves since a first approach resulting from models of degradation of Pan and Monroe on the one hand and Vazquez et al. on the other hand towards a reliability analysis approach based on experience. The coupling proposed in this work consists of an exchange of information between the two diagnostic and prognostic processes for decision support according to an architecture where decisions are made by means of a threshold test on the selected indicator, the Corrected performance Ratio. Information is returned to the user on last layer of the OSA-CBM architecture. All approaches have been validated with experimental data from different plants (UPJV Platform, Catania University…).

Abstract Most applications using network connection services require information to be transferred within specific constraints (or Quality of Service). Parlay enables applications to access functionality of underlying networks while preserving network integrity. Connection service functionality of underlying networks is currently provided to applications by Parlay in the Call Control and Data Session Control SCFs. Parlay does not however provide access to the QoS functionality of connection services although this functionality may be provided by networks. This report presents the design, specification and simulation of a QoS framework for connection services provided by Parlay. The QoS framework provides applications with access to the QoS functionality of connection services in the underlying networks. The design is divided into 3 parts (or models): the object model (defines the objects that make up the QoS framework), the information model (deals with how objects specify QoS and mappings between different QoS specifications), and the interaction model (defines how objects interact). A formal, technology-independent specification of the QoS framework is presented using UML. The specification is composed using the three parts of the design. A simulation of the QoS framework presented in this report is also described to validate the framework. The simulation is a multi-threaded, distributed CORBA application implemented in JAVA (Java SDK version 1.5) and is based on the UML specification of the QoS framework. Details about the simulation design and implementation are summarised in this report. The QoS framework provides per-application, per-connection QoS support for Parlay’s connection services, supports existing and future Parlay connection services, follows existing Parlay design paradigms, and co-exists with and makes use of existing and future Parlay infrastructure. Parlay guidelines relating to permitted changes are followed strictly in the design of the QoS framework, which ensures the backward compatibility of Parlay if the QoS framework is added to the Parlay API. All Parlay design guidelines are also followed to allow for the easy integration of the QoS framework into the Parlay API. The simulation validates that the design of the QoS framework is: complete (in terms of specification), realistic, compatible with a standard Parlay implementation (JAVA and CORBA), and scalable (easy integration of new connection services).

SAS has become an increasingly important public-health problem in recent years. It can adversely affect neurocognitive, cardiovascular, respiratory diseases and can also cause behavior disorder. Moreover, up to 90% of these cases are obstructive sleep apnea (OSA). Therefore, the study of how to diagnose, detect and treat OSA is becoming a significant issue, both academically and medically. Polysomnography can monitor the OSA with relatively fewer invasive techniques. However, polysomnography-based sleep studies are expensive and time-consuming because they require overnight evaluation in sleep laboratories with dedicated systems and attending personnel. Therefore, to improve such inconveniences, one needs to develop a simplified method to diagnose the OSA, so that the OSA can be detected with less time and reduced financial costs. Since currently there seems to be no OSA detection technique available in Taiwan, the goal of this work is to develop a reliable OSA diagnostic algorithm. In particular, via signal processing, feature extraction and artificial intelligence, this thesis describes an on-line ECG-based OSA diagnostic system. It is hoped that with such a system the OSA can be detected efficiently and accurately.

The goal of this study was to investigate the potential of breathing sounds recorded during wakefulness for Obstructive Sleep Apnea (OSA) screening and severity estimation. Breathing sounds were recorded from 189 subjects in supine and sitting postures during nose and mouth breathing. Features were extracted from power spectrum and bispectrum of the signals. Data from 70 subjects were used for training. Validation accuracy, specificity, and sensitivity for non-OSA and OSA groups were 78%, 77%, and 82%, respectively. Screening based on six OSA risk factors was less accurate. Parallel classification by both breathing sound features and risk factors had high sensitivity (94%). OSA severity estimation, by classifying subjects into three classes of OSA severity, achieved a maximum validation accuracy of 71%. The results demonstrate the potential of breathing sounds for OSA screening. The proposed method can lead to significant improvements in efficient use of resources such as sleep laboratories.

The goal of this study was to investigate the potential of breathing sounds recorded during wakefulness for Obstructive Sleep Apnea (OSA) screening and severity estimation. Breathing sounds were recorded from 189 subjects in supine and sitting postures during nose and mouth breathing. Features were extracted from power spectrum and bispectrum of the signals. Data from 70 subjects were used for training. Validation accuracy, specificity, and sensitivity for non-OSA and OSA groups were 78%, 77%, and 82%, respectively. Screening based on six OSA risk factors was less accurate. Parallel classification by both breathing sound features and risk factors had high sensitivity (94%). OSA severity estimation, by classifying subjects into three classes of OSA severity, achieved a maximum validation accuracy of 71%. The results demonstrate the potential of breathing sounds for OSA screening. The proposed method can lead to significant improvements in efficient use of resources such as sleep laboratories.

Obstructive sleep apnea (OSA) is a prevalent disease. Often resources to provide care for OSA are inadequate, leading to long patient waiting times. Simpler validated methods of care are needed. In the first study in Chapter 2, the utility of a new high-sampling rate oximeter to diagnose OSA was explored. The home oximetry data collection was robust, with few failures and the data allowed the “rule in” or “rule out” of moderate-severe OSA with high degree of certainty. It is concluded that home oximetry could replace polysomnography (PSG) as a diagnostic test in a significant proportion of patients, thus allowing limited resources available for the care of those with OSA to be re-directed e.g. towards providing therapy. In Chapter 3, the diagnostic information from the oximeter was used to underpin a study designed to demonstrate that a nurse-led model of care could produce health outcomes in moderate-severe OSA not inferior to physician-led care. A randomised controlled multi-centre non-inferiority clinical trial was performed. 1,427 patients referred to 3 sleep medical centres with possible OSA were assessed. 195 patients were randomised to 2 models of care. Model A, a simplified model, involved home oximetry to diagnose moderate-severe OSA, auto-titrating constant positive airway pressure (APAP) to set a therapeutic constant positive airway pressure (CPAP), with all care supervised by an experienced nurse. Model B involved 2 laboratory PSG’s, to diagnose OSA then titrate CPAP, supervised by a sleep physician. The primary endpoint was change in Epworth Sleepiness Score (ESS) measured before and after 3 months of CPAP. A range of other outcome measures were collected. The change in ESS for nurse-led management (Model A) was not inferior to the physician-led service (Model B) since the lower limit of the two-sided 95% CI did not include -2, the margin of equivalence (difference 0.13, 95% CI -1.52 to -1.25). 11 patients in Model A and 10 in Model B were lost to follow up during the trial. There were no significant differences between Model A and Model B after 3 months of CPAP in any of the other outcome measures, including CPAP adherence at 3 months. It is concluded that a simplified nurse-led model of care can produce non-inferior results to physician-directed care in the management of moderate-severe OSA. In Chapter 4 the efficacy of CPAP in normalising or improving subjective and objective sleepiness, quality of life and selected neurocognitive measures was explored. It was shown that only a proportion of patients (60% on ESS, 35% on FOSQ) normalised their scores after 3 months of CPAP therapy. This is important information. As new health care delivery strategies evolve as a result of the data presented in Chapter 3 and elsewhere, it will be crucially important to train new health care professionals in the complexities of OSA management, such that they are aware that the symptoms of patients presenting for OSA investigations can have multiple aetiologies, and may not always resolve by simply applying CPAP. The data presented in this thesis add to the evidence base in treatment of moderate-severe OSA and will help further evolve health care delivery for this important disease.

The OSA/Parlay architecture supports the development of applications that control network connections through an open API. This research presents a proposal on improving the rate at which applications are developed and deployed using the Parlay/OSA architecture. The work seeks to facilitate software reuse by providing logical groupings in the application layer of the Parlay/OSA architecture. This research presents a new layer to provide a higher level of abstraction for application developer using Parlay to provide telecommunication services. The layer introduced is referred to as the Generic Application Programming (GApp) interface. This document details the design and implementation of this interface.

BACKGROUND: Treatment-emergent central sleep apnea (TECSA, formerly known as complex sleep apnea) is the conversion of obstructive sleep apnea (OSA) to central sleep apnea (CSA) and/or hypopnea while undergoing treatment, especially continuous positive air pressure (CPAP) therapy. Some research suggests prevalence of TECSA on CPAP could be as high as 20.3%, while other data reports prevalence as low as 0.56%. OBJECTIVES: The objective of this study is to elucidate the prevalence, as well as natural course, of TECSA with CPAP use which currently varies widely in different studies. METHODS: This is a prospective multicenter longitudinal study to be conducted with 3276 treatment-naïve OSA patients for 36 months each under a rolling enrollment model, with an estimated study run time of 5-6 years. After undergoing full-night CPAP titration, each patient will begin CPAP therapy at home using AirView or other ResMed PAP devices, which will record apnea-hypopnea index (AHI) and central apnea index (CAI) data into a U.S. PAP device telemonitoring database. Every six months, all patients will have a follow-up appointment, and a subset will receive full-night diagnostic polysomnography to confirm CPAP data. Patients will also self-assess and report habitual daytime sleepiness using the Epworth Sleepiness Scale upon diagnosis and at follow-up every 6 months. RESULTS: Prevalence of TECSA will be assessed for statistical significance using the Chi-squared test or Fisher’s exact test if expected cell counts are <5. Statistical significance is assumed for type I error <0.05. To measure the sensitivity and specificity of telemonitoring CPAP data, continuous variables for the accuracy and precision of the overlapping measures (AHI, apnea index, and hypopnea index) will be compared to PSG. Results from the Epsworth Sleepiness Scale will be measured by computing the mean score at each 6-month follow-up. DISCUSSION: This study will provide evidence as to the true prevalence of TECSA in the U.S. for OSA patients receiving CPAP therapy. Estimated TECSA prevalence while on CPAP therapy is anticipated to be higher than is currently shown in the literature, with a conservative estimate of 15%. As the largest study evaluating for TECSA prevalence to date, any results that are found will be more accurate than previous studies. This is expected to guide future treatment of OSA as a higher prevalence of TECSA while on CPAP therapy (the mainstay of OSA treatment) may impact clinical decision-making when considering treatment options for OSA.

No description available.