Hand gesture recognition system based in computer vision and machine learning: Applications on human-machine interaction

Tese de Doutoramento em Engenharia de Eletrónica e de Computadores

Algorithms and computational tools for metabolic flux analysis

PhD thesis in Biomedical Engineering

A survey of distributed data aggregation algorithms

Distributed data aggregation is an important task, allowing the de- centralized determination of meaningful global properties, that can then be used to direct the execution of other applications. The resulting val- ues result from the distributed computation of functions like count, sum and average. Some application examples can found to determine the network size, total storage capacity, average load, majorities and many others. In the last decade, many di erent approaches have been pro- posed, with di erent trade-o s in terms of accuracy, reliability, message and time complexity. Due to the considerable amount and variety of ag- gregation algorithms, it can be di cult and time consuming to determine which techniques will be more appropriate to use in speci c settings, jus- tifying the existence of a survey to aid in this task. This work reviews the state of the art on distributed data aggregation algorithms, providing three main contributions. First, it formally de nes the concept of aggrega- tion, characterizing the di erent types of aggregation functions. Second, it succinctly describes the main aggregation techniques, organizing them in a taxonomy. Finally, it provides some guidelines toward the selection and use of the most relevant techniques, summarizing their principal characteristics.

SQL based frequent pattern mining

Data mining, frequent pattern mining, database mining, mining algorithms in SQL

Quantifying consistency of Event Related Potentials across subjects based on single-trial topographic analysis

Introduction: Non-invasive brain imaging techniques often contrast experimental conditions across a cohort of participants, obfuscating distinctions in individual performance and brain mechanisms that are better characterised by the inter-trial variability. To overcome such limitations, we developed topographic analysis methods for single-trial EEG data [1]. So far this was typically based on time-frequency analysis of single-electrode data or single independent components. The method's efficacy is demonstrated for event-related responses to environmental sounds, hitherto studied at an average event-related potential (ERP) level. Methods: Nine healthy subjects participated to the experiment. Auditory meaningful sounds of common objects were used for a target detection task [2]. On each block, subjects were asked to discriminate target sounds, which were living or man-made auditory objects. Continuous 64-channel EEG was acquired during the task. Two datasets were considered for each subject including single-trial of the two conditions, living and man-made. The analysis comprised two steps. In the first part, a mixture of Gaussians analysis [3] provided representative topographies for each subject. In the second step, conditional probabilities for each Gaussian provided statistical inference on the structure of these topographies across trials, time, and experimental conditions. Similar analysis was conducted at group-level. Results: Results show that the occurrence of each map is structured in time and consistent across trials both at the single-subject and at group level. Conducting separate analyses of ERPs at single-subject and group levels, we could quantify the consistency of identified topographies and their time course of activation within and across participants as well as experimental conditions. A general agreement was found with previous analysis at average ERP level. Conclusions: This novel approach to single-trial analysis promises to have impact on several domains. In clinical research, it gives the possibility to statistically evaluate single-subject data, an essential tool for analysing patients with specific deficits and impairments and their deviation from normative standards. In cognitive neuroscience, it provides a novel tool for understanding behaviour and brain activity interdependencies at both single-subject and at group levels. In basic neurophysiology, it provides a new representation of ERPs and promises to cast light on the mechanisms of its generation and inter-individual variability.

Epidemiology of multiple congenital anomalies in Europe: a EUROCAT population-based registry study.

BACKGROUND: This study describes the prevalence, associated anomalies, and demographic characteristics of cases of multiple congenital anomalies (MCA) in 19 population-based European registries (EUROCAT) covering 959,446 births in 2004 and 2010. METHODS: EUROCAT implemented a computer algorithm for classification of congenital anomaly cases followed by manual review of potential MCA cases by geneticists. MCA cases are defined as cases with two or more major anomalies of different organ systems, excluding sequences, chromosomal and monogenic syndromes. RESULTS: The combination of an epidemiological and clinical approach for classification of cases has improved the quality and accuracy of the MCA data. Total prevalence of MCA cases was 15.8 per 10,000 births. Fetal deaths and termination of pregnancy were significantly more frequent in MCA cases compared with isolated cases (p < 0.001) and MCA cases were more frequently prenatally diagnosed (p < 0.001). Live born infants with MCA were more often born preterm (p < 0.01) and with birth weight < 2500 grams (p < 0.01). Respiratory and ear, face, and neck anomalies were the most likely to occur with other anomalies (34% and 32%) and congenital heart defects and limb anomalies were the least likely to occur with other anomalies (13%) (p < 0.01). However, due to their high prevalence, congenital heart defects were present in half of all MCA cases. Among males with MCA, the frequency of genital anomalies was significantly greater than the frequency of genital anomalies among females with MCA (p < 0.001). CONCLUSION: Although rare, MCA cases are an important public health issue, because of their severity. The EUROCAT database of MCA cases will allow future investigation on the epidemiology of these conditions and related clinical and diagnostic problems.

Multi-scale support vector algorithms for hot spot detection and modelling

The algorithmic approach to data modelling has developed rapidly these last years, in particular methods based on data mining and machine learning have been used in a growing number of applications. These methods follow a data-driven methodology, aiming at providing the best possible generalization and predictive abilities instead of concentrating on the properties of the data model. One of the most successful groups of such methods is known as Support Vector algorithms. Following the fruitful developments in applying Support Vector algorithms to spatial data, this paper introduces a new extension of the traditional support vector regression (SVR) algorithm. This extension allows for the simultaneous modelling of environmental data at several spatial scales. The joint influence of environmental processes presenting different patterns at different scales is here learned automatically from data, providing the optimum mixture of short and large-scale models. The method is adaptive to the spatial scale of the data. With this advantage, it can provide efficient means to model local anomalies that may typically arise in situations at an early phase of an environmental emergency. However, the proposed approach still requires some prior knowledge on the possible existence of such short-scale patterns. This is a possible limitation of the method for its implementation in early warning systems. The purpose of this paper is to present the multi-scale SVR model and to illustrate its use with an application to the mapping of Cs137 activity given the measurements taken in the region of Briansk following the Chernobyl accident.

A musculoskeletal shoulder model based on pseudo-inverse and null-space optimization.

The goal of the present work was assess the feasibility of using a pseudo-inverse and null-space optimization approach in the modeling of the shoulder biomechanics. The method was applied to a simplified musculoskeletal shoulder model. The mechanical system consisted in the arm, and the external forces were the arm weight, 6 scapulo-humeral muscles and the reaction at the glenohumeral joint, which was considered as a spherical joint. The muscle wrapping was considered around the humeral head assumed spherical. The dynamical equations were solved in a Lagrangian approach. The mathematical redundancy of the mechanical system was solved in two steps: a pseudo-inverse optimization to minimize the square of the muscle stress and a null-space optimization to restrict the muscle force to physiological limits. Several movements were simulated. The mathematical and numerical aspects of the constrained redundancy problem were efficiently solved by the proposed method. The prediction of muscle moment arms was consistent with cadaveric measurements and the joint reaction force was consistent with in vivo measurements. This preliminary work demonstrated that the developed algorithm has a great potential for more complex musculoskeletal modeling of the shoulder joint. In particular it could be further applied to a non-spherical joint model, allowing for the natural translation of the humeral head in the glenoid fossa.

Cardiovascular risk estimation and eligibility for statin therapy using different scoring systems in Europe: a population-based study in Switzerland

BACKGROUND: Recommendations for statin use for primary prevention of coronary heart disease (CHD) are based on estimation of the 10- year CHD risk. We compared the 10-year CHD risk assessments and eligibility percentages for statin therapy using three scoring algorithms currently used in Europe. METHODS: We studied 5683 women and men, aged 35-75, without overt cardiovascular disease (CVD), in a population-based study in Switzerland. We compared the 10-year CHD risk using three scoring schemes, i.e., the Framingham risk score (FRS) from the U.S. National Cholesterol Education Program's Adult Treatment Panel III (ATP III), the PROCAM scoring scheme from the International Atherosclerosis Society (IAS), and the European risk SCORE for low-risk countries, without and with extrapolation to 60 years as recommended by the European Society of Cardiology guidelines (ESC). With FRS and PROCAM, high-risk was defined as a 10- year risk of fatal or non-fatal CHD>20% and a 10-year risk of fatal CVD≥5% with SCORE. We compared the proportions of high-risk participants and eligibility for statin use according to these three schemes. For each guideline, we estimated the impact of increased statin use from current partial compliance to full compliance on potential CHD deaths averted over 10 years, using a success proportion of 27% for statins. RESULTS: Participants classified at high-risk (both genders) were 5.8% according to FRS and 3.0% to the PROCAM, whereas the European risk SCORE classified 12.5% at high-risk (15.4% with extrapolation to 60 years). For the primary prevention of CHD, 18.5% of participants were eligible for statin therapy using ATP III, 16.6% using IAS, and 10.3% using ESC (13.0% with extrapolation) because ESC guidelines recommend statin therapy only in high-risk subjects. In comparison with IAS, agreement to identify eligible adults for statins was good with ATP III, but moderate with ESC. Using a population perspective, a full compliance with ATP III guidelines would reduce up to 17.9% of the 24′ 310 CHD deaths expected over 10 years in Switzerland, 17.3% with IAS and 10.8% with ESC (11.5% with extrapolation). CONCLUSIONS: Full compliance with guidelines for statin therapy would result in substantial health benefits, but proportions of high-risk adults and eligible adults for statin use varied substantially depending on the scoring systems and corresponding guidelines used for estimating CHD risk in Europe.

Creation of a carrier-grade telco based on a voice over IP backbone

En aquesta memòria l'autor, fent servir un enfoc modern, redissenya i implementa la plataforma que una empresa de telecomunicacions del segle 21 necessita per poder donar serveis de telefonia i comunicacions als seus usuaris i clients. Al llarg d'aquesta exposició es condueix al lector des d'una fase inicial de disseny fins a la implementació i posada en producció del sistema final desenvolupat, centrant-nos en solucionar les necessitats actuals que això implica. Aquesta memòria cubreix el software, hardware i els processos de negoci associats al repte de fer realitat aquest objectiu, i presenta al lector les múltiples tecnologies emprades per aconseguir-ho, fent emfàsi en la convergència actual de xarxes cap al concepte de xarxes IP i basant-se en aquesta tendència i utilitzant aquesta tecnologia de veu sobre IP per donar forma a la plataforma que finalment, de forma pràctica, es posa en producció.

Fast numerical algorithms for the computation of invariant tori in Hamiltonian systems

In this paper, we develop numerical algorithms that use small requirements of storage and operations for the computation of invariant tori in Hamiltonian systems (exact symplectic maps and Hamiltonian vector fields). The algorithms are based on the parameterization method and follow closely the proof of the KAM theorem given in [LGJV05] and [FLS07]. They essentially consist in solving a functional equation satisfied by the invariant tori by using a Newton method. Using some geometric identities, it is possible to perform a Newton step using little storage and few operations. In this paper we focus on the numerical issues of the algorithms (speed, storage and stability) and we refer to the mentioned papers for the rigorous results. We show how to compute efficiently both maximal invariant tori and whiskered tori, together with the associated invariant stable and unstable manifolds of whiskered tori. Moreover, we present fast algorithms for the iteration of the quasi-periodic cocycles and the computation of the invariant bundles, which is a preliminary step for the computation of invariant whiskered tori. Since quasi-periodic cocycles appear in other contexts, this section may be of independent interest. The numerical methods presented here allow to compute in a unified way primary and secondary invariant KAM tori. Secondary tori are invariant tori which can be contracted to a periodic orbit. We present some preliminary results that ensure that the methods are indeed implementable and fast. We postpone to a future paper optimized implementations and results on the breakdown of invariant tori.

La médecine de famille en Suisse à l'horizon 2030: quelle place pour la délégation des tâches médicales. [Family practitioners in Switzerland by 2030: which roles for medical task delegation?]

Due to population aging, by 2030 Switzerland may face a demand of 24 million family practitioner visits, a growth of 13 percent from the 2005 level. This result is based on the assumption that the per capita demand for doctor visits remains what was observed in 2005 by age groups and sex. During the same period, the total number of practitioners may decrease by 14 percent whereas the female proportion of such practitioners may double. These changes may cause a 33 percent decrease in the supply of physician visits to reach only 14 millions. The comparison of the demand and supply of family doctor visits reveals that by 2030, 10 million visits may be unmet which represents 40 percent of the demand. On the supply side, a full scale implementation of task delegation may partially reduce that gap (minus 2 millions). On the demand side, improved health status may bring in a larger decrease in the needs for visits (minus 4 million).

A Survey of Active Learning Algorithms for Supervised Remote Sensing Image Classification

Defining an efficient training set is one of the most delicate phases for the success of remote sensing image classification routines. The complexity of the problem, the limited temporal and financial resources, as well as the high intraclass variance can make an algorithm fail if it is trained with a suboptimal dataset. Active learning aims at building efficient training sets by iteratively improving the model performance through sampling. A user-defined heuristic ranks the unlabeled pixels according to a function of the uncertainty of their class membership and then the user is asked to provide labels for the most uncertain pixels. This paper reviews and tests the main families of active learning algorithms: committee, large margin, and posterior probability-based. For each of them, the most recent advances in the remote sensing community are discussed and some heuristics are detailed and tested. Several challenging remote sensing scenarios are considered, including very high spatial resolution and hyperspectral image classification. Finally, guidelines for choosing the good architecture are provided for new and/or unexperienced user.

Le traitement de l'échinococcose alvéolaire humaine: une approche multidisciplinaire [Treatment of alveolar echinococcosis: a multidisciplinary task]

Alveolar echinococcosis is characterized by a long asymptomatic period but, without treatment, up to 80% of patients may die within ten years of diagnosis. Owing to a lack of fast-acting and fully effective chemotherapy, partial radical hepatic resection is the only chance of cure. One-third of patients are now treated in this way, and complex vascular and biliary reconstruction procedures are sometimes necessary. Liver transplantation may also be indicated for highly selected patients (about 5%) with life-threatening complications after failure of other treatments. Interventional radiology and endoscopy can be used to drain liver abscesses and/or infected and obstructed bile ducts, either as palliative procedures or as a bridge to radical resection. Parasitostatic benzimidazole therapy, especially based on continuous albendazole administration, is mandatory for at least two years after radical resection, and for life in inoperable patients.