Cardiovascular diseases are the leading cause of death worldwide, and are estimated to kill over 17 million people each year, about 31% of all deaths. In the clinic, the first diagnostic procedure for a suspected cardiac abnormality is often acquisition of an electrocardiogram (ECG), which measures the electrical potential of the heart at the body surface. Understanding the mechanisms underlying generation of the ECG waveforms is crucial for optimal clinical benefit. Computer simulations possess several strengths as a tool to gain this understanding, particularly in terms of human-specificity, flexibility, repeatability, and ethics. The ventricles make up the majority of the cardiac volume and are therefore responsible for the majority of ECG waveforms. Ventricular disorders are the most life-threatening, because the ventricles are responsible for pumping blood to the body. Due to their size it has only recently become possible to perform biophysically detailed simulations of the ventricles and torso using supercomputers. In this thesis, multiscale, mathematical models of the ventricles and torso using the Chaste software library are simulated on high performance computing systems. A description is included of the performance enhancements made in Chaste to improve resource efficiency and accelerate job turnaround, particularly in data storage and the auxiliary tasks of post-processing and data conversion. A novel model of ventricular activation is presented and parametrized using multi-modal human data, and successfully used to simulate normal and pathological QRS complexes. Similarly, repolarization gradients are imposed based on the literature and result in a variety of T waves. Finally, the developed human whole-ventricular and torso models are utilized to gain new insights into possible ionic mechanisms underlying the clinical manifestations of the early repolarization syndrome. Overall, this thesis presents a novel framework for simulation of the human ECG using high performance computers, with possible applications in basic science and computational medicine.

nÃo hà / A tarefa mais importante em processamento de sinais de eletrocardiograma (ECG) à a determinaÃÃo exata do complexo de QRS, em particular, a detecÃÃo dos picos de onda R atravÃs de sistemas e anÃlises computadorizadas. à essencial, especialmente, para uma medida correta da variabilidade do ritmo cardÃaco (HRV). Um grande obstÃculo a ser superado para uma detecÃÃo confiÃvel à a sensibilidade do eletrocardiograma a diversas fontes de distÃrbio, tais como, a interferÃncia à rede elÃtrica, os artefatos do movimento, flutuaÃÃo da linha base e o ruÃdo dos mÃsculos. Este trabalho utiliza as propriedades matemÃticas da transformaÃÃo de Hilbert sobre wavelets para desenvolver um novo algoritmo capaz de diferenciar as ondas R das demais (P, Q, S, T e U) e facilitar a detecÃÃo dos complexos QRS. Uma taxa de detecÃÃo do complexo QRS de 99,92% à alcanÃada para a base de dados de arritmias do MIT-BIH. A tolerÃncia a ruÃdo do mÃtodo proposto foi tambÃm testada usando os registros padrÃo da base de dados MIT-BIH Noise Stress Test. A taxa da detecÃÃo do detector ficou aproximadamente 99,35% mesmo para as relaÃÃes sinal-ruÃdo (SNR) tÃo baixo quanto 6dB. / The most important task in the ECG signal processing is the accurate determina-tion of QRS complex, in particular, accurate detection of the R wave peaks, is essential in computer-based ECG analysis especially for a correct measurement of Heart Rate Variability (HRV). A great hurdle to be overcome in reliable detection is the sensibility of the electrocar-diogram to several disturbance sources such as powering source interference, movement arti-facts, baseline wandering and muscle noise. This study uses the Hilbert Transform pairs of wavelet bases for QRS detection. From the properties of these mathematical tools it was pos-sible to develop an algorithm which is able to differentiate the R waves from the others (P, Q, S, T and U waves).The performance of the algorithm was verified using the records MIT-BIH arrhythmia and normal databases. A QRS detection rate of 99.92% was achieved against MIT-BIH arrhythmia database. The noise tolerance of the proposed method was also tested using standard records from the MIT-BIH Noise Stress Test Database. The detection rate of the detector remains about 99.35% even for signal-to-noise ratios (SNR) as low as 6dB.

The aim of this thesis is to analyse high-frequency ECG to detect some heart diseases. This is performed with averaging of selected QRS complexes for each lead of the signal; these are thenfilteredin range 500-1 000 Hz. After that the envelope of the signal is done and here the peaks are detected. Based on mutual positions of this peaks, it is possible to detectwhat kind od signal we treat.

The aim of this thesis is to analyse high-frequency ECG to detect some heart diseases. This is performed with averaging of selected QRS complexes for each lead of the signal; these are then filtered in range 500-1 000 Hz. After that the envelope of the signal is done and here the peaks are detected. Based on mutual positions of this peaks, it is possible to detect what kind od signal we treat.

This thesis deals with implementation of detector QRS complex with use of wavelet transform. The first part is focused on formation and possibility to measure cardiac activity. The other part of thesis we will familiarise with the different possibilities of detection QRS complex and we intimately focus on wavelet transform that will be used for a project of detection QRS complex. The practical part of thesis focuses on the project of detector in programming language Matlab and his different setting. This projected detector has been tested with CSE database. Achieved results of projected detector are evaluated with the results of others authors.

In this thesis, many machine learning algorithms were applied to electrocardiogram (ECG), spectral analysis, and Field Programmable Gate Arrays (FPGAs). In ECG, QRS complexes are useful for measuring the heart rate and for the segmentation of ECG signals. QRS complexes were detected using WaveletCNN Autoencoder filters and ConvLSTM detectors. The WaveletCNN Autoencoders filters the ECG signals using the wavelet filters, while the ConvLSTM detects the spatial temporal patterns of the QRS complexes. For the spectral analysis topic, the detection of chemical compounds using spectral analysis is useful for identifying unknown substances. However, spectral analysis algorithms require vast amounts of data. To solve this problem, B-spline neural networks were developed for the generation of infrared and ultraviolet/visible spectras. This allowed for the generation of large training datasets from a few experimental measurements. Graphical Processing Units (GPUs) are good for training and testing neural networks. However, using multiple GPUs together is hard because PCIe bus is not suited for scattering operations and reduce operations. FPGAs are more flexible as they can be arranged in a mesh or toroid or hypercube configuration on the PCB. These configurations provide higher data throughput and results in faster computations. A general neural network framework was written in VHDL for Xilinx FPGAs. It allows for any neural network to be trained or tested on FPGAs. / Graduate

nÃo hà / Sistemas automÃticos de auxÃlio ao diagnÃstico visam à extraÃÃo de mÃtricas especÃficas, podendo ser por algoritmos computacionais, de forma a subsidiar a anÃlise por parte do especialista de condiÃÃes orgÃnicas e fisiolÃgicas do paciente. No contexto da cardiologia, referidos sistemas sÃo particularmente importantes quando aplicados no processamento de sinais de longa duraÃÃo, como o eletrocardiograma (ECG) de 24 horas. As tÃcnicas para segmentaÃÃo e extraÃÃo automÃtica de parÃmetros do sinal ECG propostas nesta tese abrangem diversos campos de pesquisa. Inicialmente, o sistema realiza a detecÃÃo e a segmentaÃÃo do complexo QRS, relacionado à despolarizaÃÃo ventricular. Como metodologia, utiliza-se a combinaÃÃo das tÃcnicas do limiar adaptativo, das transformadas de Hilbert e Wavelet e do filtro derivativo com uma nova abordagem de reduÃÃo de prÃ-processamento e de seleÃÃo do fator de escala da Wavelet. Ao final desta etapa, obtÃm-se a sÃrie de intervalos RR, a sÃrie de duraÃÃes de cada complexo QRS e de suas amplitudes. No segundo momento, tem-se a detecÃÃo e a segmentaÃÃo da onda T, relacionada à repolarizaÃÃo ventricular. PropÃe-se um novo modelo matemÃtico do comportamento morfolÃgico da onda T baseado na funÃÃo Gaussiana, modificada por um procedimento matemÃtico de inserÃÃo de assimetria. Uma vez obtidos os parÃmetros de modelagem para uma dada morfologia predominante de onda T, a funÃÃo de correlaÃÃo cruzada à utilizada para a detecÃÃo do pico e uma tÃcnica baseada no cÃlculo da Ãrea de trapÃzios à utilizada para a localizaÃÃo do final da forma de onda. Dentre as mÃtricas derivadas das informaÃÃes extraÃdas, destaca-se a sÃrie de intervalos QT, segmento que vai do inÃcio de cada complexo QRS ao final de cada onda T. Finalizado o processo de segmentaÃÃo, dois estudos de caso sÃo realizados: subtraÃÃo da atividade ventricular em sinais eletrogramas atriais de pacientes com fibrilaÃÃo atrial (FA) e anÃlise de sÃries de variabilidade da frequÃncia cardÃaca (VFC) de um conjunto de pacientes idosos selecionados pelo AmbulatÃrio de Geriatria do Hospital UniversitÃrio WÃlter CantÃdio. A partir de experimentos de validaÃÃo em bases de dados diversas com anotaÃÃes manuais dos batimentos, obtÃm-se as seguintes taxas de detecÃÃo e erros de delineamento para o complexo QRS: sensibilidade de 99,51%, preditividade positiva de 99,44%, erro mÃdio de inÃcio (QRS onset) de 2,85  9,90 ms e erro mÃdio de final (QRS offset) de 2,83  12,26 ms. Com relaÃÃo à detecÃÃo e segmentaÃÃo da onda T, obtÃm-se os seguintes resultados: sensibilidade de 99,48%, preditividade positiva de 99,53%, erro mÃdio de localizaÃÃo de pico de 0,51  8,06 ms e erro mÃdio de localizaÃÃo de final da forma de onda de 0,11  11,73 ms. Quanto ao primeiro estudo de caso de uso dos pontos fiduciais detectados, a potÃncia mÃdia dos sinais eletrogramas atriais, apÃs a subtraÃÃo da atividade ventricular, à significativamente reduzida para frequÃncias acima de 10 Hz, predominantemente associadas ao complexo QRS, bem como para frequÃncias na faixa de 3 a 5 Hz, relacionadas à atividade elÃtrica de repolarizaÃÃo ventricular. Para o segundo estudo, a anÃlise do comportamento de mÃtricas no domÃnio da frequÃncia associadas à atividade do sistema nervoso simpÃtico permite o reconhecimento de tendÃncias prÃprias e caracterÃsticas, no que tange a aspectos de funcionamento/disautonomia do sistema nervoso autonÃmico, de cada classe prÃ-determinada de idosos segundo os conceitos de fenÃtipo de fragilidade: idosos frÃgeis, prÃ-frÃgeis e robustos. Os resultados obtidos sugerem que o conjunto de metodologias desenvolvidas para a segmentaÃÃo do sinal ECG apresenta altas taxas de precisÃo, repetibilidade e robustez a uma ampla gama de morfologias, podendo ser aplicado em diversos contextos de auxÃlio ao diagnÃstico. Dadas as mÃtricas e sÃries temporais que podem ser extraÃdas, os referidos mÃtodos tambÃm podem dar suporte a processos de investigaÃÃo clÃnica e desenvolvimento de marcadores/indicadores de eventos cardiovasculares adversos.

This project considers methods of construction QRS detectors. It focus in detection complexes of QRS single leads and space speed, which are calculated from three orthogonal leads. In theory was refer to various methods, which lead to design detector. It were designed two algoritms (constant and adaptive detecting threshold), which were implemented into detector and the signal was preprocessed by Hilbert transformation. Toward algoritms were completed by modification, which improved detection effectivity. Function of algoritms were tested in all signals of CSE (V2,V5,aVF).

The Hilbert transform is a widely used transform in signal processing. In this thesis we explore its use for three different applications: electrocardiography, the Hilbert-Huang transform and modulation. For electrocardiography, we examine how and why the Hilbert transform can be used for QRS complex detection. Also, what are the advantages and limitations of this method? The Hilbert-Huang transform is a very popular method for spectral analysis for nonlinear and/or nonstationary processes. We examine its connection with the Hilbert transform and show limitations of the method. Lastly, the connection between the Hilbert transform and single-sideband modulation is investigated.

This project investigates methods of construction the wavelet-based QRS-complex detector. QRS-complex detection is very important, because it helps automatically calculate heart rate and in some cases it is used for compression ECG signal. The design of QRS detector can be made with many methods, in this project were mentioned and consequently tested only a few variants. The principle of designed detector used a wavelet-based decomposition of the original ECG signal to several frequency-coded bands. These bands are consequently transformed to absolute values and with the help of the threshold value are marked positions of assumed QRS complexes. Then are these assumed positions from all bands compared between themselves. If the position is confirmed at least at one nearby band, then is this position marked as true QRS complex. To increase efficiency of designed detector, two modifications were additionally mentioned. The first one, using the envelope of the signal, had rather negative effect on detectors efficiency. The second modification, using combined signal from three pseudoorthogonal leads, had reversely very good effect on detectors efficiency. In the end, the designed detector and all its modifications were tested on signals from CSE library (exactly on leads II, V2 and V6).