Cardiologists can gain useful insight into a patient's condition when they are able to correlate the patent's symptoms and activities. For this purpose, a Holter Monitor is often used - a portable electrocardiogram (ECG) recorder worn by the patient for a period of 24-72 hours. Preferably, the monitor is not cumbersome to the patient and thus it should be designed to be as small and light as possible; however, the storage requirements for such a long signal are very large and can significantly increase the recorder's size and cost, and so signal compression is often employed. At the same time, the decompressed signal must contain enough detail for the cardiologist to be able to identify irregularities. "Lossy" compressors may obscure such details, where a "lossless" compressor preserves the signal exactly as captured.<p>The purpose of this thesis is to develop a platform upon which a Holter Monitor can be built, including a hardware-assisted lossless compression method in order to avoid the signal quality penalties of a lossy algorithm. <p>The objective of this thesis is to develop and implement a low-complexity lossless ECG encoding algorithm capable of at least a 2:1 compression ratio in an embedded system for use in a Holter Monitor. <p>Different lossless compression techniques were evaluated in terms of coding efficiency as well as suitability for ECG waveform application, random access within the signal and complexity of the decoding operation. For the reduction of the physical circuit size, a System On a Programmable Chip (SOPC) design was utilized. <p>A coder based on a library of linear predictors and Rice coding was chosen and found to give a compression ratio of at least 2:1 and as high as 3:1 on real-world signals tested while having a low decoder complexity and fast random access to arbitrary parts of the signal. In the hardware-assisted implementation, the speed of encoding was a factor of between four and five faster than a software encoder running on the same CPU while allowing the CPU to perform other tasks during the encoding process.

Cardiologists can gain useful insight into a patient's condition when they are able to correlate the patent's symptoms and activities. For this purpose, a Holter Monitor is often used - a portable electrocardiogram (ECG) recorder worn by the patient for a period of 24-72 hours. Preferably, the monitor is not cumbersome to the patient and thus it should be designed to be as small and light as possible; however, the storage requirements for such a long signal are very large and can significantly increase the recorder's size and cost, and so signal compression is often employed. At the same time, the decompressed signal must contain enough detail for the cardiologist to be able to identify irregularities. "Lossy" compressors may obscure such details, where a "lossless" compressor preserves the signal exactly as captured.<p>The purpose of this thesis is to develop a platform upon which a Holter Monitor can be built, including a hardware-assisted lossless compression method in order to avoid the signal quality penalties of a lossy algorithm. <p>The objective of this thesis is to develop and implement a low-complexity lossless ECG encoding algorithm capable of at least a 2:1 compression ratio in an embedded system for use in a Holter Monitor. <p>Different lossless compression techniques were evaluated in terms of coding efficiency as well as suitability for ECG waveform application, random access within the signal and complexity of the decoding operation. For the reduction of the physical circuit size, a System On a Programmable Chip (SOPC) design was utilized. <p>A coder based on a library of linear predictors and Rice coding was chosen and found to give a compression ratio of at least 2:1 and as high as 3:1 on real-world signals tested while having a low decoder complexity and fast random access to arbitrary parts of the signal. In the hardware-assisted implementation, the speed of encoding was a factor of between four and five faster than a software encoder running on the same CPU while allowing the CPU to perform other tasks during the encoding process.

O sistema Holter é considerado padrão ouro para detecção de arritmias cardíacas, possibilitando o seu diagnóstico e sua relação com os sintomas do paciente. As novas gerações de marcapassos (MP) detectam e registram eletrogramas intracavitários e poderiam dispensar o Holter no seguimento dos pacientes. Os diferentes MP dispõem de algoritmos de detecção de arritmias ventriculares (AV) que variam de acordo com o modelo e sua correlação diagnóstica com Holter foi pouco estudada. O objetivo desse estudo foi avaliar 1. a correlação entre o monitor de eventos dos MP e do Holter na detecção de AV e 2. a concordância das AV detectadas nos dois sistemas. Foram estudados129 pacientes portadores de MP com função de detecção de arritmias, com idade média de 68,6+19.1 anos (entre 19 e 94), 54,8% do sexo feminino. Os pacientes foram submetidos à monitorização com Holter por 72 horas. Assim que o sistema foi ligado, os contadores de eventos dos marcapassos foram reiniciados e os relógios sincronizados, para que ambos os sistemas detectassem os eventos simultaneamente. Os MP foram programados para detecção de eventos com o menor valor de frequência ventricular (FV) e o menor número de batimentos sequenciais (BT) possíveis. Depois de 72 horas, os sistemas Holter foram retirados e os registros analisados, assim como os registros simultâneos dos gravadores dos MP. Foram qualificados como eventos arrítmicos no Holter e MP, respectivamente: EV isoladas e \"PVC\"; extra-sístoles em pares e \"couplets\"; TVNS (> 3BT) e \"triplets\"(3BT), \"runs\"(3 a 8 ou > 8 BT) e \"HVR\" (3 ou 4 BT). As correlações de Spearman foram utilizadas para avaliar se o marcapasso acompanha a detecção dos parâmetros do Holter. Os coeficientes de correlação intraclasse e os respectivos intervalos com 95% de confiança, calculados para avaliar a concordâncias entre os parâmetros equivalentes do marcapasso e do Holter. Foram calculados os coeficientes Kappa para avaliar a concordância na detecção de > 10 \"PVC\"/h com > 10 EV/h. Resultados: Os monitores dos marcapassos subestimaram o Holter. Os registros de \"PVC\", \"triplet\" e \"HVR\" por TVNS apresentaram correlações positivas em relação aos parâmetros do Holter, sendo a mais alta aquela entre \"PVC\" e EV (r=0,501). Porém, a concordância entre os tipos de arritmias detectadas foi baixa (CCI < 0,5), exceção feita à concordância de \"triplet\" com TVNS de três batimentos (CCI=0.984). A concordância na detecção de mais de 10 PVC/h e mais de 10 EV/h foi moderada (kappa= 0,483), embora para os MP da Medtronic, o coeficiente de concordância foi alto (kappa=0.877). Para os MP com algoritmo de detecção de sequências de três batimentos com FV menor que 140bpm ( < 140/3), a correlação entre HVR e TVNS foi expressiva (r = 1), sendo a concordância entre esses parâmetros também bastante alta (CCI = 0,800). Conclusões: A correlação e a concordância na detecção de AV registradas nos MP e no Holter foram inconsistentes. Padronização dos algoritmos de detecção de AV, semelhantes os do sistema Holter, é necessária para que os pacientes portadores de dispositivos implantáveis possam se beneficiar dessa função para seguimento clínico e estratificação de risco / The Holter monitoring is considered the gold standard method for detection of cardiac arrhythmias, enabling its diagnosis and its correlation with the patient\'s symptoms. New generations of artificial pacemakers can detect and record intracavitary electrograms and, theoretically, could discard the Holter during the follow-up of the patients. Pacemakers have different ventricular arrhythmias (VA) detection algorithms, varying according to the model. Their diagnosis capacity comparing to Holter monitoring has been poorly studied. The aim of this study was to evaluate 1. the correlation between the event monitor of the pacemaker and the Holter in VA detection and 2. the agreement between the VA detected in both systems. We studied 129 patients with implanted pacemakers, which had arrhythmia detection function, mean age of 68.6 + 19.1 years (19 to 94), 54.8% female. The patients underwent Holter monitoring for 72 hours. Once the system was connected, event counters of pacemakers were reset and the clocks synchronized, so that both systems could detect the arrhythmic events simultaneously. The pacemakers were programmed to detect events with the lowest ventricular rate and lowest number of sequential beats allowed for each model. After 72 hours, Holter systems were removed and the records analyzed, as well as the simultaneous records of the pacemakers. Were considered as arrhythmic events: PVC isolated (in the pacemaker described as \"PVC\"); premature beats in pairs (pacemaker described as \"couplets\"); NSVT (pacemaker described as \"triplets\"- 3beats, \"runs\"- 4 to 8 or > 8 beats and \"HVR\"- 3 to 4 beats). Spearman rank correlations were used to assess whether the pacemaker and Holter identified the same parameters. The intraclass correlation coefficients and the respective intervals with 95% confidence were calculated to evaluate the concordance between the equivalent parameters of the pacemaker and Holter. Kappa coefficients were calculated to assess the agreement in the detection of > 10 PVC/h by the pacemakers and by the Holter. Results: The pacemakers underestimated the arrhythmias detection of Holter. Records of \"PVC\", \"triplet\" and \"HVR\" by NSVT showed positive correlations with the Holter parameters, and the highest one was among \"PVC\" and EV (r = 0.501). The agreement between the types of arrhythmias detected was quite low (CCI < 0.5), except for \"triplet\" detected by pacemakers and three beats NSVT by Holter (ICC = 0.984). The correlation detection for more than 10 PVC /h was moderate (kappa = 0.483), except for Medtronic pacemakers (kappa=0.877). When the pacemaker was programmed to detect sequences of three beats with heart rate lower than 140bpm ( < 140/3), the correlation between HVR and NSVT was perfect (r = 1) and the agreement between these parameters was also quite high (ICC = 0.800). Conclusions: The correlation and agreement between pacemakers and Holter monitoring in the detection of VA were not consistent. A standardization of the pacemakers\' detection algorithms is necessary before using this function for clinical follow-up and risk stratification of the patients

O sistema Holter é considerado padrão ouro para detecção de arritmias cardíacas, possibilitando o seu diagnóstico e sua relação com os sintomas do paciente. As novas gerações de marcapassos (MP) detectam e registram eletrogramas intracavitários e poderiam dispensar o Holter no seguimento dos pacientes. Os diferentes MP dispõem de algoritmos de detecção de arritmias ventriculares (AV) que variam de acordo com o modelo e sua correlação diagnóstica com Holter foi pouco estudada. O objetivo desse estudo foi avaliar 1. a correlação entre o monitor de eventos dos MP e do Holter na detecção de AV e 2. a concordância das AV detectadas nos dois sistemas. Foram estudados129 pacientes portadores de MP com função de detecção de arritmias, com idade média de 68,6+19.1 anos (entre 19 e 94), 54,8% do sexo feminino. Os pacientes foram submetidos à monitorização com Holter por 72 horas. Assim que o sistema foi ligado, os contadores de eventos dos marcapassos foram reiniciados e os relógios sincronizados, para que ambos os sistemas detectassem os eventos simultaneamente. Os MP foram programados para detecção de eventos com o menor valor de frequência ventricular (FV) e o menor número de batimentos sequenciais (BT) possíveis. Depois de 72 horas, os sistemas Holter foram retirados e os registros analisados, assim como os registros simultâneos dos gravadores dos MP. Foram qualificados como eventos arrítmicos no Holter e MP, respectivamente: EV isoladas e \"PVC\"; extra-sístoles em pares e \"couplets\"; TVNS (> 3BT) e \"triplets\"(3BT), \"runs\"(3 a 8 ou > 8 BT) e \"HVR\" (3 ou 4 BT). As correlações de Spearman foram utilizadas para avaliar se o marcapasso acompanha a detecção dos parâmetros do Holter. Os coeficientes de correlação intraclasse e os respectivos intervalos com 95% de confiança, calculados para avaliar a concordâncias entre os parâmetros equivalentes do marcapasso e do Holter. Foram calculados os coeficientes Kappa para avaliar a concordância na detecção de > 10 \"PVC\"/h com > 10 EV/h. Resultados: Os monitores dos marcapassos subestimaram o Holter. Os registros de \"PVC\", \"triplet\" e \"HVR\" por TVNS apresentaram correlações positivas em relação aos parâmetros do Holter, sendo a mais alta aquela entre \"PVC\" e EV (r=0,501). Porém, a concordância entre os tipos de arritmias detectadas foi baixa (CCI < 0,5), exceção feita à concordância de \"triplet\" com TVNS de três batimentos (CCI=0.984). A concordância na detecção de mais de 10 PVC/h e mais de 10 EV/h foi moderada (kappa= 0,483), embora para os MP da Medtronic, o coeficiente de concordância foi alto (kappa=0.877). Para os MP com algoritmo de detecção de sequências de três batimentos com FV menor que 140bpm ( < 140/3), a correlação entre HVR e TVNS foi expressiva (r = 1), sendo a concordância entre esses parâmetros também bastante alta (CCI = 0,800). Conclusões: A correlação e a concordância na detecção de AV registradas nos MP e no Holter foram inconsistentes. Padronização dos algoritmos de detecção de AV, semelhantes os do sistema Holter, é necessária para que os pacientes portadores de dispositivos implantáveis possam se beneficiar dessa função para seguimento clínico e estratificação de risco / The Holter monitoring is considered the gold standard method for detection of cardiac arrhythmias, enabling its diagnosis and its correlation with the patient\'s symptoms. New generations of artificial pacemakers can detect and record intracavitary electrograms and, theoretically, could discard the Holter during the follow-up of the patients. Pacemakers have different ventricular arrhythmias (VA) detection algorithms, varying according to the model. Their diagnosis capacity comparing to Holter monitoring has been poorly studied. The aim of this study was to evaluate 1. the correlation between the event monitor of the pacemaker and the Holter in VA detection and 2. the agreement between the VA detected in both systems. We studied 129 patients with implanted pacemakers, which had arrhythmia detection function, mean age of 68.6 + 19.1 years (19 to 94), 54.8% female. The patients underwent Holter monitoring for 72 hours. Once the system was connected, event counters of pacemakers were reset and the clocks synchronized, so that both systems could detect the arrhythmic events simultaneously. The pacemakers were programmed to detect events with the lowest ventricular rate and lowest number of sequential beats allowed for each model. After 72 hours, Holter systems were removed and the records analyzed, as well as the simultaneous records of the pacemakers. Were considered as arrhythmic events: PVC isolated (in the pacemaker described as \"PVC\"); premature beats in pairs (pacemaker described as \"couplets\"); NSVT (pacemaker described as \"triplets\"- 3beats, \"runs\"- 4 to 8 or > 8 beats and \"HVR\"- 3 to 4 beats). Spearman rank correlations were used to assess whether the pacemaker and Holter identified the same parameters. The intraclass correlation coefficients and the respective intervals with 95% confidence were calculated to evaluate the concordance between the equivalent parameters of the pacemaker and Holter. Kappa coefficients were calculated to assess the agreement in the detection of > 10 PVC/h by the pacemakers and by the Holter. Results: The pacemakers underestimated the arrhythmias detection of Holter. Records of \"PVC\", \"triplet\" and \"HVR\" by NSVT showed positive correlations with the Holter parameters, and the highest one was among \"PVC\" and EV (r = 0.501). The agreement between the types of arrhythmias detected was quite low (CCI < 0.5), except for \"triplet\" detected by pacemakers and three beats NSVT by Holter (ICC = 0.984). The correlation detection for more than 10 PVC /h was moderate (kappa = 0.483), except for Medtronic pacemakers (kappa=0.877). When the pacemaker was programmed to detect sequences of three beats with heart rate lower than 140bpm ( < 140/3), the correlation between HVR and NSVT was perfect (r = 1) and the agreement between these parameters was also quite high (ICC = 0.800). Conclusions: The correlation and agreement between pacemakers and Holter monitoring in the detection of VA were not consistent. A standardization of the pacemakers\' detection algorithms is necessary before using this function for clinical follow-up and risk stratification of the patients