Rhythm analysis of body surface potential mapping recordings from atrial fibrillation patients based on autocorrelation function / 生物医学工程学杂志

The study of atrial fibrillation (AF) has been known as a hot topic of clinical concern. Body surface potential mapping (BSPM), a noninvasive electrical mapping technology, has been widely used in the study of AF. This study adopted 10 AF patients' preoperative and postoperative BSPM data (each patient's data contained 128 channels), and applied the autocorrelation function method to obtain the activation interval of the BSPM signals. The activation interval results were compared with that of manual counting method and the applicability of the autocorrelation function method was verified. Furthermore, we compared the autocorrelation function method with the commonly used fast Fourier transform (FFT) method. It was found that the autocorrelation function method was more accurate. Finally, to find a simple rule to predict the recurrence of atrial fibrillation, the autocorrelation function method was used to analyze the preoperative BSPM signals of 10 patients with persistent AF. Consequently, we found that if the patient's proportion of channels with dominant frequency larger than 2.5 Hz in the anterior left region is greater than the other three regions (the anterior right region, the posterior left region, and the posterior right region), he or she might have a higher possibility of AF recurrence. This study verified the rationality of the autocorrelation function method for rhythm analysis and concluded a simple rule of AF recurrence prediction based on this method.

Totais decendiais de precipitação pluvial em Campinas, SP: persistência temporal, periodicidades e tendências climáticas / Ten-day rainfall amounts at Campinas, State of São Paulo, Brazil: temporal persistence, periodicities, and climate trends

O objetivo do estudo foi descrever as componentes tendências, periodicidades e persistência temporal presentes nas séries decendiais de precipitação pluvial (PRE) na localidade de Campinas/SP, entre os anos de 1890 e 2009. A aplicação da função auto-correlação e do teste Run permitiu inferir a inexistência de significativas correlações seriais entre os totais decendiais de PRE. Análises espectrais conduzidas no domínio tempo-frequencia não possibilitaram a indicação de significativas periodicidades nos picos de variância do sinal temporal PRE. Com base em métodos não paramétricos e considerando as probabilidades máximas de ocorrência do erro tipo I, usualmente admitidas na literatura (10 e 5 por cento), não houve indicação de marcantes tendências climáticas nas séries analisadas. Contudo, observou-se concentração de casos de tendências de queda a partir do final da época seca regional (mês de agosto) até o início do mês de novembro (pertencente à estação chuvosa). Essa característica pode ser considerada um indício de mudança nos padrões do clima regional no sentido de atraso da retomada da estação chuvosa.

The aim of the study was to evaluate the presence of temporal persistence, periodicals components, and trends within the ten-day rainfall series (PRE), observed during 1890-2009 at the weather station of Campinas, State of São Paulo, Brazil. Based on the autocorrelation function and on the Run test it was observed no significant temporal persistence within the data sample. The spectral analysis carried on the time-frequency domain has shown no remarkable periodicity associated with the variance peaks in the temporal signal under evaluation. Based on non parametric methods and considering the significance levels commonly adopted in the scientific literature (10 and 5 percent) it was detected no important climate trend within the PRE series. However, during the end of the dry season and the beginning of the rainy season it was observed an unexpected high number of decreasing (negative) trends cases that may be an indication of changes in the climatic conditions observed during 1890-2009. This last feature may reveal a delay in the resumption of the rainy season.

Rede neural de base radial aplicada na predição de séries temporais epidemiológicas: uso da validação cruzada com K dobras na determinação dos parâmetros da rede / Radial basis neural network applied in the prediction of time series epidemiologic use of K-fold cross validation with the determination of network parameters

This paper presents a radial basis function neural network (RBF) approach for the modeling and forecasting of public health survellance data. It extends the k-fold cross validation procedure for training neural networks with independent observations to handle time-series data...