Application of neural networks for the prediction of cardiovascular signals

Abstract

The parameters of the cardiovascular system in general provide information related to normal physiological functioning and can be used in the prediction of disease-specific singularities. In particular, electrical and biomechanical records were studied in relation to the integrated cardio-respiratory-vascular system. Deep learning methods were applied to these records, in data sequences from the systemic vascular network through the use of a dynamic neural network: Long-Short Term Memory, for the prediction of time series such as aortic pressure and electrocardiogram using computational analysis tools. Errors were compared between the predictions and the training and validation sets of the real signals obtained from a clinical measurement protocol, obtaining results with optimizers such as adam or adagrad.