Abstract
INTRODUCTION: Among the causes of cardiovascular diseases, abnormal blood pressure is especially significant. Blood pressure is a crucial hemodynamic biomarker of the cardiovascular health. Central aortic blood pressure correlates more closely with cardiovascular disease than peripheral arterial blood pressure. It can reflect the status of coronary arteries and aortas more directly and accurately, making it a significant tool for assessing cardiovascular risks. Invasive central aortic blood pressure measurement is considered the "gold standard" for evaluating left ventricular and coronary artery loads. However, due to the invasive and high cost of consumables, the widespread use of central aortic blood pressure is hindered in primary medical institutions and among large populations. Traditional non-invasive methods also have some limitations. METHODS: This paper proposes reconstructing central aortic pressure based on the TCN-Attention model, which primarily extracts local patterns from the time series. Simultaneously, the attention mechanism focuses on extracting global patterns to compensate for the shortcomings of the TCN model, which cannot perform global feature extraction. It efficiently extracts local patterns in the time series data that characterize mutations and other key time points and global patterns that indicate trends and periodicity, thus enabling the efficient reconstruction of central aortic pressure. RESULTS: The experimental results demonstrate that the improved TCN-Attention model presented in this paper is more accurate than the TCN model. DISUSSION: The precise measurement of central aortic pressure has significant clinical value in preventing, diagnosing, and treating cardiovascular diseases.