基于GRU-Attention多跳特征融合网络的血压预测模型建立

Establishment of Blood Pressure Prediction Model Based on GRU-Attention Multi-Hop Feature Fusion Network

摘要: 针对当今临床医学对高精度连续无创血压监测方法的需求，本文提出一种基于光电容积脉搏波(PPG)信号与心电图(ECG)信号的多传感器信号特征提取及融合的深度神经网络血压预测模型。通过对经过预处理的信号以基于多跳问答推理机制设计的多跳GRU-Attention网络进行特征融合来实现对动脉血压的预测。预测结果的各项评估指标与Bland-Altman一致性分析表明，该模型的预测效果良好，对临床医学上连续无创血压预测技术的发展具有积极意义。

Abstract: In response to the demand for high-precision continuous non-invasive blood pressure monitoring methods in today’s clinical medicine, this paper proposes a deep neural network blood pressure prediction model based on multi-sensor signal feature extraction and fusion of photoplethysmog-raphy (PPG) signals and electrocardiogram (ECG) signals. The prediction of arterial blood pressure is realized by performing feature fusion on the preprocessed signal with a multi-hop GRU- Attention network designed based on a multi-hop question answering reasoning mechanism. The Bland-Altman consistency analysis of the evaluation indicators of the prediction results shows that the prediction effect of the model is good, and it has positive significance for the development of continuous non-invasive blood pressure prediction technology in clinical medicine.