Physiology-Informed Real-Time Mean Arterial Blood Pressure Learning and Prediction for Septic Patients Receiving Norepinephrine

Abstract

Objective: Septic shock is a life-threatening manifestation of infection with a mortality of 20-50% [1]. A catecholamine vasopressor, norepinephrine (NE), is widely used to treat septic shock primarily by increasing blood pressure. For this reason, future blood pressure knowledge is invaluable for properly controlling NE infusion rates in septic patients. However, recent machine learning and data-driven methods often treat the physiological effects of NE as a black box. In this paper, a real-time, physiology-informed human mean arterial blood pressure model for septic shock patients undergoing NE infusion is studied.

Methods: Our methods combine learning theory, adaptive filter theory, and physiology. We learn least mean square adaptive filters to predict three physiological parameters (heart rate, pulse pressure, and the product of total arterial compliance and arterial resistance) from previous data and previous NE infusion rate. These predictions are combined according to a physiology model to predict future mean arterial blood pressure.

Results: Our model successfully forecasts mean arterial blood pressure on 30 septic patients from two databases. Specifically, we predict mean arterial blood pressure 3.33 minutes to 20 minutes into the future with a root mean square error from 3.56 mmHg to 6.22 mmHg. Additionally, we compare the computational cost of different models and discover a correlation between learned NE response models and a patient's SOFA score.

Conclusion: Our approach advances our capability to predict the effects of changing NE infusion rates in septic patients.

Significance: More accurately predicted MAP can lessen clinicians' workload and reduce error in NE titration.

Fig. 1.

Demonstration of baroreflex model

Fig. 2.

Physiological NE-MAP model

Fig. 3.

Graphical demonstration of each measurement

Fig. 4.

MAP prediction flow chart

Fig. 5.

Prediction results comparison where the red curves in the figures are predicted 6 from data collected minutes prior. We compare four methods: (a) our method, (b) reduced rank mean squares, (c) ARMAX, and (d) IIR

Fig. 6.

The root mean square error of MAP prediction as a function prediction time. The RMSE is measured by comparing the predicted MAP and observed MAP from patient data. The horizontal axis represents the prediction time, how far into the future we are predicting.

Fig. 7.

The comparison of the computational cost of each method in log(sec) versus data memory size. The computational time cost is measured in how long the computer needs to complete one prediction sample in seconds. The vertical axis represents this time length in log scale. The horizontal axis represents how many minutes of past data is used in the prediction process. One minute of data is equivalent to 300 samples when the sampling frequency is 5Hz.

Fig. 8.

NE response of HR over the time for two patients (rows) at three different times (columns)

Fig. 9.

NE response of PP over the time for two patients (rows) at three different times (columns)

Fig. 10.

NE response of KR over the time for two patients (rows) at three different times (columns)

Fig. 11.

The Relative Standard Deviation of NE related filter coefficient

Fig. 12.

The correlation between NE-PP and NE-HR mean response coefficients

Fig. 13.

The correlation between NE-PP mean response coefficient and SOFA score