Deep learning approach for long term, cuffless, and continuous arterial blood pressure estimation

Patent metadata

Field	Value
Publication number	US-10973468-B2
Application number	US-201816033744-A
Country	US
Kind code	B2
Filing date	Jul 12, 2018
Priority date	Jul 12, 2018
Publication date	Apr 13, 2021
Grant date	Apr 13, 2021

Abstract

Methods and devices for long term, cuffless, and continuous arterial blood pressure estimation using an end-to-end deep learning approach are provided. A deep learning system comprises three modules: a deep convolutional neural network (CNN) module for learning powerful features; a one- or multi-layer recurrent neural network (RNN) module for modeling the temporal dependencies in blood pressure dynamics; and a mixture density network (MDN) module for predicting final blood pressure value. This system takes raw physiological signals, such as photoplethysmogram and/or electrocardiography signals, as inputs and yields arterial blood pressure readings in real time.

First claim

What is claimed is:

1. A method for long-term, cuffless, and continuous arterial blood pressure estimation, the method comprising:
extracting features from a physiological signal using a deep convolutional neural network (CNN);
modeling temporal dependencies in blood pressure dynamics using a recurrent neural network (RNN); and
outputting an estimate of arterial blood pressure using a mixture density network (MDN).

2. The method of claim 1 , wherein the physiological signal comprises a single physiological signal.

3. The method of claim 2 , wherein the single physiological signal comprises a photoplethysmogram (PPG) signal.

4. The method of claim 1 , wherein the physiological signal comprises multiple physiological signals.

5. The method of claim 4 , wherein the multiple physiological signals comprise a PPG signal and an electrocardiography (ECG or EKG) signal.

6. The method of claim 1 , wherein the physiological signal is filtered to remove noise and redundant information.

7. The method of claim 1 , further comprising training the CNN, the RNN, and the MDN through a backpropagation algorithm.

8. A system for long-term, cuffless, and continuous arterial blood pressure estimation, the system comprising:
a deep convolutional neural network (CNN) configured to extract features from a physiological signal;
a recurrent neural network (RNN) configured to model temporal dependencies in blood pressure; and
a mixture density network (MDN) configured to output an estimate of arterial blood pressure.

9. The system of claim 8 , wherein the CNN comprises a multi-layered convolutional neural network.

10. The system of claim 8 , wherein the CNN is a one-dimensional (1D) CNN, a two-dimensional (2D) CNN, or a three-dimensional (3D) CNN.

11. The system of claim 8 , wherein the CNN is configured to use generic convolution or dilated convolution.

12. The system of claim 8 , wherein the CNN is configured to be a cascade multi-layer structure incorporated with skip connections, and wherein the skip connection merge lower layer feature maps with higher layer feature maps to capture different time-scale variation patterns.

13. The system of claim 8 , wherein the CNN module is configured to work on a plurality of time-scales of input data.

14. The system of claim 8 , wherein the RNN further comprises a one-layer recurrent neural network or a multi-layer recurrent neural network.

15. The system of claim 8 , wherein the RNN comprises at least one of a standard RNN, a gated recurrent unit (GRU), and a long-short term memory (LSTM) network.

16. The system of claim 8 , wherein the RNN comprises a standard RNN or a bidirectional RNN.

17. The system of claim 8 , wherein the RNN comprises an attention mechanism.

18. The system of claim 8 , wherein the MDN is configured to model a blood pressure prediction as a classification problem, and wherein an output comprises a mixture of Gaussian distribution.

19. A system for long-term, cuffless, and continuous arterial blood pressure estimation, the system comprising:
a deep convolutional neural network (CNN) configured to extract features from a physiological signal and modeling temporal deficiencies, and comprising at least one of an attention mechanism and an algorithm configured to position-embed inputs; and
a mixture density network (MDN) configured to output an estimate of arterial blood pressure.

20. The system of claim 19 , wherein the attention mechanism enhances CNN feature maps with attention masks, and wherein the attention masks are generated from the feature maps.

Assignees

• Univ Hong Kong Chinese

Inventors

• Zhao Ni
• Su Peng
• Zhang Yuanting

Classifications

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title