Estimation of a physiological parameter using a neural network

• US 8,457,706 B2
• Filed: 05/15/2009
• Issued: 06/04/2013
• Est. Priority Date: 05/16/2008
• Status: Active Grant

• Current Assignee: Covidien LP (Medtronic Plc)
• Original Assignee: Covidien LP (Medtronic Plc)
• Inventors: Baker, Clark R. Jr.
• Primary Examiner(s): SZMAL, BRIAN SCOTT
• Application Number:
  US12/467,013
• Publication Number:
  US 20090287070A1
• Time in Patent Office: 1,481 Days
• Field of Search: 600/309, 600/310, 600/300, 600/481, 600/502, 600/323, 600/324, 367/155
• US Class Current: 600/323
• CPC Class Codes: A61B 5/14551 for measuring blood gases
• CPC Class Codes: A61B 5/7221 Determining signal validity, reliability or quality A61B5/7207 and A61B5/7217 take precedence
• CPC Class Codes: A61B 5/725 using specific filters therefor, e.g. Kalman or adaptive filters A61B5/316 take precedence
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A neural network is used to combine one or more estimates of a physiologic parameter with one or more associated signal quality metrics, creating a more accurate estimate of said physiologic parameter, as well as a second estimate of the accuracy of said physiologic parameter estimate.

Claims

1. A method for determining the oxygen saturation of blood of a patient comprising:

estimating a physiologic parameter from an electronic signal;

calculating one or more signal quality metrics of the electronic signal;

inputting information into a neural network, wherein said input information includes one or more estimates of the physiologic parameter and at least one signal quality metric; and

calculating with the neural network at least one output datum, wherein the output datum is representative of the oxygen saturation of the blood of the patient.

2. The method of claim 1 wherein the physiological parameter is pulsatile oxygen saturation of the blood of the patient.

3. The method of claim 1 wherein the input information comprises:

i) the estimated physiologic parameter;

ii) at least one signal quality metric;

iii) data processed from the estimated physiologic parameter; and

/or

iv) data processed from the one or more signal quality metrics.

4. The method of claim 1 further comprising:

displaying the output datum from the neural network on a display device.

5. The method of claim 1 further comprising:

generating an alarm if the output datum from the neural network is not within an acceptable range of oxygen saturation.

6. The method of claim 1 further comprising:

selecting the information to be input into the neural network; and

training the neural network using the selected information.

7. The method of claim 1 wherein the signal quality metrics comprise a Kalman saturation, an adaptive comb filter (ACF) rate, a Kalman saturation calculated using Kalman Cardiac Gated Averaging, a pattern match rate, a long-term pulse period variability metric, an ACF confidence metric, a pattern match rate confidence metric, a Kalman saturation variability metric calculated without Kalman Cardiac Gated Averaging, a Kalman saturation variability metric calculated using Kalman Cardiac Gated Averaging, an ACF signal-to-noise metric, an ACF skewness (pulse shape) metric, an age metric based on the Kalman saturation calculated without Kalman Cardiac Gated Averaging, an age metric based on the Kalman Saturation calculated using Kalman Cardiac Gated Averaging, a mean AC amplitude, an AC amplitude variability metric, a IR/Red light correlation metric, a DC variability metric, a DC slope, a skewness (pulse shape) metric of IR derivative waveform, a maximum autocorrelation metric based on an IR waveform, a persistent IR-Red phase difference metric, a frequency ratio, an amplitude ratio, and/or a low DC signal metric.

8. The method of claim 1 further comprising:

selecting the neural network from a set of neural networks based on current conditions.

9. The method of claim 8 wherein selecting further comprises:

selecting the neural network from the set of neural networks based on a current value of one or more of the calculated signal quality metrics.

10. The method of claim 8 wherein selecting further comprises:

selecting the neural network from the set of neural networks based on a characteristic of the patient.

11. A pulse oximeter comprising:

a microprocessor configured for calculating an estimated value of oxygen saturation of a patient'"'"'s blood based at least in part upon an electronic signal received from a sensor;

a neural network module configured for receiving the estimated value as an input and calculating a revised value of oxygen saturation of a patient'"'"'s blood using a neural network; and

a display configured for displaying the revised value of the oxygen saturation of the patient'"'"'s blood.

12. The pulse oximeter of claim 11 further comprising:

non-transitory memory configured for storing at least one neural network including the neural network used by the neural network module to calculate the revised value.

13. The pulse oximeter of claim 12, wherein the neural network module is further configured for receiving the estimated value and at least one signal quality metric describing the electronic signal.

14. The pulse oximeter of claim 12 further comprising:

an accuracy module configured for calculating an accuracy of the estimated value.

15. The pulse oximeter of claim 14, wherein the neural network module is further configured for receiving the estimated value and the accuracy.

16. The pulse oximeter of claim 14, wherein the neural network module is further configured for calculating a revised accuracy associated with the revised value of the oxygen saturation of the patient'"'"'s blood.

17. The pulse oximeter of claim 11, wherein the neural network is a backpropagation neural network.