Minimax filtering for pulse oximetry
Methods and systems are provided for filtering a pulse oximetry signal without making assumptions on the noise statistics of the signal. A pulse oximeter may receive an energy signal proportional to an amount of light detected at the sensor. The energy signal may be converted and digitally processed to estimate physiological data. The pulse oximeter may include a processor configured to execute H ∞ filtering algorithms to estimate physiological data without requiring the variances and/or distributions of measurement and process noise in the signal. In one or more embodiments, the pulse oximeter may also be configured to execute other filtering algorithms which update H ∞ filtering algorithms based on the pulse oximetry signal.
1. A method of filtering a pulse oximetry signal, comprising: inputting the pulse oximetry signal into filtering algorithms, wherein the filtering algorithms comprise pre-programmed weighting matrices that depend on one or more of a characteristic of a sensor, a patient measurement site, or a type of physiological data to be collected, and wherein the filtering algorithms comprise H ∞ filter algorithms adapted to filter the pulse oximetry signal.
2. The method of claim 1 , comprising executing the filtering algorithms without a priori noise statistics.
3. The method of claim 2 , comprising:
outputting a filtered pulse oximetry signal; and
estimating physiological data from the filtered pulse oximetry signal, wherein the worst possible estimation errors are minimized in estimating the physiological data.
4. The method of claim 1 , comprising:
determining noise statistics from the pulse oximetry signal; and
updating the filtering algorithms based on the determined noise statistics.
5. The method of claim 1 , wherein the physiological data comprises one or more of blood-oxygen saturation, pulse rate, and volume of blood pulsation in tissue.
6. A pulse oximetry system comprising:
a sensor configured to transmit light into and receive light from tissue, and further configured to output an analog signal based on the received light;
an analog-to-digital converter configured to digitize the analog signal; and
a digital filter configured to filter the digitized signal with H ∞ filtering algorithms, wherein the filtering algorithms comprise pre-programmed weighting matrices that depend on one or more of a sensor characteristic, a patient measurement site, or a type of physiological data to be collected.
7. The pulse oximetry system of claim 6 , comprising a memory component accessible by the digital filter, and wherein the H ∞ filtering algorithms are stored in the memory component.
8. The pulse oximetry system of claim 7 , wherein the memory component comprises weighting matrices, and wherein the digital filter uses the weighting matrices to filter the digitized signal with the H ∞ filtering algorithms.
9. The pulse oximetry system of claim 6 , wherein the digital filter is configured to use adaptive filtering algorithms to detect changes in the digital signal.
10. The pulse oximetry system of claim 9 , wherein the H ∞ filtering algorithms are substantially modifiable based on the detected changes in the digital signal.
11. The pulse oximetry system of claim 6 , wherein the digital filter is configured to filter the digitized signal with the H ∞ filtering algorithms when the digitized signal comprises noise that does not have a stationary zero-mean noise distribution.
12. The pulse oximetry system of claim 6 , wherein the digital filter is configured to filter the digitized signal with the H ∞ filtering algorithms when the digitized signal comprises noise that does not have a known variance.
13. The pulse oximetry system of claim 6 , comprising a processor configured to calculate physiological data based on the filtered signal.
14. The pulse oximetry system of claim 13 , wherein the physiological data comprises at least one of blood-oxygen saturation, pulse rate, or volume of blood pulsation in tissue.
15. A pulse oximetry monitor comprising:
a processor configured to determine whether a pulse oximetry system has a stationary zero-mean noise distribution, to access one or more stored filtering algorithms, to select an H ∞ filtering algorithm from the one or more stored filtering algorithms when the pulse oximetry system is determined not to have a stationary zero-mean noise distribution, and to execute the H ∞ filtering algorithm on a pulse oximetry signal to minimize a maximum error in the pulse oximetry signal.
16. The pulse oximetry monitor of claim 15 , wherein the processor is configured to select and to execute a Kalman filtering algorithm when the pulse oximetry system is determined to have a stationary zero-mean noise distribution.
17. The pulse oximetry monitor of claim 15 , wherein the processor is configured to use an adaptive filtering algorithm to detect changes in the pulse oximetry signal, and wherein the processor is configured to modify the selected H ∞ filtering algorithm based on the detected changes.
18. The pulse oximetry monitor of claim 15 , wherein the processor is configured to estimate physiological data based on a filtered pulse oximetry signal.
19. The pulse oximetry monitor of claim 15 , wherein the processor is configured to select the H ∞ filtering algorithm when the pulse oximetry signal is determined to comprise noise with a non-Gaussian distribution.