METHODS AND SYSTEMS FOR CIRCADIAN PHYSIOLOGY PREDICTIONS

• US 20100138379A1
• Filed: 11/27/2009
• Published: 06/03/2010
• Est. Priority Date: 05/29/2007
• Status: Active Grant

• Current Assignee: McGill University, Pulsar Informatics, Inc.
• Original Assignee: McGill University
• Inventors: BOIVIN, Diane, MOLLICONE, Daniel, MOTT, Christopher, DUMONT, Guy
• Granted Patent: US 8,484,153 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: A61B 5/4857 Indicating the phase of biorhythm clocks or watches with indicators for biological cycles G04B19/264
• 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: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

Systems and methods are provided for predicting a circadian state of an individual. The methods comprise: providing a model representative of the response of the circadian state to light stimulus, the model comprising at least one model variable representative of a probability distribution function (PDF) of a phase offset of the circadian state of the individual; and using the model to estimate an updated PDF of the phase offset, wherein using the model to estimate the updated PDF of the phase offset comprises performing a Bayesian estimation process commencing with an initial PDF of the phase offset and iterating toward the updated PDF of the phase offset.

Claims

1. A system for predicting a circadian state of an individual, the system comprising:

a processor connected to receive light stimulus information related to light stimulus to which the individual is exposed;

wherein the processor is configured to;

provide a model representative of circadian response to the light stimulus information, the model comprising one or more model variables and at least one model variable representative of a probability distribution function (PDF) of a phase offset of the circadian state of the individual; and

use the model and the light stimulus information to estimate an updated PDF of the phase offset, wherein using the model and the light stimulus information to estimate the updated PDF of the phase offset comprises performing a Bayesian estimation process commencing with an initial PDF of the phase offset and iterating toward the updated PDF of the phase offset.

2. A system according to claim 1 comprising a prediction updator configured to estimate one or more PDFs of the one or more model variables at a first time step based on the light stimulus information and one or more known PDFs of the one or more model variables at a second time step, the first time step subsequent to the second time step, and wherein the prediction updator is used by the processor to iterate toward the updated PDF of the phase offset.

3. A system according to claim 2 comprising a measurement updator connected to receive measurement information representative of the phase offset based on one or more physiological parameters of the individual and to receive the one or more PDFs of the one or more model variables at the first time step from the prediction updator, the measurement updator configured to use the measurement information and the one or more PDFs of the one or more model variables at the first time step to estimate one or more measurement-updated PDFs of the one or more model variables at the first time step, wherein the one or more measurement-updated PDFs of the one or more model variables at the first time step take into account the measurement information and wherein the measurement updator is used by the processor to iterate toward the updated PDF of the phase offset.

4. A system according to claim 2 comprising:

one or more physiological sensors connected to measure one or more corresponding physiological parameters from the individual and to provide one or more corresponding physiological input signals; and

a physiological phase estimator connected to receive the one or more physiological input signals and to estimate one or more corresponding physiological feature PDFs of the phase offset based on the one or more physiological input signals, each physiological feature PDF expressing a degree of confidence in a phase offset prediction based on the corresponding physiological input signal.

5. A system according to claim 4 comprising a measurement updator connected to receive the one or more physiological feature PDFs of the phase offset from the physiological phase estimator and to receive the one or more PDFs of the one or more model variables at the first time step from the prediction updator, the measurement updator configured to use the one or more physiological feature PDFs and the one or more PDFs of the one or more model variables at the first time step to estimate one or more measurement-updated PDFs of the one or more model variables at the first time step, wherein the one or more measurement-updated PDFs of the one or more model state variables at the first time step take into account the one or more physiological feature PDFs of the phase offset and wherein the measurement updator is used by the processor to iterate toward the updated PDF of the phase offset.

6. A system according to claim 4 wherein the physiological phase estimator comprises, for each physiological input signal, a marker component for:

extracting a marker from the physiological input signal, the marker comprising at least one of;

a feature of the physiological input signal; and

a time of occurrence of the feature of the physiological input signal; and

determining a marker PDF representing an uncertainty present in extracting the marker from the physiological input signal.

7. A system according to claim 6 wherein the physiological phase estimator comprises, for each physiological input signal, a marker-to-phase converter component for transforming the marker PDF into a domain of the phase offset of the individual to thereby obtain the physiological feature PDF of the phase offset.

8. A system according to claim 1 wherein the processor is configured to perform the Bayesian estimation process using a particle filtering procedure which comprises:

converting the initial PDF into an initial set of discrete particles representative of the initial PDF, each of the initial particles comprising a point;

iteratively propagating the initial set of discrete particles through the Bayesian estimation process to obtain an updated set of discrete particles; and

converting the updated set of discrete particles into the updated PDF of the phase offset.

9. A system according to claim 8 wherein the processor is configured to iteratively propagate the initial set of discrete particles through the Bayesian estimation process by, for each iteration:

obtaining a first set of discrete particles, the first set of discrete particle comprising one of;

the initial set of discrete particles or an output set of discrete particles from a previous iteration;

performing a prediction update operation on the first set of discrete particles, the prediction update operation based on application of a state transition equation of the model to the first set of discrete particles and the prediction update operation outputting a second set of discrete particles.

10. A system according to claim 9 comprising

one or more physiological sensors connected to measure one or more corresponding physiological parameters from the individual and to provide one or more corresponding physiological input signals;

and

a physiological phase estimator connected to receive the one or more physiological input signals and to determine one or more corresponding physiological feature PDFs of the phase offset based on the one or more physiological input signals, each physiological feature PDF expressing a degree of confidence in a phase offset prediction based on the corresponding physiological input signal; and

wherein the processor is configured to iteratively propagate the initial set of discrete particles through the Bayesian estimation process by, for each iteration in which there is a new physiological feature PDF;

performing a measurement update operation on the second set of discrete particles, the measurement update operation based on application of the new physiological feature PDF to the second set of discrete particles to obtain a third set of discrete particles; and

outputting the third set of discrete particles.

11. A system according to claim 10 wherein the physiological phase estimator comprises, for each physiological input signal, a marker component for:

extracting a marker from the physiological input signal, the marker comprising at least one of;

a feature of the physiological input signal; and

a time of occurrence of the feature of the physiological input signal; and

determining a marker PDF representing an uncertainty present in extracting the marker from the physiological input signal.

12. A system according to claim 11 wherein the physiological phase estimator comprises, for each physiological input signal, a marker-to-phase converter component for transforming the marker PDF into a domain of the phase offset of the individual to thereby obtain the physiological feature PDF of the phase offset.

13. A system according to claim 9 comprising

one or more physiological sensors connected to measure one or more corresponding physiological parameters from the individual and to provide one or more corresponding physiological input signals;

and

a physiological phase estimator connected to receive the one or more physiological input signals and to estimate one or more corresponding physiological feature PDFs of the phase offset based on the one or more physiological input signals, each physiological feature PDF expressing a degree of confidence in a phase offset prediction based on the corresponding physiological input signal; and

wherein the processor is configured to iteratively propagate the initial set of discrete particles through the Bayesian estimation process by, for each iteration in which there is a new physiological feature PDF;

performing an importance weight operation on the second set of discrete particles, the importance weight operation based on application of the new physiological feature PDF to the second set of discrete particles to obtain a third set of discrete particles, wherein the third set of discrete particles are each assigned a weight based at least in part on the new physiological feature PDF;

performing a resampling operation on the third set of discrete particles to obtain a fourth set of particles, the resampling operation comprising;

removing individual ones of the third set of discrete particles if the weights of the individual ones of the third set of discrete particles are below a first threshold; and

dividing individual ones of the third set of discrete particles into multiple particles if the individual ones of the third set of discrete particles are above a second threshold;

performing a move operation on the fourth set of particles to obtain a fifth set of discrete particles, the move operation redistributing the fourth set of particles; and

outputting the fifth set of discrete particles.

14. A system according to claim 4 wherein the one or more physiological parameters comprise one or more of:

core body temperature (CBT) of the individual;

cortisol level of the individual;

melatonin level of the individual;

an activity level of the individual;

a rate of cell proliferation of the individual;

a parameter related to a cardiac regulatory system of the individual;

a parameter related to the chemoreceptive respiratory feedback system of the individual; and

a parameter related to the cognitive performance of the individual.

15. A system according to claim 6 wherein, for each physiological input signal, the marker component is configured to extract the marker from the physiological input signal by:

fitting a curve to the physiological input signal; and

identifying the feature in the curve, the feature comprising at least one of;

a minimum of the curve;

a maximum of the curve;

a point where the curve crosses a threshold; and

a region where the curve is located on one side of a threshold.

16. A system according to claim 15 wherein for each physiological input signal, the marker PDF reflects a degree of certainty in one or more of:

the fitted curve; and

the identified feature.

17. A system according to claim 7 wherein the marker PDF reflects a degree of confidence in the time of occurrence of the feature within the physiological input signal and wherein the marker-to-phase converter component is configured to shift the marker PDF in the time domain to obtain the physiological feature PDF.

18. A system according to claim 8 wherein the model provided by the processor is configured to accommodate statistical uncertainty in the light stimulus information.

19. A system according to claim 18 wherein the model provided by the processor is configured to model one or more of:

one or more times corresponding to changes in light stimulus using one or more corresponding light-time PDFs representative of a confidence in the one or more times; and

one or more levels of light stimulus intensity using one or more corresponding light-level PDFs representative of a confidence in the one or more levels.

20. A system according to claim 1 comprising one or more light sensor devices for measuring light intensity and for estimating the light stimulus to which the individual is exposed to thereby provide the light stimulus information.

21. A system according to claim 1 comprising one or more light estimation devices which estimate the light stimulus to which the individual is exposed based on criteria other than light measurement to thereby provide the light stimulus information.

22. A method for predicting a circadian state of an individual, the method comprising:

receiving light stimulus information related to light stimulus to which the individual is exposed;

providing a model representative of circadian response to the light stimulus information, the model comprising one or more model variables and at least one model variable representative of a probability distribution function (PDF) of a phase offset of the circadian state of the individual; and

using the model to transform the light stimulus information into an estimate of an updated PDF of the phase offset, wherein using the model to transform the light stimulus information into an estimate of the updated PDF of the phase offset comprises performing a Bayesian estimation process commencing with an initial PDF of the phase offset and iterating toward the updated PDF of the phase offset.

23. A method according to claim 22 wherein performing the Bayesian estimation processing comprises:

converting the initial PDF into an initial set of discrete particles representative of the initial PDF, each of the initial particles comprising a point;

iteratively propagating the initial set of discrete particles through the Bayesian estimation process to obtain an updated set of discrete particles; and

converting the updated set of discrete particles into the updated PDF of the phase offset.

24. A method according to claim 23 wherein iteratively propagating the initial set of discrete particles through the Bayesian estimation process comprises, for each iteration:

obtaining a first set of discrete particles, the first set of discrete particle comprising one of;

the initial set of discrete particles or an output set of discrete particles from a previous iteration;

performing a prediction update operation on the first set of discrete particles, the prediction update operation based on application of a state transition equation of the model to the first set of discrete particles and the prediction update operation outputting a second set of discrete particles.

25. A method according to claim 24 comprising:

measuring one or more physiological parameters of the individual to obtain one or more physiological input signals;

processing the one or more physiological input signals to obtain one or more corresponding physiological feature PDFs of the phase offset, each physiological feature PDF expressing a degree of confidence in a phase offset prediction based on the corresponding physiological input signal; and

wherein iteratively propagating the initial set of discrete particles through the Bayesian estimation process comprises, for each iteration in which there is a new physiological feature PDF;

performing a measurement update operation on the second set of discrete particles, the measurement update operation based on application of the new physiological feature PDF to the second set of discrete particles to obtain a third set of discrete particles; and

outputting the third set of discrete particles.

26. A method according to claim 24 comprising:

measuring one or more physiological parameters of the individual to obtain one or more physiological input signals;

processing the one or more physiological input signals to obtain one or more corresponding physiological feature PDFs of the phase offset, each physiological feature PDF expressing a degree of confidence in a phase offset prediction based on the corresponding physiological input signal; and

wherein iteratively propagating the initial set of discrete particles through the Bayesian estimation process comprises, for each iteration in which there is a new physiological feature PDF;

performing an importance weight operation on the second set of discrete particles, the importance weight operation based on application of the physiological feature PDF to the second set of discrete particles to obtain a third set of discrete particles, wherein the third set of discrete particles are each assigned a weight based at least in part on the new physiological feature PDF;

performing a resampling operation on the third set of discrete particles to obtain a fourth set of particles, the resampling operation comprising;

removing individual ones of the third set of discrete particles if the weights of the individual ones of the third set of discrete particles are below a first threshold; and

dividing individual ones of the third set of discrete particles into multiple particles if the individual ones of the third set of discrete particles are above a second threshold;

performing a move operation on the fourth set of particles to obtain a fifth set of discrete particles, the move operation redistributing the fourth set of particles; and

outputting the fifth set of discrete particles.

27. A method according to claim 26 wherein performing the resampling operation comprises ensuring that the fourth set of particles incorporates at least approximately the same statistical information as the third set of discrete particles and wherein performing the move operation comprises ensuring that the fifth set of discrete particles incorporates at least approximately the same statistical information as the fourth set of particles.

28. A method according to claim 25 wherein processing the one or more physiological input signals to obtain one or more corresponding physiological feature PDFs of the phase offset comprises, for each physiological input signal:

extracting a marker from the physiological input signal, the marker comprising at least one of;

a feature of the physiological input signal; and

a time of occurrence of the feature of the physiological input signal; and

determining a marker PDF representing an uncertainty present in extracting the marker from the physiological input signal.

29. A method according to claim 28 wherein processing the one or more physiological input signals to obtain one or more corresponding physiological feature PDFs of the phase offset comprises, for each physiological input signal, transforming the marker PDF into a domain of the phase offset of the individual to thereby obtain the physiological feature PDF of the phase offset.

30. A method according to claim 28 wherein, for each physiological input signal, extracting the marker from the physiological input signal comprises:

fitting a curve to the physiological input signal; and

identifying the feature in the curve, the feature comprising at least one of;

a minimum of the curve;

a maximum of the curve;

a point where the curve crosses a threshold; and

a region where the curve is located on one side of a threshold.

31. A method according to claim 30 wherein for each physiological input signal, the marker PDF reflects a degree of certainty in one or more of:

the fitted curve; and

the identified feature.

32. A method according to claim 29 wherein the marker PDF reflects a degree of confidence in the time of occurrence of the feature within the physiological input signal and wherein transforming the marker PDF into a domain of the phase offset to thereby obtain the physiological feature PDF comprises shifting the marker PDF in the time domain to obtain the physiological feature PDF.

33. A method according to claim 25 wherein the one or more physiological parameters comprise one or more of:

core body temperature (CBT) of the individual;

cortisol level of the individual;

melatonin level of the individual;

an activity level of the individual;

a rate of cell proliferation of the individual;

a parameter related to a cardiac regulatory system of the individual;

a parameter related to the chemoreceptive respiratory feedback system of the individual; and

a parameter related to the cognitive performance of the individual.

34. A method according to claim 23 wherein providing the model representative of circadian response to the light stimulus information comprises providing a model which accommodates statistical uncertainty in the light stimulus information.

35. A method according to claim 34 wherein providing a model which accommodates statistical uncertainty in the light stimulus information comprises one or more of:

modeling one or more times corresponding to changes in light stimulus using one or more corresponding light-time PDFs representative of a confidence in the one or more times; and

modeling one or more levels of light stimulus intensity using one or more corresponding light-level PDFs representative of a confidence in the one or more levels.

36. A method according to claim 22 comprising measuring light intensity to estimate the light stimulus to which the individual is exposed and to thereby provide the light stimulus information.

37. A method according to claim 22 comprising estimating the light stimulus to which the individual is exposed based on criteria other than light measurement to thereby provide the light stimulus information.

38. A method according to claim 22 wherein performing the Bayesian estimation process comprises incorporating process noise into the model, the process noise additive, such that it may be added to a state transition equation of the model.

39. A method according to claim 22 wherein performing the Bayesian estimation process comprises incorporating measurement noise into the model, the measurement noise additive, such that it may be added to a measurement equation of the model.

40. A computer program product comprising computer-readable instructions, which when executed by a suitably configured processor, cause the processor to perform a method for predicting a circadian state of an individual, the method comprising:

receiving light stimulus information related to light stimulus to which the individual is exposed;

providing a model representative of circadian response to the light stimulus information, the model comprising one or more model variables and at least one model variable representative of a probability distribution function (PDF) of a phase offset of the circadian state of the individual; and

using the model to estimate an updated PDF of the phase offset, wherein using the model to estimate the updated PDF of the phase offset comprises performing a Bayesian estimation process commencing with an initial PDF of the phase offset and iterating toward the updated PDF of the phase offset.