U.S. patent number 8,615,291 [Application Number 12/724,162] was granted by the patent office on 2013-12-24 for method, system and computer program method for detection of pathological fluctuations of physiological signals to diagnose human illness.
This patent grant is currently assigned to N/A, National Institutes of Health (NIH), The United States of America as Represented by the Department of Health and Human Services. The grantee listed for this patent is John B. Delos, Abigail Flower, Douglas E. Lake, Randall Moorman. Invention is credited to John B. Delos, Abigail Flower, Douglas E. Lake, Randall Moorman.
United States Patent |
8,615,291 |
Moorman , et al. |
December 24, 2013 |
**Please see images for:
( Certificate of Correction ) ** |
Method, system and computer program method for detection of
pathological fluctuations of physiological signals to diagnose
human illness
Abstract
Method, system, and computer program method for detecting
pathological fluctuations of physiological signals to diagnose
human illness. The method comprises performing a sliding window
analysis to find sequences in physiological signal data that match
amplitude- and duration-adjusted versions of a template function to
within a specified tolerance.
Inventors: |
Moorman; Randall
(Charlottesville, VA), Lake; Douglas E. (Charlottesville,
VA), Flower; Abigail (Stevenson, MD), Delos; John B.
(Williamsburg, VA) |
Applicant: |
Name |
City |
State |
Country |
Type |
Moorman; Randall
Lake; Douglas E.
Flower; Abigail
Delos; John B. |
Charlottesville
Charlottesville
Stevenson
Williamsburg |
VA
VA
MD
VA |
US
US
US
US |
|
|
Assignee: |
National Institutes of Health
(NIH) (Washington, DC)
The United States of America as Represented by the Department of
Health and Human Services (Washington, DC)
N/A (N/A)
|
Family
ID: |
42731279 |
Appl.
No.: |
12/724,162 |
Filed: |
March 15, 2010 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20100234748 A1 |
Sep 16, 2010 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
61160024 |
Mar 13, 2009 |
|
|
|
|
Current U.S.
Class: |
600/509 |
Current CPC
Class: |
A61B
5/02405 (20130101); A61B 5/412 (20130101); G16H
10/20 (20180101); A61B 5/7275 (20130101); G06N
7/02 (20130101); A61B 5/35 (20210101); G16H
50/50 (20180101) |
Current International
Class: |
A61B
5/04 (20060101) |
Field of
Search: |
;600/515-518,521,509 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
Ary L. Goldberger, Luis A. N. Amaral, Jeffrey M. Hausdorff, Plamen
Ch. Ivanov, C.-K. Peng, and H. Eugene Stanley, Fractal dynamics in
physiology: Alterations with disease and aging, PNAS 2002 99 (Suppl
1) 2466-2472; doi:10.1073/pnas.012579499. cited by applicant .
Buchman, T. G. (2004) Curr. Opin. Crit Care. 10, 378-382. cited by
applicant .
Griffin P., Moorman R., Toward the Pediatrics; Early Diagnosis of
Neonatal Sepsis and Sepsis-Like Illness Using Novel Heart Rate
Analysis; Official Journal of the American Academy of Pediatrics
107; 2001 pp. 97-104; Elk Grove Village IL. cited by applicant
.
Griffin P., et al., Pediatrics; Abnormal Heart Rate Characteristics
Preceding Neonatal Sepsis and Sepsis-Like Illness; International
Pediatric Research Foundation; Wake Forest School of Medicine,;
2003 Winston Salem, NC; pp. 920-926. cited by applicant .
M. Pamela Griffin, et al.; Pediatrics; Abnormal Heart Rate
Characteristics Are Associated with Neonatal Mortality; Departments
of Pediatrics [M.P.G.], internal Medicine [D.E.L., J.R.M.], Health
Evaluation Sciences [E.A.B., F.E.H.], and Internal Medicine and
Molecular Physiology and Biological Physics [J.R.M.], and the
Cardiovascular Research Center, University of Virginia Health
System, Charlottesville, Virginia; Winston-Salem, North Carolina;
2004 pp. 782-788, U.S.A. cited by applicant .
Griffin P., et al; Pediatrics; Heart Rate Characteristics and
Laboratory Tests in Neonatal Sepsis; International Pediatric
Research Foundation; Wake Forest School of Medicine; 2004 Winston
Salem, NC. cited by applicant .
Griffin P., et al; Heart Rate Characteristics: Novel Physiomarkers
to Predict Neonatal infection and death; International Pediatric
Research Foundation; Wake Forest School of Medicine,; pp.
1070-1074, 2004 Winston Salem, NC. cited by applicant .
Griffin P., et al; Heart Rate Characteristics and Clinical Signs in
Neonatal Sepsis; International Pediatric Research Foundation; Wake
Forest School of Medicine; pp. 222-227; 2007; Winston Salem, NC.
cited by applicant .
Boris P. Kovatchev, et al; Sample Asymmetry Analysis of Heart Rate
Characteristics with Application to Neonatal Sepsis and Systemic
Inflammatory Response Syndrome; 2003, vol. 54 pp. 892-898;
Charlottesville, VA. cited by applicant .
Lake D. et al; The American Journal of Physiology--Regulatory,
Integrative and Comparative Physiology; pp. 789-797; 2002; Bethesda
MD. cited by applicant .
Lake D.; Renyi Entropy Measures of Heart Rate Gaussianity;
Transactions on Biomedical Engineering, Vol. 53, No. 1, Jan. 2006;
pp. 21-27; U.S. cited by applicant .
Moorman R. et al.; Heart Rate Characteristics Monitoring for
Neonatal Sepsis; Transactions on Biomedical Engineering, vol. 53,
No. 1, Jan. 2006; pp. 126-132; U.S. cited by applicant .
Richman J. et al; Physiological time-series analysis using
approximate entropy and sample entropy; pp. 2039-2049; 2000; U.S.
cited by applicant .
Richman J. et al.; Sample Entropy; Methods in Enzymology, vol. 384;
pp. 172-184; 2004; U.S. cited by applicant .
Flower A. et al.; Abstract 1769: Storms of Heartrate Decelerations
in Asymptomatic infants Prior to Neonatal Sepsis; Univ. of VA;
Charlottesville, VA; 2006 pp. 1-2. cited by applicant .
Griffin P. et al.; Heart rate characteristics monitoring to detect
neonatal sepsis Univ. of VA; Charlottesville, VA; 2006 pp. 1-39.
cited by applicant.
|
Primary Examiner: Patton; Amanda
Attorney, Agent or Firm: McGuire; Brian M. Frommer Lawrence
& Haug LLP
Parent Case Text
CROSS REFERENCE TO RELATED APPLICATIONS
This application claims priority to U.S. Provisional Patent
Application Ser. No. 61/160,024, filed on Mar. 13, 2009, the
entirety of which is incorporated by reference herein.
Claims
What is claimed is:
1. A method comprising: accepting time series data of a cardiac
signal including a heart beat or an RR interval; obtaining a
pattern for the time series data; choosing a template function that
corresponds to the pattern of the time-series data; performing an
analysis of the time series data to match sequences in the time
series data to the template function, wherein one or more of the
sequences comprises fluctuations; calculating one or more
characteristics of the fluctuations based on the analysis;
identifying a risk of clinical condition associated with one or
more of the characteristics.
2. The method of claim 1, wherein the fluctuations include
accelerations or decelerations.
3. The method of claim 1, wherein the template function is a
wavelet function.
4. The method of claim 3, wherein the wavelet function is selected
from a group consisting of exponential, Gaussian, Lorentzian, and
.chi.(n), wherein .chi.(n) is
.chi..function..times..times..function..function..times..function.
##EQU00008## wherein the amplitude a is calculated as described by
Equation .function..times..times..chi..chi. ##EQU00009## and b is
the width parameter, the numerator RR.chi..sub.b being
.times..times..chi..times..function..times..chi..function..function..time-
s..chi..function..chi..times..times. ##EQU00010##
.chi..chi..function..ltoreq..ltoreq..times..times. ##EQU00011##
wherein the derivation of .chi.(n) is based on decelerations in RR
interval time series considered as the sum of .chi.(n) and some
remainder as illustrated in Equation
RR(n)=a.chi..sub.b(n-n.sub.0)+G(n), RR(n) being the time between
beat n and beat n+1, and .chi..sub.b(n-n.sub.0) is a function
describing a deceleration of width b centered about the point
n.sub.0, and G(n) represents the remainder of the signal.
5. The method of claim 4, wherein the template function is
.chi.(n).
6. The method of claim 1, wherein the analysis comprises a
sliding-window analysis, the analysis including sweeping the
template function through the time series data in sequence
increments and calculating a correlation coefficient between the
template function and any sequence increment, wherein the sequence
increment may comprise one or more fluctuations.
7. The method of claim 6, wherein different analytic forms of the
template function are swept through the time series data.
8. The method of claim 7, wherein the different analytic forms of
the template function that are swept through the time series data
vary in amplitude, duration, or a combination thereof.
9. The method of claim 8, wherein the different analytic form of
the template function that is swept through the time series data
increase in duration from about 8 to about 100 beats.
10. The method of claim 7, further comprising determining the
different analytic forms of the template function that highly
correlate with the sequence increments.
11. The method of claim 10, wherein the correlation between the
different analytic forms of the template function and the sequence
increments is at a minimum threshold.
12. The method of claim 10, further comprising removing overlapping
fluctuations.
13. The method of claim 12, wherein removal of overlapping
fluctuations comprises identifying and keeping the overlapping
fluctuation that has the highest correlation coefficient.
14. The method of claim 1, further comprising removing the
fluctuations from the cardiac signal and determining baseline
signal variability.
15. The method of claim 14, wherein baseline signal variability is
measured through sample asymmetry, smoothness, residual
variability, or a combination thereof.
16. The method of claim 1, wherein the characteristics of the
fluctuations are selected from the group consisting of number of
fluctuations; amplitude of the fluctuations; widths of the
fluctuations; R.sup.2 between the template function and
fluctuation; and baseline signal variability.
17. The method of claim 1, wherein the clinical condition is an
illness.
18. The method of claim 17, wherein the illness is sepsis or a
condition associated with sepsis.
19. The method of claim 1, wherein the method further comprises:
identifying information about the illness based on the
analysis.
20. The method of claim 18 wherein the illness is sepsis, and the
information includes identifying the risk that the sepsis is that
of a gram-positive organism or gram-negative organism.
21. The method of claim 1, wherein method comprises: choosing from
a plurality of templates that corresponds to the pattern of the
time-series data.
22. The method of claim 21, wherein the method further includes:
applying each of the plurality of templates individually and,
applying each of the plurality of templates for every possible
combination of the plurality of templates that fit to the
time-series data; and choosing the individual template, or the
combination of templates, that maximizes a predetermined
variable.
23. A method comprising: accepting time series data, said data of a
cardiac signal including a heart beat or an RR interval; analyzing
the time series data using a pattern matching algorithm to identify
pathological fluctuations in the signal; and identifying a risk of
clinical condition based on the analysis.
24. A system for identifying or monitoring a risk of a clinical
condition, the system comprising at least one computer, a
processor, at least one storage device, and at least one computer
readable medium storing thereon a program, wherein the system
comprises: an input for accepting time series data, said data of a
cardiac signal including a heart beat or an RR interval; and the
program is configured to, when executed by the processor, cause the
system, to at least: accept time series data of a cardiac signal;
obtain a pattern for the time series data; choose a template
function that corresponds to the pattern of the time-series data;
perform an analysis of the time series data to match sequences in
the time series data to the template function, wherein one or more
of the sequences comprises fluctuations; calculate one or more
characteristics of the fluctuations based on the analysis; and
identify a risk of the clinical condition associated with one of
more of the characteristics.
25. The system of claim 24, wherein the fluctuations include
accelerations or decelerations.
26. The system of claim 24, wherein the template function is a
wavelet function.
27. The system of claim 26, wherein the wavelet function is
selected from a group consisting of exponential, Gaussian,
Lorentzian, and .chi.(n), wherein .chi.(n) is
.chi..function..times..times..function..function..times..function.
##EQU00012## Wherein the amplitude a is calculated as described by
Equation .function..times..times..chi..chi. ##EQU00013## and b is
the width parameter, the numerator RR.chi..sub.b being
.times..times..chi..times..function..times..chi..function..function..time-
s..chi..function..chi..times..times. ##EQU00014##
.chi..chi..function..ltoreq..ltoreq..times..times. ##EQU00015##
wherein the derivation of .chi.(n) is based on decelerations in RR
interval time series considered as the sum of .chi.(n) and some
remainder as illustrated Equation
RR(n)=a.chi..sub.b(n-n.sub.0)+G(n), RR(n) being the time between
beat n and beat n+1, and .chi..sub.b(n-n.sub.0) is a function
describing a deceleration of width b centered about the point
n.sub.0, and G(n) represents the remainder of the signal.
28. The system of claim 27, wherein the template function is
.chi.(n).
29. The system of claim 24, wherein the analysis comprises a
sliding-window analysis, the analysis including sweeping the
template function through the time series data in sequence
increments and calculating a correlation coefficient between the
template function and any sequence increment, wherein the sequence
increment may comprise one or more fluctuations.
30. The system of claim 29, wherein different analytic forms of the
template function are swept through the time series data.
31. The system of claim 30, wherein the different analytic forms of
the template function that are swept through the time series data
vary in amplitude, duration, or a combination thereof.
32. The system of claim 31, wherein the different analytic forms of
the template function that is swept through the time series data
increase in duration from about 8 to about 100 beats.
33. The system of claim 30, wherein the system is configured to at
least: determine the different analytic forms of the template
function that highly correlate with the sequence increments.
34. The system of claim 33, wherein the correlation between the
different analytic forms of the template function and the sequence
increments is at a minimum threshold.
35. The system of claim 33, wherein the computer program is
configured to cause the system to at least: remove overlapping
fluctuations.
36. The system of claim 35, wherein removal of overlapping
fluctuations comprises identifying and keeping the overlapping
fluctuation that has the highest correlation coefficient.
37. The system of claim 24, wherein the system is configured to at
least: remove the fluctuations from the cardiac signal and
determining baseline signal variability.
38. The system of claim 37, wherein baseline signal variability is
measured through sample asymmetry, smoothness, residual
variability, or a combination thereof.
39. The system of claim 24, wherein the characteristics of the
fluctuations are selected from the group consisting of number of
fluctuations; amplitude of the fluctuations; widths of the
fluctuations; R.sup.2 between the template function and
fluctuation; and baseline signal variability.
40. The system of claim 24, wherein the clinical condition is an
illness.
41. The system of claim 40, wherein the illness is sepsis or
conditions associated with sepsis.
42. The system of claim 24, wherein the system is configured to at
least: identify information about the illness based on the
analysis.
43. The system of claim 42 wherein the illness is sepsis, and the
information includes identifying the risk that the sepsis is that
of a gram-positive organism or gram-negative organism.
44. The system of claim 24, wherein the system is configured to
cause the system to identify a risk of the illness occurring in the
subject.
45. The system of claim 24, wherein the system is configured to
cause the system to: choose from a plurality of templates that
corresponds to the pattern of the time-series data.
46. The system of claim 45, wherein the system is configured to
cause the system to: apply each of the plurality of templates
individually and, apply each of the plurality of templates for
every possible combination of the plurality of templates that fit
to the time-series data; and choosing the individual template, or
the combination of templates, that maximizes a predetermined
variable.
47. The system of claim 24, wherein the system is operatively
connected to a cardiac signal monitoring system.
48. A system for monitoring a risk of a clinical condition, the
system comprising at least one computer, a processor, at least one
storage device, at least one computer readable medium storing
thereon a program, wherein the system comprises: an input for
accepting time series data, said data of a cardiac signal including
a heart beat or an RR interval; and the program is configured to,
when executed by the processor, cause the system to at least:
analyze the time series data using a pattern matching algorithm to
identify pathological fluctuations in the signal; and identify a
risk of clinical condition based on the analysis.
49. A computer program product comprising a computer usable medium,
comprising at least one computer, a processor, at least one storage
device, and the computer readable medium storing thereon the
program, wherein the program is configured to, when executed by the
processor, cause a system to at least: analyze time series data of
a cardiac signal including a heart beat or an RR interval using a
pattern matching algorithm to identify pathological fluctuations in
the signal; and identify a risk of clinical condition based on the
analysis.
Description
FIELD OF THE INVENTION
The present disclosure relates to methods, systems, and computer
programs for predicting pathologies based on fluctuations of
physiological, periodic signals.
DISCUSSION OF RELATED ART
A longstanding and widely accepted concept of human physiology is
that of complexity (1, 2). The healthy interplay amongst systems
leads to complex appearances of physiological signals.
Understandably, this complexity has been difficult to describe and
to capture accurately using physiological models. For example,
previous methods of analyzing physiological signals rely on
indirect statistical measures to make predictive models for
clinical use (3-8), but fail to capture important features of
individual fluctuations.
SUMMARY OF THE INVENTION
Disclosed are methods for monitoring a clinical condition and
identifying risks of a clinical condition. In one aspect, the
disclosure refers to detecting pathological fluctuations in a
physiological signal. The method comprises accepting time series
data of a physiological signal; analyzing the time series data
using a pattern matching algorithm to identify pathological
fluctuations in the signal; and identifying a risk of clinical
condition based on the analysis. The method can include choosing a
template function that corresponds to the pattern of the time
series data. The analysis of the time series data can include
matching sequences in the time series data to the template
function, in which one or more of the sequences comprise the
fluctuations. The identification of a risk of clinical condition
can include calculating one or more characteristics of the
fluctuations based on the analysis, and identifying a risk of
clinical condition associated with one or more of the
characteristics.
The physiological signal can include a signal that is cyclical in
whole or in part. The physiological signal can include a cardiac
signal, and the cardiac signal can include a heart beat or an RR
interval over time.
The fluctuations can include accelerations or decelerations.
According to some embodiments of the method, the template function
is a wavelet function. The wavelet function can be selected from a
group consisting of exponential, Gaussian, Lorentzian, and
.chi.(n).
According to some embodiments of the method, the analysis comprises
a sliding-window analysis, such that the analysis includes sweeping
the template function through the time series data in sequence
increments and calculating a correlation coefficient between the
template function and any sequence increment. The sequence
increment may comprise one or more fluctuations. Different versions
of the template function can be swept through the time series data.
Also, the different versions of the template function that can be
swept through the time series data can vary in amplitude, duration,
or a combination thereof. Further, the different versions of the
template function that can be swept through the time series data
can increase in duration from about 8 to about 100 beats.
According to some embodiments, the method further comprises
determining the different versions of the template function that
highly correlate with the sequence increments. The correlation
between the different versions of the template function and the
sequence increments can be at a minimum threshold. The method can
further comprise removing overlapping fluctuations, which can
comprise identifying and keeping the overlapping fluctuation that
has the highest correlation coefficient.
According to some embodiments, the method further comprises
removing fluctuations from the physiological signal and determining
baseline signal variability. Baseline signal variability can be
measured through sample asymmetry, smoothness, residual
variability, or a combination thereof.
The characteristics of the fluctuations can be selected from the
group consisting of number of fluctuations; amplitude of the
fluctuations; widths of the fluctuations; R2 between the template
function and fluctuation; and baseline signal variability.
The clinical condition can be an illness. The illness can be sepsis
or a condition associated with sepsis.
According to some embodiments, the method further comprises
identifying information about the illness based on the analysis.
The illness can be sepsis, and the information can include
identifying the risk that the sepsis is that of a gram-positive
organism or a gram-negative organism.
According to some embodiments, the method comprises choosing from a
plurality of templates that corresponds to the pattern of the
time-series data. The method can further include applying each of
the plurality of templates individually and, applying each of the
plurality of templates for every possible combination of the
plurality of templates that fit to the time-series data; and
choosing the individual template, or the combination of templates,
that maximizes a predetermined variable.
Also disclosed is a system for monitoring a clinical condition and
identifying risks of a clinical condition, in which the system
comprises at least one computer, a processor, at least one storage
device, and at least one computer readable medium that stores a
program. The system can comprise an input for accepting time series
data, in which the data is of a physiological signal. The program
can be configured to, when executed by the processor, cause the
system to at least: analyze the time series data using a pattern
matching algorithm to identify pathological fluctuations in the
signals; and identify a risk of clinical condition based on the
analysis. The computer program can be configured to cause the
system to accept time series data of the physiological signal. The
computer program can also be configured to cause the system to
obtain a pattern for the time series data and choose a template
function that corresponds to the pattern of the time-series data.
The analysis of the time series data may comprise matching
sequences in the time series data to the template function, in
which one or more of the sequences comprises the fluctuations. The
identification of a risk of clinical condition may comprise
calculating one or more characteristics of the fluctuations based
on the analysis, and identifying a risk of the clinical condition
associated with one of more of the characteristics.
The physiological signal can include a signal that is cyclical in
whole or in part. The physiological signal can include a cardiac
signal, and the cardiac signal can include heart beats or RR
intervals over time.
The fluctuations can include accelerations or decelerations.
According to some embodiments of the system, the template function
is a wavelet function. The wavelet function can be selected from a
group consisting of exponential, Gaussian, Lorentzian, and
.chi.(n).
According to some embodiments of the system, the analysis comprises
a sliding-window analysis, such that the analysis includes sweeping
the template function through the time series data in sequence
increments and calculating a correlation coefficient between the
template function and any sequence increment. The sequence
increment may comprise one or more fluctuations. Different versions
of the template function can be swept through the time series data.
Also, the different versions of the template function that can be
swept through the time series data can vary in amplitude, duration,
or a combination thereof. Further, the different versions of the
template function that can be swept through the time series data
can increase in duration from about 8 to about 100 beats.
According to some embodiments, the system is configured to
determine the different versions of the template function that
highly correlate with the sequence increments. The correlation
between the different versions of the template function and the
sequence increments can be at a minimum threshold. The system can
be further configured to remove overlapping fluctuations, which can
comprise identifying and keeping the overlapping fluctuation that
has the highest correlation coefficient.
According to some embodiments, the system is configured to remove
fluctuations from the physiological signal and determining baseline
signal variability. Baseline signal variability can be measured
through sample asymmetry, smoothness, residual variability, or a
combination thereof.
The characteristics of the fluctuations can be selected from the
group consisting of number of fluctuations; amplitude of the
fluctuations; widths of the fluctuations; R2 between the template
function and fluctuation; and baseline signal variability.
The clinical condition can be an illness. The illness can be sepsis
or a condition associated with sepsis.
According to some embodiments, the system is configured to identify
information about the illness based on the analysis. The illness
can be sepsis, and the information can include identifying the risk
that the sepsis is that of a gram-positive organism or a
gram-negative organism.
According to some embodiments, the system is configured to identify
a risk of the illness occurring in the subject.
Furthermore, according to some embodiments, the system is
configured to choose from a plurality of templates that corresponds
to the pattern of the time-series data. The system can be
configured to include applying each of the plurality of templates
individually and, applying each of the plurality of templates for
every possible combination of the plurality of templates that fit
to the time-series data; and choosing the individual template, or
the combination of templates, that maximizes a predetermined
variable.
According to some embodiments, the system is operatively connected
to a physiological signal monitoring system.
In addition, disclosed is a system for monitoring a clinical
condition and identifying risks of a clinical condition, in which
the system comprises at least one computer, a processor, at least
one storage device, and at least one computer readable medium that
stores a program. The system can comprise a means for accepting
time series data, in which the data is of a physiological signal; a
means for analyzing the time series data using a pattern matching
algorithm to identify pathological fluctuations in the signals; and
a means for identifying a risk of clinical condition based on the
analysis. The system can further comprise a means for accepting
time series data of the physiological signal, as well as a means
for obtaining a pattern for the time series data and a means for
choosing a template function that corresponds to the pattern of the
time-series data. The means for analyzing the time series data may
comprise a means for matching sequences in the time series data to
the template function, in which one or more of the sequences
comprises the fluctuations. The means for identifying a risk of
clinical condition may comprise a means for calculating one or more
characteristics of the fluctuations based on the analysis, and a
means for identifying a risk of the clinical condition associated
with one of more of the characteristics.
The physiological signal can include a signal that is cyclical in
whole or in part. The physiological signal can include a cardiac
signal, and the cardiac signal can include a heart beat or an RR
interval.
The fluctuations can include accelerations or decelerations.
According to some embodiments of the system, the template function
is a wavelet function. The wavelet function can be selected from a
group consisting of exponential, Gaussian, Lorentzian, and
.chi.(n).
According to some embodiments of the system, the means for
analyzing the time series data comprises a means for sweeping the
template function through the time series data in sequence
increments and a means for calculating a correlation coefficient
between the template function and any sequence increment. The
sequence increment may comprise one or more fluctuations. Different
versions of the template function can be swept through the time
series data. Also, the different versions of the template function
that can be swept through the time series data can vary in
amplitude, duration, or a combination thereof. Further, the
different versions of the template function that can be swept
through the time series data can increase in duration from about 8
to about 100 beats.
According to some embodiments, the system comprises a means for
determining the different versions of the template function that
highly correlate with the sequence increments. The correlation
between the different versions of the template function and the
sequence increments can be at a minimum threshold. The system can
further comprise a means for removing overlapping fluctuations,
which can comprise a means for identifying and keeping the
overlapping fluctuation that has the highest correlation
coefficient.
According to some embodiments, the system comprises a means for
removing the fluctuations from the physiological signal and
determining baseline signal variability. Baseline signal
variability can be measured through sample asymmetry, smoothness,
residual variability, or a combination thereof.
The characteristics of the fluctuations can be selected from the
group consisting of number of fluctuations; amplitude of the
fluctuations; widths of the fluctuations; R2 between the template
function and fluctuation; and baseline signal variability.
The clinical condition can be an illness. The illness can be sepsis
or a condition associated with sepsis.
According to some embodiments, the system comprises a means for
identifying information about the illness based on the analysis.
The illness can be sepsis, and the information can include a means
for identifying the risk that the sepsis is that of a gram-positive
organism or gram-negative organism.
According to some embodiments, the system is configured to identify
a risk of the illness occurring in the subject.
Furthermore, according to some embodiments, the system comprises a
means for choosing from a plurality of templates that corresponds
to the pattern of the time-series data. The system may comprise a
means for applying each of the plurality of templates individually
and, a means for applying each of the plurality of templates for
every possible combination of the plurality of templates that fit
to the time-series data; and a means for choosing the individual
template, or the combination of templates, that maximizes a
predetermined variable.
According to some embodiments, the system is operatively connected
to a physiological signal monitoring system.
Disclosed is a computer program product comprising a computer
usable medium comprising at least one computer, a processor, at
least one storage device, and the computer readable medium storing
thereon the program. The program can be configured to, when
executed by the processor, cause a system to at least: analyze time
series data of a physiological signal using a pattern matching
algorithm to identify pathological fluctuations in the signal; and
identify a risk of clinical condition based on the analysis.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1(a)-(d) shows examples of approximately four minutes of
continuous RR intervals from four different neonatal intensive care
unit (NICU) patients. FIG. 1(a) is an RR interval series for a
healthy NICU patient, while FIG. 1(b) is an RR interval series for
a NICU patient showing reduced heart rate variability prior to
diagnosis of sepsis. FIG. 1(c) is an RR interval series for a NICU
patient showing decelerations, and FIG. 1(d) is an RR interval
series showing part of a long cluster of periodic
decelerations;
FIG. 2 is a schematic depicting how the RR interval time series is
considered to be the sum of decelerations of various widths and
heights and some remainder or residual signal;
FIG. 3(a)-(b) shows how decelerations have a common shape that can
be represented by a template function for use in a deceleration
detector. FIG. 3(a) shows three decelerations from a record that
was scaled to width of one and height of one. FIG. 3(b) shows
different standard functions that were optimized to fit a
deceleration. Blue asterisks=data; red line=exponential; green
line=Gaussian; cyan line=Lorentzian; black line=.chi.;
FIG. 4 shows correlation coefficient a(n.sub.0,n) at each scale (b)
and translation (n) resulting from a sweep .chi..sub.b(n) through
the time series RR(n) at widths ranging from 8 to 100 beats;
FIG. 5(a)-(b) shows the results of correcting over-counting
decelerations. FIG. 5(a) shows data (blue line) that has four
decelerations but calculated to have 6 templates (red line),
wherein the first two decelerations each have 2 templates. FIG.
5(b) shows the data (blue line) and a corrected calculation of 4
templates (red line);
FIG. 6(a)-(b) shows instances wherein correction from over-counting
decelerations is not necessary. FIG. 6(a) shows that the templates
(green line) are representative of the decelerations (blue line)
and that there is one template (green line) for each deceleration.
FIG. 6(b) shows that each template (green line) is representative
of each deceleration (blue line);
FIG. 7(a)-(b) shows RR interval data that has 12 tall decelerations
(FIG. 7(a)) and 13 tall decelerations (FIG. 7(b)). The RR interval
data of FIG. 7(a) has an R.sub.1 measure of 14.4193 and a
smoothness of 5.2733, while the RR interval data of 7(b) has an
R.sub.1 measure of 6.7810 and a smoothness of 1.97772;
FIG. 8(a)-(d) shows that a noisy Hopf bifurcation model produces
behavior similar to that of observed study. FIG. 8(a) shows
oscillations induced in a noisy hard Hopf bifurcation model
transformed into RR intervals via a Fourier series representation
of a deceleration, while FIG. 8(b) shows bursts of periodic
decelerations created by allowing the parameter .mu. to vary near
zero. FIG. 8(c) shows data from neonatal RR interval record showing
bursts of periodic decelerations (top) and simulation of data
produced by noisy Hopf mode (bottom). FIG. 8(d) shows close-up of
real data (top) and corresponding simulation created by forcing the
parameter .mu. to cross respective critical points at times when
bursts are observed to begin and terminate (bottom);
FIG. 9 is a functional block diagram for a computer system for
implementation of the invention.
FIG. 10(a)-(d) shows decelerations that may occur in clusters but
demonstrate periodicity. FIG. 10(a) shows a cluster of tall
decelerations in a half-hour record of RR interval data (top) and
the number of tall decelerations in a half-hour record as a
function of days since birth (bottom). FIG. 10(b) shows a burst of
decelerations arising from a state of low variability. FIG. 10(c)
shows periodic bursts of decelerations for six NICU patients. FIG.
10(d) shows a fold-increase risk of sepsis as a function of number
of tall decelerations occurring within a half-hour;
FIG. 11(a)-(d) shows that gram-positive and gram-negative organisms
have different effects of transient decelerations in neonatal
sepsis. The organisms studied were coagulase-negative gram positive
(CONS) (FIG. 12(a)), other gram-positive bacteria (FIG. 12(b)),
gram-negative bacteria (FIG. 12(c)), and fungus (FIG. 12(d));
FIG. 12(a)-(d) shows that gram-positive and gram-negative organisms
have different effects of reduced variability in neonatal sepsis.
The organisms studied were coagulase-negative gram positive (CONS)
(FIG. 13(a)), other gram-positive bacteria (FIG. 13(b)),
gram-negative bacteria (FIG. 13(c)), and fungus (FIG. 13(d)).
DETAILED DESCRIPTION OF THE INVENTION
During illness, physiological dynamics can show reduced complexity
that is more readily described and modeled. For example, for heart
rates, heart rate decelerations and otherwise low heart rate
variability that often precede acute neonatal illness appear to
arise from a simpler dynamical system. These large pathological
decelerations are often subclinical and unnoticed by clinical
personnel, are remarkably similar in appearance among infants, and
can appear in clusters in which they can repeat periodically over
epochs as long as two days. Detection of these decelerations is
useful in, for example, clinical monitoring strategies to make the
early detection of clinical conditions.
As such, the decelerations are exemplary of the phenomena of
pathological fluctuations in a physiological signal that may be
cyclical in whole or in part. General methods for detecting the
decelerations and other pathological fluctuations can be useful in
the clinical arena for risk assessment, diagnosis, screening and
evaluation of treatment of clinical conditions.
Fluctuations in a physiological signal may share common
characteristics, especially when the fluctuations are indicative of
an underlying pathology. These fluctuations may be distinguishable
from the normal variability found in physiological signals due to
the common characteristics. To this end, disclosed herein are
embodiments directed toward direct detection of a physiological
signal by characterizing the shape and duration of the
fluctuations. Provided is an algorithm, method, system, and
computer program product for characterizing a physiological signal
that may serve as signs of clinical conditions. Moreover, detection
of specific variations in the non-linear dynamical properties of
continuously monitored physiological data can be used to
differentiate among clinical conditions that present with similar
clinical syndromes.
The fluctuations in the physiological signal are sometimes
periodic, i.e., occurring at fixed or somewhat variable intervals.
The periodic occurrences of the fluctuations may occur throughout
the signal or may occur in clusters.
An example of a physiological signal having fluctuations is that of
heart rate, which can be measured as heartbeats or, inversely, as
the time between heartbeats ("RR intervals" or "interbeat").
Accordingly, the physiological signal recorded may be a cardiac
signal such as a heart rate recorded as time series data, as for
example, in an electrocardiogram. The signal may be obtained from a
subject and recorded using devices or machinery known in the art,
e.g., heart monitors, such as PHILIPS INTELLIVUE and GE SOLAR
monitors, etc. The recorded physiological signal may be further
processed after it is recorded. For example, for a heart rate, a
heart beat time series recorded in the form of a heart-beat may be
converted into an RR interval time series, or vice versa, depending
on, inter alia the method adopted for recording the signal and the
method adopted to process it.
In one example, RR intervals may exhibit pathological fluctuations
in the form of decelerations. These decelerations have common
features of shape and duration, and can be distinguished from the
random fluctuations that are part of normal heart rate variability
as well as from heart rate periodicities known to those of ordinary
skill in the art (e.g., respiratory sinus arrhythmia, which is the
coupling of heart rate to breathing, produces fluctuations in RR
intervals and Mayer waves, which are correlated with blood pressure
cycles).
FIG. 1(a)-(d) compares RR interval times series of a healthy infant
from the neonatal intensive care unit (NICU) (FIG. 1(a)); a time
series of a NICU infant with sepsis (FIG. 1(b)); a time series with
decelerations (FIG. 1(c)); and a time series with periodic
decelerations (FIG. 1(d)). The time series of FIG. 1(d) is from the
heart rate records of a neonate who suffered intracranial
hemorrhage with concomitant sepsis, wherein the periodic
decelerations lasted over 48 hours. Comparing FIGS. 1(a) and 1(d),
the time series with decelerations represents reduced-dimensional
dynamics. The period of the decelerations is nearly constant at
about 45 beats, or 15 seconds, and, while the decelerations have a
variety of heights (which can be anywhere from 20 to 300 ms), they
have a common shape. Moreover, they are more organized than random
fluctuations, and have slower periods and larger amplitudes than
respiratory sinus arrhythmia or Mayer waves. The algorithm, method,
system, or computer program product disclosed herein can be used to
detect and analyze the distinguishing characteristics of these
periodic decelerations, or fluctuations from other types of
physiological phenomena, that may serve as an indication of an
underlying clinical condition, such as sepsis.
In one aspect, the method, system, and computer program product can
be used to apply a template-based pattern-matching algorithm that
may be based on a template function that provides an adequate
description of the shape of a fluctuation. The template function is
used in a sliding window analysis to find sequences in the
physiological signal that match amplitude- and duration-adjusted
versions of the template to within a specified tolerance.
Analytical functions that may serve as the template function
include, but are not limited to, exponential, Gaussian, and
Lorentzian. Alternatively, the analytical function .chi.(n), which
was developed for the present embodiment, may be applied. This
function closely matches the features of individual clinically
observed decelerations of neonatal heart rate.
The derivation of .chi.(n) was based on decelerations in RR
interval time series. The RR interval time series was considered as
the sum of .chi.(n) and some remainder, as illustrated in equation
(S.1) below, and in FIG. 2. i. RR(n)=a.chi..sub.b(n-n.sub.0)+G(n)
(S.1)
RR(n) is the time between beat n and beat n+1,
.chi..sub.b(n-n.sub.0) is a function describing a deceleration of
width b centered about the point n.sub.0, and G(n) represents the
remainder of the signal. The value a is the height or amplitude of
the deceleration and is:
.times..function..times..times..chi..chi..times. ##EQU00001##
The numerator RR.chi..sub.b is:
.times..times..times..chi..times..times..times..function..times..chi..fun-
ction..times..function..times..chi..function..times..chi..times..times..ti-
mes..times..times..chi..chi..function..ltoreq..ltoreq..times.
##EQU00002##
In other words, when calculating the convolution RR.chi..sub.b, a
portion of the signal RR(n) is projected onto a deceleration
waveform, .chi..sub.b(n-n.sub.0). If G(n) from Equation (S.1) was
presumed to have the properties of Gaussian white noise, then
formula (S.2) would be a maximum likelihood estimate of a.
Function .chi.(n) meets common characteristics of decelerations.
Decelerations, as shown in FIG. 3, are relatively symmetric and,
compared to a Gaussian wave, are narrower at the peak and wider in
the wings. Hence, function .chi.(n) was devised to match these
features and reads as follows:
.times..chi..function..times..times..function..function..times..times..fu-
nction..times. ##EQU00003## where the amplitude a is calculated as
described by Equation (S.2), and b is the width parameter. Compared
with visually identified decelerations, the function .chi.(n) was
determined to be an adequate match; the median R-squared value
between the function .chi.(n) and the decelerations was 0.93. In
addition, the .chi.(n) function consistently provided a reasonable
estimate of the height of a deceleration relative to its
baseline.
In some embodiments, the method, system, and computer program
product can be used to apply the sliding-window analysis to sweep
.chi..sub.b (n) through the time series RR(n) at varying widths.
For example, the width of .chi..sub.b(n) during the analysis may
increase from about eight to about 100 beats, such that the first
sweep involves function .chi..sub.b(n) having a width of eight
beats, the second sweep involves function .chi..sub.b(n) having a
width of nine beats, etc.
The correlation coefficient a(n.sub.0,b) between the function
.chi..sub.b(n) and any fluctuation present at the particular width
("scale," b) and heart beat around which the function
.chi..sub.b(n) is centered ("translation," n.sub.0) in the time
series RR(n) is calculated. For instance, if the scale of
.chi..sub.b(n) is eight beats, then a(n.sub.0,b) is calculated for
any deceleration present at beats 1-8, then beats 2-9, etc, until
the calculation continues ("sweeps") through the desired number of
beats (e.g., 300 beats) of the RR(n) time series. Then, if the
second scale of .chi..sub.b(n) is 9 beats, a(n.sub.0, b) is
calculated for any deceleration present at beats 1-9, then beats
2-10, etc.
Once a(n.sub.0,b) is calculated at each desired scale and
translation, a surface of a(n.sub.0,b) values can be generated, as
shown in FIG. 4. From this analysis, the points (n.sub.0, b), which
is where the correlation between .chi..sub.b(n) and the signal was
the strongest, i.e., where a(n.sub.0,b) has a local maximum, can be
identified.
It is to be understood that the sliding-window analysis described
herein may be applied to other template functions, such as
exponential, Gaussian, and Lorentzian functions.
In some embodiments, the method, system, and computer program
product can be used to apply additional measures to ensure that the
match between the template function and the decelerations is
accurate. For example, if there is more than one local maximum at a
particular translation, only the largest of the maxima (i.e.,
highest correlation) may be recognized so that one deceleration was
not erroneously represented by two waveforms, as described below.
Also, to avoid spurious correlations, a minimum correlation
coefficient (R.sup.2) that characterizes the fit between the
wavelet .chi..sub.b(n-n.sub.0) and the RR interval time series may
be required. For instance, the requirement may be that R.sup.2 is
at least greater than some minimum, such as 0.74.
Moreover, another measure to ensure accuracy is to adopt a
threshold for the amplitude of the deceleration. In general,
pathological decelerations are much larger in amplitude than nearby
heart rate accelerations. Therefore, the amplitude of the
deceleration may be required to surpass a minimum value or measure
of the accelerations. For example, the amplitude of the
deceleration may be required to be greater than the R.sub.1 as
determined by sample asymmetry, such as by 5- or 7- or 10-fold.
Sample asymmetry analysis is a method of determining asymmetry of
frequency histograms. Generally, a power function is used to weigh
the deviation of each RR interval in the series from a certain RR
interval value. The average weighted deviation for intervals lower
than this reference value is calculated, and is lower for RR
interval records showing reduced variability and transient
decelerations. The average weighted deviation for intervals higher
than the median is also calculated, and is higher for RR interval
records showing reduced variability and transient decelerations.
The ratio right/left weighted deviation (R.sub.1 for accelerations,
R.sub.2 for decelerations) is computed as an indicator of asymmetry
of each RR sample.
In certain embodiments, the method, system, and computer program
product can be used to analyze two or more potential decelerations
that overlap. This can be achieved by testing and calculating
R.sup.2 (commonly called the coefficient of determination) for each
template separately. R.sup.2 is also calculated for every possible
combination of the two or more templates that fit to the data. The
individual template, or the combination of templates, that
maximizes R.sup.2 is chosen as the correct representation of the
data. This last step ensures that an individual deceleration is
only counted once.
The reasons for including a step of distinguishing overlapping
potential decelerations are illustrated in FIGS. 5 and 6. FIG. 5(b)
shows two templates identified for each of the first two
decelerations shown in the data as potential. If all of the
identified potential decelerations are counted as actual
decelerations, six decelerations would be identified, rather than 4
decelerations as shown in FIG. 5(b). In a different scenario, FIG.
6(a) shows two templates whose tails overlap that fit the two
decelerations in the data. The present method determined that both
of these templates are needed to correctly represent this portion
of the data as containing two decelerations, as shown in FIG.
6(b).
In certain embodiments, the method, system, and computer program
product can be used to remove decelerations from RR interval time
series in order to reduce residual heart rate variability and
improve the applicability of conventional heart rate variability
measures such as variance or standard deviation (3). While
pathological decelerations have the effect of increasing the
overall variability of the RR intervals, there is reduced baseline
variability in-between the decelerations. This may be detected by
calculating the standard deviation of the RR interval time series
that remains after decelerations have been identified and
subtracted away. In fact, there is enhanced diagnostic accuracy if
both findings--decelerations and reduced baseline variability--are
present. For example, false positive identification of
decelerations may be identified, because such sequences differ from
pathological decelerations by failing to rise and fall from a
baseline of reduced heart rate variability. Thus, detection of a
quiet baseline is an important diagnostic factor.
Baseline variability may be measured through various techniques
including, but are not limited to, determining R.sub.1 sample
asymmetry, which is greater in normal records, and measuring
smoothness (that is, low variability) of the residual signal.
Sample asymmetry may be determined as described above, while
smoothness may measured by methods known in the art, e.g.,
calculate the variance, or the absolute value of the first
difference (difference of consecutive RR intervals) of the
remaining segments of signal, and determine the mean. This shown by
of FIG. 2. For example, one method for calculating smoothness, or
the degree of residual fluctuations, is (1) identify sufficiently
well-matched template waveforms in the signal; (2) remove these
segments of the signal; (3) calculate the variability of the
remaining signal using, for example, the root-mean-squared of the
first differences, or the mean of the absolute values of the first
differences. For instance, FIG. 7(a) shows an RR interval time
series from a healthy infant that demonstrates a high degree of
normal heart rate variability. In this particular case, the
decelerations arise from an active parasympathetic nervous system
mediated through the vagus nerve. After decelerations are removed,
the remaining baseline heart rate is largely variable, which is
normal. This is in contrast to the RR interval time series shown in
FIG. 7(b), which is severely abnormal and is from an infant with
sepsis. Applying sample asymmetry analysis as described above,
R.sub.1 is greater in the RR interval time series of FIG. 7(a) as
compared to FIG. 7(b). Also, smoothness analysis shows that the RR
interval from beat-to-beat between decelerations has low
variability after decelerations are removed.
Some embodiments relate to deceleration detectors, which comprise a
method, system, or computer program product used to distinguish the
"true" decelerations from the false positives, applying the means
described above.
The method, system, and computer program product disclosed herein
may be applied to short data recordings, as little as 30 minutes or
less.
It is understood that one of ordinary skill in the art could apply
the methods described herein to a signal other than heartbeats, and
to subjects other than infants. For example, the signal may be
temperature readings over time, which can fluctuate when
temperature increases or decreases, and may be used to predict
sepsis. Also, the subjects may be adults or the aged, or may be
non-human subjects.
Furthermore, it is understood that the skilled artisan may apply
the method of the invention to determine the risk of clinical
conditions other than sepsis. For example, the method may be used
to determine the risk of heart blockage in subjects. Moreover, the
skilled artisan would recognize that clinical sepsis is correlative
with other clinical conditions, and that these other clinical
conditions may be predicted by the method of the invention.
In another embodiment, the Hopf Bifurcation Theory may be applied
to simulate heart rate based on the characterization of the time
between beats as described herein. Hopf Bifurcation Theory is based
on very general assumptions, for example, that the pacemaking
system of the heart has feedback loops that can be modeled by a
large number of dynamical variables (denoted u=[u.sub.1 . . .
u.sub.n]) governed by an equally large number of differential
equations (equation S.5).
.times.dd.function..times. ##EQU00004##
These governing equations are assumed to contain many parameters
p=p.sub.1 . . . p.sub.m, which may vary with time as follows: on
short time-scales, the parameters may have small, rapid, noisy
fluctuations but generally they do not have large changes; on long
time-scales, the parameters may have substantial slow variation.
The functions f(u;p), governing the rate of change of u, are not
presumed to be known. However, it is assumed that these functions
have Taylor expansions that converge in the range of interest, and
that these functions have a zero (steady state) which can be taken
to occur at u=0. For some values of the parameters p, the steady
state is assumed to be stable.
If the Taylor expansions are truncated at first degree, a set of
linear equations may be obtained:
.times.dd.function..times..times. ##EQU00005##
The eigenvalues of the matrix M(p) associated with these linear
equations must either be real or occur in complex conjugate pairs.
When all real parts of these eigenvalues are negative, the steady
state is stable. The steady state can go unstable if, as the
parameters p change, one real eigenvalue, or a pair of
complex-conjugate eigenvalues, crosses the imaginary axis. Hopf
Bifurcation Theory examines the latter case.
Modern versions of Hopf theory are much more general than Hopf's
original version, and they provide two powerful theorems.
(1) "Center Manifold Theorem": In the state space of dynamical
variables u, there is a two-dimensional surface (called a center
manifold) which is an invariant surface and an attractor. That
means: (a) if u(t) lies initially on this two-dimensional surface,
it stays on this surface; (b) if u(t) lies initially off the
surface, it moves toward the surface in the manner of exponential
decay. Furthermore, the surface is analytic (Taylor-expandable) and
the evolution in the surface can be described by two differential
equations.
(2) "Normal Form Theorem": there exists a smooth change of
variables to new coordinates (x,y) such that the governing pair of
equations can be reduced to a standard "normal" form.
dd.mu..times..times..omega..times..times..function..function..function..t-
imes..times..function..times..times..times..times.dd.mu..times..times..ome-
ga..times..times..function..function..function..times..times..function..ti-
mes..times..times. ##EQU00006##
In polar coordinates, that normal form is
.times.dd.mu..times..times..times..times..times.d.theta.d.omega..times.
##EQU00007##
The new parameters .mu., a, b, and .omega. depend on the parameters
p. If, .mu. changes from negative to positive as the parameters p
vary with time, the stable steady state goes unstable, and one of
two things happens: (1) if a<0, a stable cycle is created (this
is a "soft Hopf" bifurcation); (2) if a>0, an unstable cycle is
destroyed; in this case, if b<0, the radius r(t) can go quickly
to a large value while the angle .theta.(t) increases steadily, so
the system "jumps" to a large cycle of frequency .omega. (this is a
"hard Hopf" bifurcation).
As indicated above, this theory can be used to simulate the heart
rate where it can be postulated that the time between beats RR(t)
is some function of one of these unknown variables
x(t)=r(t)cos(.theta.(t)). Since the shape of the decelerations is
known, a Fourier representation of that shape to get
RR(r(t),.theta.(t)) can be used. a.
RR(r(t),.theta.(t))=.SIGMA.c.sub.nr(t).sup.n cos(n.theta.(t))
(S.9)
To generate a simulation, equations (S.7) were integrated, adding
random noise to both variables after every time step, .DELTA.t.
This noise term was multiplied by a coefficient, k=.xi. {square
root over (.DELTA.t)}. The factor .xi. controls the strength of the
noise and, for the simulations, generally assumed a value between
0.01 and 0.1. The factor {square root over (.DELTA.t)} ensures that
the statistical properties of the noise fluctuations are
independent of the step size .DELTA.t. (The width parameter of the
distribution resulting from a random walk is proportional to the
square of the size of each step times the number of steps.) The
integration was carried out for various values of parameters .mu.,
a, and b (though a was always kept negative and b positive in order
to keep the bifurcation "hard") and for cases in which these
parameters fluctuated in time. This integration process yielded
output like that shown in FIG. 4. This output in (x(t),y(t)) was
then transformed into RR(t) as defined by Equation (S.9), using
only the first 11 terms in this series, as these are sufficient to
provide a good approximation to the deceleration shape. This
process and result are shown in FIG. 8(a)-(d).
FIG. 8(a) show oscillations induced in a noisy hard Hopf
Bifurcation Theory model transformed into RR intervals via a
Fourier series representation of a deceleration. Oscillations arise
when .mu. passes through zero in the positive sense and terminate
when .mu. passes through its critical value, in this case -1, in
the negative sense. FIG. 8(b) shows bursts of periodic
decelerations created by allowing the parameter .mu. to vary near
zero. Such simulation results resemble observed data (.mu. plotted
in black, RR intervals in blue). FIG. 8(c) presents data from
neonatal RR interval record that show bursts of periodic
decelerations (top) and simulation of data produced by noisy Hopf
model. FIG. 8(d) shows a close-up of real data (top) and
corresponding simulation created by forcing the parameter .mu. to
cross respective critical points at times when bursts are observed
to begin and terminate (bottom). The noisy precursors in both the
data and the model occur in a noisy Hopf model when the parameter
.mu. lingers just below the bifurcation point.
The success of the simulation from Hopf Bifurcation Theory supports
the general approach of detection of pathological fluctuations in
human illness, and allows the possibility of extracting previously
unavailable diagnostic information from clinical time series and
waveform data. Parameters of dynamical systems such p in the Hopf
model contain in themselves such information, so that fitting
dynamical models to observed data yields screening and diagnostic
test information.
The present invention is also directed to a system for implementing
the method described herein. The system may comprise hardware,
software or a combination thereof and may be implemented in one or
more computer systems or other processing systems, such as personal
digit assistants (PDAs) equipped with adequate memory and
processing capabilities. In an example embodiment, the method of
the invention may be implemented in software running on a general
purpose computer 1100 as illustrated in FIG. 9. The computer system
1100 may includes one or more processors, such as processor 1104.
The Processor 1104 is connected to a communication infrastructure
1106 (e.g., a communications bus, cross-over bar, or network). The
computer system 1100 may include a display interface 1102 that
forwards graphics, text, and/or other data from the communication
infrastructure 1106 (or from a frame buffer not shown) for display
on the display unit 1130. Display unit 1130 may be digital and/or
analog.
The computer system 1100 may also include a main memory 1108,
preferably random access memory (RAM), and may also include a
secondary memory 1110. The secondary memory 1110 may include, for
example, a hard disk drive 1112 and/or a removable storage drive
1114, representing a floppy disk drive, a magnetic tape drive, an
optical disk drive, a flash memory, etc. The removable storage
drive 1114 reads from and/or writes to a removable storage unit
1118 in a well known manner. Removable storage unit 1118,
represents a floppy disk, magnetic tape, optical disk, etc. which
is read by and written to by removable storage drive 1114. As will
be appreciated, the removable storage unit 1118 includes a computer
usable storage medium having stored therein computer software
and/or data.
In alternative embodiments, secondary memory 1110 may include other
means for allowing computer programs or other instructions to be
loaded into computer system 1100. Such means may include, for
example, a removable storage unit 1122 and an interface 1120.
Examples of such removable storage units/interfaces include a
program cartridge and cartridge interface (such as that found in
video game devices), a removable memory chip (such as a ROM, PROM,
EPROM or EEPROM) and associated socket, and other removable storage
units 1122 and interfaces 1120 which allow software and data to be
transferred from the removable storage unit 1122 to computer system
1100.
The computer system 1100 may also include a communications
interface 1134. Communications interface 1134 allows software and
data to be transferred between computer system 1100 and external
devices. Examples of communications interface 1134 may include a
modem, a network interface (such as an Ethernet card), a
communications port (e.g., serial or parallel, etc.), a PCMCIA slot
and card, a modem, etc. Software and data transferred via
communications interface 1134 are in the form of signals 1128 which
may be electronic, electromagnetic, optical or other signals
capable of being received by communications interface 1134. Signals
1138 are provided to communications interface 1134 via a
communications path (i.e., channel) 1126. Channel 1126 (or any
other communication means or channel disclosed herein) carries
signals 1128 and may be implemented using wire or cable, fiber
optics, blue tooth, a phone line, a cellular phone link, an RF
link, an infrared link, wireless link or connection and other
communications channels.
In addition, the computer system 1100 may include a physiological
monitoring system (not shown). The physiological signal monitoring
system may be linked to the computer system 1100 via the
communications interface 1134. The physiological monitoring system
may include devices such as heart monitors, e.g., PHILIPS.RTM.
IntelliVue and GE.RTM. Solar monitors, etc.
In this document, the terms "computer program medium" and "computer
usable medium" are used to generally refer to media or medium such
as various software, firmware, disks, drives, removable storage
drive 1114, a hard disk installed in hard disk drive 1112, and
signals 1128. These computer program products ("computer program
medium" and "computer usable medium") are means for providing
software to computer system 1100. The computer program product may
comprise a computer useable medium having computer program logic
thereon. The invention includes such computer program products. The
"computer program product" and "computer useable medium" may be any
computer readable medium having computer logic thereon.
Computer programs (also called computer control logic or computer
program logic) are may be stored in main memory 1108 and/or
secondary memory 1110. Computer programs may also be received via
communications interface 1124. Such computer programs, when
executed, enable computer system 1100 to perform the features of
the present invention as discussed herein. In particular, the
computer programs, when executed, enable processor 1104 to perform
the functions of the present invention. Accordingly, such computer
programs represent controllers of computer system 1100.
In an embodiment where the invention is implemented using software,
the software may be stored in a computer program product and loaded
into computer system 1100 using removable storage drive 1114, hard
drive 1112 or communications interface 1124. The control logic
(software or computer program logic), when executed by the
processor 1104, causes the processor 1104 to perform the functions
of the invention as described herein.
In another embodiment, the invention is implemented primarily in
hardware using, for example, hardware components such as
application specific integrated circuits (ASICs). Implementation of
the hardware state machine to perform the functions described
herein will be apparent to persons skilled in the relevant
art(s).
In yet another embodiment, the invention is implemented using a
combination of both hardware and software.
In an example software embodiment of the invention, the methods
described above may be implemented in SPSS control language or C++
programming language, but could be implemented in other various
programs, computer simulation and computer-aided design, computer
simulation environment, MATLAB, or any other software platform or
program, windows interface or operating system (or other operating
system) or other programs known or available to those skilled in
the art.
It is understood that one of ordinary skill in the art can
implement the above method and system into various devices, such as
bedside monitors, in order to record and provide predictive
information for illnesses.
The devices, systems, computer program products, and methods of
various embodiments of the invention disclosed herein can also
utilize aspects disclosed in the following references,
applications, publications and patents and which are hereby
incorporated by reference herein in their entirety:
International Application No. PCT/US2009/33082 filed Feb. 4, 2009.
"System, Method and Computer Program Product for Detection of
Changes in Health Status and Risk of Imminent Illness."
International Application No. PCT/US2008/60021 filed on Apr. 11,
2008. "Method, System and Computer Program Product for Non-Invasive
Classification of Cardiac Rhythm."
US Application Publication No. 2006/0074329 filed on Aug. 10, 2005.
"Quantitative Fetal Heart Rate and Cardiotocographic Monitoring
System and Related Method Thereof."
US Application No. PCT2002/0052557 filed on Feb. 27, 2001. "Method
and Apparatus for the Early Diagnosis of Subacute, Potentially
Catastrophic Illness."
U.S. Pat. No. 6,923,763 filed on Feb. 22, 2004. "Method and
Apparatus for Predicting the Risk of Hypoglycemia." License
terminated by Medical Predictive Science Corporation (MPSC) on Jun.
12, 2007.
International Application No. PCT/US2000/22886 filed on Aug. 21,
2000. "Method and Apparatus for Predicting the Risk of
Hypoglycemia."
International Application No. PCT/US2005/0137484 filed on Dec. 1,
2004. "Method and apparatus for the early diagnosis of subacute,
potentially catastrophic illness."
U.S. Pat. No. 6,856,831 filed on Jan. 29, 2001. "Method for the
early diagnosis of subacute, potentially catastrophic illness."
U.S. Pat. No. 6,330,469 filed on Sep. 25, 2000. "Method and
apparatus for the early diagnosis of subacute, potentially
catastrophic illness."
U.S. Pat. No. 5,216,032 filed on Mar. 17, 1999. "Method and
apparatus for the early diagnosis of subacute, potentially
catastrophic illness."
EXAMPLES
Example 1
To test the hypothesis that heart rate deceleration frequency and
characteristics improve the early detection of neonatal sepsis,
heart rate records of 479 very low birth weight (<1500 g)
neonates from NICUs at the University of Virginia, the University
of Alabama at Birmingham, and Wake Forest University, with a total
of 513,193 half-hour heart rate records collected between August
2005 and May 2006 were tested. Sepsis was defined as signs of
illness leading to a positive blood culture, or to a negative blood
culture that was followed by a course of antibiotic treatment, in
the 6 hours prior to or the 18 hours following the heart rate
record. All the rhythms were sinus in origin, and there were no
variations in the P-wave morphology. There was no large change in
the PR interval during the decelerations.
The sliding-window analysis was applied to these records, using
.chi.(n). The results revealed that the decelerations were
isolated, or, less commonly, were in clusters lasting up to two
days.
FIG. 10(a) shows a 25-minute excerpt of such a cluster, the whole
of which lasted approximately one day (top). This cluster occurred
several hours before a clinical diagnosis of sepsis was made. For
this particular infant, the number of tall (greater than 100 ms
from peak to baseline) decelerations in a half-hour record as a
function of days since birth is provided (FIG. 10, bottom). Each
data point represents one half-hour record. Through most of the
infant's stay, there were few occurrences of tall decelerations,
but a cluster occurred around Day 23. This cluster began six hours
before the infant showed clinical signs of sepsis.
FIG. 10(b) shows a burst of decelerations arising from a state of
low variability. The abrupt onset of the decelerations is
indicative of a "hard" Hopf bifurcation.
FIG. 10(c) shows that, within such extended clusters of
decelerations, shorter intervals of time were sometimes found,
lasting up to several hours, in which the decelerations showed
remarkable periodicity.
Generally, clusters of large decelerations, even those not
exhibiting striking periodicity, were associated with the onset of
sepsis, as shown in FIG. 10(d). This was evaluated by calculating
the fold-increase of risk of sepsis. At any moment, an infant in
the study population has a 3.9% chance of being within 48 hours of
a clinical diagnosis of sepsis. In the population of half-hour
records containing 15 decelerations or more, we about 20% occurred
near sepsis--a more than 5-fold increased risk. Overall, the number
of "large" decelerations in a neonatal HR record was proportional
to the fold-increase in risk of sepsis ("large" is defined as
greater than 100 ms in height from base to peak). The new
deceleration metrics added independent information to the existing
heart rate characteristics analysis in predicting neonatal sepsis
(p<0.0001).
Example 2
To determine whether specific illnesses may lead to characteristic
features of pathophysiological dynamics, heart rate characteristics
at the time of diagnosis of sepsis with blood cultures positive for
gram-negative (n=42), coagulase-negative gram-positive (CONS,
n=78), other gram-positive (n=37), and fungal (n=22) sepsis during
the years 2000 to 2004 was examined. The results, as shown in FIGS.
11(a)-(d) and 12(a)-(d), reveal that Gram-negative sepsis led to a
more than 2-fold increase in rate of decelerations (p<0.001),
and gram-positive organisms led to smaller increases (p<0.05).
All gram-positive organisms, on the other hand, led to more reduced
variability (p<0.005). Fungal infections, interestingly, led to
smaller and non-significant changes in both parameters. These
results point to the possibility that organism-specific toxins
alter control of heart rate in similar but not identical ways, with
gram-negative toxins leading to more frequent decelerations,
gram-positive toxins leading to larger reductions in heart rate
variability, and fungal toxins having the smallest effects.
Together, these results suggest that the dissection of abnormal RR
interval time series into components of reduced heart rate
variability and decelerations should allow more informed selection
of initial antibiotic therapy
Accordingly, while the invention has been described and illustrated
in connection with preferred embodiments, many variations and
modifications as will be evident to those skilled in this art may
be made without departing from the scope of the invention, and the
invention is thus not to be limited to the precise details of
methodology or construction set forth above, as such variations and
modification are intended to be included within the scope of the
invention. Therefore, the scope of the appended claims should not
be limited to the description and illustrations of the embodiments
contained herein.
REFERENCES
The following patents publications as listed below and throughout
this document are hereby incorporated by reference in their
entirety herein. 1. Goldberger, A. L., Amaral, L. A. N., Hausdorff,
J. M., Ivanov, P. C., Peng, C. K. & Stanley, H. E. (2002) Proc.
Natl. Acad. Sci. (USA) 99, 2466-2472. 2. Buchman, T. G. (2004)
Curr. Opin. Crit Care. 10, 378-382. 3. Griffin, M. P. &
Moorman, J. R. (2001) Pediatrics 107, 97-104. 4. Kovatchev, B. P.,
Farhy, L. S., Cao, H., Griffin, M. P., Lake, D. E. & Moorman,
J. R. (2003) Pediatr. Res. 54, 892-898. 5. Lake, D. E., Richman, J.
S., Griffin, M. P. & Moorman, J. R. (2002) Am. J. Physiol. 283,
R789-R797. 6. Richman, J. S. & Moorman, J. R. (2000) Am. J.
Physiol. 278, H2039-H2049. 7. Richman, J. S., Lake, D. E. &
Moorman, J. R. (2004) Methods Enzymol. 384, 172-184. 8. Lake, D. E.
(2006) IEEE Trans. Biomed Eng. 53, 21-27.
* * * * *