U.S. patent application number 13/252640 was filed with the patent office on 2012-04-12 for method for detecting and discriminating breathing patterns from respiratory signals.
Invention is credited to Jeffrey ARMITSTEAD.
Application Number | 20120088992 13/252640 |
Document ID | / |
Family ID | 36601272 |
Filed Date | 2012-04-12 |
United States Patent
Application |
20120088992 |
Kind Code |
A1 |
ARMITSTEAD; Jeffrey |
April 12, 2012 |
METHOD FOR DETECTING AND DISCRIMINATING BREATHING PATTERNS FROM
RESPIRATORY SIGNALS
Abstract
A Cheyne-Stokes (CS) diagnosis system classifies periods of
CS-like breathing by examining a signal indicative of a respiratory
parameter. For example, nasal flow data is processed to classify it
as unambiguously CS breathing or nearly so and to display the
classification Processing may detect and display: apnoeas,
hypopnoeas, flow-limitation and snore. The signal may be split into
equal length epochs and event features are extracted. Statistics
are applied to these primary feature(s) to produce secondary
feature(s) representing the entire epoch. Each secondary feature is
grouped with other feature(s) extracted from the entire epoch
rather than from the epoch events. This final group of features is
the epoch pattern. The epoch pattern is classified to produce a
probability for possible event classes (e.g., Cheyne-Stokes
breathing, OSA, etc.). The epoch is assigned to the class with the
highest probability, which may both be reported as an indication of
disease state.
Inventors: |
ARMITSTEAD; Jeffrey; (Bella
Vista, AU) |
Family ID: |
36601272 |
Appl. No.: |
13/252640 |
Filed: |
October 4, 2011 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
11576210 |
Mar 28, 2007 |
8066647 |
|
|
PCT/AU2005/001942 |
Dec 21, 2005 |
|
|
|
13252640 |
|
|
|
|
60638169 |
Dec 23, 2004 |
|
|
|
Current U.S.
Class: |
600/323 ;
600/538 |
Current CPC
Class: |
A61B 5/7267 20130101;
A61M 16/021 20170801; A61B 5/726 20130101; A61M 2230/40 20130101;
A61B 5/145 20130101; A61B 5/4818 20130101; A61B 5/087 20130101;
A61B 5/7232 20130101 |
Class at
Publication: |
600/323 ;
600/538 |
International
Class: |
A61B 5/087 20060101
A61B005/087; A61M 16/00 20060101 A61M016/00; A61B 5/1455 20060101
A61B005/1455 |
Claims
1-51. (canceled)
52. A method for a controlled ventilatory apparatus to
automatically determine the absence or the presence of
Cheyne-Stokes breathing in a person comprising the steps of:
capturing a signal indicative of respiratory flow of a person;
determining the absence or presence of an apnea or an hypopnea from
said signal and the rate of increase of flow following the
determined presence of an apnea or an hypopnea; determining any
peak in a frequency spectrum of said signal; and determining the
absence of Cheyne-Stokes breathing from a sudden increase in flow
following an apnea or an hypopnea, or determining the presence of
Cheyne-Stokes breathing from a combination of a peak in the
frequency spectrum and a slow rate of increase in flow following an
apnea or an hypopnea.
53. The method of claim 52 wherein said signal indicative of
respiratory flow is a nasal flow signal.
54. The method of claim 52 wherein said step of capturing a signal
indicative of respiratory flow of the person encompasses pulse
oximetry.
55. The method of claim 52 wherein the step of determining a peak
in a frequency spectrum of said signal comprises a step of
calculating a spectrogram of said signal and determining any peak
from said spectrogram.
56. The method of claim 52 further comprising the step of
determining whether the signal has a peak in a frequency range of 0
Hz to 0.075 Hz.
57. The method of claim 52 wherein the step of determining said
peak in a frequency spectrum of said signal includes calculating a
Fourier transform of the signal.
58. The method of claim 52 wherein said step of determining a rate
of increase of flow is determined from the signal of the
respiratory flow in a region of hypopnea to hyperpnoea.
59. The method of claim 52 further comprising the step of
calculating a shape of an envelope of said signal and utilizing
said shape as an indication of central apnea.
60. A ventilatory apparatus for automatically detecting the absence
of the presence of Cheyne-Stokes breathing in a person comprising:
a sensor for sensing a signal indicative of the respiratory flow of
a person; a detector for determining the presence of apnea or
hypopnea in said signal; and a processor for determining (a) any
peak in a frequency spectrum of said signal, (b) any sudden
increase in said signal, (c) the presence of any slow increase in
said signal; and (d) the absence of Cheyne-Stokes breathing from a
sudden increase in flow following an apnea or an hypopnea, or the
presence of Cheyne-Stokes breathing from the combination of a peak
in the frequency spectrum and a slow rate of increase in flow
following an apnea or an hypopnea.
61. The apparatus of claim 60 wherein said respiratory flow is
nasal flow.
62. The apparatus of claim 60 wherein said signal indicative of
respiratory flow encompasses pulse oximetry.
63. The apparatus of claim 60 wherein said processor determines a
peak in a frequency spectrum of said signal by calculating a
spectrogram of said signal.
64. The apparatus of claim 60 wherein said processor determines
whether the signal has a peak in a frequency range of 0 Hz to 0.075
Hz.
65. The apparatus of claim 60 wherein said processor determines a
peak in a frequency spectrum of said signal by calculating a
Fourier transform of said signal.
66. The apparatus of claim 60 wherein said processor determines a
rate of increase in flow in a region of hypopnea to hyperpnoea.
67. The apparatus of claim 60 wherein said processor calculates a
shape of an envelope of said signal and utilizes said shape as an
indication of central apnea.
Description
CROSS-REFERENCES TO RELATED APPLICATIONS
[0001] This application claims the priority of U.S. Provisional
Patent Application 60/638,169, filed Dec. 23, 2004.
FIELD OF THE INVENTION
[0002] The present invention relates to methods, algorithms and
diagnostic apparatus for the detection of sleep disordered
breathing patterns and the discrimination between patterns of
different disease states such as obstructive sleep apnea, central
sleep apnea and Cheyne-Stokes breathing and mixed sleep apnea.
BACKGROUND OF THE INVENTION
Breathing Disorders
[0003] Sleep-disordered breathing (SDB) encompasses a group of
disorders where the breathing pattern or quality of ventilation is
abnormal during sleep. Obstructive sleep apnea (OSA), the most
common such disorder (effecting possible 4-5% of the adult
population), is characterized by repetitive closing or collapse of
the upper airway and partial or complete diminution of breathing.
The obstruction is normally ended by the patient arousing briefly
when the muscles of the upper airway act to clear the obstruction.
During the repetitive cycle of obstruction and arousal, the OSA
patient will always continue to make "efforts" to breath; in other
words there is no central or brain-mediated disruption to the
breathing cycle.
[0004] Conversely, in central sleep apnea (CSA), there is a
disruption to breathing which is brain or control-centre in origin.
Cheyne-Stokes breathing (CSB) is one of the more common forms of
CSA. It is caused by an abnormal limit-cycle instability of the
patient's respiratory controller in which there are rhythmic
alternating periods of waxing and waning ventilation. Patients with
cardiac failure (a condition where the heart fails to pump
adequately) often have CSA, especially as the condition
deteriorates or where therapy has ceased to allow effective
compensation by the heart. Cheyne-Stokes breathing appears as a
cyclical variation in tidal volume seen in heart failure patients.
The cycle consists of an apnoea or hypopnoea followed by an
overshooting hyperpnoea which often (but not always) has a
characteristic hump backed morphology s a.k.a. a "Sydney Harbor
Bridge" shape. The exact cause of CS breathing is not fully
understood. However, the characteristic waxing and waning cycle is
strongly reminiscent of limit cycles in a poorly adjusted control
system with a maladjusted gain or destabilizing feedback-loop
delay.
[0005] Sleep-disordered breathing is undesirable in all its forms
because it disrupts sleep architecture (the pattern and proportion
of the different forms of sleep) leading to daytime somnolence. The
repetitive cessation or diminution of ventilation causes (sometimes
dramatic) drops in blood oxygenation levels. These and other
complications are probably responsible for the now established
sequelae of cardiovascular conditions.
[0006] The treatment of choice for OSA is continuous positive
airway pressure (CPAP) as first described by Sullivan [Sullivan CE,
et al. Reversal of obstructive sleep apnea by continuous positive
airway pressure applied through the nares. Lancet 1981 Apr. 18;
1(8225):862-5]. CPAP is also used to treat some heart-failure
patients with CSA and congestive heart failure (fluid on the
lungs). However, Cheyne-Stokes breathing is ineffectively treated
by CPAP and may require the application of servo-ventilation
[Teschler H et al. Adaptive pressure support servo-ventilation: a
novel treatment for Cheyne-Stokes respiration in heart failure. Am
J Respir Crit Care Med. 2001 Aug. 15; 164(4):614-9. Berthon-Jones
Ventilatory assistance for treatment of cardiac failure and
Cheyne-Stokes breathing. U.S. Pat. No. 6,532,959].
Diagnosis From Multiple Signals
[0007] The gold standard for the diagnosis of SDB and sleep apnea
is the polysomnograph (PSG): the measurement and recording of a
multitude of physiological signals during a stay overnight in a
sleep laboratory. Briefly, the PSG signal ensemble normally
includes one or more signals indicative of a respiratory parameters
such as patient airflow rate (for the calculation of ventilation
and the detection of apneas and hypopnoeas), multiple
electroencephalogram (EEG), electrooculogram (EOG) and
electromyogram (EMG) signals (for the determination of patient
sleep state, position and the detection of arousals from sleep),
breathing effort signals (either chest and abdominal distension
bands or an esophageal pressure-measuring catheter), snore
amplitude, and oxygen saturation. Another method of diagnosing SDB
is polygraphy (PG) whereby a reduced number of parameters are
recorded while the patient sleeps. These parameters include:
nasal/oral airflow rate, snore amplitude, oxygen saturation,
respiratory effort (thoracic and abdominal) and body position.
[0008] In both the PSG and the PG a breathing-effort signal is
recorded to enable the discrimination of OSA events from CSA or
Cheyne-Stokes breathing. (A third type of event is also
possible--the mixed apnea--where the event is initiated by a
centrally-mediated lack of breathing drive and ends with an airway
obstruction and subsequent arousal). It is impossible for the
inexperienced observer to reliably determine the type of apnea
without reference to at least the flow signal and a measure of
breathing effort. However, an experienced and trained observer
(expert) can often readily detect patterns in a run of events
(apneas/hypopnoeas) allowing a reliable determination of the type
of underlying disease. This is especially true of Cheyne-Stokes
breathing which has a very characteristic waxing and waning pattern
of ventilation.
Simple Recording Devices
[0009] The performance of either a PSG or PG requires trained
technicians, is expensive, is time consuming and can itself
introduce sleep disturbances. Also, it is well known that a
shortage of sleep laboratories is hampering the diagnosis and
treatment of current SDB patients, let alone what is considered a
vast undiagnosed population. For these reasons a type of
"screening" device (e.g., the microMesan.RTM. from MAP of Germany,
or the ApneaLink.TM. from ResMed) is available to test patients
suspected of having sleep-disordered breathing. Such devices are
small, recording just one or two physiological signals, and can be
readily sent home with the patient for a screening study. For
example: patients' nasal airflow can be recorded and later examined
by a physician using a personal computer and a connection to the
device. A software package would then be available to read the data
from the device, show statistics and make recommendations regarding
suspected sleep-related pathology.
Diagnosis Classifier
[0010] The calculation of the apnea-hypopnoea index (AHI, number of
such events per hour on average) is a measure regularly used to
guide the direction of either treatment or further investigation
with a full PSG or PG. A computer program or algorithm which
further enables the discrimination between different underlying
disease states based on the recorded breathing patterns provides
added guidance to the clinical pathway. A strong indication of
Cheyne-Stokes disease, for example, would suggest completely
different follow-up compared to the more common forms of sleep
apnea.
[0011] The concept of a classifier is common to many fields where
it is desirable to assign an object or an underlying state of an
object to one of a number of classes. This concept is used, for
example, in the fields of voice recognition (where sound bytes are
classified as different words or syllables), radar detection (where
visual signals are classified as enemy/friendly targets) and
medical diagnosis (where test results are used to classify a
patient's disease state). The design of a classifier falls under
the field of Pattern Recognition and a classifier can be of the
supervised type (the classifier is built from training data which
has been pre-classed by a supervisor or "expert") or unsupervised
type (where the natural ordering or clustering of the data
determines the different classes). Time signal classification
usually relies on representing the signal at particular time points
with "features". Features are simply numbers that distill the
essence of the signal at a point in time, a form of compression. A
set (or vector) of features is called a "pattern". A classifier
takes a pattern and manipulates it mathematically with a suitable
algorithm to produce a probability value for each of a number of
classes. The pattern is assigned to the class with the highest
probability.
[0012] In U.S. Pat. No. 6,839,581 there is disclosed a method for
detecting CS respiration in patients with congestive heart failure
by performing spectral analysis of overnight oximeter recordings to
obtain a set of parameters that can be used in the construction of
a classification tree and a trained neural network.
[0013] In summary, sleep-disordered breathing is a common syndrome
with different underlying disease types requiring very different
treatment options. There is a need for a small and relatively
inexpensive screening devices that can help unblock the treatment
bottleneck that currently exists at the sleep laboratory. An
algorithm and diagnostic apparatus that can replicate the expert's
ability to detect breathing patterns associated with particular
disease states will enhance the diagnosis and treatment of patients
being screened for sleep-disordered breathing, or for monitoring
patients already undergoing therapy. What is needed is an algorithm
for flow data in the form of classifier.
[0014] What is particularly desirable is a method and apparatus for
diagnosing Cheyne-Stokes breathing from flow readings or oximeter
readings by use of appropriate software in conjunction with a small
hand-held device for use in a home setting.
BRIEF SUMMARY OF THE INVENTION
[0015] The CS diagnosis system of the present invention uses
pattern classification techniques on a digital computer to identify
periods of CS-like breathing by examining the flow signal alone.
Ordinarily the definitive diagnosis of CS breathing relies on an
"effort" signal, either esophageal pressure or an elastic band
signal from the abdomen or thorax. An absence of effort denotes a
central apnoea which may otherwise be difficult to distinguish from
an obstructive apnoea or a mixed apnoea. A mixed apnoea is
comprised of a central beginning (without effort) followed by a
section of obstructed breaths once drive returns.
[0016] ApneaLink.TM. nasal flow data without other channels is
processed to classify it as unambiguously Cheyne-Stokes (CS)
breathing or nearly so and to then display a likely record to the
physician for quick expert confirmation. An ApneaLink.TM. recorder
is a single channel battery-powered respiratory pressure sensor
system and provides recordings of respiratory pressure during
sleep. The ApneaLink.TM. is a small (hand held) device manufactured
by ResMed, designed for use in a home setting where it is worn
strapped to the patient's chest. The device only records nasal flow
(indirectly) using a nasal pressure-sensing catheter. All relevant
respiratory information during sleep will be collected via nasal
pressure cannula. This will allow cardiologists to manage such
patients more expediently. For example, CS patients would go on to
a full polysomnogram (PSG) workup for possible AutoSet CS therapy
as appropriate. Non-CS patients might just go on to AutoSet to
treat the underlying OSA as appropriate. After suitable offline
processing of the nasal flow signal using a PC, the following
events can be detected and displayed: apnoeas, hypopnoeas,
flow-limitation and snore.
Cheyne-Stokes Detection Algorithm
[0017] The CS-detection algorithm uses the nasal flow signal from a
device such as ResMed's ApneaLink.RTM. or other signal indicative
of at least one respiratory parameter together with pattern
recognition techniques to assign a probability of CS breathing to
each 30 minute epoch of flow recorded. This invention details the
initial filtering and "event" detection, where events are defined
as regions of hypopnoea-hyperpnoea sequence characteristic of CS
breathing. The detection of such events may be determined from the
duration of one or more regions of hyperpnoea when the duration of
the hyperpnoea exceeds a threshold. or a statistic of the duration
of regions of hyperpnoea exceeds a threshold. Such a statistic may
be an average or a standard deviation or other statistics specified
below.
[0018] Pattern classification techniques are statistical and rely
on a so-called "training" data set by which a "classifier" can be
trained to recognize certain "patterns", in this case CS breathing.
A pattern is a group or vector of features. A feature is a number
which represents some aspect of the signal being examined. An
example of a feature is apnoea length. A pattern might be the group
comprising apnoea length, hyperpnoea length and a number
representing the closeness of the shape of the hyperpnoea to a
"harbor bridge".
[0019] One aspect of the invention is directed to a method and
apparatus or system capable of better diagnosing the presence of
sleep disorders, preferably with a higher level of confidence. The
diagnosis may comprise analyzing a signal indicative of a
respiratory parameter to determine a rate of increase of the signal
in the region from hypopnea to hyperpnoea and where the rate of
increase is a slow increase, concluding that Cheyne-Stokes
breathing is present and where the rate of increase is a sudden
increase, concluding that Cheyne-Stokes breathing is absent.
[0020] According to one aspect of the invention, a signal
representative of a patient's respiration is split into equal
length epochs which can be as long (the entire record) or as short
(the length of a representative hypopnoea-hyperpnoea sequence) as
desired. Preferably, the signal will be subject to a number of
pre-processing steps in order to filter out noise and zero the
baseline, for example.
[0021] Preferably, from each epoch one or more primary features is
extracted from the signal that act as a compressed representation
of the signal events. By events it is meant: e.g., apneas,
hypopnoeas and hyperpnoeas. Statistics are applied to the primary
feature(s) to produce one or more secondary features which
represent the entire epoch. Each secondary feature is grouped with
one or more other features that is extracted from the entire epoch
rather than from the epoch events. This final group of features is
the epoch pattern.
[0022] The epoch pattern is preferably manipulated with a suitable
classifier algorithm to produce a probability for each possible
class that the signal may be representative of (e.g. Cheyne-Stokes
breathing, OSA etc.). The epoch is assigned to the class with the
highest probability and the class and the strength of the
probability can be reported as an indication of the underlying
disease state.
[0023] The classifier algorithm is preferably learned from a
training data set which has been pre-classified by a human expert.
In this sense the classifier is of the supervised machine learning
type.
BRIEF DESCRIPTION OF THE DRAWINGS
[0024] FIG. 1 is a block diagram of the signal processing pathway
leading from the patient's respiratory signal through
pre-processing, feature extraction based on epochs, through to
classification.
[0025] FIG. 2 shows a typical respiratory signal epoch including a
number of "events" (in this case apnea-hyperpnoea sequences).
Several primary features are either shown explicitly (2.1
apnea/hypopnoea length, 2.2 hyperpnoea length, 2.3 cycle length) or
inferred (2.4 shape features of the hyperpnoea, 2.5 a feature
representing the initial "jump" at the beginning of the
hyperpnoea).
[0026] FIG. 3 shows details of the calculation of primary features.
3.1 shows the basis functions used in the determination of the
hyperpnoea shape features. 3.2 shows an hyperpnoea typical of OSA
together with the calculation of the jump feature displayed
graphically. 3.3 shows a similar depiction of an hyperpnoea more
typical of CS breathing. In both cases the calculated jump features
and shape features are tabled.
[0027] FIG. 4 shows examples of epoch classification, e.g., using
bar charts.
[0028] FIG. 5 shows the distribution of the normalized max jump in
a hyperpnoea signal.
[0029] FIG. 6 shows a cluster analysis of CS and OSA patients.
[0030] FIG. 7 shows results from a LD (linear) classifier.
[0031] FIG. 8 shows results from a QD (quadratic) classifier.
[0032] FIG. 9 shows the correction of data for baseline offset.
[0033] FIG. 10 shows a Cheyne-Stokes flow waveform, the long-term
ventilation and the left-shifted 10-second ventilation for a
typical patient.
[0034] FIG. 11 depicts those parts of the flow waveform that
correspond to hyperpnoeas.
[0035] FIG. 12 shows a Cheyne-Stokes patient's nasal flow signal
over about 15 minutes.
[0036] FIG. 13 shows a patient's Spo2 signal (saturation) and
ventilation signal (low-pass filtered absolute value of flow).
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
Process Description
[0037] FIG. 1 shows one embodiment of the classification process.
While the following may be explained in terms of a sequential
process, it is understood that the process can be carried out using
a non-linear, non-sequential, or non-staged process, or the order
of the process may be changed. Also, while FIG. 1 describes an
entire process, aspects of the invention may relate to only a
subset of that process. A signal representative of respiration is
first recorded from a patient using a logging device which includes
a data-acquisition system and a memory. The respiratory signal is
then processed either on-board by the recording device or off-line
using a computer.
[0038] Preferably, the signal is initially pre-processed. For
example, the signal is filtered to remove unwanted noise and, where
appropriate, the baseline is zeroed. The signal may also be
linearised depending on the transducer used to detect the
respiration.
[0039] In the next stage the signal is divided into n epochs of
equal length. The epoch length can be as long as the entire record
or as short as is practicable to enable detection of respiratory
patterns. In one preferred embodiment the epoch length is 30
minutes.
[0040] FIG. 2 shows a typical epoch recording in a patient with
Cheyne-Stokes breathing. The shape of the curve is reminiscent of
the shape of the Sydney Harbor Bridge and is sometimes referred to
by that name. See also FIG. 10. The recording consists of eight
"events", each event consisting of a hypopnoea (in this case also
an apnea) followed by a hyperpnoea. For each event an algorithm is
used to detect the beginning and end points such that event lengths
can be calculated: e.g., apnea/hypopnoea length and hyperpnoea
length. A further algorithm may be used to reject events if they do
not follow the correct sequence of hypopnoea/apnea-hyperpnoea.
Another further algorithm may be used to reject events that fall
outside sensible length scale limits.
Determination of Shape Features
[0041] Each hyperpnoea is further processed to derive four
so-called "shape features". These features indicate different
shaped hyperpnoeas (bell-shaped versus triangle-shaped for
example). The shape features are calculated using singular value
decomposition of the hyperpnoea ventilation signal as follows:
First, the hyperpnoea is extracted from the respiratory signal and
the absolute value is taken of the respiratory signal, giving a
ventilation signal. The ventilation signal is scaled by its mean
value to give a vector of values V.sub.hyperp. For mathematical
convenience the time base of the hyperpnoea [0 . . . T], where T is
the end of the hyperpnoea, is mapped to the interval [0 . . .
2.pi.]. A set of four orthogonal functions are calculated and
arranged as a 4.times.m matrix (where m is the number of values in
the hyperpnoea signal). A convenient set of orthonormal function
are:
M Basis = ( 1 .pi. sin ( t 2 ) 1 .pi. cos ( t 2 ) ( 3 .pi.sin ( t )
- 8 cos ( t 2 ) ) .pi. ( 9 .pi. 2 - 64 ) ( 3 .pi. cos ( t ) + 4 sin
( t 2 ) ) .pi. ( 9 .pi. 2 - 16 ) ) ##EQU00001##
where t is the time base over the hyperpnoea from 0 to 2.pi.. The
basis functions are shown plotted in FIG. 3.1. The four shape
features are then calculated as:
F.sub.P(1-4)=V.sub.hyperp.times.PseudoInverse(M.sub.Basis),
and are normalized by:
F P ( 1 - 4 ) = F P ( 1 - 4 ) L 2 F P ( 1 - 4 ) ##EQU00002##
where L.sub.2.parallel. is the L2 or Euclidean norm.,
i n x i 2 ##EQU00003##
[0042] The pseudoinverse M.sup.+ of a matrix M is a generalization
of a matrix inverse, and exists for any (m,n) matrix M (for
convenience assume m>n). If such a matrix M has full rank (n)
one defines: M.sup.+=(M.sup.TM).sup.-1M.sup.T. The solution of Mx=b
is then x=M.sup.+b. (Pseudoinverses are useful because of a general
theorem stating that F=M.sup.+v is the shortest length least
squares solution to the problem MF=v.)
Jump Determination
[0043] Since sudden jumps in the ventilation/flow at the beginning
of an hyperpnoea are characteristic of OSA, (see FIG. 2) each
hyperpnoea is further processed to derive the so-called "jump"
feature, indicative of the extent of any sudden increase in flow,
as follows: Again, the hyperpnoea is extracted from the respiratory
signal, the absolute value is taken of the respiratory signal,
giving a ventilation signal, a droopy peak-detector is used to
approximate the envelope of the ventilation signal:
TABLE-US-00001 e[1] = v[1] for i = 2 . . . m if v[i] .gtoreq. e[i -
1] e[i] = v[i] else e [ i ] = e [ i - 1 ] + 1 2.5 f s ( v [ i ] - e
[ i - 1 ] ) ##EQU00004## end
where e[i] is the approximate envelope, f.sub.s is the sampling
frequency and .nu.[i] is the ventilation signal. The envelope is
interpolated over a new two-second time base (chosen to be roughly
the time-length of a breath) to give e.sub.1 (between non-breathing
intervals). The maximum positive difference e.sub.1-e.sub.1(i-1)
(over the two second interval) is found in the interpolated signal
in the interval between the beginning of the envelope and the point
at which the envelope attains its maximum value. Finally, the
maximum difference is scaled by the mean value of the ventilation
signal to give the "jump feature". FIGS. 3.2 and 3.3 show this
process graphically for two representative hyperpnoeas.
Secondary Feature Determination
[0044] Secondary features are calculated from primary features
using the (measure of variation) statistics detailed below. (Note
log denotes the logarithm to base e.) First we define the standard
deviation as:
STD ( F P ) = 1 n - 1 i = 1 n ( F Pi - F P _ ) 2 , where F P _ = 1
n i = 1 n F Pi ##EQU00005##
For length measures (e.g. hypopnoea length) and the jump feature
the four features are::
1. 1 n i = 1 n log ( F Pi ) ##EQU00006## 2. log ( STD ( log ( F P )
) ) ##EQU00006.2## 3. log ( 1 n i = 1 n ( F Pi - F P ( i - 1 ) ) 2
) ( To get a normed deviation ) ##EQU00006.3## 4. log ( STD ( F Pi
- F P ( i - 1 ) ) ) ##EQU00006.4##
[0045] For hyperpnoea shape features the four features are:
1. 1 n i = 1 n ( F Pi ) ##EQU00007## 2. log ( STD ( ( F P ) ) )
##EQU00007.2## 3. log ( 1 n i = 1 n ( F Pi - F P ( i - 1 ) ) 2 )
##EQU00007.3## 4. log ( STD ( F Pi - F P ( i - 1 ) ) )
##EQU00007.4##
Additional Feature Determination
[0046] Additional features can be calculated using the entire (e.g.
30 minute) epoch signal. One such feature is derived from the
spectrogram of the epoch signal and determining that Cheyne-Stokes
breathing is present if the spectrogram indicates that the signal
has a peak. This feature is calculated as follows: First, the mean
of the respiratory signal is calculated and subtracted from the
respiratory signal and the resulting signal is chopped into n
slices which overlap each other by exactly half the slice length.
Each slice is next windowed, preferably using a Hanning window (to
reduce edge effects).
[0047] The use of a Hanning window to prepare the data for a FFT is
as follows: The FFT function treats the N samples that it receives
as though they formed the basic unit of a repetitive waveform: It
assumes that if one took more samples they would repeat exactly,
with the (N+1) sample being identical to the first sample, and so
on. The usual case is that if one's N samples start at one point in
a cycle, they end at some other point, so that if one really did
play these back in a loop one would get a discontinuity between the
last sample and the first. Hence one tapers both ends of the set of
samples down to zero, so they always line up perfectly if looped.
The formal name for this process is "windowing", and the "window
function" is the shape that we multiply the data by. When the
window function is the "raised cosine" 1+cos t the window is termed
a Hanning window. Other periodic functions can be used, yielding
other windows.
[0048] Next, since CS data appears periodic, a fast Fourier
transform is applied to each windowed slice, yielding a complex
vector result for each slice. The absolute value is taken of each
complex result yielding a real valued vector per slice. The mean is
taken of the resulting vectors to yield one vector. The natural log
is taken of the subsequent vector and the values in the frequency
range 0 Hz to 0.075 Hz are extracted to form a sub-vector, which is
then de-trended. Cheyne-Stokes behavior is present if the
spectrogram indicates the signal has a peak in the range 0 Hz to
0.075 Hz.
[0049] Briefly, the method of detrended fluctuation analysis is
useful in revealing the extent of long-range correlations in time
series, where the time series is a vector of data pairs (t.sub.i,
x.sub.i), where t represents time and x represents the variable
being measured. De-trending consists of subtracting from the x
values, values that have been calculated using a polynomial of
order n that has been fitted to the data. For example, for order
zero the polynomial is simply the mean of all the x values, and
that mean is subtracted from the original values. For order one,
the polynomial is simply a linear fit to the data (t.sub.i,
x.sub.i). Values calculated using the best linear fit are then
subtracted from the original values (so removing any linear
"trend"). For order two the fitted polynomial is a quadratic, for
order three a cubic etc.
[0050] The feature is then calculated as the maximum minus the mean
of the de-trended vector. Alternatively one could calculate the
entropy of the FFT instead of its peak.
[0051] Additional features can be derived by applying wavelet
analysis to each epoch. In this case wavelet coefficients or
statistics derived from wavelet coefficients are used as features
for the epoch. This yields the location of the peak in time. In
wavelet analysis a wave packet, of finite duration and with a
specific frequency, is used as a window function for an analysis of
variance. This "wavelet" has the advantage of incorporating a wave
of a certain period, as well as being finite in extent. A suitable
wavelet (called the Morlet wavelet) is a sine wave multiplied by a
Gaussian envelope.
Classification
[0052] A subset of features is then selected for use by the
classifier. It is known that a particular subset of features can
provide more accurate classification than the full set of features.
This is caused in part by the so-called "curse of dimensionality",
whereby the required number of training samples increases with the
number of features used. The curse of dimensionality causes
networks with lots of irrelevant inputs to behave relatively badly:
Where the dimension of the input space is high, the network uses
almost all its resources to represent irrelevant portions of the
space.
[0053] An algorithm is employed to select the best subset based on
the training data. Ideally every subset of features should be
tested for accuracy and the best subset chosen. The number of
subsets is 2.sup.n-1 where n is the number of features. Unless
there is a small number of features the exploration of all subsets
is impractical and, in any case, accuracy measures tend to be noisy
which further hampers the search for the best subset. Alternative
algorithms that enable selection of "good" feature subsets include
"best first", "remove worst", "random start with add and remove",
"simulated annealing" and genetic algorithms.
[0054] A method often used to measure accuracy is 10-fold
cross-validation. The training data are split into ten groups or
folds and ten different accuracy tests are performed. In each case
9 tenths of the folds are used for training and the resulting
classifier is tested for accuracy on the remaining tenth.
Statistics are performed on the 10 results to give a measure of
accuracy.
Training the Classifier
[0055] Once a feature subset is chosen, the classifier is trained
using the entire training data set. A number of classifier types
are available including: Baysean maximum likelihood linear and
quadratic discriminants, neural networks and support vector
machines. In each case a discriminant function is calculated which,
when applied to features calculated from new data to be classified,
provides probability estimates for different classes. The data
(epoch) is assigned to the class with the highest probability.
[0056] In one particular embodiment the discriminant function
includes or preferably consists of two weight vectors (of the same
length as the feature subset) and two constants. When the desired
feature subset has been extracted from the respiratory epoch, the
discriminant functions and probability are calculated as
follows:
d 1 = W 1 .quadrature. F + C 1 ##EQU00008## d 2 = W 2 .quadrature.
F + C 2 ##EQU00008.2## probability = ( d 1 - d 2 ) 1 + ( d 1 - d 2
) ##EQU00008.3## [0057] where W.sub.1, W.sub.2 are vectors and
C.sub.1, C.sub.2 are constants.
[0058] The probability cutoff may be set at 0.5 in which case a
probability of 1.0 would equate to class A and a probability of 0.0
to class B. The cutoff can be adjusted to suit the desired
sensitivity and specificity. This is a two-way classification. With
suitable training data, a three-way classification is also possible
as are even higher n-way classifications.
[0059] In one particular embodiment the classification of each
epoch could be displayed in a bar chart as in FIG. 4. Frame 4.1
shows a bar chart where many epochs show a high probability of a
class of respiration (in this case CS-like breathing). This
provides an "at-a-glance" indication of a patient record. Frame 4.2
shows a bar chart where only a single epoch displays strong CS-like
tendency. This provides an indication of where in the patient's
record a more detailed investigation is warranted.
Cheyne-Stokes Classifier Based on a Flow Signal or an Spo2 Signal
or Both
[0060] The ApneaLink device is capable of measuring an estimate of
a patient's flow signal which can be used as an input to the
algorithm described herein. Equally there are similar portable
devices that can measure and log SpO2, the saturation of
oxyhemoglobin in the blood as estimated by pulse oximetry. Pulse
oximetry is a simple non-invasive method of monitoring the
percentage of haemoglobin (Hb) which is saturated with oxygen. The
pulse oximeter consists of a probe attached to the patient's finger
or ear lobe which is linked to a computerised unit.
[0061] SpO2 is typically reported as a percentage, values above 95%
being normal and those below 95% indicating some degree of hypoxia
(lack of oxygen in the blood). Should a patient undergo an apnoea
or hypopnoea, it is usual for the SpO2 signal to fall concomitantly
with the ventilation, albeit after some delay. During Cheyne-Stokes
breathing the SpO2 signal will undergo the classic waxing and
waning pattern also characteristic of the ventilation.
[0062] Hence, it is conceivable that the algorithm described herein
might use a flow signal estimate (ventilation) or an SpO2 signal or
both signals to classify breathing patterns as being typical of
Cheyne-Stokes, OSA, mixed apnoeas etc.
[0063] FIG. 12 shows a Cheyne-Stokes patient's nasal flow signal
over about 15 minutes. FIG. 13 shows the same patient's Spo2 signal
(saturation) and ventilation signal (low-pass filtered absolute
value of flow). The signals have been normalized and shifted to
display them n the same graph. The same pattern recognition
techniques may be applied to both signals. For example: segment the
signal into hypopnoeas/hyperpnoeas; analyze the shape of the
hypopnoeas; determine the cycle lengths and space ratios; perform a
spectrogram (average of absolute value of a number of FFTs);
determine peaks in the spectrogram at the CS frequency; determine a
morphologic feature in both signals such as the jump feature;
perform a continuous wavelet transform on both signals and use
ridge finding techniques t follow any CS frequency component over
time.
Example 1
[0064] A set of data for testing the ability of flow data to be
classified into OSA and CS instances consisted of 90 patient
studies of approximately 8 hours each. For purposes of the test,
both nasal pressure, flow and two effort signals (abdomen &
thorax) were recorded, making a confirming diagnosis of the
underlying disease possible. The set was divided into 3 groups of
30 patients: OSA (obstructive apnoea), CS and Mixed. The data were
further classified (initially) on a 30-minute time-bin basis. The
time periods were classified into the following categories: No
apnoeas or hypopnoeas (<5) within the time period; Primarily CS
breathing (>90%); Primarily OSA (>90%); Primarily (>90%)
apnoeas and hypopnoeas of mixed type (i.e. having a central
component followed by a number of obstructed breaths); A
combination of different events, typically brief periods of CS or
mixed apnoeas interspersed between OSA; Patient is moving and the
signal is of too low a quality to be useable.
[0065] Typically if CS disease is present, CS breathing will occur
in large blocks of at least 20-30 minutes. The data set contained
very few periods of "pure" mixed apnoeas. Rather, the mixed group
of 30 patients contained periods of OSA, CS breathing or a mixed
picture.
Feature Analysis
[0066] All features were analyzed by calculating distributions for
the different groups (OSA, Mixed, and CS). The distribution was
normalized by application of an appropriate function, for example
FIG. 5 shows the distribution of the normalized max jump in
hyperpnoea signal between beginning of hyperpnoea and time of peak
flow after application of log-to-base-e. The leftmost curves
represent a "normal" or Gaussian distribution. It can be seen that
the application of the log function has normalized the
distributions and, further, that this feature shows good separation
between the CS (left) and OSA (rightmost) groups.
Cluster Analysis
[0067] Both k-means and fuzzy k-means clustering techniques were
utilized to visualize feature separation power. The features were
first averaged on a per-patient basis and then cluster analysis was
used to demonstrate a natural clustering into correct groups. FIG.
6 shows such an analysis. The Euclidean 2-norm distance from each
of two cluster centers is plotted one against the other. The CS and
OSA patients naturally fall into two groups except for one CS
patient. The Mixed patients fall into one group or the other
depending on the length of time spent during the night in different
breathing patterns. The separating diagonal in the figure
represents a naive classifier suitable for per-patient grouping.
What such a classifier cannot do is find a short period of CS
breathing from amongst an otherwise OSA-dominated night.
Feature Temporal Averaging
[0068] The training of a classifier using patterns assigned to
individual events is problematical. Temporal averaging was used to
reduce the amount of calculation, while also (potentially)
increasing statistical power. A 30-minute time-bin was chosen as a
best first-guess. After temporal averaging, a new set of
per-time-bin patterns is created. The raw features used (visible
separation of groups) were: hypopnoea length; hyperpnoea length;
1.sup.st Fourier shape feature; 2.sup.nd Fourier shape feature; and
normalized max jump. The time-averaged 30-minute bin features
tested were (std=standard deviation, meansq=mean of square of
values, sqrt=square root, shift=allows calculation of temporal
difference).
Classifier Training and Testing
[0069] Once the data had been processed and the "expert" diagnosis
made, a group of 1440 30-minute bins was available for classifier
training (90 patients.times.16 bins).
Classifier Selection
[0070] Numerous statistical methods exist for the training of a
classifier from n-dimensional data, e.g.: nearest neighbor, neural
nets, cluster analysis. However, because the data "appeared"
linearly separable, Bayesian decision theory was used. This theory
(which relies on underlying normal probability density functions)
uses the minimization of the Bayes error to calculate a
discriminant surface. Such a surface separates the data into one of
n categories (in this case 2). Both linear and quadratic
discriminant functions were utilized. The former separates the data
with a hyperplane in m dimensions (where m is the number of
features) while the latter separates the data with a hyperquadric.
A hyperplane discriminant is always preferred (assuming accuracy of
the same order), as it will tend to be well behaved in areas of
minimal data coverage.
Over-Optimistic Train and Test
[0071] The classifier was trained using the training set and then
the classifier was tested using the same data. This results in
over-optimistic values of sensitivity and specificity, as one would
intuitively expect. However, again this is an insightful process
and one can use a minimal features set (.ltoreq.3 features) in
order to visualize the result. FIG. 7 shows an LD classifier (plane
shows equi-probability surface). FIG. 8 shows a QD classifier
(quadric shows equi-probability surface).
Results
[0072] During each test the accuracy, sensitivity and specificity
were noted as was the current features set (or group of feature
sets) with the best accuracy. Estimates of accuracy, sensitivity
and specificity resulted of the order of 91%, 91% and 96%
respectively.
Example 2
Flow Filtering
[0073] The flow is filtered first to remove unwanted and
uninteresting high-frequency content. The filter used is a digital
FIR (finite impulse response) filter designed using the Fourier
method using a rectangular window. The filter has a pass-band from
0 to 0.9 Hz, a transition band from 0.9 to 1.1 Hz and a stop band
above 1.1 Hz. The number of terms in the filter varies with
sampling frequency. The flow signal is filtered by convolving the
time series point-wise with a filter vector.
Ventilation Calculation
[0074] A long-term ventilation signal y.sub.long is calculated
using a simple (first order) low-pass filter applied to the flow
signal. A time constant of 200 seconds is used (longer than the
longest possible cycle of Cheyne-Stokes breathing). In order to
measure ventilation (and not mean flow), the filter is applied to
the square of the flow signal and the square root is taken of the
filter output. Next, a ten-second ventilation y.sub.10 is
calculated (a more "recent" measure). This measure is created by
convolving the square of the flow signal with a 10-second square
wave with an area of one, i.e. a 10-second-long moving average, and
then taking the square-root of the result. This filter will have a
five second delay constant over the frequency range of interest.
For this reason the signal is shifted left by five seconds so that
it "lines up" with the original signal for timing purposes. FIG. 10
shows a Cheyne-Stokes flow waveform (large amplitude rapid varying
curve), the long-term ventilation (low amplitude slowly varying
curve) and the left-shifted 10-second ventilation (moderately
varying curve) for a typical patient.
Event Detection from Ventilation Signals
[0075] The 10-second ventilation signal is used to create low and
high thresholds for detection of events (hypopnoea-hyperpnoea
sequences). The thresholds are:
thresh.sub.low=0.5.times.y.sub.long;
thresh.sub.high=0.75.times.y.sub.long;
The timing of events is calculated using the following algorithm:
1. Find all points where y.sub.10<thresh.sub.low. 2. Find all
contiguous sections of the above points. These are provisional
hypopnoeas. 3. Find all points where y10>thresh.sub.high. 4.
Iterate over all of the hypopnoeas identified in step 2. If no
points identified in step 3 (hyperpnoeas) fall between hypopnoea n
and hypopnoea n+1, then the hypopnoeas n & n+1 are joined
together (because no hyperpnoea has been identified between them)
to form one hypopnoea. Repeat for all iterations. 5. The hypopnoeas
are now confirmed. All inter-hypopnoea regions are considered
hyperpnoeas. Each hypopnoea-hyperpnoea "event" constitutes one
possible Cheyne-Stokes cycle. E.g. in FIG. 10 there are five cycles
shown.
Calculate Event Timings
[0076] Event timings are calculated for each event as follows:
.tau..sub.hypopnoea=t(end_of_hypopnoea)-t(beginning_of_hypopnoea);
.tau..sub.cycle=t(beginning_of_next_hypopnoea)-t(beginning_of_hypopnoea)-
;
.tau..sub.hyperpnoea=t.sub.cycle-t.sub.hypopnoea;
[0077] Obviously the above events will include some unwanted
"garbage". For example, a one-hour-long period of normal breathing
bracketed on each side by Cheyne-Stokes breathing will look like a
one-hour-long hyperpnoea! (y10 always greater than threshold).
Hence, the following sensible limits are applied to the events:
.tau..sub.hypopnoea: minimum=10 seconds, maximum=100 seconds,
.tau..sub.cycle: minimum=15 seconds, maximum=250 seconds,
.tau..sub.hyperpnoea: minimum=5 seconds.
[0078] All events outside these limits are rejected and not
processed. We now have event timings and the ability to extract
parts of the flow waveform for further analysis. For example, we
can iterate over all the events and select out those parts of the
flow waveform that correspond to hyperpnoeas. FIG. 11 is an example
where we have selected out an hyperpnoea from the above sequence
and plotted it separately. In all further processing it is the 1 Hz
filtered flow signal that is used for feature extraction.
[0079] While the invention has been described in connection with
what are presently considered to be the most practical and
preferred embodiments, it is to be understood that the invention is
not to be limited to the disclosed embodiments, but on the
contrary, is intended to cover various modifications and equivalent
arrangements included within the spirit and scope of the
invention.
* * * * *