Method and apparatus for improved flow limitation detection of obstructive sleep apnea
View Patent ↗In a respiratory apparatus for treatment of sleep apnea and other disorders associated with an obstruction of a patient's airway and which uses an airflow signal, an obstruction index is generated which detects the flattening of the inspiratory portion of the airflow. The flattening index serves as an obstruction index used to differentiate normal and obstructed breathing. The obstruction index is based upon comparison of values of airflow in different sectors of the inspiratory peak of the wave function and is particularly suitable for distinguishing M shaped or square shaped respiratory patters indicative of partially obstructed airways.
1. A method for detecting inspiratory and expiratory flow-time curves in an airflow, comprising:
generating an airflow signal representing a respiratory airflow curve based on measurements of the airflow over time, and
processing the airflow signal using a processor by:
provisionally taking a start of inspiration as a time that the airflow exceeds a predetermined threshold,
rejecting the start of inspiration if the airflow then falls below zero before a total volume inspired reaches a first predetermined value,
provisionally taking a start of expiration as a time that the airflow falls below zero, and
rejecting the start of expiration if the airflow then goes above a second predetermined value in an inspiratory direction before a total volume expired reaches a predetermined value; and
based on the processed airflow signal, adjusting a treatment pressure provided to a patient.
2. The method of claim 1 further comprising trimming, using the processor, any end-expiratory pause that has been included in the inspiratory flow-time curve.
3. The method of claim 2 , wherein the trimming comprises finding an earliest time at which inspiration could have started by
locating a time where flow first reaches a predetermined percentage of peak inspiratory flow,
searching backwards to the time where the flow last reached a second predetermined percentage of the peak inspiratory flow, and
extrapolating backwards to find a time where flow should be zero.
4. The method of claim 1 further comprising reversing, using the processor, an effect of a given low pass filter on an original flow-time curve by the steps of
(a) passing the original flow-time curve through a second low pass filter that is identical to the given low pass filter a second time to obtain a doubly filtered flow-time curve,
(b) subtracting the doubly filtered flow-time curve from the original flow-time curve to give a difference representing a first-order approximation of the effect of the given low pass filter, and
(c) adding the difference back to the original flow-time wave.
5. The method of claim 1 further comprising normalizing, using the processor, an effect of overshoot on the inspiratory flow-time curve.
6. The method of claim 5 wherein normalizing the effect of overshoot comprises
estimating a peak flow in a middle portion of the inspiratory flow-time wave,
resetting each point of the inspiratory flow-time curve outside the middle portion that exceeds the peak flow.
7. The method of claim 6 further comprising continuing the resetting until no point of the inspiratory flow-time curve outside the middle portion exceeds the peak flow.
8. The method of claim 5 wherein normalizing the effect of overshoot comprises:
dividing the inspiratory flow-time curve into n parts,
calculating mean amplitudes for each part,
determining what sort of wave a pattern of the mean amplitudes indicates, and
if overshoot is indicated, marking flattening as low.
9. The method of claim 5 wherein normalizing the effect of overshoot comprises:
calculating an estimate of flattening for every n samples in a middle half of the inspiratory flow-time curve; and
marking a flattening index as a value that corresponds to the lowest value obtained for all sample permutations.
10. The method of claim 1 further comprising calculating, using the processor, a mean deviation of the inspiratory flow-time curve from a reference amplitude (M), for a middle portion of the inspiratory flow-time curve.
11. The method of claim 10 where the middle portion is one quarter of the inspiratory flow-time curve and is centered, wherein calculating the mean deviation uses the formula
MeanDeviation
=
∑
t
=
24
40
(
fs
(
t
)
-
M
)
16
wherein fs(t) is a sample of a patient's inspiratory airflow.
12. The method of claim 10 further comprising determining a flattening Index (FI) as
FI
=
MeanDeviation
M
.
13. The method of claim 1 further comprising updating, using the processor, a multiple breath pointwise moving average flow-time curve.
14. The method of claim 1 further comprising determining, using the processor, whether an inspiratory flow-time curve has an M shape by the steps of:
normalizing a time interval so that inspirations begins at zero and ends at B radians,
decomposing the inspiratory flow-time curve into a vector of features and a pair of basis vectors that are orthogonal, wherein the decomposing comprises
finding a pseudo-inverse of a basis made up of the pair of basis vectors, and
multiplying the inspiratory flow-time curve by the pseudo-inverse to find the vector of features.
15. The method of claim 14 further comprising
mapping a linear space for many patients and noting clusters into which points representing the individual patients fall, and
determining whether a breath is M shaped by a Euclidean measure of distance of the vector of features from an M cluster center.