Detection of sleep apnea using respiratory signals
A method and system for sleep apnea detection are disclosed. The method comprises detecting at least one respiratory signal and utilizing a detection algorithm to automatically detect at least one sleep apnea event from the at least one respiratory signal. The system includes a sensor to determine at least one respiratory signal, a processor coupled to the sensor, and a memory device coupled to the processor, wherein the memory device includes a detection algorithm and an application that, when executed by the processor, causes the processor to utilize the detection algorithm to automatically determine at least one sleep apnea event from the at least one respiratory signal.
1. A method for sleep apnea detection using a wireless sensor device, wherein the wireless sensor device comprises a sensor to detect at least one respiratory signal; a processor coupled to the sensor; and a memory device coupled to the processor, the method comprising:
detecting at least one respiratory signal by the sensor;
extracting a plurality of features from the at least one respiratory signal by the processor;
utilizing a detection algorithm by the processor to automatically determine at least one sleep apnea event from the plurality of features,
wherein determining the at least one sleep apnea event includes
determining at least one candidate sleep apnea event using the plurality of features, and
confirming the at least one sleep apnea event by comparing:
a mean during the at least one candidate sleep apnea event to a mean threshold times a mean during baseline, and
the mean during the at least one candidate sleep apnea event to a variable times a mean dispersion metric.
2. The method of claim 1 , wherein the at least one respiratory signal is any of a nasal airflow respiratory (NAR) signal, a respiratory inductive plethysomography (RIP) effort signal and electrocardiogram (ECG) derived respiratory signal.
3. The method of claim 1 , further comprising:
preprocessing of the at least one detected respiratory signal using a plurality of low-pass filters that are elliptic.
4. The method of claim 1 , further comprising:
determining, using the detection algorithm, if a candidate sleep apnea event is a false positive; and
rejecting the candidate sleep apnea event when the candidate sleep apnea event is false positive, wherein the candidate sleep apnea event is determined to be false positive when at least one of following conditions is true:
a mean during the candidate sleep apnea event is not less than the mean threshold times the mean during baseline, and
the mean during the candidate sleep apnea event is not less than the variable times the mean dispersion metric.
5. The method of claim 1 , wherein the plurality of features include any of a low-pass filtered envelope width, a trend mean of the low-pass filtered envelope width, and a statistical dispersion of the low-pass filtered envelope width.
6. The method of claim 5 , wherein the statistical dispersion of the low-pass filtered envelope width is a difference between 90 th and 10 th percentiles over a predetermined time period.
7. The method of claim 1 , wherein determining the at least one candidate sleep apnea event further comprises:
determining whether an instantaneous amplitude of a low-pass filtered envelope width of the at least one respiratory signal is less than an amplitude threshold times a trend mean of the low-pass filtered envelope width; and
determining whether the trend mean of the low-pass filtered envelope width is less than a variable times a statistical dispersion of the low-pass filtered envelope width.
8. The method of claim 7 , wherein a counter is incremented if the instantaneous amplitude of the low-pass filtered envelope width is less than the amplitude threshold times the trend mean and the trend mean is less than the variable times the statistical dispersion, further wherein the at least one candidate sleep apnea event is determined if the counter is greater than a threshold.
9. A wireless sensor device for sleep apnea detection, the wireless sensor device comprising:
a sensor to detect at least one respiratory signal;
a processor coupled to the sensor; and
a memory device coupled to the processor, wherein the memory device includes a detection algorithm and an application that, when executed by the processor, causes the processor to:
utilize the detection algorithm to automatically determine at least one sleep apnea event from the at least one respiratory signal,
wherein the detection algorithm is optimized according to product of sensitivity and positive predictive values of apnea events among a training dataset of a predetermined number of apnea subjects by maximizing the product of the sensitivity based on true positives and false negatives, and by maximizing the positive predictive value based on true positives and false positives for apnea subjects of the training dataset of a predetermined number of apnea subjects, with a constraint on a false positive rate among a training dataset of a predetermined number of control subjects.
10. The wireless sensor device of claim 9 , wherein the at least one respiratory signal is any of a nasal airflow respiratory (NAR) signal, a respiratory inductive plethysomography (RIP) effort signal and electrocardiogram (ECG) derived respiratory signal.
11. The wireless sensor device of claim 9 , wherein the application, when executed by the processor, further causes the processor to:
preprocess of the at least one detected respiratory signal using a plurality of low-pass filters that are elliptic.
12. The wireless sensor device of claim 9 , wherein to utilize further comprises to:
extract a plurality of features from the at least one respiratory signal;
determine at least one candidate sleep apnea event using the plurality of features; and
confirm the at least one sleep apnea event using the at least one determined candidate sleep apnea event.
13. The wireless sensor device of claim 12 , wherein the application, when executed by the processor, further causes the processor to:
reduce at least one false positive from the at least one determined candidate sleep apnea event.
14. The wireless sensor device of claim 12 , wherein the plurality of features include any of a low-pass filtered envelope width, a trend mean of the low-pass filtered envelope width, and a statistical dispersion of the low-pass filtered envelope width.
15. The wireless sensor device of claim 14 , wherein the statistical dispersion of the low-pass filtered envelope width is a difference between 90 th and 10 th percentiles over a predetermined time period.
16. The wireless sensor device of claim 12 , wherein to determine further comprises to:
determine whether an instantaneous amplitude of a low-pass filtered envelope width of the at least one respiratory signal is less than an amplitude threshold times a trend mean of the low-pass filtered envelope width; and
determine whether the trend mean of the low-pass filtered envelope width is less than a variable times a statistical dispersion of the low-pass filtered envelope width.
17. The wireless sensor device of claim 16 , wherein the application, when executed by the processor, further causes the processor to:
increment a counter if the instantaneous amplitude of the low-pass filtered envelope width is less than the amplitude threshold times the trend mean and the trend mean is less than the variable times the statistical dispersion, wherein the at least one candidate sleep apnea event is determined if the counter is greater than a threshold.
18. The wireless sensor device of claim 12 , wherein the at least one sleep apnea event is confirmed when:
a mean during the at least one candidate sleep apnea event is less than a mean threshold times a mean during baseline, and
the mean during the at least one candidate sleep apnea event is less than a variable times a mean dispersion metric.