The established practice of medical sleep monitoring, poly-somnography, involves wearing multiple electrophysiolog-ical sensors for a single night, at a sleep laboratory or at home. It provides clinically valuable information, but is ex-pensive and uncomfortable. More long-term and comfort-able measurements can be done with actigraphy, where the overall sleeping patterns of a patient are measured with a wrist-worn movement sensor.

Beddit measures sleep with unobtrusive force sensors. The idea is to measure the forces caused by the body on the bed with a flexible film sensor that is placed below the bed sheet.

Each ECG heartbeat signal consists of a clear spike (the QRS complex), which is followed by an impulse in the BCG around 80 ms later. The ECG spike precedes the impulse in the BCG, because electrical activity causes the mechan-ical contraction of the heart.

Respiration

Beddit Sleep Monitor is also capable of measuring respira-tory activity as respiration causes the chest to move mea-surably. Figure 2 shows how the respiration cycles and heart rate events appear in the sensor signal:

There are three main motivations for measuring respira-tion unobtrusively during sleep. First, respiration conveys information about the general condition of the patient, so the deterioration of health can be detected with respira-tion monitoring. Second, sleep-related breathing disorders (SRBD) such as sleep apnea represent a major share of sleeping problems. Third, the structure of sleep can be an-alyzed based on respiration, because sleep stages have differing effects on respiration.

The respiration measurement is a 4-step process where first the parts of the signal that contain gross movements are discarded. Then the respiration signal is low-pass fil-tered on 4 distinct f Hz frequencies with potentially disturb-ing phenomenons taken into account at around 2x f Hz. Therefore, at least one of the filters will result in an output signal that has the respiration frequency intact but the dis-turbance removed.

Then the respiration cycles are detected from each filtered signal. A respiration cycle begins at a local maximum and ends at the next local maximum in the signal. In addition, the amplitude of each respiration cycle is calculated by taking the difference between the signal value of the local maximum that starts the cycle and the minimal signal value in the cycle. Lastly, the final sequence of respiration cycle lengths is compiled from the four signals based on the sta-bility of respiration cycle amplitudes in each signal.

The correct signal is typically selected, because the signal that contains frequencies up to the respiratory frequency is more stable in its amplitude than a signal that also contains higher-frequency disturbing phenomena.

Beddit Sleep Monitor and the Science Behind It

→

www.beddit.com | Beddit Ltd. All rights reserved 2014.

The measurement methodology poses scientific chal-lenges because physiological information (heart rate, respiration, etc.) that are vital for analyzing sleep cannot be readily extracted from the sensor’s signal, but requires sophisticated signal analysis methods.

Ballistocardiography

The measurement of mechanical cardiac activity from the platform supporting the body is called ballistocardiogra-phy (BCG). Each time the heart beats, the acceleration of blood generates a mechanical impulse that can be measured with a proper force sensor, such as the Beddit Sleep Monitor.

Heart rate

Measuring the heart rate from BCG or similar mechanical signals is much more complex than measuring the heart rate using electrocardiogram (ECG), the most common-place cardiac measurement method. Individual heartbeats can be detected in an ECG signal relatively easily, by locating a clear spike (called the QRS complex, from the consecutive named spikes Q, R, S of the ECG heartbeat) that accompanies each heartbeat.

However, with BCG, the cardiac impulses are less pro-nounced and more variable than the salient shape of the QRS complex. The differences between the ECG and BCG-signals are shown in figure 1:

Movement

As sleep correlates with a low level of motility, circadian rhythmicity can be estimated with a method called actig-raphy. An accelerometer sensor is worn on the wrist 24 hours a day, which allows estimating the daily alternation between sleep and wakefulness. Due to its limited accu-racy, actigraphy is typically used for the overall characteri-zation of sleeping patters over a period of at least a week.

The Beddit Sleep Monitor detects the gross movement of the person sleeping. Even if Beddit is not a medical device and shouldn’t be used for any kind of medical diagnoses, the excessive movement during the night could potentially be a sign of, for example, periodic limb movement disor-der. In such a case, the user should be in contact with an appropriate doctor.

The movement information is analyzed by detecting dis-crete events of movement from the BCG signal. That is done by dividing the high-pass filtered (cut-off frequency 5 Hz) signal into three-second windows. Each window is detected as movement if the difference between signal minimum and maximum in the segment is above a fixed threshold.

Validity of the measurements

The measurement data provided by the Beddit Sleep Mon-itor has been tested against results from ECG and poly-somnography (PSG).

The heart rate data has been tested in a clinical study con-ducted during 2013 in co-operation with the VitalMed Re-search Centre together with the renowed sleep researcher and expert, Prof. Markku Partinen, M.D. The study con-sisted of 46 test subjects whose ages varied between 20 to 74 years. The study focused on measuring beat-to-beat heart rate. Out of all the detected beat-to-beat intervals, a maximum precision of 99.94% was achieved with the Bed-dit Sleep Monitor compared to the reference ECG-signal.

www.beddit.com | Beddit Ltd. All rights reserved 2014.

As for the the low-frequency component of the signal re-corded by Beddit, it’s very similar to the abdominal respira-tion effort signal in PSG.

In the figure 4 above, the upmost signal is from an airflow pressure meter while the middle signal is from an abdom-inal sensor band, both commonly used in PSG. As can be seen, the signal on the bottom that is produced by the Beddit Sleep Monitor is comparable with the reference sig-nals from PSG. However, the final classification of any dis-order typically requires a PSG measurement, with oximetry and measurement of airflow.

Also the gross movement signal recorded and analyzed by the Beddit Sleep Monitor has been tested against PSG as can be seen in figure 5.

Sources

Paalasmaa, J., Toivonen, H. & Partinen, M., 2014. Adaptive Heartbeat Modeling for Beat-to-Beat Heart Rate Measurement in Ballistocardio-grams. IEEE Journal of Biomedical and Health Informatics. To appear

Paalasmaa, J., Waris, M., Toivonen, H., Leppäkorpi, L. & Partinen, M., 2012. Unobtrusive Online Monitoring of Sleep at Home. 34th Annual In-ternational Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’12).

Paalasmaa, J., Leppäkorpi, L. & Partinen, M., 2011. Quantifying respira-tory variation with force sensor measurements. 33rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’11).

Paalasmaa, J., Partinen, M., 2012. A telemedicine solution based on a piezoelectric movement sensor for the long-term monitoring of sleep dis-order patients. Poster in 21st Congress of the European Sleep Research Society. Paris, France, September 4-8, 2012.

Figure 5: PSG-signal references against the movement data recorded by the Sleep Monitor.

Figure 4: PSG-signal references against the low-frequency component recorded by the Sleep Monitor.

Figure 3: Bland-Altman plot of the difference between the detected BCG beat- to-beat intervals and the ECG reference.

Out of the 770676 detected beat-to- beat intervals, 30000 have been randomly selected to make the plot easier to read. The upper dashed horizontal line denotes the average Emean

statistic across all subjects (the other line is its negation).

Figure 2: A 12-second signal excerpt. The low-frequency phenomenon with around 4 second period is respiration (blue oval). The heartbeat is the fluctuation that recurs

around every second (red circles).

Figure 1: A 10-second excerpt of synchronized ballistocardi-ography (top) and electrocardiography (bottom) signals.

Beddit Science.indd 2-3 29/08/14 13:14

www.beddit.com | Beddit Ltd. All rights reserved 2014. www.beddit.com | Beddit Ltd. All rights reserved 2014.

The established practice of medical sleep monitoring, poly-somnography, involves wearing multiple electrophysiological sensors for a single night, at a sleep laboratory or at home. It provides clinically valuable information, but is expensive and uncomfortable. More long-term and comfortable mea-surements can be done with actigraphy, where the overall sleeping patterns of a patient are measured with a wrist-worn movement sensor.

Beddit measures sleep with unobtrusive force sensors. The idea is to measure the forces caused by the body on the bed with a flexible film sensor that is placed below the bed sheet.

Each ECG heartbeat signal consists of a clear spike (the QRS complex), which is followed by an impulse in the BCG around 80 ms later. The ECG spike precedes the impulse in the BCG, because electrical activity causes the mechanical contraction of the heart.

Respiration

Beddit Sleep Monitor is also capable of measuring respira-tory activity as respiration causes the chest to move mea-surably. Figure 2 shows how the respiration cycles and heart rate events appear in the sensor signal:

There are three main motivations for measuring respiration unobtrusively during sleep. First, respiration conveys infor-mation about the general condition of the patient, so the de-terioration of health can be detected with respiration monitor-ing. Second, sleep-related breathing disorders (SRBD) such as sleep apnea represent a major share of sleeping prob-lems. Third, the structure of sleep can be analyzed based on respiration, because sleep stages have differing effects on respiration.

The respiration measurement is a 4-step process where first the parts of the signal that contain gross movements are dis-carded. Then the respiration signal is low-pass filtered on 4 distinct f Hz frequencies with potentially disturbing phenome-nons taken into account at around 2x f Hz. Therefore, at least one of the filters will result in an output signal that has the respiration frequency intact but the disturbance removed.

Then the respiration cycles are detected from each filtered signal. A respiration cycle begins at a local maximum and ends at the next local maximum in the signal. In addition, the amplitude of each respiration cycle is calculated by taking the difference between the signal value of the local maximum that starts the cycle and the minimal signal value in the cycle. Lastly, the final sequence of respiration cycle lengths is com-piled from the four signals based on the stability of respiration cycle amplitudes in each signal.

The correct signal is typically selected, because the signal that contains frequencies up to the respiratory frequency is more stable in its amplitude than a signal that also contains higher-frequency disturbing phenomena.

Beddit Sleep Monitor and the Science Behind It

→

The measurement methodology poses scientific challenges because physiological information (heart rate, respiration, etc.) that are vital for analyzing sleep cannot be readily ex-tracted from the sensor’s signal, but requires sophisticated signal analysis methods.

Ballistocardiography

The measurement of mechanical cardiac activity from the platform supporting the body is called ballistocardiography (BCG). Each time the heart beats, the acceleration of blood generates a mechanical impulse that can be measured with a proper force sensor, such as the Beddit Sleep Monitor.

Heart rate

Measuring the heart rate from BCG or similar mechanical signals is much more complex than measuring the heart rate using electrocardiogram (ECG), the most common-place cardiac measurement method. Individual heartbeats can be detected in an ECG signal relatively easily, by locat-ing a clear spike (called the QRS complex, from the con-secutive named spikes Q, R, S of the ECG heartbeat) that accompanies each heartbeat.

However, with BCG, the cardiac impulses are less pro-nounced and more variable than the salient shape of the QRS complex. The differences between the ECG and BCG-signals are shown in figure 1:

Movement

As sleep correlates with a low level of motility, circadian rhyth-micity can be estimated with a method called actigraphy. An accelerometer sensor is worn on the wrist 24 hours a day, which allows estimating the daily alternation between sleep and wakefulness. Due to its limited accuracy, actigraphy is typically used for the overall characterization of sleeping pat-ters over a period of at least a week.

The Beddit Sleep Monitor detects the gross movement of the person sleeping. Even if Beddit is not a medical device and shouldn’t be used for any kind of medical diagnoses, the excessive movement during the night could potentially be a sign of, for example, periodic limb movement disorder. In such a case, the user should be in contact with an appro-priate doctor.

The movement information is analyzed by detecting discrete events of movement from the BCG signal. That is done by dividing the high-pass filtered (cut-off frequency 5 Hz) sig-nal into three-second windows. Each window is detected as movement if the difference between signal minimum and maximum in the segment is above a fixed threshold.

Validity of the measurements

The measurement data provided by the Beddit Sleep Moni-tor has been tested against results from ECG and polysom-nography (PSG).

The heart rate data has been tested in a clinical study con-ducted during 2013 in co-operation with the VitalMed Re-search Centre together with the renowed sleep researcher and expert, Prof. Markku Partinen, M.D. The study consisted of 46 test subjects whose ages varied between 20 to 74 years. The study focused on measuring beat-to-beat heart rate. Out of all the detected beat-to-beat intervals, a max-imum precision of 99.94% was achieved with the Beddit Sleep Monitor compared to the reference ECG-signal.

As for the the low-frequency component of the signal record-ed by Beddit, it’s very similar to the abdominal respiration effort signal in PSG.

In the figure 4 above, the upmost signal is from an airflow pressure meter while the middle signal is from an abdomi-nal sensor band, both commonly used in PSG. As can be seen, the signal on the bottom that is produced by the Bed-dit Sleep Monitor is comparable with the reference signals from PSG. However, the final classification of any disorder typically requires a PSG measurement, with oximetry and measurement of airflow.

Also the gross movement signal recorded and analyzed by the Beddit Sleep Monitor has been tested against PSG as can be seen in figure 5.

Sources

Paalasmaa, J., Toivonen, H. & Partinen, M., 2014. Adaptive Heartbeat Modeling for Beat-to-Beat Heart Rate Measurement in Ballistocardiograms. IEEE Journal of Biomedical and Health Informatics. To appear

Paalasmaa, J., Waris, M., Toivonen, H., Leppäkorpi, L. & Partinen, M., 2012. Unobtrusive Online Monitoring of Sleep at Home. 34th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’12).

Paalasmaa, J., Leppäkorpi, L. & Partinen, M., 2011. Quantifying respira-tory variation with force sensor measurements. 33rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’11).

Paalasmaa, J., Partinen, M., 2012. A telemedicine solution based on a piezoelectric movement sensor for the long-term monitoring of sleep dis-order patients. Poster in 21st Congress of the European Sleep Research Society. Paris, France, September 4-8, 2012.

Figure 5: PSG-signal references against the movement data recorded by the Sleep Monitor.

Figure 4: PSG-signal references against the low-frequency component recorded by the Sleep Monitor.

Figure 3: Bland-Altman plot of the difference between the detected BCG beat- to-beat intervals and the ECG reference.

Out of the 770676 detected beat-to- beat intervals, 30000 have been randomly selected to make the plot easier to read. The upper dashed horizontal line denotes the average Emean

statistic across all subjects (the other line is its negation).

Figure 2: A 12-second signal excerpt. The low-frequency phenomenon with around 4 second period is respiration (blue oval). The heartbeat is the fluctuation that recurs

around every second (red circles).

Figure 1: A 10-second excerpt of synchronized ballistocardi-ography (top) and electrocardiography (bottom) signals.

Beddit Science.indd 4-5 29/08/14 13:14