See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/40452228
`
`Development of the Irregular Pulse Detection Method in Daily Life using
`Wearable Photoplethysmographic Sensor
`
`Article  in  Conference proceedings: ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference ·
`September 2009
`
`READS
`2,704
`
`Ken-Ichi Kameyama
`Teijin pharma
`
`21 PUBLICATIONS   205 CITATIONS   
`
`SEE PROFILE
`
`DOI: 10.1109/IEMBS.2009.5335401 · Source: PubMed
`
`CITATIONS
`15
`
`3 authors:
`
`Takuji Suzuki
`Toshiba Corporation
`
`22 PUBLICATIONS   310 CITATIONS   
`
`SEE PROFILE
`
`Toshiyo Tamura
`Waseda University
`
`386 PUBLICATIONS   5,021 CITATIONS   
`
`SEE PROFILE
`
`Some of the authors of this publication are also working on these related projects:
`
`Heat stroke monitored by Deep Body Thermometer View project
`
`Development of Wearable and Invisible sensors View project
`
`All content following this page was uploaded by Toshiyo Tamura on 04 February 2017.
`
`The user has requested enhancement of the downloaded file.
`
`APPLE 1016
`
`1
`
`

`

`31st Annual International Conference of the IEEE EMBS
`Minneapolis, Minnesota, USA, September 2-6, 2009
`
`978-1-4244-3296-7/09/$25.00 ©2009 IEEE
`
`6080
`
`Development of the Irregular Pulse Detection Method in Daily Life
`using Wearable Photoplethysmographic Sensor
`
`Takuji Suzuki, Ken-ichi Kameyama and Toshiyo Tamura, Member, IEEE
`
`Abstract— We developed an arrhythmic pulse detection
`algorithm from photoplethysmography (PPG) measured in
`daily life using a wearable PPG sensor, in order to provide a
`simpler device than a Holter electrocardiograph (ECG).
`However, PPG is very sensitive to artifacts in daily life, e.g. body
`movement.
`First, we analyzed the correlation between the ECG and the
`PPG measured at the same time when the arrhythmic heartbeat
`occurred in daily life. Using the correlation characteristics, we
`developed a detection algorithm of the arrhythmic pulse to
`distinguish the artifacts ascribable to body movement and
`evaluated its accuracy. The algorithm detects pulse-to-pulse
`interval (PPI) and pulse amplitude by a beat to distinguish
`between irregular PPI by arrhythmic pulse and that by the
`artifact.
`
`A
`
`INTRODUCTION
`I.
`RRYTHEMIA is closely related to various cardiovascular
`diseases, including lethal ones, such as ventricular
`fibrillation, that can cause of sudden death. So it is important
`to detect and diagnose it early. In particular, real-time alert
`and diagnostic services for arrhythmia in daily life are
`required that apply recent developments in IT to provide
`preventive healthcare.
`In conventional clinical practice, electrocardiography
`(ECG) at hospital is used for diagnosis of arrhythmia.
`However, as the irregular heartbeat caused by arrhythmia
`does not necessarily occur during examination at hospital, it
`is necessary to measure ECG in daily life. A Holter ECG is a
`small device for measuring one or two channels of an ECG
`during 24 hours in daily life. The patient suspected of having
`arrhythmia can be diagnosed using the Holter ECG. There are
`some works concerning auto detection of the arrhythmia from
`Holter ECG [5][6]. However, since it is necessary to attach
`some electrodes to the patient’s chest in the case of the Holter
`ECG, it is troublesome to check arrhythmia using a Holter
`ECG for preventive healthcare in daily life. Therefore,
`development of a heart rate monitor that is simpler and easier
`to use than the Holter ECG is desirable.
`Such a heart rate monitor would also be useful for checking
`the change of a physical condition of an artificial dialysis
`patient during dialysis and to check the stress condition of a
`normal person by autonomic analysis of the heart rate
`
`Takuji Suzuki, Ken-ichi Kameyama
`Corporate Research and Development Center Toshiba Corporation,
`Kanagawa, Japan
`Email: takuji1.suzuki@toshiba.co.jp
`Toshiyo Tamura
`Graduate School of Engineering, Chiba University, Chiba, Japan
`
`variability.
`In this study, we use a PPG sensor, which is a simpler
`device than the Holter ECG. Using the PPG sensor, our goal
`is to detect the arrhythmic pulse with the same accuracy as the
`arrhythmic heartbeat is detected by the Holter ECG. First, we
`analyzed the correlation between the ECG measurement and
`the PPG measurement performed at the same time when the
`arrhythmic heartbeat occurred. Using
`the correlation
`characteristics, we developed a detection algorithm of the
`arrhythmic pulse for premature beat and evaluated its
`accuracy. PPG is more sensitive to body movement than ECG.
`The algorithm reduces the influence of the body movement.
`
`II. PHOTOPLETHYSMOGRAPHY OF IRREGULAR HEARTBEAT
`There are 8 kinds of arrhythmia according to the Minnesota
`code that is widely used in the clinical field [1]. Since atrial or
`junctional premature beat (8-1-1), ventricular premature beat
`(8-1-2), atrial fibrillation / atrial flutter (8-3), supraventricular
`tachycardia
`intermittent (8-4-2), sick sinus syndrome
`(sinoatrial
`arrest,
`sinoatrial
`block)
`(8-5),
`sinus
`tachycardia(8-7), and sinus bradycardia(8-8) are detectable
`on the basis of RR intervals, they are the target of this study.
`Pulse-to-pulse interval (PPI) by PPG can be considered the
`same as R-R interval (RRI) by ECG provided the subject’s
`physiological state is static.
`However, in the case of RRIs with irregular intervals
`brought about by these 8 kinds of arrhythmia, pulse amplitude
`of the PPG varies. When the irregular interval is shorter than
`the normal interval, the amplitude of the irregular pulse is
`smaller than that of the normal pulse. This is because the
`duration of ventricular diastole is shorter owing to the shorter
`RRI, causing blood volume congested in the left ventricle to
`decrease, and consequently, stroke volume to decrease. In
`contrast, when the irregular interval is longer than the normal
`interval, the amplitude of the irregular pulse is larger than that
`of the normal pulse [4].
`In developing the irregular pulse detection algorithm, it is
`necessary to consider the change of the pulse amplitude
`owing to the irregular pulse.
`
`III. DATA ACQUISITION
`For confirmation of the characteristics of the PPG of
`irregular pulse described above, we measured ECG and PPG
`simultaneously for a person suspected to have arrhythmia.
`
`2
`
`

`

`6081
`
`
`
`
`
`Fig.3 also shows the waveform of the irregular pulse of the
`second subject who has atrial premature beat.
`
`
`
`Fig. 3. ECG and PPG waveforms. (including irregular heartbeat by atrial
`premature beat [arrow])
`
`
`Fig.4 shows the waveform of the irregular pulse of the third
`subject who has ventricular premature beat.
`
`
`
`
`
`Fig. 4. ECG and PPG waveforms. ( including irregular heartbeat by
`ventricular premature beat [arrow])
`
`
`There are common characteristics in the three PPG
`waveforms. The characteristics are
`the same as
`the
`assumption described above. Meanwhile, the baseline of the
`PPG waveforms during the body movement fluctuates
`widely.
`
`A. Devices
`We used a polygraph system (Polymate AP1124, TEAC
`Corp., Japan) for measuring ECG, the sampling frequency of
`the ECG is 200Hz. We also used a prototype of a wearable
`sensor that is wrist-watch style and has a reflective PPG
`sensor (wavelength of 525 nm) in the sensor head worn on the
`finger for measuring pulse wave and 3-axis acceleration
`sensor (H34C, Hitachi, -3 to +3G) in the sensor body for
`measuring body movement (Fig.1)[2][3]. The sampling
`frequency of the PPG is 64Hz. The clock between ECG and
`the wearable sensor were synchronized.
`B. Procedure
`We measured the 2-lead ECG and PPG for three subjects
`suspected to have arrhythmia during one normal night of
`sleep. PPG was measured from the index finger. The
`experimental procedure was explained, and written informed
`consent was obtained from all subjects.
`
`Fig. 1. Wearable photoplethysmographic sensor
`
`
`
`IV. RESULTS OF THE DATA AQUISITION
`Fig.2 shows the waveform of the irregular pulse of the first
`subject who has atrial premature beat.
`
`
`Fig. 2. ECG and PPG waveforms. (upper: during body movement, lower:
`including irregular heartbeat by atrial premature beat[arrow])
`
`
`In the case of an irregular pulse, AIR is almost a constant
`value. In contrast, in the case of artifact, AIR is varied by an
`
`V. IRREGULAR PULSE DETECTION ALGORITHM
`Using the characteristics described above, we developed
`an irregular heartbeat detection algorithm. Fig.5 shows the
`flowchart of the algorithm.
`First, pulse trigger point (falling-edge) is detected by
`finding the crossing point of the internally dividing point
`between maximum and minimum value of the differential of
`the pulse wave during the prior 1 second, and then PPI is
`calculated by getting the interval from the previous pulse
`trigger point and the present point. Irregular PPI is detected
`from the change of the PPI. Also, we defined an index AIR
`(Amplitude / Interval Ratio) to distinguish an irregular pulse
`and an artifact as follows. AIR for the irregular PPI is
`calculated.
`
`AIR
`
`
`=
`
`
`
`tudePulseAmpli
`
`/
`
`PPI
`
` (1)
`
`3
`
`

`

`6082
`
`
`
`
`
`artifact because the artifact does not have regularity.
`The set value 1 and 2 were set empirically from measured
`three subject’s data described below.
`
`
`
`
`
`
`Fig. 5. Flowchart of the irregular heartbeat detection algorithm.
`
`
`
`
`VI. EVALUATION OF THE ALGORITHM
`We evaluated the accuracy of this algorithm using sample
`data. Fig.6 shows the detection result of atrial premature beat
`data (shown as Fig. 3). Large dot indicates the irregular pulse
`detection. This result shows that the irregular pulse detected
`by this algorithm is coincident with the irregular heartbeat
`detected from ECG. However, short-interval irregular pulse
`could not be detected because its amplitude was too small to
`detect pulse trigger.
`Fig.7 shows the detection result of body movement data
`(shown as Fig.2). Large dot indicates the error detection (not
`irregular heartbeat). This result shows that the irregular
`heartbeat detection algorithm can discriminate between pulse
`wave of irregular heartbeat and pulse wave with an artifact
`(e.g. body movement).
`Also, we evaluated the algorithm using all the data
`gathered in the acquisition described above during sleep.
`Table 1 shows the detection results for the three subjects.
`The number of the artifact detected from the PPG is 47 for
`the three subjects. Almost artifact case was rolling over
`during sleep.
`
`
`
`
`
`Fig. 6 Results of irregular heartbeat detection (upper: ECG and pulse
`waveforms; middle: ECG R wave trigger detection to get R-R intervals; Pulse
`trigger detection to get pulse intervals. DIFF ECG/Pulse means differential of
`ECG / pulse wave. ECG / Pulse threshold means internally dividing point
`between maximum and minimum of ECG / pulse waveform.)
`
`
`
`Fig. 7 Results of irregular heartbeat detection from body movement data
`(upper: ECG and pulse waveforms; middle: ECG R wave trigger detection to
`get R-R intervals; Pulse trigger detection to get pulse intervals. DIFF
`ECG/Pulse means differential of ECG / pulse wave. ECG / Pulse threshold
`means internally dividing point between maximum and minimum of ECG /
`pulse waveform.)
`
`
`
`4
`
`

`

`
`
`View publication statsView publication stats
`
`6083
`
`TABLE I
`Irregular heartbeat detection results
`Photoplethysmography
`Normal
`
`
`
`Irregular
`Normal
`Sum
`
`Irregular
`161
`123
`284
`
`15
`12214
`12229
`
`
`Accuracy
`Sensitivity
`Specificity
`False negative rate
`False positive rate
`Positive predictive value
`Negative predictive value
`
`
`
`[4]
`
`[3] T. Suzuki, K Ouchi, A. Moriya, K. Kameyama, M. Takahashi,
`“Development of a sleep-stage Estimation Method using Heart Rate
`Variability and Actigraphy measured by Wearable Sensor”, Sleep and
`Biological Rhythm, 5 Suppl 1, p.A38, 2007
`J. A. Pollard, “Cardiac arrhythmia and pulse variability: a
`plethysmographic study”, Anesthesia, vol.25, No.1, pp.63-72, 1970
`[5] K. Shin, YH. Kim, JP. Kim, JC Park, “The preliminary study on the
`clinical application of the WHAM (Wearable Heart Activity Monitor).”,
`Proceedings of the 28th IEEE EMBS Annual International Conference,
`pp.6034-6036, 2006.
`I. Mohamed Owis, H. Ahmed, M.Abou-Zied, Abou-Bakr Youssef,
`Yasser M. Kadah, “Study of Features Based on Nonlinear Dynamical
`Modeling in ECG Arrhythmia Detection and Classification”, IEEE
`TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 49, NO. 7,
`pp.733-736, 2002.
`
`[6]
`
`Sum
`
`176
`12337
`12513
`
`
`0.989
`0.915
`0.990
`0.085
`0.433
`0.567
`0.999
`
`
`
`
`
`
`
`
`
`ECG
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`VII. DISCUSSION
`Since, as shown in Table 1, this algorithm has sufficiently
`high specificity, we confirmed the feasibility of detecting
`irregular pulse without misdetection of body movement data
`using the characteristics of the pulse amplitude fluctuation.
`Pulse trigger detection of this algorithm uses pulse
`amplitude of every pulse wave employing constant width
`time window (1 second width). So, in the case that the
`fluctuation of the pulse amplitude is large, the pulse trigger
`cannot be detected. However, this algorithm can detect the
`irregular pulse of the premature beat robustly using the
`pattern that the next interval of short interval is longer
`interval.
`The index AIR includes the influence of pulse amplitude
`fluctuation by respiration and autonomic nervous system. It is
`necessary to distinguish the irregular pulse from these
`influences.
`
`VIII. CONCLUSION
`We developed an irregular pulse detection algorithm using
`photoplethysmography. We confirmed the feasibility of
`detecting irregular pulse without misdetection of body
`movement data using the characteristics of the pulse
`amplitude fluctuation. In future work, we intend to apply and
`evaluate this algorithm for data on various situations in daily
`life.
`
`REFERENCES
`[1] R. J. Prineas, R. S. Crow, Minnesota Code Manual of
`Electrocardiographic Finding, John Wright-PSG, Inc. Littleton, MA,
`June 1982.
`[2] T. Suzuki, K Ouchi, A. Moriya, K. Kameyama, M. Takahashi,
`“Development of a sleep monitoring system with wearable vital sensor
`for home use”, Proceedings of International Conference on Biomedical
`Electronics and Devices, pp.326-331, 2009
`
`5
`
`