J Med Syst (2015) 39:191 DOI 10.1007/s10916-015-0339-7
SYSTEMS-LEVEL QUALITY IMPROVEMENT
The Effect of Electrode Designs Based on the Anatomical Heart Location for the Non-Contact Heart Activity Measurement Sun Ok Gi 1 & Young-Jae Lee 2 & Hye Ran Koo 1 & Seung Pyo Lee 1 & Kang-Hwi Lee 2 & Kyeng-Nam Kim 2 & Seung-Jin Kang 2 & Joo Hyeon Lee 1 & Jeong-Whan Lee 2
Received: 14 November 2014 / Accepted: 7 September 2015 # Springer Science+Business Media New York 2015
Abstract This research is an extension of a previous research [1] on the different effects of sensor location that is relatively suitable for heart rate sensing. This research aimed to elucidate the causes of wide variations in heart rate measurements from the same sensor position among subjects, as observed in previous research [1], and to enhance designs of the inductive textile electrode to overcome these variations. To achieve this, this study comprised two parts: In part 1, X-ray examinations were performed to determine the cause of the wide variations noted in the findings from previous research [1], and we found that at the same sensor position, the heart activity signal differed with slight differences in the positions of the heart of each subject owing to individual differences in the anatomical heart location. In part 2, three types of dual-loop-type textile electrodes were devised to overcome variations in heart location that were confirmed in part 1 of the study. The variations with three types of sensor designs were compared with that with a single-round type of electrode design, by using computer simulation and by performing a t-test on the data obtained from the experiments. We found that the oval-oval shaped,
This article is part of the Topical Collection on Systems-Level Quality Improvement. * Joo Hyeon Lee
[email protected] * Jeong-Whan Lee
[email protected] 1
Research center for Textile & Fashion, College of Human Ecology, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-749, South Korea
2
Department of Biomedical Engineering, College of Biomedical & Health Science, Konkuk University, 268 Chungwondaero, Chungju-si, Chungcheongbuk-do 380-701, South Korea
dual-loop-type textile electrode was more suitable than the single round type for determining morphological characteristics as well as for measuring appropriate heart activity signals. Based on these results, the oval-oval, dual-loop-type was a better inductive textile electrode that more effectively overcomes individual differences in heart location during heart activity sensing based on the magnetic-induced conductivity principle. Keywords Heart location . Heart activity sensing . Magnetic-induced conductivity sensing method . Enhanced designs for the inductive textile electrode
Introduction Medical care costs have increased sharply as the society is aging, and people now have high expectations in terms of the quality of life they lead. To reduce medical costs while maintaining a good quality of life, numerous non-invasive and non-restraining wearable health care systems have been developed. In this context, many different textile-based electrodes and wearable platforms for daily health monitoring intended for long-term use have also been developed. In recent years, a wearable heart rate sensing systems have various forms, such as wristwatches, wristbands, electronic patches, chest bands, shirts, shoes, and even gloves, chairs, and beds [2–4]. Among these, systems in clothing form have been regarded as the least invasive and least restraining platform [2]. In this context, various types of garment-based heart activity sensing systems using textile electrodes, have been devised and evaluated in the field of u-health care, military system, sports, etc. [5–9]. Clothing designed to detect heart rates can be categorized into two types depending on the sensing method. Early studies
191
Page 2 of 17
using contact-type sensing method focused on acquiring heart activity signals through direct contact between the wearer’s skin and the textile electrode, which generally consisted of a conductive textile material such as stainless steel yarn, silver yarn, metal-plated fabric, or a fabric coated with carbon nanotubes [10–13]. Later, these electrodes evolved into a sandwich structure consisting of new materials such as conductive rubber or a conductive polymer. However, regardless of the material used, continual contact between the textile electrode and the wearer’s skin is crucial for contact-type heart-rate sensing clothing, as the signal quality declines if the contact spot shifts. Therefore, for contact-type heart-rate sensing clothing, extremely tight-fitting heart-rate sensing garments were used, in which the contact position between the textile electrode and the wearer’s skin could be maintained by the strong pressure due to the extreme tightness of the garment, [14]. However, discomfort during movement and breathing due to the excessive tightness are the reported drawbacks of the contact-type sensing, as well as has skin irritation at the contact spot. Therefore, to overcome these limitations of the contact-type electrodes, non-contact-type electrodes that measure heart rate have been considered as alternatives [1]. The non-contact type sensing methods are characterized by the principle on which they work. There are capacitive and magnetically induced conductivity sensing methods. The capacitive sensing method measures variation in the capacitance of an electrode when attached to clothing, relying on another virtual electrode in the heart of the wearer. Although the capacitive sensing method is advantageous for obtaining heart activity signals with a larger electrode, a large electrode leads to inferior sensing quality, as it is mostly accompanied by an increased amount of uncontrollable measurement noise resulting from motion artefacts created by body movements [15]. On the other hand, the magnetically induced conductivity sensing method has emerged as an alternative non-contacttype heart-rate measurement method. This method adopts a coil-shaped electrode that induces a magnetic field by permitting an alternating current. The induced magnetic flux that penetrates the volume of interest, i.e., the heart, reflects changes in the conductivity around the heart triggered by its ventricular contraction, which is finally detected as the inductance change in the coil electrode [16]. Recently, this method has been applied to clothing, chairs, and beds to measure heart rate or respiration rate in patients [1, 3, 17–19]. Our present study focused on the non-contact type heart activity measurement based on the magnetically induced conductivity sensing principle. Our previous studies based on this non-contact type method have been gradually progressed as follows: Initially, in our previous study, we examined the feasibility of the textile inductive sensor for the non-contact type heart rate sensing [20]. The results of that study showed that the use of textile inductive sensor is feasible for heart rate sensing.
J Med Syst (2015) 39:191
Besides feasibility, the physical properties of the inductive textile sensor were also suitable for heart rate sensing. Second, we observed the effects of the shape of the textile inductive sensor on the measurement of heart activity [17]. In that study, we found that the entire shape of the textile inductive sensor dominantly influenced heart rate sensing performance than the central shape of the sensor. Third, we examined the effects of the positions of the textile inductive sensor placed on the surface of a sensing garment on the quality of heart rate measurements [1]. In that study, the feasibility of eight candidate positions of an inductive sensor in a specially designed heart-rate-sensing garment was explored. Among the eight sensor positions in that study, the sensor position labelled BP3^ located 3 cm from the center front point on the chest circumference line, showed relatively high efficiency. However, the result of that study also highlighted the wide variation in the measurement efficiency at the same position BP3^ among five subjects. In other words, compared with other sensor positions, although the sensor position BP3^ was consistently superior in its sensing result in every subject, some notable variations in these results from BP3^ were found across the subjects. On this basis, we inferred that this phenomenon may be associated with individual variations in the location of the heart. On the basis of the results from these previous studies, in this study, we aimed to elucidate the causes of the wide variations in heart rate measurement using the same sensor position BP3^ highlighted in the third study [1], and to enhance designs of the inductive textile electrode to overcome these variations. To achieve these objectives, this study was performed in two parts. In part 1, we performed X-ray examinations to determine the causes of the wide variations in heart rate measurements from BP3^. In part 2, based on the results from part 1, we devised three types of the textile electrode designs to overcome individual differences in the heart location and compared the effects of the three types of the electrode designs with those of a single round type design that was used in the previous study [1]. Principle of the heart activity sensing based on magnetically induced conductivity measurements This research is based on the principle of magnetically induced currents. When AC current is applied to a spiral coil, a time-varying magnetic field is generated. The magnetic flux then penetrates the surrounding materials. If there is any conductive material near the spiral coil, the magnetic flux induces a current, also known as an eddy current, inside the material. The eddy current creates a secondary magnetic field that is opposite to the primary magnetic field (Lenz’s law) [21]. This affects the inductance of the spiral coil, as follows: As the reinduced magnetic field is related to the impedance distribution
J Med Syst (2015) 39:191
inside the body, changes in the impedance directly affect the re-induced current. The resulting induced magnetic flux can be assumed to be a linkage flux in a coupled coil composition. Therefore, heart activity can be measured as a variation of coil inductance (Fig. 1) [2, 4, 19, 22, 24]. In our research, we applied this principle to body tissue. In other words, when the magnetic field formed from the textile electrode placed on the garment, which was materialized by computerized embroidery method in flat swirl shape, penetrates the garment wearer’s body, an eddy current is induced near the heart, as the heart is the most conductive tissue in the thorax (Fig. 1). Next, a re-induced magnetic field is generated by the induced eddy current according to the aforementioned Lenz’s law [21]. Changes in impedance distribution in the body in the re-induced magnetic field triggered by the heart activity, affect the re-induced current. As a result of the linkage flux, heart activity signals can be detected by inductance changes in the inductive textile sensor on the surface of the garment [1, 19, 22–24]. Based on this measurement principle, the heart activity measurement system in this study consisted of a textile inductive sensor; a knit shirt serving as the wearable platform; hardware including a sensing module, a transmission unit, and a battery; and snap-type interconnections, the same as those used in the previous study (Fig. 2) [1]. Part 1: Inspection of individual variations in the location of the heart Background The study of Koo et al. [1] investigated the position of a new heart-rate sensing method based on the principle of magnetically induced conductivity and focused on one main factor influencing the measurement results, the effect of the sensor position. For this, Koo et al. devised a spiral-shaped inductive Fig. 1 The basic principle of magnetic-induced conductivity measurements (Lee et al., World of Electricity, 2013 [25])
Page 3 of 17 191
textile sensor made of a metal hybrid conductive yarn and applied it to a heart-rate-sensing shirt (Fig. 2). With this sensing shirt, they examined the feasibility of eight candidate positions of the inductive sensor (Fig. 3) consisting of six lead positions based on conventional electrocardiographic (ECG) measurement positions and two positions frequently used in earlier heart-rate-sensing clothing. For the quantitative evaluation of the effect of the textile-based inductive sensor positions on heart rate measurements, Koo et al. defined a new quality index (QI), which was based on the morphological repeatability of the measured signals. Their results showed that the sensor position BP3^ among the eight candidate positions, which was located 3 cm from the center front line on the chest circumference line, had the highest QI value and the highest morphological association with R-peaks on ECG readings, which were simultaneously measured in all subjects. Therefore, the position BP3^ sensor position that is based on the magnetically induced conductivity principle showed the highest efficiency. At the same time, another important point regarding the BP3^ position was noted. Compared to the other positions, the sensor position P3 consistently showed a superior sensing result in all subjects with regard to signal magnitude as well as the quality index of the signal. Meanwhile, when the sensing result data from BP3^ were analyzed in all subjects, some notable variations in the QI value were observed among the subjects. Based on these findings from the previous study, we inferred that the wide variation in the measurement at the same sensor position among subjects might stem from individual variations in the heart location. In fact, the sensor position P3 was determined by measuring 3 cm from the center front point on the chest circumference line to the left side. Thus, the BP3^ position was determined only by an external criterion that was not essentially associated with the heart position
191
Page 4 of 17
J Med Syst (2015) 39:191
Fig. 2 Experiment setup of the garment-type wearable system and signal processing block diagram (Koo et al., The effect of textile-based inductive coil sensor positions for heart rate monitoring,38:2, 2014 [1])
in vivo. Hence, we assumed that subjects showing relatively low efficiency levels during heart activity sensing according to the BP3^ position may have a slightly different heart position in vivo. Consequently, to verify this inference, we investigated individual variations in heart positions across subjects using X-ray examinations in the first stage of the present study. The method and results of X-ray examination With regard to X-ray examinations used to investigate individual variations in heart positions across subjects, a grid plate was devised for use as a standard to compare X-ray images of the subjects. On the plate, the BP3^ position in the previous study by Koo et al. [1] was marked using a copper tape in a corner of the grids, which were spaced 3 cm apart near the heart. X-ray images with the BP3^ position thus marked in the
Fig. 3 The eight candidate positions of the inductive sensor (Koo et al., The effect of textile-based inductive coil sensor positions for heart rate monitoring, 38:2, 2014 [1])
six subjects were obtained by capturing the images with the grid plate attached to the chest. As the outline of the part of the atrium in the X-ray image appeared unclear due to overlap with other organs in the body, the X-ray examination was confined to the ventricle, which was acceptably clear and recognizable with unaided eyes. In every case, a reversed BL^ shape was drawn from the partial outline of the ventricle, and the distance between the folding point in the reversed BL^ and BP3^ marked by the copper tape was then measured (Fig. 4). In this way, the relative position of BP3^ in the heart of every subject was obtained. Two analysis methods were performed on the X-ray images to extract variations in the heart location among subjects. In the first method, we used the clavicle and the cervical vertebrae as the criterion position for the analysis of the variations in heart locations among the subjects. The intersection line between the clavicle and the cervical vertebrae and the reversed BL^ line, as drawn on the X-ray image, was
Fig. 4 The grid in the X-ray inspection
J Med Syst (2015) 39:191
connected and then configured into a square. By drawing squares on the X-ray images of each subject, a total of six squares were obtained. It was assumed that in the absence of variation in the heart location among the six subjects, the six positions of BP3^ as well as the six squares would perfectly overlap when these would be superposed in the same plane. As shown in Fig. 5, the relative differences in the BP3^ positions in the vertical direction in the subjects were visualized when the six squares were superimposed at the point of the intersection between the clavicle and the cervical vertebrae (Fig. 5). The symbol B◆^ in Fig. 5 indicates the relative distance from the intersection between the clavicle and the cervical vertebrae to BP3^ in the heart of each subject. The color of the symbol matching that of the reversed BL^ line indicates that both were drawn from the same subject. However, this method is limited when used to illustrate differences in subjects’ heart positions only in the vertical direction. The second method of analysis aimed to determine the relative differences in the distance between the BP3^ position and the reversed BL^-shaped outline of the ventricle by overlapping the BP3^ positions obtained from each subject. In Fig. 6, the symbol B◆^ indicates the difference in the relative distance from the reversed BL^ to the overlapped BP3^ in the
Page 5 of 17 191
Fig. 6 An illustration of the variation in the heart position in the horizontal and vertical directions. *‘◆’ denotes the overlapped P3 position of each subject while the square-lined boxes indicate outlines of the ventricle
heart in all subjects. In this way, variations in heart locations among the subjects in the vertical and horizontal directions could be visualized, as shown in Fig. 6. Based on the results of the second analysis method, individual variations in heart locations were derived as follows: In the horizontal direction, the longest distance between BP3^ and the outline of the ventricle was 1.7 times the shortest distance, while the greatest distance was 2.6 times the shortest distance in the vertical direction. These results imply that heart activity from BP3^ in each subject was actually sensed at different positions in the heart owing to individual variations in the heart locations depending on the subject, even though BP3^ was physically determined at the same position on the body surface in all subjects. Part 2: An approach to an enhanced design of the inductive textile sensor
Fig. 5 An illustration of the individual variations in heart positions in the vertical direction. *‘◆’ denotes the P3 position of each subject, while square-lined boxes indicate outlines of the ventricle
We aimed to obtain the enhanced design of the textile inductive electrode based on the magnetic-induced conductivity principle for heart rate sensing; for this, we performed two sequential experiments. In part 1, we found out the cause of the wide variations in the sensing results from the same sensor position on the surface of a garment observed in subjects [1]. Based on the cause of the aforementioned individual variations in part 1, new types of electrode designs that were expected to resolve this problem were determined in part 2 of
191
Page 6 of 17
this research. The effects of these were evaluated by comparing the results to those of a single round type electrode that was used in the previous study [1], by using computer simulation and by performing a set of experiments in human subjects. Designs and characteristics of the inductive textile electrode Materials A type of polyester-metal hybrid thread was developed for the textile electrode through a yarn treatment process. Each yarn unit of the conductive thread is a combination of two heterogeneous components, i.e., polyester and silverplated nickel yarn. In detail, the conductive thread that was composed of multiple filaments of polyester (75 denier per thickness in a single filament) and eight lines of silver-plated nickel yarn, consisting of six lines of 30 μm yarn and two plies of 50 μm yarns, was fabricated by a twisting method. Thus, a single strand of the conductive thread consisted of nine yarn units, and the electrical resistance of the conductive thread for textile electrodes was 0.234 Ω/m. Owing to the combination of the polyester and the metal yarn, the conductive thread developed in this study had appropriate flexibility and durability to undergo a post-fabrication process. New designs of an inductive textile electrode New designs of the inductive textile electrode were devised in this study in order to overcome variations in the measurement efficiency levels in the aforementioned BP3^ position among subjects in the earlier study [1]. Three types of dual-loop-type textile electrodes were created to induce a deliberately distorted form of magnetic field aimed to include areas of individual variations of the heart positions. All of the textile electrodes in this study consisted of ten turns of spiral lines on a polyester fabric sheet by a computerized embroidery machine. The distances between the spiral lines were controlled to be within 1– 1.5 mm, as it was difficult to control the line-to-line distance and ensure that it was identical to 1 mm in all cases owing to the limitations of the computerized embroidery method. Characteristics The characteristics of the single round textile electrode used in the previous study of Koo et al. [1] and the three dual-loop textile electrodes, termed the Bround-round,^ the Bround-oval,^ and the Boval-oval^ electrodes, are shown in Table 1. Measurements were taken by a network analyzer (HP8735D, Agilent, USA). A comparison of the textile electrode types using computer simulation To measure the formed magnetic flux subjected by each electrode, a simulation program (COMSOL Multi-physics 4.4, USA) was used for calculations by considering the method in which the flux was formed in the body. The program
J Med Syst (2015) 39:191
utilized in this study is the AC/DC Model of COMSOL Multi-physics 4.4, and the specific conditions are listed as Table 2. In this simulation, the red streamlines indicate the isointensity lines of the magnetic field in the finite domain, representing only the magnetic field formed; they have no direct relationship with the intensity and density of the magnetic field. During simulation, all the domains for the finite-element analysis were split into meshes. Through this process, the curves in all domains for the finite-element analysis were converted into a combination of straight lines. Owing to the limitation of this conversion method when translate curves are used instead of a set of straight lines, some errors and misalignments of the reconstructed geometric shapes are unavoidable. As infinite alternations of the calculation are required to resolve this problem using the COMSOL Multi-physics program, the simulation is assumed to be complete when the error approaches a designated threshold. This is regarded as a limitation of this simulation method. The simulated results of each electrode are as follows: ① Single round electrode For the single round electrode, all magnetic fluxes passed out through the center of the electrode. When this occurred, a symmetric flux formation in all directions was observed from the zero point (Fig. 7). ② Round-round electrode For the round-round electrode, as the directions of the currents applied to the two electrodes are opposite, the polarity of the magnetic field is also opposite. The induced magnetic flux formed in one electrode adjacent to the other one turns to the center of the neighboring electrode. The remainder of the magnetic loop is simulated such that it turns to the origin electrode (Fig. 8). In the round-round electrode, the effective area of the induced magnetic flux is therefore expected to be broader than that of the single round electrode. ③ Round-oval electrode The round-oval electrode consists of two heterogeneously shaped electrodes: one round and one oval. The angle between the longer diameters of both electrodes was normal. Owing to the longer wire length of the oval electrode than that of the round electrode, the magnetic flux from the oval electrode was skewed to the round electrode side, where the strength of the magnetic flux is relatively weak (Fig. 9). In this way, the coupled magnetic field becomes stronger on the oval electrode side than that on the round electrode side. ④ Oval-Oval electrode As in a round-round electrode, the induced magnetic flux formed in the one part of the electrode adjacent to another part turns to the center of the neighboring
J Med Syst (2015) 39:191 Table 1
Page 7 of 17 191
The four types of embroidered textile electrodes and details: (a) Single-Round, (b) Round-Round, (c) Round-Oval, and (d) Oval-Oval
electrode. The remaining part of the magnetic loop is simulated to turn to the origin electrode (Fig. 10). In the part of the magnetic flux where the two electrodes are adjacent, the magnetic field strength increases owing to magnetic coupling compared to that on the opposite side. In this electrode, the effective area of the induced magnetic flux is simulated such that it is broadened as compared to that in the single round electrode. From the computer simulation results, it was expected that the magnetic flux in the three types of dual-looptype textile electrodes would be more widespread than that from the single round electrode, and that the magnetic flux would be deformed because they were concentrated in specific parts of the coupled magnetic field.
The experiment An experiment with human subjects was performed in a laboratory environment. Twenty-five male subjects in their twenties participated in the experiment. All were in good Table 2 Conditions of the simulation
health and were arbitrarily sampled. They were asked to sit in a chair while the heart activity was measured for 10 s. Their heart activity signals were obtained using all types of textile electrodes for each subject in this study. The effects of the four types of textile electrodes, the single-round, round-round, round-oval, and oval-oval types, were examined after being attached to the left part of the chest of a sleeveless knitted sports shirt (Fig. 11). The signals from the four types of textile electrodes were simultaneously acquired from an ECG lead II configuration, and the respiration signal was acquired from a belt-type respiration sensor (Biopac System Inc., Model MP150 RSPEC-R Wireless Transmitter). These signals were then compared. The normalized arithmetic mean and the normalized standard deviation of the heart activity signal magnitudes were derived from the original data of the heart activity from the subjects. Based on the resulting normalized arithmetic mean and the normalized standard deviation of the heart activity signal magnitudes, the QI values defined in the section of BComparison of the quality index depending upon the type of electrode^ for each of the three types of inductive textile electrodes were calculated.
Material Properties
Physical conditions of the magnetic field
Material Features Relative permeability Relative permittivity Electrical conductivity Ampere’s law Magnetic insulation Edge Current
Air Copper* 1 1 1 1 0 S/m 6e7 S/m Applied to all domains All boundaries 20 mA, 1 MHz, Opposite direction for each electrode
*Copper was chosen as the material for the simulation considering that copper is the most similar in terms of permeability to the conductivity of metal-polyester hybrid yarn used in this study
191
J Med Syst (2015) 39:191
Page 8 of 17
Fig. 7 Streamline of magnetic field of the single round electrode in a normalized unit
(a) x-z plane
(b) z-y plane
(c) Three-dimensional plane
Morphological comparison of the heart activity signals obtained from the four electrodes The morphological characteristics of the resting heart activity signals from the four textile electrodes in this study are as follows (Figs. 12, 13, 14 and 15). Overall, the signals from the four textile electrodes were generally associated with the QRS-complex of the ECG lead II signals, but with a slight delay. This delay of the signal from the textile electrodes in comparison with the QRS-complex on ECG, was assumed to be attributed to the contraction and expansion of the heart, that is caused by the electrical activity of the heart [25]. The textile electrodes in this study represent mechanical activity whereas the ECG lead II signals detect the electrical activity of the heart. As shown in Figs. 12, 13, 14 and 15, the morphological quality of the heart activity signals from the three dual-looptype textile electrodes shows comparatively sharper peaks and clearer shapes than those from the single round electrode.
Another result from Figs. 12, 13, 14 and 15 is that the fluctuating amplitudes of the measured heart activity signals from the inductive textile sensors are thought to reflect respiration rate, the same as that reported in our previous study [1]. It was inferred that the amplitude of the measured heart activity signal was enhanced when the respiratory signal increased during inhalation, whereas the amplitude of heart activity signal reduced, when the respiratory signal decreased during exhalation. Koo et al. [1] reported that the amplitude of the heart activity signals has a very high correlation with respiratory signals, and inferred that the observed fluctuation of the amplitudes in the measured heart rate signal may be associated with lung movements while breathing. Comparison of the signal magnitude depending on the electrode type Table 3 shows a comparison of heart activity signal magnitudes obtained from the four textile electrodes. In order to
J Med Syst (2015) 39:191
Page 9 of 17 191
Fig. 8 Streamline of magnetic field of the round-round electrode in a normalized unit
(a) x-z plane
(b) z-y plane
(c) Three-dimensional plane minimize the intervening effect of the individual differences, heart activity signal magnitudes were converted into normalized values, by drawing a ratio of the original signal magnitude value to the averaged value from the four electrodes in each subject. Table 3 shows the normalized arithmetic means of the signal magnitude in the order of the round-oval (1.1112 V), oval-oval (1.0561 V), round-round (0.9806 V) and single round (0.8520 V) electrodes. Therefore, three dual-loop-type textile electrodes resulted in better quality signal in terms of the normalized average of the signal magnitude, in comparison with those from the single-round electrodes. This result is associated with aforementioned computer simulation results, that show that the magnetic flux in the three types of dual-loop-type textile electrodes was more widespread than that from the single-round electrode and that the magnetic flux in the three types of dualloop-type textile electrodes are deformed because they are concentrated on specific parts in the coupled magnetic field.
Hence, this result indicates that the dual-loop-type textile electrode produces larger signal magnitudes than the single-round electrode due to the wider distribution of the coupled magnetic fields. Table 3 shows a relatively wide range in standard deviation of each electrode of each subject (Fig. 16). This indicates that the amplitude of the measured heart activity of each electrode per subject showed wide fluctuations according to any influence other than the cardio-related activity. Koo et al. [1] indicated that this unique pattern is caused by respiration, as this tendency follows the respiration cycle of each subject, as formerly mentioned in the section of BMorphological Comparison of the Heart Activity Signals Obtained from the Four Electrodes^. In order to determine a significant difference between the normalized values of the signal magnitude depending upon the electrode type, a set of two-tailed Student’s t-test was applied. Student’s t-test was used because of the small number
191
J Med Syst (2015) 39:191
Page 10 of 17
Fig. 9 Streamline of magnetic field of the round-oval electrode in a normalized unit
(a) x-z plane
(b) z-y plane
(c) Three-dimensional plane
of subjects (25), which was not enough to adopt an F-test such as ANOVA. Assuming a t distribution of data and unknown variance of the population, a series of Student’s t-test was applied to every pair across the four electrode types. In other words, Student’s t-test was applied to evaluate significant differences in the normalized signal magnitude in each of the six combinations: ① single round type and round-round type, ② single-round type and round-oval type, ③ single-round type and oval-oval type, ④ round-round type and round-oval type, ⑤ round-oval type and oval-oval type, and ⑥ round-round type and oval-oval type. For this analysis, we established a statistical hypothesis that there would be no significant difference in the normalized signal magnitude between the two electrode designs for each combination. The results of the Student’s t-tests showed a significant difference only between the single-round type and round-oval type (Table 4). No significant difference was found in the remaining cases. Based on this result, it was
inferred that the heart activity signal magnitude obtained using the round-oval type electrode was significantly enhanced compared to that in obtained with the single-round type electrode. Comparison of the quality index depending on the electrode type To evaluate the performance of each textile electrode, a QI based on frequency domain analysis established in the study of Koo [26] was used in the present study. In order to obtain the QI, the heart activity signals acquired using the four textile electrodes that were originally obtained in a time-domain, were converted into magnitudes of power spectrum in a frequency-domain. Because the spectrum is an indication of the signal with respect to the frequency, the signal frequency can be easily identified from the spectrum [26]. To determine the quality index in this study, the heart activity
J Med Syst (2015) 39:191
Page 11 of 17 191
Fig. 10 Streamline of magnetic field of the oval-oval electrode in a normalized unit
(a) x-z plane
(b) z-y plane
(c) Three-dimensional plane signals from the textile electrodes placed on a garment worn by each subject, were considered to consist of the heart activity signal and noise caused by minute movement or respiration.
Fig. 11 The dual-loop-type textile electrodes attached to a sleeveless sports shirt
Prior to the frequency-domain analysis, the raw heart activity signals were first acquired through the textile inductive sensors at a 1 kH sampling rate in the sitting position for each subject. These raw signals were then subjected to the low pass filtering procedure of 3 Hz cutoff frequency to eliminate all possible external interference or power noise. In consequence, these low pass filtered data were treated by band pass filter procedure of 1~2 Hz passband frequency, and the outcome signals from 1 to 2 Hz frequency band were regarded as the heart activity signals, which was defined empirically by averaging power spectrum of all subjects. Meanwhile, by applying another band pass filter procedure of 0.1~1 Hz frequency band to the identical data, the signals between 0.1 and 1 Hz that were considered as noise from minute movement or respiration were obtained. Using this quantitative assessment, we reduced the influence of respiration on the heart activity signal, which was formerly discussed in the section of
J Med Syst (2015) 39:191
Page 12 of 17 2
2
1.5
1.5 Voltage(V)
Voltage(V)
191
1 0.5
1 0.5
coil signal 0
0
1
2
3
4
5
6
7
8
10
0
1
2
3
4
5
6
7
8
1
1 0.5 0
9
10
ECG
ECG Voltage(V)
Voltage(V)
9
coil signal 0
0.5 0 -0.5
-0.5 0
1
2
3
4
5 time(s)
6
7
8
9
Fig. 12 Heart activity signals obtained using the single round electrode (Subject 1)
BMorphological Comparison of the Heart Activity Signals Obtained from the Four Electrodes^ and BComparison of the Signal Magnitude Depending on the Electrode Type^. Frequency analysis based on the Welch power spectral density method was applied to these two sets of filtered data. The area of heart activity signals is noted as the power spectral density of the heart activity signal (PSD heart), while the area of noise is expressed as power spectral density of noise (PSD noise) (Figs. 16 and 17). The QI in this study represents the area ratio of the PSD heart to that of the PSD noise (Eq. 1) The greater the heart activity signal power, the greater the QI value, which indicates a better quality heart activity signal from the textile electrode [26].
Z
2
7
8
9
10
2
ð1Þ
1 1
1
coil signal 3
4
5
6
7
8
1
9
0
10
0
1
2
3
4
5
6
7
8
1
ECG
0.5
Voltage(V)
Voltage(V)
6
0.5
coil signal 1
5 time(s)
*(∫21PSDheart df ): the integral value of the heart activity power spectral density *(∫21PSDnoise df ): the integral value of the noise power spectral density In order to eliminate the intervening effect of the individual difference across subjects, the QI values extracted by Eq. 1 were converted into normalized QI values, by drawing a ratio of the original QI value to the averaged QI value from the four electrodes in each subject.
1.5
0
4
PSDnoise d f
1.5
0
3
0:1
2
0.5
2
PSDheart d f QI ¼ Z
2
1
1
Fig. 14 Heart activity signals obtained from the round-oval electrode (Subject 1)
Voltage(V)
Voltage(V)
0
10
0 -0.5
9
10
ECG
0.5 0 -0.5
0
1
2
3
4
5 time(s)
6
7
8
9
10
Fig. 13 Heart activity signals obtained from the round-round electrode (Subject 1)
0
1
2
3
4
5 time(s)
6
7
8
9
10
Fig. 15 Heart activity signals obtained from the oval-oval electrode (Subject1)
J Med Syst (2015) 39:191 Table 3
Page 13 of 17 191
Normalized signal magnitudes of heart activity signals according to the four electrodes (V)
Electrode type
Single-round
Round-round
Round-oval
Oval-oval
Subject
Arithmetic means per subject
Standard deviation per subject
Arithmetic means per subject
Standard deviation per subject
Arithmetic means per subject
Standard deviation per subject
Arithmetic means per subject
1
0.7292
0.8975
0.3163
0.0919
0.8350
0.1867
2.1195
0.4654
2
0.8498
0.3132
1.3427
0.4313
0.8440
0.4425
0.9635
0.3592
3
1.1865
0.6621
1.4527
0.4096
0.1703
0.0579
1.1905
0.3409
4
0.7969
0.1750
0.9160
0.1023
1.1698
0.1466
1.1173
0.2952
5
1.7689
0.3765
0.7981
0.3929
0.8996
0.2394
0.5334
0.2853
6
1.0073
0.3616
0.5360
0.1823
1.4232
0.4076
1.0335
0.4985
7
0.3727
0.2140
1.0303
0.6975
1.4148
0.5918
1.1821
0.2781
8
0.6438
0.4446
1.0638
0.7335
1.3887
0.6703
0.9037
0.4975
9
0.8542
0.3954
1.1011
0.5218
1.6268
0.6565
0.4179
0.2289
Standard deviation per subject
10
1.1040
0.3926
0.9831
0.5696
0.9553
0.4074
0.9576
0.4791
11
1.1083
0.8025
0.7731
0.5690
1.0752
1.0700
1.0435
0.7780
12
0.8325
0.2870
0.9022
0.2370
1.5127
0.8238
0.7526
0.2177
13
1.2825
0.1499
0.7985
0.3115
1.1120
0.4968
0.8071
0.4283
14
0.5718
0.2906
1.1045
0.3779
0.7142
0.3131
1.6095
0.9084
15
0.5718
0.2906
1.1045
0.3779
0.7142
0.3131
1.6095
0.9084
16
0.3119
0.0899
1.0233
0.2004
1.4067
0.3099
1.2582
0.2895
17
0.5029
0.3680
1.1095
0.3535
1.4471
0.5456
0.9405
0.3633
18
0.6050
0.3569
0.8397
0.2917
1.3623
0.4383
1.1929
0.5751
19
1.0140
0.3310
0.7120
0.3410
1.2510
0.9200
1.0240
0.4480
20
1.1370
0.2420
0.7720
0.1340
0.7980
0.1370
1.2920
0.1180
21
0.4880
0.1430
1.9480
1.8690
1.0400
0.9370
0.5240
0.1110
22
0.4840
0.1610
0.8800
0.2410
1.2800
0.5240
1.3560
0.4490
23
0.6320
0.1230
1.2040
0.2580
1.1540
0.1570
1.0100
0.2840
24
1.3430
0.2850
0.7700
0.2260
1.0550
0.3440
0.8320
0.1770
25
1.1030
0.1180
1.0330
0.1850
1.1310
0.3600
0.7330
0.1500
Averaged normalized arithmetic means per electrode
0.8520
0.3308
0.9806
0.4042
1.1112
0.4599
1.0561
0.3973
Table 4
Results of the t-test on normalized signal magnitudes of heart activity signals of the four electrodes
Combinations for Comparison
N
Levene’s test for equality of variance
Student’s t-test
F value
Sig.
t value
Degree of freedom
Sig. (2-tailed)
Mean of difference
①
Single-round Round-round
25 25
1.125
.294
−1.367
48
.178
−.1285
②
Single round Round-oval
25 25
.342
.561
−2.713
48
.009**
−.2592
③
Single round Oval-oval
25 25
.046
.832
−1.998
48
.051
−.2041
④
Round-round Round-oval
25 25
.247
.621
−1.446
48
.155
−.1307
⑤
Round-round Oval-oval
25 25
.543
.465
−.777
48
.441
−.0756
⑥
Round-oval Oval-oval
25 25
.091
.764
.558
48
.580
.0551
**p<.01
191
J Med Syst (2015) 39:191
Page 14 of 17
For this analysis, we established a statistical hypothesis that there would be no significant difference in the normalized QI among the textile electrodes, for each comparison. The result of the Student’s t-test showed that significant difference in QI between the single-round type and the oval-oval type, and that between the round-oval type and the oval-oval type, as shown in Table 6. No significant difference in QI was found in the remaining pairs. These results indicated that the quality of heart activity signal using the oval-oval electrode is significantly better than those in both the single-round and roundoval electrodes, in terms of the normalized QI.
Conclusion
Fig. 16 Error bars of the averaged normalized arithmetic means depending on the electrode type
The normalized QI values depending on the type of textile electrode are presented in Table 5. The QI values were larger in the order of the oval-oval, round-round, round-oval and single round types, similar to the average normalized QI values. Hence, the three dual-loop-type textile electrodes, the round-round type, round-oval type and oval-oval type, showed better measurement performance than the singleround electrode in terms of the average normalized QI value. In order to observe the performance of the four textile sensors, a set of two-tailed Student’s t-test was applied to the QI data; we used this test because of the small number of subjects (25 cases), which was not enough to for performing an F-test such as ANOVA. Under the assumption of t distribution of the data and unknown variance of the population, a series of Student’s t-tests was applied to each of the six combinations across the four electrode types.
Fig. 17 Concept of frequency analysis
This research focused on individual differences in the heart location and the method of resolving these differences in heart activity sensing based on the principle of magnetically induced conductivity. The focus of this research stems from the result of a previous study by Koo et al. [1], in which the feasibility of eight candidate sensor positions was examined and the sensor position BP3^ located 3 cm to the left side from the center front point on the chest circumference line, was determined to be a suitable position for textile inductive sensors of the heart activity. In addition to this main result, several notable variations were found in the sensing results from the same position of P3 across the subjects. Hence, it was inferred that this phenomenon may be associated with individual variations in the heart location. To elucidate the cause of individual differences in the efficiency of heart activity measurements from the position P3, we performed X-ray examinations in part 1 of this study. The results of the X-ray examination indicated that the heart activities from P3 in each subject were actually sensed at different positions of the heart owing to individual differences in heart
J Med Syst (2015) 39:191 Table 5
Page 15 of 17 191
Normalized quality index of the four textile electrodes
Type Subject
Singleround
Roundround
Roundoval
Ovaloval
1 2 3 4
0.2203 1.1084 1.0926 0.7470
0.1452 1.7831 1.6599 0.7470
3.0288 0.5301 1.1346 1.0121
0.6058 0.5783 0.1129 1.4940
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
2.2621 0.2581 0.9500 0.6689 1.1294 0.7800 0.4459 0.8682 1.2632 1.4940 1.1111 0.5616 0.4498 0.2424 0.9611 1.0384 0.3261 1.4700
0.8000 0.2581 0.9500 0.8055 1.3176 1.1600 0.5813 1.1163 1.4737 1.2530 0.8333 1.2264 0.6586 0.7273 0.8666 0.9316 1.7374 1.2269
0.1931 1.3763 1.2250 0.2457 0.4706 1.2200 1.4864 0.7752 0.4737 0.4337 0.7222 1.1920 0.5141 1.0909 0.8666 0.6665 0.5585 0.3951
0.7448 2.1075 0.8750 2.2799 1.0824 0.8400 1.4864 1.2403 0.7895 0.8193 1.3333 1.0201 2.3775 1.9394 1.3057 1.3635 1.3780 0.9081
23 24 25 Average Normalized Quality Index Standard deviation of the Normalized Quality Index
0.3177 1.5363 0.7731 0.8830
0.8962 0.6516 1.0233 0.9932
0.8896 0.8237 1.0431 0.8947
1.8965 0.9884 1.1606 1.2291
0.4946
0.4132
0.5700
0.5590
Table 6
location, although BP3^ was determined to be physically in the same position on the body surface. In part 2 of the study, which was sequentially performed based on the results from part 1, we devised new textile electrode designs, i.e., three designs of dual-loop-type electrodes, to overcome with individual variations in heart locations and compared their effects with that of a single round electrode by computer simulation and by performing an experiment with human subjects. From the computer simulation results, it was expected that the magnetic flux from the three types of dualloop electrodes would be more widespread than that from the single round electrode, and that the signal would be deformed because it was concentrated on the core or on one side. The results of the heart activity measurements obtained in the experiment on human subjects indicated that the heart activity signals from the three dual-loop-types of textile electrode were superior to those from the single round electrode in terms of their morphological quality. The normalized arithmetic mean magnitude of the heart activity signals obtained from the three dual-loop electrodes was larger (in the order, round-oval, ovaloval, round-round, and single round electrodes). The finding from the computer simulation that magnetic flux in the dualloop-type textile electrodes would be more widespread than that from the single round electrode, and that the flux would be deformed and concentrated in specific parts in the coupled magnetic field, supported this result. However, in the statistical analysis, a significant difference in the signal magnitude was found only between the single round type and round-oval type electrodes. Based on this result, it was inferred that the heart activity signal magnitude using the round-oval type of electrode is significantly larger than that obtained from the single round type electrode. With regard to the QI, the normalized averaged QI were larger in the order of the oval-oval, round-round, round-oval, and single-round electrode.
Results of the t-test on the normalized QI of the four textile electrodes
Combinations for Comparison
N
Levene’s test for equality of variance
Student’s t-test
F value
Sig.
t value
Degree of freedom
Sig. (2-tailed)
Mean of difference
①
Single round Round-round
25 25
.625
.433
−.855
48
.397
−.1102
①
Single round Round-oval Single round Oval-oval Round-round Round-oval Round-round Oval-oval Round-oval Oval-oval
25 25 25 25 25 25 25 25 25 25
.001
.974
−.077
48
.939
−.0117
.314
578
−2.318
48
.025*
−.3461
.459
.501
.700
48
.488
.0985
1.789
.187
−1.697
48
.096
−.2359
.198
.198
−2.094
48
.042*
−.3344
② ③ ④ ⑤
* p<.05
191
J Med Syst (2015) 39:191
Page 16 of 17
However, the statistical analysis showed significant differences in QI values between the single-round and oval-oval electrodes, and those between the round-oval and oval-oval electrodes. These results indicated that the heart activity signal using the oval-oval electrode is significantly superior to that in both cases of single-round and round-oval electrodes, with respect to the QI. Overall, among the four designs of the textile electrodes, the oval-oval shaped, dual-loop-type textile electrode was superior compared with the single-round electrode in terms of the quality index as well as in terms of the morphology of the heart activity signal. On the other hand, with regard to signal magnitude, the round-oval shaped, dual-loop-type textile electrode was found to have better performance. However, in consideration of the collateral results of our present and precedent research that the magnitude of the heart activity signal may be affected by respiration, we placed to greater weight to the effect of the oval-oval type than that of round-oval type electrode. Based on these results and inferences, we suggest that the oval-oval shaped, dual-loop-type is the better inductive textile electrode, as it more effectively overcomes individual differences in heart location, in the heart activity sensing based on the magnetic-induced conductivity principle. This research was limited to the effect of the four dualloop-type electrode designs on non-contact heart activity sensing. Further research should analyze the effect of additional types of textile electrodes. In order to obtain better quality heart activity signals, the influence of respiration on heart activity measurement needs to be concretely examined. Third, further research should consider the influence of motion artifact on the heart activity measurement. Throughout our previous research and this study, we observed that motion artifact resulting from bodily movement declines signal quality. This phenomenon has been designated as one of the main obstacles in the field of smart sensing wear [9, 27].
2. 3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14. 15.
Acknowledgments This research was supported by the Mid-Career Researcher Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning (NRF2012R1A2A2A04045455).
16.
Ethical Standards The experiments comply with the current laws of the country in which they were performed. 17. Conflict of Interest The authors declare that they have no conflicts of interest.
References 1.
Koo, H., Lee, Y., Gi, S., Khang, S., Lee, J., Lee, J., Lim, M., Park, H., and Lee, J., The effect of textile-based inductive coil sensor positions for heart rate monitoring. J. Med. Syst. 38(1–12):2014, 2014. doi:10.1007/s10916-013-0002-0.
18.
19.
Bonfiglio, A., and Rossi, D., Wearable monitoring system. Springer Science+Business Media, LLC, Springer US, 2011. Teichmann, D., Foussier, J., and Leonhardt, S., Respiration monitoring based on magnetic induction using a single coil. Biomed. Circ. Syst. Conf. 2010. doi:10.1109/BIOCAS.2010.5709565. Zheng, Y., Yan, B., Zhang, Y., and Poon, C., An armband wearable device for overnight and cuff-less blood pressure measurement. (2014). IEEE Trans. Biomed. Eng. 61:2179–2185, 2014. doi:10. 1109/TBME.2014.2318779. Singh, M., and Jain, N., A survey on integrated wireless healthcare framework for continuous physiological monitoring. Int. J. Comput. Appl. 86:0975–8887, 2014. Trindade, I. G., Martins, F., Miguel, R., and Silva, M. S., Design and integration of wearable devices in textiles. Sensors & transducers. 183:42–47, 2014. Stoppa, M., and Chiolerio, A., Wearable electronics and smart textiles: A critical review. Sensors 14:11957–11992, 2014. doi:10. 3390/s140711957. Lim, Y. G., Lee, J. S., Lee, S. M., Lee, H. J., and Park, K. S., Capacitive measurement of ECG for ubiquitous healthcare. Annal. Biomed. Eng. 42:2218–2222, 2014. doi:10.1007/s10439-0141069-6. Kumara, P. S., Oha, S., Kwona, H., Raib, P., and Varadana, V. K., Smart real-time cardiac diagnostic sensor systems for football players and soldiers under intense physical training. Proc. Biosens. Info-Tech. Sens. Syst. 2013. doi:10.1117/12.2009861. Gi, S., A design of the non-contact type textile electrode for heart rate monitoring clothing based on Magnetic- induced conductivity measurement. Yonsei University, Seoul, 2014. Song, H., Lee, J., Kang, D., Cho, H., Cho, H., Lee, J., and Lee, Y., Textile electrodes of jacquard woven fabrics for biosignal measurem e n t . Te x t . I n s t . 1 0 1 : 7 5 8 – 7 7 0 , 2 0 1 0 . d o i : 1 0 . 1 0 8 0 / 00405000903442086. Stefan, K. J., Wiklund, U., Berglin, L., Östlund, N., Karlsson, M., Bäcklund, T., Lindecrantz, K., and Sandsjö, L., Wireless monitoring of heart rate and electromyographic signals using a smart Tshirt. International Workshop on Wearable, Micro and Nano Technologies for the Personalised Health. pHealth, 2008. Cho, H., Koo, S., Lee, J., Cho, H., Kang, D., Song, H., Lee, J., Lee, K., and Lee, Y., Heart monitoring garments using textile electrodes for healthcare applications. J. Med. Syst. 35:189–201, 2009. doi:10. 1007/s10916-009-9356-8. Kinen, M., Introduction to phase change materials. In: Intelligent Textiles and Clothing. Woodhead, 2006. Ueno, A., Akabane, Y., Kato, T., Hoshino, H., Kataoka, S., and Ishiyama, Y., Capacitive sensing of electrocardiographic potential through cloth from the dorsal surface of the body in a supine position: A preliminary study. IEEE Trans. Biomed. Eng. 54:759–766, 2007. doi:10.1109/TBME.2006.889201. Steffen, M., Aleksandrowicz, A., and Leonhardt, S., Mobile noncontact monitoring of heart and lung activity. IEEE Trans. Biomed. Eng. Circ. Syst. 1:250–257, 2007. doi:10.1109/TBCAS.2008. 915633. Gi, S., Lee, Y., Koo, H., Khang, S., Park, H., Kim, K., Lee, J., and Lee, J., An analysis on the effect of the shape features of the textile electrode on the non-contact type of sensing of cardiac activity based on the magnetic-induced conductivity principle. Trans. Korea Electr. Eng. 62:803–810, 2013. doi:10.5370/KIEE.2013. 62.6.803. Teichmann, D., Foussier, J., Jia, J., Leonhardt, S., and Walter, M., Noncontact monitoring of cardiorespiratory activity by electromagnetic coupling. IEEE Trans. Biomed. Eng. 60(2142–2152):2013, 2013. doi:10.1109/TBME.2013.2248732. Steffen, M., and Leonhardt, S., Non-contact monitoring of heart and lung activity by magnetic induction measurement. Acta Polytechnol. 48:71–78, 2009.
J Med Syst (2015) 39:191 20.
21. 22.
23.
Gi, S., Lee, Y., Koo, H., Khang, S., Park, H., Kim, K., Lee, J., and Lee, J., Application of a textile-based inductive sensor for the vital sign monitoring. J. Electr. Eng. Technol. 10(364–371):2015, 2015. doi:10.5370/JEET.2015.10.1.364. Park, G., Electromagnetics. Books Hill, Seoul, 2010. Oum, J., Lee, S., Kim, D., and Hong, S., Non-contact heartbeat and respiration detector using capacitive sensor with colpitts oscillator. Electron. Lett. 44:87–89, 2008. doi:10.1049/el:20082336. Vedru, J., Gordon, R., Hummal, L., and Trolla, J., Modelling of an inductive sensor of the Foucault cardiogram. Est. J. Eng. 17:252– 270, 2011. doi:10.3176/eng.2011.3.06.
Page 17 of 17 191 24.
Lee, J. W., and Lee, J. H., Smart wear: Heart activity monitoring system using inductive sensors. World of electricity. Trans. Korea Electr. Eng. 62:21–27, 2013. 25. Mohrmann, D. E., and Heller, H. J., Cardiovascular physiology. Mcgraw-Hill International Edition, New York, 1997. 26. Koo, H., Design methods for non-contact type heart activitysensing clothing for the reduction of motion artifacts. Yonsei University, Seoul, 2015. 27. Peng, G. C., and Bocko, M. F., Non-contact ECG sensing employing gradiometer electrodes. IEEE Trans. Biomed. Eng. 60: 179–813, 2013.