Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model

G Model

ARTICLE IN PRESS

SNA-9877; No. of Pages 11

Sensors and Actuators A xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna

Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model Yu-Jen Wang a,∗ , Tzung-Yu Chen a , Meng-Chiao Tsai a , Chih Hsun Wu b a b

Mechanical and Electromechanical Engineering Department, National Sun Yat-sen University, 70 Lienhai Road, Kaohsiung 80424, Taiwan Orthopedics Department, Tainan Municipal Hospital, 670 Chongde Road, Tainan 701, Taiwan

a r t i c l e

i n f o

Article history: Received 6 May 2016 Received in revised form 4 October 2016 Accepted 13 October 2016 Available online xxx Keywords: Noninvasive blood pressure monitor Piezoelectric actuator Wrist elasticity model Dual strain sensor

a b s t r a c t This paper proposes a novel continuous noninvasive blood pressure estimation method that is based on the variation of wrist skin strain occurring with changes in diastolic and systolic blood pressure. The wrist elasticity model used to estimate blood pressure was constructed according to an ultrasonic probe indentation method used for measuring and calculating the thicknesses and material properties of skin layers. The strain value of the wrist skin was measured using a dual strain sensor to avoid temperature drift. The dual strain sensor was fabricated using a screen-printing process on a polyimide film. To cause the dual strain sensor to indent the skin for ascertaining the depth to measure skin strain, the wristband is driven by a compact ultrasonic linear motor consisting of a piezoelectric slab and a flap clip. The compact ultrasonic linear motor provides an adequate driving distance, which was estimated using the energy method, and a constant holding force used for strain measurement without input power. In blood pressure measurement experiments performed with 30 subjects, the estimation mean error against a validated cuff-type blood pressure monitor was 1.6 mmHg at rest for diastolic pressure. The error increased when the subjects exercised, but returned to near the rest value when the subjects rested after exercise. These results showed that the noninvasive blood pressure monitor can estimate blood pressure with a small fastening force. This device can be used for bedridden patients who require continuous blood pressure monitoring. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Blood pressure, which reflects cardiovascular system conditions, is one of the most critical physiological parameters. The well-known method for noninvasive blood pressure measurement is based on the auscultatory method, employing an inflatable pressure cuff that compresses a limb to intermittently measure blood pressure. Several continuous noninvasive methods for blood pressure monitoring have been developed. One approach is to detect the pulse wave velocity using a tonometer to evaluate central arterial pressure [1]. The correlations between pulse wave velocity and arterial pressure are modified by age and gender [2]. Another noninvasive device consisting of a cushioned sensor provides blood pressure measurements every 14 beats and demonstrates excellent correlation with measurements obtained through invasive methods [3]. In addition, impedance plethysmography for determining blood volumetric changes associated with the cardiovascular cycle

∗ Corresponding author. E-mail address: [email protected] (Y.-J. Wang).

is used to estimate blood pressure [4]. Kaniusas et al. proposed a method that involves using mechanical plethysmography in combination with standard electrocardiography (ECG) and subsequently discussed the measured blood pressure waves [5]. However, these methods [4,5] require the use of ECG electrodes. Another combination method based on a mechanical strain sensor and the pulse transit time minimizes the inconvenience imposed on patients during continuous blood pressure monitoring [6]. To estimate blood pressure using wrist skin strain without ECG electrodes, this study developed a novel continuous noninvasive blood pressure monitor that is based on a wrist elasticity model and involves a fastening band combined with strain gauges. This novel approach is suitable for bedridden patients who require noninvasive continuous blood pressure monitoring. The auscultatory method, which involves using an inflated cuff to occlude the blood flow, is based on a pressure balance concept. Systolic pressure is the cuff pressure at which blood starts to flow when the cuff is deflated. The diastolic pressure is the maximum cuff pressure that has no effect on the blood flow. Blood pressure refers to the pressure exerted by the blood on the walls of the arteries. Thus, systolic and diastolic blood pressures have an obvious

http://dx.doi.org/10.1016/j.sna.2016.10.021 0924-4247/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model

ARTICLE IN PRESS

SNA-9877; No. of Pages 11

Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

2

Subjects

Progressing ultrasonic probe indentation method

Building wrist

Measuring wrist

Estimating blood

elasticity model

skin strains

pressure

Wearing strain-type blood pressure monitor Fig. 1. Flowchart of the blood pressure estimation concept.

relation to skin strain. To estimate blood pressure from the strain on the wrist skin, knowledge of the mechanical properties of the skin is essential. Human skin is a complex living material composed of three main layers [7,8]: the epidermis, dermis, and hypodermis. The thickness of each skin layer varies according to race, location in the body, and age [9]. Several noninvasive methods for evaluating the mechanical properties of the skin are based on indentation [10], suction [11,12], and torsion [13] measurements. According to the relevant literature [9–15], the material properties and dimensions of human skin vary depending on individuals and skin zones. Consequently, in a mechanics model of wrist skin, the thickness and Young’s modulus of skin layers should be measured individually to estimate blood pressure. This paper proposes a new method involving an ultrasound device that is used for measuring the thicknesses and displacements of skin layers. Simultaneously, a load cell is used to measure the applied force on the skin, and the Young’s moduli of the skin layers are then acquired. A flowchart of the blood pressure estimation concept is shown in Fig. 1. The thicknesses and Young’s moduli of a subject’s wrist skin were acquired using the ultrasonic probe indentation method. A wrist elasticity model that relates the wrist strain to the blood pressure was constructed using the wrist skin parameters. For measuring skin strain above the radial artery, a strain-type blood pressure monitor that consisted of a dual strain gauge with temperature drift suppression and was mounted on a lead zirconate titanate (PZT) actuator fastening band was worn by the subject. Finally, the estimated systolic and diastolic blood pressures were compared with those obtained from a validated cuff-type blood pressure monitor. 2. Wrist elasticity model The wrist dermis, which is a soft layer covered by the epidermis and located above the radial artery, consists of a grid of collagen and lymphatic elements [8]. The subcutaneous tissue between the radial artery and the radius is composed of the hypodermis and muscle. A double-layer elastic model is suitable for describing the different Young’s moduli E and Poisson’s ratios  of the two layers. The principle of skin indentation used in this research is based on a uniform load being applied to the wrist; the load causes vertical (z-axis) and tangential (x-axis) displacements of the skin, as shown in Fig. 2. The proposed skin model neglects the viscous part of the skin and assumes isotropic elastic material, consistent with prior studies [13–15]. The global stiffness of skin tissues is the equivalent stiffness ke of a double-layer material. For an elastic material in contact with a uniform load, the global stiffness is given by 1 le 1 1 l lS = = + = d∗ + AEe∗ AEd AEs∗ ke kd ks

(1)

where k is the stiffness, A is the contact area between the ultrasonic transducer probe and the skin, l is the displacement caused by the resultant force, and E ∗ is the reduced Young’s modulus defined by E/(1 − 2 ). The suffixes d, s, and e indicate the wrist dermis, subcutaneous tissue, and equivalence, respectively. The stiffnesses of the

Load cell

Connection structure Ultrasonic transducer probe a

x

a

Dermis

Kd

Radial artery z

Subcutaneous tissue Ks

Radius Fig. 2. Schematic representation of the measurement apparatus, and the two-layer model composed of the wrist dermis and subcutaneous tissue.

dermis and subcutaneous tissue are determined from the applied forces and the resultant vertical displacements of the skin layers. The global stiffness ke and reduced Young’s modulus are calculated from Eq. (1) by using the values of kd and ks . To obtain the equivalent Poisson’s ratio, the tangential and vertical displacements associated with the uniform load were analyzed. Because the equivalent Poisson’s ratio was calculated using the tangential displacement on the skin surface, the value of the equivalent Poisson’s ratio approached that for the wrist dermis. The assumption of the Poisson’s ratio of the subcutaneous tissue being equal to that of the wrist dermis is acceptable because the skin strain caused by blood pressure is insensitive to the Poisson’s ratio of the subcutaneous tissue. Wrist skin indentation can be considered a uniform normal pressure load p over the contact area where normal tractions exist. The pressure load is obtained by dividing the resultant force by the contact area between the ultrasound probe and the skin. The displacements resulting from contact stress in the linear elastic half-space model were derived in [16] on the assumption that the interface is frictionless. The vertical displacement of a point in the loaded region (−a ≤ x1 ≤ a) is given by



uz = −



1 − ve 2 p1 Ee



(a + x1 ) ln

 a + x1 2 a

+ (a − x1 ) ln

 a − x1 2  a

+C

(2)

An ultrasound transducer probe of width 2a is centered on the y-axis, and C is the constant of integration determined by the vertical displacement at the probe center x1 = 0. By determining uz at x1 = / 0, the relationship between e and Ee in Eq. (2) is acquired. On the assumption that the origin is not displaced, the tangential displacement in the loaded region (−a ≤ x2 ≤ a) is provided by ux = −

(1 − 2ve ) (1 + ve ) p2 x2 Ee

(3)

The suffixes 1 and 2 of x and p are used to indicate that uz and ux correspond to different locations and loads. Substituting Eq. (2) into Eq. (3) yields the equivalent Poisson’s ratio, which is formulated as ve =

p2 (C − uz )x2 + p1 ux (a − x1 ) ln (

a−x1 2 ) a

2p2 (C − uz )x2 + p1 ux (a − x1 ) ln (

+ p1 ux (a + x1 ) ln (

a−x1 2 ) a

a+x1 2 ) a

+ p1 ux (a + x1 ) ln (

a+x1 2 ) a

(4)

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model

ARTICLE IN PRESS

SNA-9877; No. of Pages 11

Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

3

Fig. 3. Schematic of the wrist elasticity model constructed using COMSOL 4.2.

The parameters d and s are substituted as e into the expression for the reduced Young’s modulus to obtain the Young’s moduli Ed and Es , respectively. In the preceding paragraph, we discussed how most of the dimensions and material properties of the radial artery and skin layers related to blood pressure can be obtained, besides the Young’s modulus of the radial artery. The Poisson’s ratio of the radial artery is assumed to be 0.45 on the basis of values presented in the literature [17]. The wrist elasticity model shown in Fig. 3 was constructed using the finite element software COMSOL 4.2 by referring to magnetic resonance images. The dimensions of the wrist contour and bones in the wrist elasticity model were scaled according to the wrist width of each subject. This is reasonable because these parameters are insensitive to skin strain. The radius and ulna were considered rigid bodies. A semicylindrical bump was subjected to a known preload pressure to cause an initial vertical displacement. The unknown Young’s modulus of the radial artery was obtained from the measurement strain of the wrist skin when the wrist blood pressure difference was measured. The blood pressure difference is defined as the difference between systolic pressure and diastolic pressure. 3. Dual strain sensor and readout circuit The skin strain caused by wrist blood pressure was measured using a novel dual strain sensor that consisted of measurement and dummy gauges mounted on the semicylindrical bump, as shown in Fig. 4(a). The measurement gauge was oriented tangential to the wrist because the tangential strain is greater than the axial strain owing to the radial artery being aligned in the axial direction. Between the measurement gauge and the semicylindrical bump, a flexible silicone rubber strip was used for providing a flexible boundary to increase the strain value recorded by the measurement gauge. The dummy strain gauge was used to eliminate the

Fig. 4. Dual strain sensor and readout circuit used for strain measurement. (a) Designs of the measurement and dummy strain gauges. (b) Half-Wheatstone bridge circuit.

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

4

strain error resulting from the temperature difference between the resistors in the Wheatstone bridge and the measurement gauge. The dummy gauge was rigidly bonded to the semicylindrical bump, which was made of polyoxymethylene and oriented in the axial direction of the wrist; the axial direction strain is insensitive to the blood pressure. The aforementioned method causes the strain value of the dummy gauge to be unaffected by the blood pressure, but introduces drifts related to the temperature effect in the same way as the measurement gauge does. A schematic circuit diagram for the dual strain sensor is shown in Fig. 4(b). When no strain is applied to the dual strain sensor, adjusting the variable resistor R2 to obtain a the balanced bridge yields R1 ∗ (Rg + Rw ) = (R g + Rw ) ∗ R2

(5)

, R

where Rw g , and Rg are the resistances of the wire, dummy gauge, and measurement gauge, respectively. When the strain sensors and wires are at temperatures different from those of R1 and R2 and a strain is applied to the measurement gauge, Eq. (5) can be written as R1 ∗ (Rg + Rg + RTg + Rw + RTw ) = / (R g +R Tg + Rw + RTw ) ∗ R2

(6)

R denotes resistance change and the suffix T indicates resistance change caused by a temperature difference. Because the process and material of the dummy and measurement gauges are identical, we obtain RTg ≈ R Tg and R1 ≈ R2 . Eq. (6) can be rewritten as R1 ∗ (Rg + Rg + Rw ) = / (R g + Rw ) ∗ R2

(7)

Thus, the temperature difference between the measurement gauge and resistors, which introduces a drift in the skin strain value, is eliminated. The screen-printing technique is a cost-effective process, and was used to fabricate the dual strain gauge on a 100 ␮m thickness polyimide (PI) film in this research. A conductive silver paste was employed as the ink for low-resistance gauges. Sensing wires were printed using a screen printer; the 420-mesh stainless steel screen of the screen printer was positioned at an angle of 22.5◦ relative to the film. A finer linewidth and space lead to a longer equivalent length and a larger resistance variation of the strain gauge for a given strain. However, a finer mesh implies a lower aperture opening ratio of the screen, which causes the silver flow to be unstable. The silver paste was preheated to 55 ◦ C in the screen-printing process to reduce its viscosity. The emulsion thickness of the screen was 15 ␮m. The film was baked at 130 ◦ C for 30 min to form a conductive film with a thickness of 10 ␮m. The PI film was used as the substrate because of its high flexibility, high chemical resistance properties, and high temperature resistance (400 ◦ C). To improve its anti-abrasion properties, the surface of the dual strain gauge was coated with a mixture of a polyurethane solution and ethyl ester acetate as a protective layer. The minimum linewidth and space of the dual strain gauge that we successfully applied were 73 and 85 ␮m, respectively, and the sensing area was 2.0 mm × 4.2 mm. Fig. 5 shows an optical micrograph of the screen-printed strain gauge. The gauge factor (defined as the ratio of the relative change in the resistance to the strain) of the fabricated strain gauge was 1.95 according to the experiment and calibration results. The output voltage of the Wheatstone bridge induced by the resistance variation of the strain gauge for determining the wrist blood pressure was in the tens of microvolts range. A general purpose instrumentation amplifier, INA-128, was used to magnify the output voltage with a gain of 1300. To suppress high-frequency noise, a second-order low-pass RC filter consisting of two passive low-pass filters chained together was used. Because the human

Fig. 5. Screen-printed strain gauge. (a) Schematic of the dual strain gauge. (b) Optical micrograph of the dual strain gauge.

pulse rate ranges from 0.8 to 2.5 Hz, the cutoff frequency of the low-pass RC filter was set as 25 Hz. The gain value of the low-pass RC filter for the wrist strain wave was evaluated in experiments. 4. Strain distribution in the measurement region A simulation illustration of the strain on the strain gauge resulting from the wrist blood pressure is shown in Fig. 6(a). The strain gauge has a sliding contact with the wrist skin for a coefficient of friction of 0.4 and Young’s modulus of the strain gauge substrate (PI film) is 250 GPa. The mesh grid is constructed from free tetrahedral elements, and the mesh elements are arrayed densely in the strain gauge and radial artery for enhancing the calculation accuracy. Fig. 6(b) shows the x-component (tangential to the wrist) strain map when the dual strain sensor made an indent of 2.30 mm at a diastolic blood pressure of 90 mmHg. At the center of the strain map, the x-component strain tensor is negative because of a surface line-load resulting from the indentation of the bump. The average strain value in Fig. 6(b) represents the DC part of the strain signal corresponding to the diastolic blood pressure for the indentation of 2.30 mm. Fig. 6(c) presents the x-component strain difference value resulting from strain values of the systolic and diastolic blood pressures with 30 mmHg difference. The average strain difference value in Fig. 6(c) represents the AC part of the strain signal. The strain map in Fig. 6(c) shows the strain distribution according to the location of the radial artery. The y-component strain value of the dummy gauge averages 10−7 that is twenty times fewer than the x-component strain value, indicating that the dummy gauge is adequate. The diastolic strain for various indentation pressures was estimated on the basis of the wrist elasticity model as shown in Fig. 7(a). In the low indentation pressure region, the nonlinear variation of

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

5

Fig. 6. Simulation illustration and results for the plane strain on the strain gauge associated with the wrist blood pressure. (a) Local illustration for the strain map simulation. (b) Strain map (x-component) for a dual strain sensor that indented the wrist skin by 2.30 mm at a diastolic blood pressure of 90 mmHg. (c) Strain map (x-component) for a strain difference at 30 mmHg blood pressure difference between the systolic and diastolic pressures.

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model

ARTICLE IN PRESS

SNA-9877; No. of Pages 11

Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

6

3.0E-03

Diastolic strain

2.0E-03 1.0E-03 0.0E+00

Diastolic pressure:

-1.0E-03 -2.0E-03

60 mmHg

90 mmHg

150 mmHg

160 mmHg

120 mmHg

-3.0E-03 -4.0E-03 0

1000

2000

3000

4000

5000

6000

7000

8000

Indentation pressure (Pa)

(a) 5.0E-04

Strain difference

Indentation pressure: 4.0E-04

500 Pa

2000 Pa

3000 Pa

6000 Pa

7000 Pa

8000 Pa

20

30

40

3.0E-04

2.0E-04

1.0E-04

0.0E+00 0

10

50

60

Blood pressure difference (mmHg)

(b) Fig. 7. Variation of the estimated strain on the basis of the wrist elasticity model with (a) indentation pressure for various diastolic pressures and with (b) blood pressure difference for various indentation pressures.

the strain was caused by the geometric effect of the semicylindrical bump and to the partial contact between the bump and the wrist skin. For indentation pressures exceeding 4000 Pa, the strain is inversely proportional to the indentation pressure. The linear variation region is suitable for diastolic blood pressure measurement. Because the strain varies with the indentation pressure, maintaining a constant indentation pressure (force) during the blood pressure measurement process is necessary. The strain difference in Fig. 7(b) is defined as the difference between the diastolic strain (corresponding to a diastolic pressure of 60 mmHg) and the systolic strain, and reflects the AC component of the strain wave generated by the blood pressure. Fig. 7(b) shows that the strain difference is directly proportional to the blood pressure difference; the slopes increase with the indentation pressure. The increase in strain difference becomes saturated at a high indentation pressure (e.g., greater than 7000 Pa); in other words, at this indentation pressure, the limit of the indentation distance of the skin is reached. An extremely high indentation pressure can interrupt blood flow, leading to the strain difference becoming zero. For blood pressure measurement, a moderate indentation pressure is necessary for maintaining a strain difference. The material properties of the skin and the dimensions for the simulation results in Fig. 7 were obtained from the average values of the 30 subjects.

5. Design of ultrasonic motor for wristband A wristband is used for binding the dual strain sensor to the wrist skin. The necessary binding force for the indentation depth and indentation pressure was calculated using the wrist elasticity model. Fig. 8 shows the simulation results for a constant actuating force of 0.16 N exerting an indentation pressure of 5230 Pa and resulting in a driving distance of 3.25 mm being exerted at the wristband end; an indentation depth of 2.41 mm was generated on the skin. As a result of the light indentation depth (1.5–2.5 mm) and small local indentation area caused by the semi-cylindrical bump, the compression necessary for blood pressure measurement can be achieved with minimal discomfort. The simulation defined the necessary output force and driving distance of the actuator required to constrict the wristband for wrist blood pressure measurement. Because the indentation depth should be consistently maintained for wrist strain measurement, an actuator that has a constant holding force without input power, as well as an adequate output force and driving distance for constricting the wristband provides an adequate solution. Accordingly, an ultrasonic linear motor capable of providing a constant holding force at the strain measurement location without input power was considered [18]. Ultrasonic linear motors in which a traveling wave or standing wave is used for the driven components to produce a linear motion have been pro-

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

7

Fig. 8. Pressing displacement on the skin, generated by the actuating force.

posed in [19,20]. In standing-wave-type motors, a chain of angled microimpulses provides elliptical paths to drive components; these motors are simpler to assemble than traveling-wave motors are [21,22]. A compact ultrasonic linear motor (CULM) and its components are shown in Fig. 9. The main body was sintered using piezoceramic powders fabricated by Eleceram Technology Co., Ltd. The top and bottom surfaces of the piezoceramic slab were covered by two symmetrical electrode plates (to control the driving direction) and a single common electrode plate, respectively. When the piezoceramic slab is actuated, the vibration of the slab sidewalls drives the flap clip connected to the wristband. The dimensions of the piezoceramic slab are 22 mm × 9 mm × 1.3 mm, contributing to the compactness of the wristband driving motor. The piezoceramic slab was driven at a resonance frequency of 215 kHz, which corresponds to the expansion mode, for generating large in-plane oscillation displacements; the power consumption was 180 mW. The vibration trajectory of the long-sidewall range from 3.9 to 16.1 mm was inclined in the positive x direction when the left electrode was applied. In this effective sidewall region, the moving direction of the flap clip reverses when the applied electrode is switched. The

vibration velocities in the x and y directions of the piezoceramic slab at different sidewall locations are shown in Fig. 10. Fig. 11 shows a free-body diagram of the flap clip when it is driven by the piezoceramic slab. The frictional driving force F is calculated as the sum of the kinetic friction forces in the region of contact between the flap clip and the piezoceramic slab. F (x) =

q 

k ·

N · sgn (Vn (x) − V (x)) q

where k is the kinetic friction coefficient between the piezoceramic slab and the flap clip (experimentally determined), N is the normal contact force exerted by the flap clip, q is the number of sections into which the flap clip is divided, and n is the location index of the piezoceramic slab. The velocity of the flap clip and the x-direction velocities of the vibration trajectories at different locations of the piezoceramic slab are denoted by V(x) and Vn (x), respectively. At the location where Vn (x) is greater than V(x), a positive friction force drives the flap clip. Vn (x) is shown in Fig. 10 for different locations. The normal contact force is determined using the dimensions and Young’s modulus of the flap clip. When the input electric current is zero, the holding force provided by the flap clip is a static friction force that is greater than the maximum frictional driving force. The work and energy method was used to analyze the driven distance of the wristband. The wristband can be actuated when the sum of the kinetic energy of the flap clip and the work of the frictional driving force is greater than the elastic potential energy of the wristband and wrist skin. This condition can be represented as 1 1 · m · V (x)2 + F (x) · (x − x0 ) > · k¯ (x) · (x − x0 )2 2 2

Fig. 9. Graphic representation of the ultrasonic motor and its components.

(8)

n=1

(9)

where m is the mass of the flap clip, x0 is the initial position for zero deformation of the wristband and wrist skin, and k¯ is the equivalent stiffness determined from the wristband stiffness and the wrist elasticity model. The wristband is made of silicon rubber with a

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11 8

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

Fig. 10. Simulation of piezoceramic slab sidewall vibration velocities in the x and y directions denoted by red and blue lines respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

0.9 Entire elastic potential energy

Work or Energy (mJ)

0.8

Work of friction driving force

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

1

2 3 Driving distance (mm)

4

4.38

5

Fig. 12. Variation of work and elastic potential energy with driving distance. Fig. 11. Free-body diagram of the flap clip.

thickness of 0.8 mm and a Young’s modulus of 5 MPa. Substituting Eq. (8) into (9) yields V(x), which is an unknown parameter. Because the flap clip is lightweight, its kinetic energy is negligible compared with the other terms in Eq. (8). Once the operation region of the flap clip is chosen in the in-phase region, Vn (x) is positive. Because the flap clip is connected to the wristband, V(x) is generally smaller than Vn (x), resulting in F being positive in the in-phase region. Fig. 12 shows the variation of the work of the frictional driving force as well as the entire elastic potential energy of the wristband and the wrist skin, calculated using the wrist elasticity model for the driving distance x. The parameters were determined to be k = 0.28 and N = 0.57 N; the skin parameters are discussed in Section 2. The input peak to peak voltage of 20 V ensures the vibration of the piezoceramic slab sidewall, which is required for driving the flap clip. In Fig. 12, the work and energy curve intersection is at the x value of 4.38 mm, indicating an achievable driving distance for the wristband. A stopper is used to limit the driving distance of the flap clip for achieving an adequate indentation distance on the wrist. To account for most of the wrist skin elasticity, the piezoceramic slab and flap clip are designed to make the work of the frictional driving force greater than the entire elastic potential energy at the stopper position. Once the flap clip is driven to the stopper position, the

flap clip can hold the wristband without any electric current input, providing a steady skin indentation for blood pressure measurement. That is because the static friction coefficient is greater than the kinetic friction coefficient. 6. Experimental results The thickness and the resultant displacements of the dermis and subcutaneous tissue were measured using an ultrasonic transducer probe, as shown in Fig. 13(a). The resultant force applied on the wrist was the sum of the weight of the ultrasonic device and the external force measured by the load cell. The Young’s modulus of skin varies with depth [13]. Therefore, reduced Young’s modulus measurements were performed at an initial indentation depth from 2.0 to 2.6 mm where evident skin strain caused by systolic blood pressure was observed to enhance the accuracy of blood pressure estimation. An ultrasonograph indicating skin layer displacements in a subject is shown in Fig. 13(b). The subjects were 18 healthy Asian men aged between 25 and 35 years and 12 healthy Asian women aged between 26 and 34 years. In the 30 subjects, the Young’s moduli of the wrist dermis (E1 ) and subcutaneous tissue (E2 ) were 41.1 ± 7 kPa and 82.5 ± 12 kPa, respectively, and the Poisson’s ratio was 0.39 ± 0.06. These properties were calculated using Eqs. (1)–(4), with tangential and vertical displacements in

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

9

Fig. 13. Ultrasonic probe indentation setup. (a) Measurement devices and a subject’s wrist. (b) Wrist ultrasonography for observing the vertical displacement.

the ultrasonic probe indentation. The minimal thicknesses of the wrist dermis and subcutaneous tissue in the left inner wrist were 1.7 ± 0.4 mm and 8.9 ± 1.8 mm, respectively, and the diameter of the radial artery was 2.2 ± 0.5 mm. The Young’s modulus of the radial artery was 227 ± 53 kPa; this value was acquired from the measured strain and known blood pressure differences. A prototype of the proposed strain-type blood pressure monitor is shown in Fig. 14(a). It consists of a dual strain sensor, a strain readout circuit, and a CULM fastening band. The experimental value of the wrist blood pressure was calculated using the wrist elasticity model and the measured strain value, and it was compared with the value obtained using a validated cuff-type blood pressure monitor (Terumo ES-P402GN). The wrist skin signals measured by the skin strain sensor are shown in Fig. 14(b), where the crest and trough of the wave correspond to the diastolic and systolic blood pressures, respectively. The dicrotic notch representing closure of the aortic valve and subsequent retrograde flow is readily apparent in the strain signal. The location of the dicrotic notch varies according to the timing of aortic closure in the cardiac cycle and defines the systolic and diastolic phases. The slopes of diastolic and systolic decline reflect the overall cardiovascular performance of the subjects. Future studies will compare the strain wave with the direct blood pressure measurement to analyze the pressure signal in detail. The location of the semicylindrical bump on the wrist skin was adjusted to ensure that a noticeable strain value difference was induced by the changes in blood pressure. Each strain reading was sustained for 5 s; the left wrist was used in measurements for the two blood pressure sensors. The blood pressure values were the averages from three sets of measurements, and the device was removed and aligned after each measurement. Table 1 presents the estimation errors and averages of the estimated blood pressures for the 30 subjects during an initial rest, after stationary bike riding, and during a final rest (5-min break after rid-

Fig. 14. Photographs of the (a) prototype of the strain-type blood pressure monitor and (b) strain signal obtained in the experiment. Table 1 Averages and errors of the estimated blood pressures. (Mean ± standard deviation; Unit: mmHg), (Average value of estimation; Unit: mmHg). State

Diastolic pressure error

Systolic pressure error

Resting on seat Bike riding Resting 5 min after bike riding

1.6 ± 2.1 (82.6) −3.1 ± 4.9 (74.9) −2.3 ± 3.6 (87.7)

−2.0 ± 3.9 (124.0) −6.9 ± 9.6 (156.1) −4.4 ± 5.3 (127.6)

ing). After the blood pressure was measured in the initial rest state to estimate the Young’s modulus of the radial artery, the wristband was reworn. The diastolic blood pressure exhibited minor negative changes from the rest state to the exercise state as observed in the cuff-type blood pressure monitor. The estimation errors for systolic pressure were greater than those for diastolic pressure. One cause of this is that absolute value of the error increases as a function of the pressure level. The standard deviation errors of the systolic pressure being higher than those of diastolic pressure was attributed to a slight inaccuracy in the skin elasticity model, namely the nonlinear Young’s moduli of the radial arteries and skin. In the bike riding state, the mean values of the estimation errors were negative; this was attributed to the slight misalignment of the strain sensor. The negative mean errors were observed even upon return to the resting status. In the experiments, increasing the stiffness of the flap clip to reduce the slight sway of the semicylindrical bump could reduce the errors in diastolic blood pressure in bike riding. The standard deviation errors were lower in the final rest state compared with those in the bike riding state, confirming the repeatability of the strain-type blood pressure monitor and the estimation method. Overall, the estimated blood pressure was consistent with that measured using the validated cuff-type

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11 10

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

blood pressure monitor. This observation supports the use of the strain-type blood pressure monitor and the wrist elasticity model for estimating blood pressure.

7. Conclusions

A novel noninvasive blood pressure monitor based on measuring strain in the wrist skin is proposed. The blood pressure measured using the proposed approach exhibits a preliminary correlation with that measured using a validated blood pressure monitor, except for small variations. The absence of some morphological variations in the ultrasonic probe indentation (e.g., radius and tendon dimensions) from the estimation model could induce slight blood pressure errors. Another source of the slight errors was that measurements by the two types of sensors were not obtained in the same test. In addition, errors existed in the measurements provided by the validated blood pressure monitor. A quantitative assessment showed low errors in the resting state (maximum mean error: −2.0 mmHg). During exercise, the standard deviation errors in systolic pressure increased to ±9.6 mmHg; this error is expected to be reduced by developing a dynamic model for cardiovascular circulation of the upper limbs in the future. The skin strain was measured using a dual strain sensor equipped with a dummy gauge to suppress temperature drift. The measurement gauge was mounted on a compliant semicylindrical bump instead of a flat plate to enhance the measured strain value. Strain gauges fabricated using a screen-printing process are cost effective for dual strain sensors and suitable for compliant substrates. With adequate fabrication parameters, the minimum linewidth was 73 ␮m. To fasten the wristband to produce a known skin indentation for an ascertained strain, a CULM was used to drive the wristband. The maximum driving distance of the CULM for the wristband was estimated according to the work and energy principle. The CULM had a small volume and low power consumption, and provided a constant holding force without input power. This research reveals that the proposed method involving the use of dual strain gauges, a PZT actuator fastening band, and a wrist elasticity model can be applied for blood pressure measurement. The novel noninvasive blood pressure monitor is easy to use and minimizes the inconvenience imposed on bedridden patients who require continuous blood pressure monitoring.

Acknowledgments

The authors appreciate the support from Ministry of Science and Technology, R.O.C under the grant no. NSC 104-2221-E-110-030 and NSC 104-2622-E-110-007-CC3. The authors are grateful to YuHong Li and Tai-Yu Hou for their help during the testing of devices and Tainan municipal hospital for related equipment.

References [1] P. Salvi, G. Lio, C. Labat, E. Ricci, B. Pannier, A. Benetos, Validation of a new non-invasive portable tonometer for determining arterial pressure wave and pulse wave velocity: the PulsePen device, J. Hypertens. 22 (2004) 2285–2293. [2] Y. Matsui, J. Ishikawa, K. Eguchi, S. Hoshide, H. Miyashita, K. Shimada, K. Kario, The influence of wave reflection on left ventricular hypertrophy in hypertensive patients is modified by age and gender, Hypertens. Res. 31 (4) (2008) 649–656. [3] K. Belani, M. Ozaki, J. Hynson, T. Hartmann, H. Reyford, J.M. Martino, M. Poliac, R. Miller, A new noninvasive method to measure blood pressure, Anesthesiology 91 (1999) 686–692. [4] A.V. Blinov, E.P. Selivanov, Plethysmographic impedance device for measuring blood pressure, Meas. Tech. 40 (2) (1997) 188–192.

[5] E. Kaniusas, H. Pfützner, L. Mehnen, J. Kosel, J. Tellez-Blanco, G. Varoneckas, A. Alonderis, T. Meydan, M. Vazquez, M. Rohn, A.M. Merlo, B. Marquardt, Method for continuous non-disturbing monitoring of blood pressure by magnetoelastic skin curvature sensor and ECG, IEEE Sens. J. 6 (3) (2006) 819–828. [6] T.Y. Tu, Y.H. Kao, P.C. Chao, Y.P. Lee, Optimizing a new blood pressure sensor for maximum performance based on finite element model, IEEE Sens. Conf. 2014 (2014) 1873–1876, http://dx.doi.org/10.1109/ICSENS.2014.6985394. [7] J.M. Pereira, J.M. Mansour, B.R. Davis, Dynamic measurement of the viscoelastic properties of skin, J. Biomech. 24 (1991) 157–162. [8] R. Reihsner, B. Balogh, E.J. Menzel, Two-dimensional elastic properties of human skin in terms of an incremental model at the in vivo configuration, Med. Eng. Phys. 17 (1995) 304–313. [9] E. Berardesca, J. de Rigal, J.L. Leveque, H.I. Maibach, In vivo biophysical characterisation of skin physiological differences in races, Dermatologica 182 (1991) 89–93. [10] C. Pailler-Mattei, S. Bec, H. Zahouani, In vivo measurements of the elastic mechanical properties of human skin by indentation tests, Med. Eng. Phys. 30 (2008) 599–606. [11] S. Diridollou, F. Patat, F. Gens, L. Vaillant, D. Black, J.M. Lagarde, Y. Gall, M. Berson, In vivo model of the mechanical properties of the human skin under suction, Skin Res. Technol. 6 (2000) 214–221. [12] F.M. Hendriks, D. Brokken, J. Van Eemeren, C. Oomens, F. Baaijens, J. Horsten, A numerical–experimental method to characterize the nonlinear mechanical behavior of human skin, Skin Res. Technol. 9 (2003) 274–283. [13] C. Pailler-Mattei, S. Bec, H. Zahouani, In vivo measurements of the elastic mechanical properties of human skin by indentation tests, Med. Eng. Phys. 30 (2008) 599–606. [14] R. Sanders, Torsional elasticity of human skin in vivo, Pflugers Arch.-Eur. J. Physiol. 342 (1973) 255–260. [15] S. Diridollou, F. Patat, F. Gens, L. Vaillant, D. Black, J.M. Lagarde, In vivo model of the mechanical properties of the human skin under suction, Skin Res. Technol. 6 (2000) 214–221. [16] K.L. Johnson, Contact Mechanics, Cambridge University Press, England, 2001. [17] D.W. Park, Arterial elasticity imaging: comparison of finite-element analysis models with high-resolution ultrasound speckle tracking, Cardiovasc. Ultrasound 8 (2010) 1–10, 1476-7120-8-22. [18] T. Hemsel, J. Wallaschek, Survey of the present state of the art of piezoelectric linear motor, Ultrasonics 38 (2000) 37–40. [19] S.L. Sharp, J.S.N. Paine, J.D. Blotter, Design of a linear ultrasonic piezoelectric motor, J. Intell. Mater. Syst. Struct. 21 (10) (2010) 961–973. [20] Y.J. Wang, K.C. Fu, C.C. Wang, A smart pinless ejection mechanism using dual-resonance excitation Langevin piezoelectric transducers, Smart Mater. Struct. 25 (2016) 015014. [21] Y. Shi, C. Zhao, A new standing-wave-type linear ultrasonic motor based on in-plane modes, Ultrasonics 51 (4) (2011) 397–404. [22] Y.J. Wang, Y.C. Chen, S.C. Shen, Design and analysis of a standing-wave trapezoidal ultrasonic linear motor, J. Intell. Mater. Syst. Struct. 26 (2015) 2295–2303.

Biographies

Yu-Jen Wang was born in Tainan, Taiwan, in 1977. He received his Ph.D. from Department of Power Mechanical Engineering at National Tsing Hua University, Taiwan, in 2011. Currently, he is an assistant professor of Mechanical and Electromechanical Engineering Department, National Sun Yat-sen University, Taiwan. His major research interests include biomedical devices, machine dynamics, actuator design and energy harvesters.

Tzung-Yu Chen received his Bachelor’s degree from Mechanical and Electromechanical Engineering Department, National Sun Yat-sen University, Taiwan. Currently, he is studying for master’s degree at the same university. His major research interests include Hemodynamics and strain sensor design.

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021

G Model SNA-9877; No. of Pages 11

ARTICLE IN PRESS Y.-J. Wang et al. / Sensors and Actuators A xxx (2016) xxx–xxx

Meng-Chiao Tsai received his Bachelor’s degree from Department of Mechanical Design Engineering, National Formosa University, Taiwan. Currently, he is studying for master’s degree at National Sun Yat-sen University. His major research interests include Biomechatronics and Biomechanics.

11

Chih Hsun Wu received Doctor of Medicine degree from Medical Collage of National Cheng Kung University, Taiwan, in 2004. He is a visiting staff of Orthopedics department, Tainan Municipal Hospital, Taiwan since 2009. His focus of practice is in Orthopaedic trauma. He is a member of Taiwan Orthopaedic Trauma association and AO Foundation Trauma.

Please cite this article in press as: Y.-J. Wang, et al., Noninvasive blood pressure monitor using strain gauges, a fastening band, and a wrist elasticity model, Sens. Actuators A: Phys. (2016), http://dx.doi.org/10.1016/j.sna.2016.10.021