Biomimetic radial pulse sensor design
Human fingertip physiological structure
Finger palpation methods have been performed many centuries, and they are still widely applied in the clinic process. The mechanism of the palpation is that the physical properties containing the physiological and pathological information of the object that is contacted with the fingertip can be purely transferred to mechanoreceptors inside of the fingertip through its fine physiological structure with mechanical filtering functions [8]. This indicates that human fingertip owns the biomechanical advantages for achieving high-performance touch/force sensing and inspires us to mimic it for designing our radial pulse sensing structure. To start the mimicking, we firstly review the physiological structure of the human fingertip. Figure 1 shows the anatomy of the fingertip and nail, and there are several layers for the whole fingertip structure constituted with different organisms. The epidermis is the first layer from the bottom to the top, which works as a protective layer isolating the interior of the fingertip from outside. Dermis is the interlayer next to epidermis and is distributed with various types of mechanoreceptors that can sense touch, pressure, and vibrations from 0.5 to 500 Hz [8]. Subcutaneous tissue is mainly a fat layer with 60% to 72% fluid in volume [9] and works like a damper to remove the unwanted information to the mechanoreceptors. The bone controlled by extensor and flexor tendons can move up and down to provide both passive and initiative movements for the effective touch and supports the fingertip as well. Due to the complex mechanical behavior and structures of the skin and subcutaneous, two simplified biomechanical models of the fingertip were reported [10]. These models include mechanical modeling of combined bio-components such as the bone, fat, soft tissues, and skin membrane. The results validating biomechanical advantages inspire us to design a radial pulse sensor that mimics the physiological structure of the human fingertip. As shown in Figure 1, several bio-components are arrowed and they are mimicked in designing our sensor components as shown in Figure 2b.
Mechanical design of biomimetic sensor
Figure 2 illustrates the mechanical design of the biomimetic radial pulse sensor using SolidWorks. Figure 2a displays the three-view diagram of the sensor structure. The length of the sensor is 50 mm, the width is 30 mm, and the height is 25 mm (without the adjustable screw); two grooves at two sides of the sensor can attach an adjustable belt to help wear it at the radial, or other body locations such as brachial, carotid, popliteal, and even superficial temporal comfortably. Figure 2b details the sensor components that functionally mimic the fingertip bio-components and their functions. All components except the sensing film and polydimethylsiloxane (PDMS) damper were made by the 3D printer using the solid plastics. The diameter of the skin contact ball is 10 mm, back side of the ball is glued with the sensing film, and the open side directly contacts with the radial pulse location of wrist. The rigidity of the ball is carefully chosen so that it can transfer the radial pulse signal to the sensing film without significant mechanical loss. The rigid contact ball design can also help user to quickly find the exact pulse points. The sensing film is one of the key components that mimic the mechanoreceptors of fingertip for the developed pulse sensor. It is an ultrasensitive hybrid carbon/polymer-based piezoresistive (HCP) film with thickness of 0.127 mm, width of 3 mm, and length of 12 mm. The film has high sensitivity, very low thermal drift, and low hysteresis that were reported in our previous study [12]. The sensing film is glued on the elastic PDMS damper, the film will be deformed, and its resistance will change following the strength of transferred radial pulse signals.
The elastic PDMS damper is another key component in the sensor. It mimics subcutaneous soft tissues and fat in the human fingertip. To reach or close to the function of the soft tissues and fat, the hardness of the damper is an important factor. In this work, a digital shore hardness tester (TYPE A, range: 0 to 100 HA, resolution: 0.5 HA, accuracy: ≤±1% HA) was firstly used to estimate the hardness of the human fingertips. The average hardness of the fingertips measured by the digital shore hardness tester is around 30 HA. To make an elastic damper with the similar hardness, Sylgard 184 Silicone Elastomer (Dow Corning Corp., Midland, MI, USA) (temperature range −45°C to 200°C) was selected. Five different samples with different mixing ratios between base liquid (part A) and curing agent (part B) were tested in order to achieve the hardness that is close to the hardness of the soft fingertip. Vacuumed Thermo Scientific Nalgene Chambers (Thermo Fisher Scientific, Waltham, MA, USA) was then used to store all the mixed silicone elastomer liquids for 48-hour curing process at the room temperature. After cured, five different hardness PDMS elastic dampers were made using different mixing ratios between base liquid and curing agent such as 1:2, 2:3, 1:1, 3:2, and 2:1. Through a long-time aging test, we found that all hardness of all damper samples gradually reaches the constant values after 33 days, as illustrated in Figure 3a. Especially for the sample with mixing ratio 1:2 (base liquid 1: curing agent 2), it can be found in Figure 3b, its hardness was measured around 30 HA that is the most close to the average hardness of the human fingertips. This PDMS sample was then chosen to be an elastic damper for the biomimetic sensor. In our design, the damper is a cuboid with a thickness of 3 mm (matches the thickness between the skin and the bone of fingertip), width of 3 mm, and length of 15 mm. In addition, the ball-shaped bone in the sensor mimics the arterial bone in the fingertip. The adjustable screw serves as the regulating unit of the bone. It functions as the extensor and flexor tendons in the fingertip. As shown in Figure 2b, all centers of the screw, bionic ball-shaped bone, and skin contact ball are aligned on a line in order to reduce the shear force and twist distortion during the operation.
Electronics design
The electronic circuit interfacing with the HCP sensing film was designed and custom-built for the sensor. It is used to process and to transform the signals that are from the film resistance changes caused by the radial pulse deformation. Since a preload force is applied on the sensing film when the sensor is worn at the radial, the gain and sensitivity of the circuit needs to be carefully adjusted to guarantee high sensitivity and high accuracy of the sensor, as well as to avoid saturation of the measured signals. To meet the requirement, as shown in Figure 4, a custom-built Wheatstone quarter bridge circuit with an AD620 amplifier is employed.
In the schematic, R6 is the resistor that represents the value of the resistive layer on the HCP sensing film. Its resistance is changed following the micro deformation caused by the radial pulse. At the top of R6, a potentiometer is used to trim the bridge circuit through removing the zero offset. R1, R2, and R5 are the constant resistors with the same resistance as the undeformed resistive layer. R3 is the resistor used to modify the gain of the amplifier AD620, and the maximum gain of the amplifier can reach 10,000. The circuit output V
o
is the analog voltage signal that is input to a PC for further display and processing through a 16-bit and 200 kS/s DAQ system PCI-DAS6013 made by Measurement Computing Corp.
Sensor calibration, identification, and verification
Sensor calibration setup
As shown in Figure 5, all calibration and verification processes were implemented on a high-performance Newport vibration-isolated workstation. The calibration and verification procedures were automatically performed by the computer program in order to minimize the environmental noises and human effects.
Two high-performance sensors, a Shimpo FGV-1XY digital force gauge (Shimpo Instruments, Itasca, IL, USA) (capacity: 5 N, resolution: 1 mN, high accuracy: ± 0.2% FS) and a Baumer OADM 20I6X41/S14F laser distance sensor (Baumer Electric, Southington, CT, USA) (range: 70 mm, resolution: 20 µm), were employed in the calibration process. The laser distance sensor OADM 20I6X41/S14F was set up on the vibration-isolated workstation and its laser beam perpendicularly shoot on the moving stage of the micromanipulator for tracking the displacement change of the skin contact ball in the developed radial pulse sensor, when the digital force gauge is controlled to push/release the contact ball during the calibration. The pushing/releasing process was controlled by a Sutter MPC 385 3-D micromanipulator (Sutter Instruments, Novato, CA, USA) (range: 25,000 µm, resolution: 1 µm) where the digital force gauge is attached on the micromanipulator. The process can generate 150-µm height square wave pushing/releasing movement by the program-controlled micromanipulator. Meanwhile, the force information was recorded by the force gauge. In the calibration process, the developed pulse sensor was fixed on the vibration-isolated workstation through a clamping stander. Note that, to meet the real condition that the preload force is applied to the developed sensor during wearing it, an average 1.6 N preload force to the contact ball was initially set by the force gauge FGV-1XY before and during the calibration. In addition, the designed electronic circuit with the gain of 50 times was interfaced with the developed sensor. All signals were recorded to a PC through the DAQ board PCI-DAS6013 (Measurement Computing Corporation, Norton, MA, USA) (resolution: 16 bits, range: ± 10 V, number of channels: 16, accuracy: ± 8.984 mV, speed: 200 kS/s) with sampling rate of 1 kS/s.
Identification and calibration
Figure 6 plots all experimental data including pushing/releasing force, the developed sensor output, and pushing/releasing displacement on the skin contact ball of the sensor. The first row is the force curve measured by the digital force gauge FGV-1XY. The second row plots the voltage output from our developed pulse sensor, and the last row shows the displacement detected by the laser distance sensor OADM 20I6X41/S14F. The data demonstrates the developed sensor which has the fast response speed and good dynamic behavior.
Based on the experimental data shown above, the transfer function model between the developed pulse sensor voltage output and the force measurement by the force gauge FGV-1XY can be found through the system identification method developed in MATLAB. Here, the sensor output is used as the input, and the force measurement is used as the output (reference). The Laplace transfer function is identified as shown in the following equation:
(1)
where the gain K = 0.20998, the coefficients Tp 1 = 0.001, Tp 2 = 9.7751, Tp 3 = 0.0023316, and T
z
= 11.063. By applying the identified model in Equation 1 to the raw sensor voltage output, as shown in Figure 7, we can see that the fitting rate between the reference force measurement (black line) and the calibrated sensor force output (red line) is 90.75% high and the maximum error at peaks is 3.25%.
Further, Figure 8 plots the comparison results on frequency-amplitude relationship between the calibrated sensor output signals and the reference signals. The consistence between the two results further indicates the calibration performance of the sensor, that is, the sensor is well calibrated and can be used for the radial pulse measurement.
Biomimetic performance verification of sensing structure
The calibration results of the sensor have demonstrated the good mechanical to electronic conversion performance of the developed sensor. To further verify the biomimetic performance of the sensing structure that is close to the structure advantage of the human fingertip, an experiment was conducted to investigate the force-displacement relationship of serially connected components in the sensing structure consisting of a rigid skin contact ball, a sensing film, a silicone elastic damper, and a ball-shaped bone structure. In this experiment, the setup is similar to the calibration one described in Subsection ‘Sensor calibration, identification, and verification’. Instead of using 150-µm height square wave motion, 150-µm height triangle wave movement was generated by the computer program to push and release the sensing structure through the contact ball. The laser distance sensor OADM 20I6X41/S14F was set to a higher resolution but with a short range (range: 30 mm, resolution: 4 µm) to accurately detect displacement/deformation of the sensing structure caused by pushing and releasing operations.
Figure 9 plots the experimental force-displacement data points of the sensing structure, the mean value data points from real human fingertips conducted by Serina et al. in [9], and two curves fitted using the above two data point groups. Note that, a preload force of 1.6 N was applied to the sensing structure before conducting experiments. Some small spines in our experimental force-displacement data curves were induced by the movement of the step driven motors of the micromanipulator. It is clear that our sensing structure force-displacement data points demonstrate a good linear relationship between the force and the displacement in several pushing and releasing experimental loops. The small overlap between these data points (curves) in several experimental loops indicates a very low hysteresis effect of the designed sensing structure. More importantly, our experimental result agrees well with the force-displacement behavior of the human fingertips conducted by Serina et al. [9]. In the figure, the brown linear fitted curve from the data points of our biomimetic sensor is very close to the green linear fitted curve from the human fingertip data measured by Serina et al. [9] and it experimentally verifies the successful biomimetic design of our sensor.