Early detection of cardiac complications after heart surgery is a crucial demand in clinical surveillance facilities. The use of accelerometer devices fixated to the heart has been demonstrated as a promising method for the detection of myocardial ischemia with high sensitivity and specificity. One of the most important challenges in developing heart implantable accelerometer devices is to minimize any tissue trauma. In this study, a new approach in making a heart muscle implantable accelerometer sensor is described. An overall diameter of 2 mm is achieved by encapsulating a miniaturized three-axis MEMS accelerometer (similar to the BMA355) from Bosch Sensortec GmbH, Germany. The design of the sensor features a simple implantation procedure, which aims to be similar to the medical procedure of temporary pacing wire attachment. The device remained in place in the heart muscle without the need for additional attachment or fixation. The essential clinical requirements for implantable devices, especially heart implantable sensors, are also taken into account in this study. Additional safety enhancement is realized by adding a galvanic isolator, which ensures electrical safety.
Continuous monitoring of patients who have undergone heart surgery (e.g., coronary artery bypass grafting surgery) is vital for early detection of complications and has been shown to improve survival and patient outcome . Myocardial ischemia has been reported in up to 38% of postoperative coronary bypass grafting surgery patients [2, 3]. Electrocardiography (ECG) is the conventional and invaluable tool for continuous monitoring of heart activity [4, 5], but its sensitivity to detect coronary occlusion is not sufficient [6, 7]. Hence, there is a need for continuous monitoring of cardiac function that can detect postoperative complications faster and with higher sensitivity, specificity, and accuracy than existing systems.
We have previously demonstrated how heart activity can be monitored continuously and how myocardial ischemia can be detected by using a three-axis accelerometer attached to the epicardium (on the outer surface of the heart) [8, 9]. These studies were done using sensors 14.5 mm (length) × 11 mm (width), and the device was sutured to the surface of the heart muscle . The device was used during open-thorax surgery and had to be removed before closing the thorax. The second-generation heart sensors demonstrated were further miniaturized and could be implanted into the myocardium (into the heart muscle wall) [11, 12]. This device had outer dimensions of 10 mm length and 3.2 mm diameter.
In this study, an improved concept for manufacturing the implantable heart muscle accelerometer device is introduced, which allows development of smaller devices. The devices can be implanted by using the clinically conventional implantation procedure familiar to cardiac surgeons. An integrated function of pacing/sensing as a unipolar pacing wire is also introduced.
Design and Assembly
A. System Design
Implantation of temporary pacing wires is standard procedure after cardiac bypass surgery [13, 14]. These implants normally consist of an insulated wire with an electrode in one end and a connector to the stimulator device in the other end. Attached to the electrode is a curved needle that is used to make the implantation channel in the heart muscle. It is advantageous for the implantable heart muscle accelerometer device to have the same implantation procedure as the temporary pacing wire. The device is, therefore, intentionally fabricated to be similar in shape to a temporary pacemaker wire and can be implanted using the same procedure. The device has two output connectors: One is connected to a stimulator device for pacing and the other to a system for recording and processing accelerometer data. The miniaturized metal capsule acts as pacemaker electrode, equipped with a needle for implantation.
The overall structure of the combined pacing and accelerometer device is shown in Fig. 1. The device consists of four main parts: connector, cable with insulation, accelerometer with metal capsule, and curved needle on a metal wire.
A three-axis accelerometer from Bosch Sensortec GmbH (Stuttgart, Germany) is used. This accelerometer is similar in form, fit, and function as the BMA355 (Bosch Sensortec GmbH). The accelerometer was built on wafer-level chip scale package technology with 10 lead-free solder balls (∅︀200 μm and 0.4 mm pitch). The small size of this accelerometer (1.2 × 1.5 × 0.8 mm) is the key factor in achieving the miniaturized design. The volume reduction from the CMA-3000A accelerometer (2 × 2 × 0.95 mm) used in a previous design [11, 12, 15] to this accelerometer is 62.1% (see Fig. 2). The BMA supports both I2C and SPI interfaces. The acceleration range can be digitally programmed from ±2 g to ±16 g. In this study, we have selected to use the ±4 g range with the sensitivity of 512 LSB/g.
The new implantable accelerometer device is built on a single-polyimide-based flexible circuit board and serves as both substrate and cable. The advantages of polyimide-based material in terms of biomedical applications, such as the miniaturization of circuit layout, flexibility, biocompatibility, and reliability, are described in [16–19]. The accelerometer supports both I2C and SPI protocol. In this study, the I2C protocol is used because it requires fewer data transmission wires than SPI, enabling a more compact circuit layout. Additional wires are needed for interrupt control and pacing/sensing function. The circuit layout of the device is shown in Fig. 3.
The flexible substrate-cable was fabricated by Dyconex AG (Bassersdorf, Switzerland). For short-term implantable application (1–4 d), the flexible circuits are fabricated in accordance to the requirement of the IPC Class 2 with better inspection and higher acceptability level than that for a commercial product. The substrate-cable is confirmed to comply with the RoHS 2011/65/EU Directive. Table I describes the cross-sectional structure of the flexible substrate cable. Two additional conductive layers were added to the sensor end and connector end, as shown in Fig. 4. According to the multilayers fabrication process of Dyconex AG, the extra layers eliminate the need for additional solder mask or cover layer and give a higher production yield.
The connector end is pulled through the chest wall before it is connected to a data acquisition system. This is done by inserting the connector in a standard introducer, which is commonly used in cardiac surgery. Then the introducer is pulled through the chest wall with the connector attached. Hence, a cylindrical shaped connector as shown in Fig. 5 was selected.
B. Device Assembly Process
To ensure proper interconnection between the flexible substrate and the accelerometer, a flip chip bonding technique was selected due to the components' size and the precision required. The assembling process was carried out by flip chip bonder FinePlacer Pico (Finetech GmbH, Germany), which can provide a placement accuracy of 5 μm and handle components down to 0.125 × 0.125 mm. Precise alignment between the accelerometer and substrate is shown in Fig. 6. The parameters of the flip chip bonding process are shown in Table II. A small amount of no-clean flux was manually dispensed on each contact pad before doing “pick-and-place” of the accelerometer to improve interconnection yield. Extra pure isopropanol was use for post soldering cleaning, which helps to enhance the adhesion of underfill and steps followed.
To avoid high-frequency noise from the power supply, a decoupling capacitor was used. The assembly of the low-pass filter capacitor was done manually. A surface-mounted capacitor, imperial code 0201-100 nF, was connected in parallel to the power supply terminals. The BMA accelerometer and the low-pass filter capacitor bonded to the flexible substrate are shown in Fig. 7.
Nonconductive adhesive EPO-TEK 353ND (Epoxy Technology, Inc.) was used as underfill to improve the mechanical strength of the bonding between the accelerometer and the flexible substrate. The curing time was set to 5 min at 100°C. Coplanarity between the bonded accelerometer and the substrate is very important, especially for the gravity calibration process of the device. In this study, a flip chip bonding technique was used and provided good alignment, suitable bonding force, precise curing temperature profile, repeatability, and productivity.
C. Encapsulation of Cable Part
Previous studies on the implantable heart monitoring device pointed out that using a polyimide-based flexible substrate-cable can provide low-level leakage current (10−8 A). The disadvantage is that the flex cable has sharp edges, which may cut into surrounding tissue . To overcome this issue, the cable part of the flexible substrate-cable was covered with a biocompatible silicone in a molding process. This step was done prior to any other assembly. The surface of the flexible substrate-cable was activated by oxygen plasma treatment. The small end of the flexible substrate-cable was pulled through a ∅︀1.5-mm silicone tube that worked as a cable mold. Two-component biocompatible silicone MED-4211 (NuSil Silicone Technology, Carpinteria, California) was mixed and filled into this tubed mold with the support of vacuum. Vacuum also released any air bubbles trapped in the silicone. Fig. 8 describes how the flexible cable was centralized in the silicone-filled tube.
The silicone molding tube was placed on a U-shaped platform with the same radius as the outer radius of silicone tube. The positions of clamp A and B (see Fig. 8) were adjusted until the cable was in the center of the silicone tube. Sufficient force was applied to keep the flexible cable in horizontal position. Precuring took place at room temperature for 1 d followed by a post curing step at 100°C in 3 h.
A tube remover was used to remove the molding tube. The molding tube with molded cable inside was pulled through the tube remover, which has a cutting mechanism to separate the molding tube. The two molded halves on the tube were then easily removed.
D. Encapsulation of Sensor Part
A metal capsule housing of the sensor part was chosen to enhance mechanical strength and incorporate a pacing function. The capsule is hollow, with an outer diameter of 2.0 mm and wall thickness of 0.1 mm. The material selected was 316L type stainless steel, a material that is widely used for implants [20, 21].
The metal capsules were made by additive manufacturing (three-dimensional [3-D] printing) using a Concept Laser M2 cusing. Since the surface of the printed capsules is rough, it was necessary to polish the capsules. First, the capsules were coarsely polished using the Struers Knuth Rotor grinding station, with grit SiC grinding papers P1500 and subsequently P2100. The last fine polishing step was carried out using the precise grinding/polishing equipment (MultiPre™ System; Allied High Tech Products Inc., Compton, California) using dense and low-napped silk (Red final C Polishing Cloths; Allied High Tech Product, Inc.) with a mixture of colloidal silica and 0.05 μm alumina (Colloidal Silica/Alumina Suspensions; Allied High Tech Product, Inc.) at 350 rpm. A comparison between a polished capsule and an as-fabricated capsule is shown in Fig. 9.
The interior of the capsule was mechanically polished by a cylindrical fine-grinding tool tip to have a smooth channel for the next assembly steps.
The encapsulation of the accelerometer, which is attached to the flex, was carried out in several steps. Initially the substrate with the bonded accelerometer and capacitor was insulated by two layers of nonconductive adhesive both on the top and bottom side. The contact pad used for ECG sensing/pacing function was kept open. The metal wire was fed through the tip of the metal capsule and a knot effectively retained the wire in place, as shown in Fig. 10. A micropipette was used to transfer some adhesive into the tip part to block the opening channel and to keep the knot in place. The same technique was used to dispense conductive adhesive into the capsule. The conductive adhesive is used to create an electrical connection between the bottom contact pad for pacing/sensing function and the metal capsule. Next, the insulated device was covered by another adhesive layer, except for the pacing/sensing pad, which should be in contact with the conductive adhesive applied in advance. The assembled device was then placed in a vacuum chamber to remove the bubbles trapped inside.
Fig. 11 shows a prototype with 3-D printed metal encapsulation fabricated by these techniques to be used in animal trials.
Test, Results, and Discussion
A. Cross Talk
1) Cause of Cross Talk in the Device
The accelerometer used in this study communicates over the I2C protocol, specified by NXP Semiconductors . I2C requires four wires; these are the signal lines serial data (SDA) and serial clock (SCL) in addition to the power supply and ground wires (Fig. 12). SDA and SCL are bidirectional lines, connected to the positive supply via pull-up resistors. Data can be transferred at rates from 100 up to 400 kbps.
This study used a 300-mm-long flexible cable to connect the sensor to a computer, and the cross section of the cable is illustrated in Fig. 13. As a thin and flexible cable is beneficial for the application, the internal wires are tightly spaced with no shielding. Such a cable can be prone to cross talk between the signal lines. To investigate and minimize this, an experiment was set up to measure cross talk for clock frequencies in the relevant range, between 100 kHz and 400 kHz.
2) Cross Talk Measurement
The routing of the signals in the flex cable should be chosen to minimize cross talk between the bus lines. This was investigated by measuring the interference between the conductors in an arrangement illustrated in Fig. 14.
The output from a function generator (TTi TG5011; Thurlby Thandar Instrument, Cambridgeshire, UK) was connected to one conductor in the substrate-cable. This was denoted as the “active” conductor. Both ends of a second conductor were grounded, whereas the third conductor, denoted as “passive,” was connected to an oscilloscope (Tektronix TDS 2012B; Tektronix, Inc., Beaverton, Oregon). The other two ends of the first and the third conductors were grounded via resistors equal to the pull-up resistance Rp of the I2C configuration. Rp can be varied from 1 kΩ to 4 kΩ, and the supply voltage was V = 3.3 V . To investigate how the routing of the signals influenced the cross talk, four different wiring configurations were tried (Fig. 15). Cross talk between the SDA and SCL should be avoided; hence, the positions of the active and passive wires corresponded to the SCL and SDA during actual data transfer.
The output from the signal generator was set to a square wave with amplitude of 4 V and duty cycle of 50% [24, 25]. The voltage ratio between the active and passive wires was measured for all configurations illustrated in Fig. 15, at frequencies from 100 kHz to 400 kHz.
The results shown in Fig. 16 were carried out on a single flex cable with 5,000 samples for each data point. The results pointed out different cross talk levels in different wiring configurations. Measurements on similar flexes provided comparable cross talk results.
These results show that the cross talk level depends strongly on the routing of the signal lines inside the flex print cable, as the difference between the best suppression, Routing 1, and the worst, Routing 3, was approximately 40 dB. The lowest interference was found for the active and passive lines placed as far apart as possible with a ground wire between them. This result seems logical if capacitive coupling is assumed to be the main source of cross talk.
Cross talk measurements for various values of the pull-up resistor are shown in Fig. 17 for the wire configuration giving the best suppression, Routing 1. The results show that the suppression is improved by 5 dB when reducing the resistor value from 4 kΩ to 1 kΩ.
Based on these results, wiring pattern Routing 1 was selected inside the flex print cable, minimizing the interference between the signal lines. The effect of the pull-up resistors was smaller, but in addition to increased cross talk, a large pull-up resistor will also increase the rise time of the signals (Fig. 17), which might influence the stability of the data acquisition. On the other hand, safety considerations make us want to keep the current as low as possible. For this reason, the higher value, 4 kΩ, was chosen for the pull-up resistors.
B. Leakage Current Measurement
1) Experimental Setup
To comply with the requirements for maximum permissible leakage current for implantable medical devices (specified as cardiac floating—CF type) set by the International Electrotechnical Commission (IEC-60601-1), the leakage current from the device should not exceed 0.01 mA under normal condition or 0.05 mA under a single fault condition . Leakage current measurement setups were used to measure leakage currents from the whole device. The measurement was carried out in phosphate-buffered saline (PBS) with pH 7.4 (P3813 0.01M PBS; Sigma-Aldrich Co., St. Louis, Missouri) to simulate the physiological environment . A thermal chamber (Heraeus T6200; Thermo Fisher Scientific, Schwerte, Germany) was used to maintain stable temperature of 37°C. The measurements were carried out by the electrometer system (model 6430 Sub-Femtoamp SourceMeter Instrument; Keithley Instruments, Inc., Cleveland, Ohio). A remote preamplifier (Keithley Remote PreAmp) was used to enhance the precision of the measurements and obtain low-current sensitivity. All the contact pads (except the ECG pacing/sensing contact pad) of the connector terminal were connected to the negative channel of the remote pre-amplifier with a platinum electrode connected to the positive channel of the preamplifier. The measurement setup is shown in Fig. 18.
2) Leakage Current Test at 37°C
As mentioned previously, a device suitable for short-term implantation was the goal of this study. The concept was that the sensor would be removed from the patient's body after 1–4 d. Therefore, leakage current measurement was carried out continuously over a period of 80 h at 37°C. Figs. 19 and 20 present a comparison of the leakage current between bare flexible cable (previously reported in ) and the complete encapsulated device proposed in this study.
C. Tensile Strength (Pulling Test)
This is one of the essential compliance requirements set by the European Standard EN 45502-1: 1997 for implantable medical devices. The standard requires that an implantable device withstands the tensile forces that may occur during and after implantation, without fracture of conductors or cracking of electrical insulation or cracking of the body . For the implantable heart sensor, the tensile strength tests provide valuable information about the maximum force that can be applied along the length of the prototype device while implanting or extracting the device out of the body. It is especially important for closed-chest application, as the sensor is to be extracted through the chest wall using the cable when monitoring is completed. The compliance is confirmed by comparison between the tensile strength of the completed device and the critical values of pull-in/pull-out force observed in animal trials. The measurement setups are illustrated in Fig. 21.
The tensile strength tests were performed using a LLOYD LS100 Universal testing system (Lloyd Instruments™ Lloyd Instruments Ltd., West Sussex, UK) at room temperature. Fig. 22 illustrates typical results for tensile strength tests on two flex cables and a single complete device: (1) A flexible cable without silicone overmold, (2) a flexible cable with silicone overmold, and (3) a complete device with metal capsule encapsulation and connector attached. The test lengths were 200, 200, and 300 mm, respectively.
Based on our measurements, the complete device can withstand a pulling force up to 12 N. This is two times larger than the measured pull-in/pull-out force when implanting/removing the device from the heart, which was presented in an earlier study . In a recent animal trial, the observed value of implantation force was <2.5 N.
D. Signal Acquisition and Pacing/Sensing Function in Animal Trials
The main goals of the animal trial carried out in this study were to demonstrate the advantages of the design for improving the implantation process, to corroborate the implantation stability in both open thorax and closed thorax during the initial hours, to investigate the difference between closed thorax acceleration data and that of opened thorax, and to affirm the additional pacing/sensing functionality.
The animal trial was approved by the Norwegian Animal Research Authority (FOTS id6085) and carried out at the Intervention Center, Oslo University Hospital. All researchers involved in the animal experiment were certified with Federation of Laboratory Animal Science Associations' category C (European Convention for the protection of Vertebrate Animals [ETS No. 123]).
A median sternotomy (a type of surgical procedure where the breastbone is divided vertically) was performed and the pericardium (thin membrane layer that covers the heart) opened. The device was inserted subepicardially (layer beneath the outermost layer of the heart) in the apical region (close to the lowest part of the heart) on the left ventricle. The chest and the pericardium were left open, and the pig was placed in dorsal supine position (lying with the face up).
Fig. 23 shows the implantation position of two accelerometer devices and pacing wires as reference electrodes for the comparison of pacing/sensing threshold between the accelerometer device and the pacing wire.
Fig. 24 presents the recoded acceleration signals from a myocardial implantable accelerometer device in an open thorax, whereas Fig. 25 shows results when the thorax is closed. The sampling rate used in these measurements was 500 samples/s. During closing of the thorax, the heart is slightly moved, and, therefore, a bias of acceleration level in x-direction was observed. There was no significant reduction of the absolute signal amplitude. The acceleration signals are unaffected. Data shown in Fig. 24 and Fig. 25 are recorded from a single device in one animal trial.
Pacing/sensing function was carried out in this study to demonstrate an additional functionality of the myocardial implantable accelerometer device. The electrical response of the heart detected by ECG signal was used to compare the performance between the original pacing lead and the built-in pacing function of the accelerometer device.
The pacing threshold is the minimal amount of energy required to produce the depolarization of the myocardium in contact with the electrode . A reference pacing/sensing threshold test was done using two original pacing wires (Ethicon 3-0 TPW10 temporary cardiac pacing wire). Those wires were implanted as shown in Fig. 23. Both of the wires were connected to a pacing generator (Medtronic 5388 dual chamber temporary pacemaker). The ECG and acceleration signal were synchronized in time. Fig. 26 presents the recorded data when the heart was paced at 80 Hz, sensing threshold was 4 mV, and pacing output current was 10 mA. Stimulus spikes can be seen in Fig. 26.
A similar pacing/sensing threshold test was repeated, but one of the two pacing leads was replaced with accelerometer device 2, as shown in Fig. 23. The pacing setup included a pulse generator, one original pacing lead, and the accelerometer device. Fig. 27 presents the synchronized ECG and acceleration signals, which were recorded when the heart was paced at 130 Hz, a sensing threshold of 1 mV, and pacing output current of 5 mA. Stimulus spikes can be seen in Fig. 27.
The relative difference between the amplitude of the pacing spikes in the ECG signals shown in Fig. 26 and Fig. 27 may be caused by several factors. Most likely the impedance difference, the surface area of the electrode, and the sensing/pacing parameters are the most important factors . Different impedance is caused by the different distances between the two electrodes in the two tests. The implantation distance between the two pacing leads was larger than between the pacing lead and the accelerometer device. This yields lower impedance between the accelerometer device and reference lead.
The electrode surface area of the accelerometer device is provided by the outer shield metal capsule, which is significantly larger than the surface area of the pacing lead. According to a study on the relationship between the surface area of a pacing lead and the polarization of ions around the lead given by de Voogt  and Timmis et al. , increased surface area can reduce the polarization, which means decreased pacing threshold. This is a sufficient explanation for having a lower pacing threshold (5 mA) but achieving stronger pacing effects as shown in Fig. 27.
A monitoring scenario is illustrated in Fig. 28, where there are four main components. The first is the implanted accelerometer device. The second component is the signal isolator, which galvanically isolates the patient from the data acquisition instrument. A computer with a data acquisition program built on a Labview platform was used to record and process data. The last is the pulse generator (e.g., Medtronic), which is commonly used as stimulator for temporary pacing wires.
The galvanic isolator includes the I2C data acquisition USB-8451 (National Instruments) and a high frequency digital signal isolation circuit. Galvanic isolation ensures that the patient's heart is not exposed to any unexpected current that otherwise may leak from the computer power supply through the device . This requirement of protection against electrical hazards is set by the IEC 60601-1: 2005, Chapter 8. The acceleration signal is optically transferred by a low-power bidirectional I2C isolator ISO1540 (Texas Instruments); the insulation and safety-related specifications are listed in Table III. Schematic illustration and layout of the galvanic isolation circuit is shown in Fig. 29.
The accelerometer device and the front end (left part) of the ISO1540 isolator are operated at 3.3 V, which is regulated from a 9 V battery supply. Battery-powered supply is considered as a suitable power supply for most medical applications  and commonly used in the pulse generator of pacing equipment (e.g., Medtronic Model 5388 Dual Chamber Temporary Pacemaker). Bidirectional I2C signal is optically transmitted to the signal acquisition part (on the right side) to have two decoupled grounds for the accelerometer device and the computer system.
This study describes the development of an MEMS-based multifunctional implantable heart device. A packaging method based on the smallest accelerometer commercially available sensor has been proposed. The overall diameter of the device is 2.0 mm, which enables direct implantation into the heart muscle. Processes used to fabricate and assemble the device (cable/substrate and metal capsule) are chosen with respect to mass production that feature less failures due to manual handling.
The miniaturized device is based on an implantation procedure similar to temporary pacing wires, a well-known procedure for surgeons and cardiologists. Simple implantation procedure saves valuable time during the operation. The device provided very good implantation stability and flexibility during animal trials. Due to reduced device size and simplified implantation procedure, the tissue trauma was greatly minimized. The integrated pacing/sensing capability adds functionality to the design. Our experiments show that the multifunctional accelerometer device can yield a pacing threshold in comparable range to the original pacing wire. However, a qualification study on stimulus functionality of the device needs to be carried out.
The device complied with the leakage current demand for heat implantable medical devices, which is regulated by the IEC (IEC 60601-1). The maximum leakage current level that was observed during 80 h was 1,000 times below the requirements (0.01 mA) under normal condition. Tensile strength tests of the device show that the device withstands a force of 12 N that may exist during implantation and extraction. The crucial requirement of galvanic isolation set by the IEC 60601-1 was compiled in this study. A galvanic isolator was proposed and fabricated, which provides electrical isolation up to 4,000 V.
This study was funded by Oslofjord Regional Research Fund under Grant No. 208933. The authors thank medical doctors and surgeons who have contributed to this study and carried out animal trials at Intervention Centre—Oslo University Hospital, Norway. We also thank Bosch Sensortec GmbH and Texas Instruments for offering samples of the new generation accelerometer BMA and I2C isolator ISO1540, respectively.
1IMST—Department of Micro and Nano Systems Technology, HBV: Buskerud and Vestfold University College, Raveien 215, 3184 Borre, Vestfold, Norway
2Intervention Centre, Oslo University Hospital, P.O. Box 9450 Nydalen, 0424 Oslo, Norway