CELL MICROINJECTION SYSTEM WITH FORCE FEEDBACK

Abstract

A novel piezo-driven cell injection system with force feedback overcomes the unsatisfied force interaction between the pipette needle and embryos in conventional position control. By integrating semiconductor strain-gage sensors for detecting the cell penetration force and the micropipette relative position in real time, the developed cell microinjection system features high operation speed, confident success rate, and high survival rate. The effectiveness of the developed cell injection system is experimentally verified by penetrating zebrafish embryos. The injection of 100 embryos are conducted with separate position control and force control. Results indicate that the force control enables a survival rate of 86%, which is higher than the survival rate of 82% produced by the position control in the same control environment. The experimental results quantitatively demonstrate the superiority of force control over conventional position control for the first time.

Claims

1. A microinjection system comprising a micro injector and a control device comprising a linear actuator which controls the movement of the injector towards and from a pierceable microstructure target and position and force strain gauge sensors configured to provide position and force information to a controller which is operatively linked to said linear actuator.

2. A microinjection system as claimed in claim 1, wherein information from the sensors is fed to a proportional-integral-derivative controller.

3. A device as claimed in claim 1, wherein said linear actuator is a piezoelectric actuator with a displacement amplifier.

4. A device as claimed in claim 3, wherein the displacement amplifier is a bridge displacement amplifier.

5. A device as claimed in claim 1, wherein said force-strain gauge sensors are semiconductor strain gauge sensors.

6. A device as claimed in claim 5, wherein said force-strain gauge sensor is configured to control the force applied by the linear actuator to be sufficient to effect puncture of a target.

7. A microinjection system as claimed in claim 1, wherein the injector is a micropipette and the system further comprises a flexure guiding mechanism for driving the micropipette to its intended location and for ensuring withdrawal of the pipette with minimal damage to the target.

8. A microinjection system as claimed in claim 7, wherein the controller is configured to exercise control over movement of the micropipette according to the equation:
u(t)=u(tT)+K.sub.p[e(t)e(tT)]+K.sub.ie(t)+K.sub.d[e(t)2e(tT)+e(t2T)](3) where x is position/force variable, x.sub.r denotes the desired position/force trajectory, T is the sampling time interval, and u(tT) denotes the control variable in the previous time step and, K.sub.p, K.sub.i and K.sub.d are tunable positive gains and e(t)=x(t)x.sub.r(t).

9. A method for microinjection of a substance into a pierceable microstructure which comprises positioning a mocroinjector relative to a pierceable microstructure by use of a charge coupled device camera using a computer vision algorithm and then effecting puncture of said pierceable microstructure by the micro injector and injecting a substance into said microstructure wherein motion of said microinjector is controlled by a control device comprising a linear actuator which controls the movement of the injector towards and from said pierceable microstructure target and position and force strain gauge sensors configured to provide position and force information to a controller which is operatively linked to said linear actuator whereby sufficient force is applied to the microinjector to effect piercing of the wall of the pierceable microstructure.

10. A method as claimed in claim 9, wherein control over the movement of the microinjectured is effected by the controller according to the equation: the controller is configured to exercise control over movement of the micropipette according to the equation:
u(t)=u(tT)+K.sub.p[e(t)e(tT)]+K.sub.ie(t)+K.sub.d[e(t)2e(tT)+e(t2T)](3) where x is position/force variable, x.sub.r denotes the desired position/force trajectory, T is the sampling time interval, and u(tT) denotes the control variable in the previous time step and, K.sub.p, K.sub.i and K.sub.d are tunable positive gains and e(t)=x(t)x.sub.r(t).

11. A method as claimed in claim 10, wherein the sensors are calibrated to provide position and force information to the controller to determine variable x in equation (3).

12. A method as claimed in claim 9, wherein said pierceable microstructure is selected from the group consisting of such as a biological cells, nuclear envelopes and viral capsids and envelopes using such a piezoelectric actuator.

13. A method as claimed in claim 12, wherein DNA is introduced into a cell nucleus.

14. A method as claimed in claim 13, wherein said introduction is effected as part of a gene therapy regimen.

15. A method as claimed in claim 9, wherein DNA is introduced into a virus.

16. A method as claimed in claim 9, wherein said pierceable microstructure is an ovum or embryo.

17. A method as claimed in claim 16, wherein said introduction is effected to effect in vitro fertilization of an ovum, cloning or insertion of cells, such as stem cells, into an embryo.

18. A control device suitable for use with an injector for injecting biological material into a pierceable microstructure said control device comprising a linear actuator which controls the movement of the injector towards and from a pierceable microstructure target and position and force strain gauge sensors configured to provide position and force information to a controller which is operatively linked to said linear actuator.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0018] FIG. 1 is a schematic diagram of the flexure-based injector, which is driven by a PZT with a bridge-type displacement amplifier.

[0019] FIG. 2 shows FEA simulation results of the injector. (a) FEA analysis configuration. (b) FEA analysis result.

[0020] FIG. 3 shows a CAD model of the developed piezo-driven injector with strain-gage position and force sensors. The moropipette is fixed on the injector tube through the force sensor and the injector tube id mounted on the top of the linear actuator by two supports.

[0021] FIG. 4 shows an experimental setup for cell injection.

[0022] FIG. 5 shows an experimental setup for force sensor calibration,

[0023] FIG. 6 shows force sensor calibration result (a) The output of loadcell sensor and strain gages; (b) Force calibration error; (c) Strain gage force sensor noise; (d) Histogram of strain gage sensor noise.

[0024] FIG. 7 shows position sensor calibration result. (a) The output of laser sensor and strain gages; (b) Position calibration error; (c) Strain gage position sensor noise; (d) Histogram of strain-gage sensor noise.

[0025] FIG. 8 depicts motion sequence of the automatic cell injection system. (a)-(d) indicate different cell injection phases, and * denotes that loop is executed all the time.

[0026] FIG. 9 shows a comparison between different injection scenarios. (a) Micropipette is puncturing the embryo; (b) micropipette slips over the embryo.

[0027] FIG. 10 depicts force sensor output signals during the cell injection. (a) Force variation of pierced cell; (b) force variation of unpierced cell.

[0028] FIG. 11 shows injection phases of the zebrafish embryo. (a) Approaching phase; (b) penetration phase; (c) injecting phase; (d) retracting phase.

[0029] FIG. 12 shows injection results of one zebrafish embryo with position control. (a) position trajectory; (b) force trajectory.

[0030] FIG. 13 shows injection results of one zebrafish embryo with force control. (a) position trajectory; (b) force trajectory.

DETAILED DESCRIPTION OF THE INVENTION

[0031] Microinjection has two main requirements namely that the injection needle is aligned with the target and that the needle remains within the target while the desired material is being injected into the target.

[0032] The present invention provides a new injector system with both position and force sensors. Such system typically further includes a micromanipulator and optical microscope to assist in aligning the injector with the target.

[0033] Suitable piezoelectric actuators for use in the present invention include piezoelectric stacks in rectangular, parallelepiped or cylindrical shapes.

[0034] Suitable displacement amplifiers include bridge-type and compound bridge type flexure-based amplifiers.

[0035] Suitable integrated position sensors include resistance based and capacitive-based sensors.

[0036] Suitable force sensors include resistance-based and capacitive-based sensors. Suitable injection pipettes include glass micropipettes and metal micropipettes.

[0037] The components of the system are configured so that the piezoelectric actuator provides an accurate location of the target for injection such as a biological cell, for example by use of a bridge-type displacement amplifier of the type shown in FIG. 1. Furthermore the sensors are configured to enable control of the force used to puncture the target and effect retraction from the target after injection has been completed. Information from the sensors is used to control the movement of the pipette, for example by use of a proportional-integral-derivative (pid) controller.

[0038] Operation of the system of the present invention may conveniently be carried out on accordance with the flow chart shown in FIG. 8.

[0039] FIG. 3 illustrates a CAD model of the designed flexure-based injector along with PZT actuator and strain-gage sensors of the present invention. A prototype of the flexure-based injection mechanism is developed as depicted in FIG. 4 later. It is fabricated with Al-7075 alloy using wire electrical discharge machining. The mechanism is driven by a piezoelectric actuator (PEA) (model: P-885.91, from Physik Instrumente, Co.). The PEA is actuated by a high-voltage amplifier (model: EPA-104, from Piezo System, Inc.), by which the voltage will be enlarged by ten times to drive the flexure-based compliant injector and deliver a maximum displacement of 200 m. Moreover, FEA simulation study shows that an output displacement of 301 m is achieved when the payload load of the injector is applied in the vertical out-of-plane direction. As compared with FEA simulation result, the experimental result is 33% smaller. The discrepancy is mainly caused by the manufacturing error of the device parameters and estimation error of the preloading.

[0040] Two half-bridge circuits are constructed by semiconductor strain gages (model: TP-3.8-350, from Bengbu Tianguang Sensor, Ltd.), which are adopted to measure the output displacement of this stage and applied force on the injection needle. The laser sensor (model: LK-H055, from Keyence Corp.) and load cell (model: GSO-10, from Transducer Technology, Inc.) are used for the calibration of the position and force sensing strain gages, respectively. A thick-walled glass capillary (borosilicate glass B100-50-10 from Sutter Instrument Co.) was pulled using micropipette puller (P-1000, Sutter Instrument Co.) to get an appropriate outer diameter (3 m) of the injection needle. The inverted microscope (model: IX81, from Olympus, Inc.) has an XY stage for positioning the petri dish. Zebrafish embryos are placed in parallel V-grooves, which are made from the 1.5% agarose gel in a petri dish. The injector can move in plane for the embryos in each groove, which simplifies 3D movement to plane motion in each groove. The calibration and experiments with zebrafish embryos are conducted at the room temperature of 24 C.

[0041] The cell injection control algorithm is implemented on Natural Instruments (NI) cRIO-9022 real-time controller integrated with cRIO-9118 reconfigurable chassis, which contains a Field-Programmable Gate Array (FPGA) module. In addition, NI-9263 analog output module with 16-bit resolution and NI-9237 bridge analog input module with 24-bit resolution are employed to produce the excitation voltage and obtain the strain gage sensor signals, respectively. The sampling rate of the system is selected as 4 kHz.

[0042] In cell injection, majority of force sensors are designed with Polyvinylidene Fluoride (PVDF) film [8,11,12,23], which have a remarkable dynamic force detection ability. However, the dynamic sensor is not suitable for the sensing of quasi-static signals. Alternatively, with the advantages of higher unit resistance and sensitivity, semiconductor strain gages are used as force sensor in this work. It works based on the piezoresistive effects of silicon and measures the change in resistance due to the accompanied strain.

[0043] Two strain gages are mounted on a cantilever-beam mylar (see FIG. 4) to construct a half-bridge circuit. It is adopted to measure the injection force in this work. The experimental setup for the calibration of the developed force sensor is shown in FIG. 5. The commercial loadcell (model: GSO-10, from Transducer Techniques Corp.) with a maximum measurement range of 98.1 mN and resolution of 50 N is adopted. It is mounted on a 3-DOF high precision positioning stage (model: HTCL25-X, from Huntington Optics, Inc.). An approximate linear relation between the loadcell and strain gage signal values is observed. The sensitivity of strain gages is estimated as 14.204 mN/mV.

[0044] FIGS. 6(a) and (b) show that the output of the calibrated strain gage force sensor matches well with the readings of loadcell sensor. However, their greater sensitivity to temperature variations and tendency to drift are disadvantages in comparison to PVDF sensors. The temperature drift is notable, as shown in FIG. 6(c), and its 2 resolution [24] can be obtained as 0.218 mN (see FIG. 6(d)).

[0045] To construct a half-bridge circuit for the position measurement, two strain gages are mounted on the flexure of displacement amplifier, as shown in FIG. 4. For the position sensor calibration, the displacement of the pipette is measured by a laser displacement sensor (model: LKH055, from Keyence Corp.), and the corresponding output value of strain-gage sensor is recorded. The linear relation between the laser output and strain gage values is found, and the sensitiveness of strain gages is estimated as 14.42 m/mV. It can be found from in FIGS. 7(a) and (b) that the calibrated strain-gage sensor matches well with the laser sensor. When a zero voltage input is applied, the noise of the strain-gage sensor is sampled as shown in FIG. 7(c). The noise follows a normal distribution as shown in FIG. 7(d). Therefore, the 2 resolution of the strain gage position sensor can be obtained as 0.054 m.

[0046] By inspecting the results in FIGS. 6(c) and 7(c), it is observed that the temperature drift is decreased in the position sensor in comparison with the force sensor using the same half-bridge signal conditioning circuit. The reason may lie in that the temperature on position sensor is more stable than that on force sensor. The displacement amplifier, which is made from aluminum material, has a better thermal conductivity than the flexure material of mylar for the force sensor. As a result, the imperceptible temperature change has less influence on the position strain-gage sensor.

[0047] The measurement range and sampling rate of position and force sensors are limited by the equipment used. For example the NI-9237 bridge module, has 25 mV analog input range and 50 Ks/s sampling rate. Hence, although a force measurement range with 355 mN and position measurement range with 360 m can be achieved in theory, these ranges may be more limited in due to the initial installing deformation.

[0048] To compare the performances of cell injection system with the traditional position control and position/force control during cell puncturing process, the same controller structure is adopted. In particular, owing to its popularity and model-free nature, the incremental PID controller is employed as the position and force controllers for cell injection system. By defining e(t)=x(t)x.sub.r(t) as the position/force tracking error, the control action is derived as follows.

u(t)=u(tT)+K.sub.p[e(t)e(tT)]+K.sub.ie(t)+K.sub.d[e(t)2e(tT)+e(t2T)](3)

[0049] where x is position/force variable, x.sub.r denotes the desired position/force trajectory, T is the sampling time interval, and u(tT) denotes the control variable in the previous time step, K.sub.p, K.sub.i and K.sub.d are the positive gains to be tuned.

[0050] For cell microinjection testing, the fabricated injector is mounted on an XYZ micromanipulator (model: MP-285, from Sutter Instrument Corp) to develop an automatic cell injection system. In the experimental study, a batch of zebrafish embryos are immobilized on the V-groove agarose gel in a petri dish, which is placed on the platform of an inverted microscope. The injector tip can move in plane for embryos in each V-groove. After the initialization and calibration of the CCD camera, the relative position between the pipette and target embryo can be recognized with computer vision algorithm. The micropipette moves from the initial position to the target cell controlled by the XYZ micromanipulator first. Once the pipette needle contacts with the embryo, which can be detected by the integrated force sensor, the coarse movement of micromanipulator is switched to the precision control of the piezo-driven injector. Two control schemes are realized in this work, i.e. (a) the traditional injection sequence with position control and (b) the improved operation flow with force control intervened.

[0051] A predefined position/force trajectory is used as a reference based on the trigger value for pipette puncturing until a dramatic force drop is detected. Then, a time interval is reserved to execute practical injection task, i.e. injecting the desired material. After that, a gentle retracting movement is applied under the precise position control to avoid throwing out the injected material by the retracting pipette. Finally, when the piezoelectric actuator returns to its home position, the micromanipulator translates back with a safe distance to prevent the pipette needle from crashing into the embryo during moving to the next embryo. When the next embryo comes into camera view, the above injection process is repeated. The motion sequence is shown in FIG. 8.

[0052] Experimental Study and Results

[0053] In this section, experimental study on cell injection is carried out using the developed injector with the foregoing control schemes.

[0054] Cell Injection Experiments

[0055] In practice, majority of unsuccessful cell injections are caused by the fact that, the computer believes that the target cell had been punctured while the pipette needle did not do that in reality. Actually, it is difficult to distinguish with the visual feedback whether the target embryo has contacted with micropipette or just slipped over.

[0056] For example, two injection scenarios are depicted in FIG. 9. FIG. 9(a) shows that the pipette is puncturing the embryos while FIG. 9(b) illustrates the case when the pipette slips over. There is no clear difference between these two scenarios in the pictures. Adding another camera to monitor the vertical movement of pipette can be a possible solution [25]. But it increases the system complexity at the same time. Alternatively, with the force feedback, the problem is easier to solve. Specifically, when the pipette needle contacts with target embryo, the force sensor will give a force feedback. Moreover, it is easy to identify the cell injection state, i.e. whether pierced or unpierced. The force feedback will drop dramatically once the cell membrane is punctured, which can be detected by the integrated force sensor in the proposed injection system. For instance, FIG. 10 shows the force sensor output signals for the pierced and unpierced cells. The difference between the pierced and unpierced situations can be identified easily from the force curves.

[0057] The collected embryos are spread on the V-grooves and aligned with manual assist carefully. A total of 100 zebrafish embryos are divided into to two equal groups, which are continuously injected using the proposed injection system with position and force control, respectively. The average injection time for each embryo is reduced to less than 3 s, which is better than that of a proficient human operator. The injection speed is set as 2 mm/s and 20 mN/s for position and force control, respectively. A gentle retracting speed is set as 220 m/s. The force trigger threshold from coarse to fine movement is 500 N, the punctured force velocity trigger is 10 mN/s. In addition, once the embryo is pierced, a period of 500 ms is reserved to inject the desired material. The safe distance of coarse movement is set as 5 mm. By trial and error, the parameters of PID position controller are tuned as K.sub.p=210.sup.3, K.sub.i=610.sup.4, and K.sub.d=1.510.sup.3. The parameters of PID force controller are adjusted as K.sub.p=610.sup.3, K.sub.i=910.sup.4, and K.sub.d=110.sup.3. FIG. 11 shows the sequence pictures during the microinjection process of a zebrafish embryo. The injected embryos are cultured at room temperature of 24 C.

[0058] It is found that the change rate of the temperature drift is quite low (10 mN/s). In practice, an efficient and quick manipulation behavior (20 mN/s) during the cell injection is preferred. Hence, the velocity of variation between the noise and working force can be differentiated conveniently. In this work, a change rate limit of 10 mN/s is used to extract the force signal from noise during algorithm implementation, which means that only a rapid change signal (i.e. puncturing force) can pass this gate while the noise is resisted.

[0059] To reduce the temperature drift of strain-gage force sensor further, the motion sequence of the automatic cell injection system is specifically designed. The force output is set as zero before the signal value reaches contact force threshold. The puncturing time is quite short (0.1 s), which reduces the drift impact further. The force is reset to zero again after each cell membrane is punctured.

[0060] Results

[0061] The experimental results of position and force control are shown in FIGS. 12 and 13, respectively. The force trajectory is smoother with force control (see FIG. 13(b)) while there are some burrs in the force trajectory when position control only is used for cell injection (see FIG. 12(b)). In order to quantitatively evaluate the performance of the automated injection system, the success rate and survival rate are usually adopted as evaluation criteria to characterize a cell injection system [26]. The success rate is defined as the ratio between the number of embryos with a dramatic drop force curve and the total number of injected embryos. With the force feedback, the unsuccessfully injected embryos can be detected and eliminated during the cell injection process. In this work, all of the manipulated cells are injected successfully.

[0062] The survival rate is defined as the ratio between the number of injected embryos that are capable of developing into larva and the total number of embryos injected. It essentially represents the severity and frequency of cell damage due to the injection. Based on the 100 injected zebrafish embryos with position control and force control, the survival rates are calculated as 82 and 86%, respectively.

TABLE-US-00002 TABLE 2 Comparison of experimental results. Survival rate Method Cell type (number) (%) Manual operation Zebrafish embryo 50-70 Piezo-drill system [14] Mouse oocytes (322) 80 Piezo-driven system [27] Intracytoplasmic Sperm 88 (25) Piezo-driven ultrasonic method [19] Zebrafish embryo (200) 80.7 Piezo-driven with position control Zebrafish embryo (50) 82 (this work) Piezo-driven with force control Zebrafish embryo (50) 86 (this work)

[0063] The survival rates obtained by the existing approaches are compared as shown in Table 2. It is observed that the force control produces a higher survival rate than conventional position control in the same operation environment. With force control involved, the achieved survival rate of 86% is better than majority of existing works. It is notable that the survival rate can be further improved with more delicate care during the embryo culturing and precision control algorithm design. Even so, it is better than the survival rate of 50-70% in manual operation which is caused by the proficiency difference and fatigue problems. The improvement of survival rate with force control is dominantly contributed by the smoother force interaction and the greatly reduced embryo deformation.

[0064] With the force-controlled injection system, another remarkable improvement is the injection speed. With the advantages of force feedback and PZTs, quick response and assured pierced state are achieved without time delay and misjudgment. Injection time of 3 s per embryo can be achieved, i.e. 20 embryos/min. The injection speed is much quicker than the manual injection speed of about 8-15 embryos/min. In common practical experiments, the cell number to be processed generally is about 300-500. The improvement of the cell manipulation speed can reduce the individual difference effectively for practical applications, such as drug development and gene therapy.

[0065] Further improvement of the proposed injection system will be focused on higher-throughput automatic cell injection and force sensor resolution.

[0066] Experimental results confirm the superiority of the force control vs. conventional position control in the cell injection system. Despite the difference size between the commonly used objects in biological injection experiments, this proposed system can be easily applied to other cell injection applications, such as mouse oocytes, graine, and other types of suspended cells.

[0067] To demonstrate the superiority of the developed system over skilled human operator, practical experimental studies have been conducted. Different from the previous experimental study which mimics the material delivery process, the sgRNA and Cas9 mixture is injected into Zebrafish embryos for gene editing. In this work, the drug dose of 2.2 nL is injected to every cell. Zebrafish embryos are placed in parallel V-grooves in a petri dish, and the V-grooves are made from the 1.5% agarose gel. The petri dish is mounted on an XY stage, which is used along with the microscope. The tip diameter of the adopted injector is around 1 m and the depth of injection is set as 700 m.

[0068] In each experimental setup, a batch of 69 cells are injected by the automated microinjection system. For more convincing comparison, the same operation is also conducted by a skilled human operator who has more than three years' manual injection experience. The experiment has been repeated by 5 times. Results indicate that the developed microinjection system provides the average injection speed of 3.55 s per cell, which is 42% slower than the skilled manual speed of 2.50 s per cell. On the other hand, the results reveal that the automated system achieves the average cell viability of 46.0%, which is 18% higher than the manual operation result of 39.0%. Moreover, concerning the cell viability, the standard deviation of the microinjection system's result is 0.37, which has been reduced by 91% in comparison with the manual operation result of 4.24.

[0069] Considering that the injection speed of the automated system can be greatly enhanced by employing multiple machines working in parallel, the developed microinjection system is superior to manual manipulation in terms of consistency and viability of the injected cells, which are the key indicators of the cell injection quality. Therefore, the experimental results confirm the effectiveness of the developed force-sensing microinjection system for multi-cell injection task. It is notable that the injection speed of the system can be improved by employing more advanced image acquisition and processing hardware.

REFERENCES

[0070] [1] Howe K, Clark M D, Torroja C F, et al. The zebrafish reference genome sequence and its relationship to the human genome. Nature. 2013;496(7446):498-503.

[0071] [2] Sun Y, Nelson B J. Biological cell injection using an autonomous microrobotic system. Int J Rob Res. 2002;21(10-10:861-868.

[0072] [3] Yan X, Sun D. Multilevel-based topology design and cell patterning with robotically controlled optical tweezers. IEEE Trans Control Syst Technology. 2015;23(1):176-185.

[0073] [4] Permana S, Grant E, Walker G M, et al. A review of automated microinjection systems for single cells in the embryogenesis stage. IEEE/ASME Trans Mechatron. 2016;21(5):2391-2404.

[0074] [5] Karimirad F, Chauhan S, Shirinzadeh B. Vision-based force measurement using neural networks for biological cell microinjection. J Biomech. 2014;47(5):1157-1163.

[0075] [6] Zhang Y, Tan K K, Huang S. Vision-servo system for automated cell injection. IEEE Trans Indus Electron. 2009;56(1):231-238.

[0076] [7] Ammi M, Ferreira A. Biological cell injection visual and haptic interface. Adv Rob. 2006;20(3):283-304.

[0077] [8] Pillarisetti A, Pekarev M, Brooks A D, et al. Evaluating the effect of force feedback in cell injection. IEEE Trans Autom Sci Eng. 2007;4(3):322-331.

[0078] [9] Shen Y, Fukuda T. State of the art: micro-nanorobotic manipulation in single cell analysis. Rob Biomimetics. 2014;1(1):1-13.

[0079] [10] Lu Z, Chen P C, Nam J, et al. A micromanipulation system with dynamic force-feedback for automatic batch microinjection. J Micromech Microeng. 2007;17(2):314-321.

[0080] [11] Xie Y, Sun D, Tse H Y G, et al. Force sensing and manipulation strategy in robot-assisted microinjection on zebrafish embryos. IEEE/ASME Trans Mechatron. 2011;16(6):1002-1010.

[0081] [12] Huang H B, Sun D, Mills J K, et al. Robotic cell injection system with position and force control: Toward automatic batch biomanipulation. IEEE Trans Rob. 2009;25(3):727-737.

[0082] [13] Xu Q. Digital integral terminal sliding mode predictive control of piezoelectric-driven motion system. IEEE Trans Indus Electron. 2016;63(6):3976-3984.

[0083] [14] Ron-El R, Liu J, Nagy Z, et al. Intracytoplasmic sperm injection in the mouse. Human Reprod. 1995;10(10:2831-2834.

[0084] [15] Huang T, Kimura Y, Yanagimachi R. The use of piezo micromanipulation for intracytoplasmic sperm injection of human oocytes. J Assisted Reprod Genet. 1996;13(4):320-328.

[0085] [16] Ediz K, Olgac N. Effect of mercury column on the microdynamics of the piezo-driven pipettes. J Biomech Eng. 2005;127(3):531-535.

[0086] [17] Fan M, Feng Z, Agca Y, et al. Vibration study of the piezodriven pipettes immersed in viscous liquids. J Appl Phys. 2006;100(7):074701.

[0087] [18] Gan Y, Chen Z. A study of the zona piercing process in piezodriven intracytoplasmic sperm injection. J Appl Phys. 2008;104(4):044702.

[0088] [19] Huang H, Mills J K, Lu C, et al. A universal piezodriven ultrasonic cell microinjection system. Biomed Microdevices. 2011;13(4):743-752.

[0089] [20] Avci E, Hattori T, Kamiyama K, et al. Piezo-actuated parallel mechanism for biological cell release at high speed. Biomed Microdevices. 2015;17(5):1-10.

[0090] [21] Xu Q. Precision motion control of piezoelectric nanopositioning stage with chattering-free adaptive sliding mode control. IEEE Trans Autom Sci Eng. 2017;14(1):238-248.

[0091] [22] Xu Q, Li Y. Analytical modeling, optimization and testing of a compound bridge-type compliant displacement amplifier. Mech Mach Theory. 2011;46(2):183-200.

[0092] [23] Wei Y, Xu Q. Design of a PVDF-MFC force sensor for robot-assisted single cell injection. IEEE Sens J. 2017;17(13):3975-3982.

[0093] [24] Xu Q. Design and smooth position/force switching control of a miniature gripper for automated microhandling. IEEE Trans Indus Inf. 2014;10(2):1023-1032.

[0094] [25] Wang W, Liu X, Sun Y. Contact detection in microrobotic manipulation. Int J Rob Res. 2007;26(8):821-828.

[0095] [26] Wang W H, Liu X Y, Sun Y, et al. High-throughput automated injection of individual biological cells. IEEE Trans Autom Sci Eng. 2009;6(2):209-219.

[0096] [27] Yi Z, KokKiong T, Sunan H. Software based vision system for automated cell injection. Proceedings of International Conference on BioMedical Engineering and Informatics (BMEI 2008); Sanya, China; 2008. p. 718-722.