Next Article in Journal
Monte Carlo Sampling of Inverse Problems Based on a Squeeze-and-Excitation Convolutional Neural Network Applied to Ground-Penetrating Radar Crosshole Traveltime: A Numerical Simulation Study
Previous Article in Journal
A New Last Mile Delivery Approach for the Hybrid Truck Multi-Drone Problem Using a Genetic Algorithm
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 14
Issue 2
10.3390/app14020617
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

High-Speed Cell Assembly with Piezo-Driven Two-Finger Microhand

by
Yue Zhao
1,
Yan Deng
1,
Junnan Chen
1,
Masaru Kojima
2,
Qiang Huang
1,
Tatsuo Arai
1,3 and
Xiaoming Liu
1,*
1
School of Mechatronical Engineering, Beijing Institute of Technology, Beijing 100081, China
2
Department of Materials Engineering Science, Osaka University, Osaka 560-8531, Japan
3
Center for Neuroscience and Biomedical Engineering, The University of Electro-Communications, Tokyo 182-8585, Japan
*
Author to whom correspondence should be addressed.
Appl. Sci. 2024, 14(2), 617; https://doi.org/10.3390/app14020617
Submission received: 24 December 2023 / Revised: 9 January 2024 / Accepted: 9 January 2024 / Published: 11 January 2024
Browse Figures
Review Reports Versions Notes

Abstract

:
In the past few decades, researchers have conducted extensive studies on cell micromanipulation methods. However, there has consistently been a lack of a micromanipulation system that excels in both precision and speed. Additionally, many of these methods rely on manual control, thus significantly reducing efficiency. In this paper, a robotized micromanipulation system employing a two-finger microhand is proposed. The microhand has a 3-DoF parallel mechanism driven by three piezoelectric actuators, enabling high-precision micromanipulation. Replacing the needle-tip end-effector with a hemispherical end-effector makes cell grasping easier and more stable. In addition, a vibration-based release method combined with gel coating is proposed to reduce the release difficulty caused by adhesion forces. Through multiple sets of experiments, we have determined the optimal grasping and releasing conditions while balancing precision, stability, and damage degree to cells. An automated cell assembly strategy based on microscopic visual feedback and pick-and-place path planning is proposed to achieve the robotized high-speed cell array. Hela cells were chosen as the operation objects, achieving a 95% success rate in grasping and a 97% success rate in releasing. A “T” letter array formed by cells was successfully assembled with an average grasp and release time of less than 0.8 s and an assembly accuracy of 4.5 μm for a single cell. This study holds significant implications for the fields of biology and medicine, presenting potential applications in tissue engineering.

1. Introduction

Cells are the fundamental building blocks of the human body, and in-depth research on cells is crucial for revealing life’s mechanisms and conquering major diseases like cancer. Due to the unique actuation precision, micromanipulation systems have been used for single-cell operations and analyses, such as measurement of cellular mechanical properties [1,2] and in vitro fertilization (IVF) [3]. Furthermore, the manipulation of multiple cells also receives widespread attention, such as bottom-up tissue engineering, which is expected to regenerate human tissues and organs for transplantation [4,5,6,7,8]. Cell manipulation techniques are gradually translating from laboratory to clinical applications, but the current mainstream methods still fall short in terms of precision and efficiency due to the reliance on manual control. Researchers are exploring methods that can achieve high-precision and high-speed micromanipulation at the cellular level.
The dimensions of cells typically range from a few micrometers to several tens of micrometers, posing a challenge to the precision of micromanipulation systems. Micromanipulation systems come in various types, primarily categorized as contact and non-contact. The latter includes optical tweezers [9,10], magnetic control [11,12], acoustical tweezers [13,14], microfluidics [15,16,17], and others. Among different types of micromanipulation systems, mechanical contact manipulation occupies a significant position due to its high flexibility, controllable force, and ease of automation [18,19,20,21,22]. Firstly, these systems offer controllable forces, thus avoiding impact on the biological target’s activity [23,24]. Secondly, robotic contact manipulations exhibit high precision, stability, and repeatability. Thirdly, they are easily programmable and the position and force feedback information can be obtained through sensors and image processing algorithms for automation [25]. Traditional solutions, such as precision lead screws driven by servo motors, face difficulties in achieving spatial positioning at the micro or nanometer scale due to issues of thread clearance, friction, and dynamic response speed. Piezoelectric ceramics, characterized by fast response, high resolution, and substantial thrust, are suitable for high-precision and small-range micromanipulations [26,27]. Moreover, in-series mechanisms exhibit low stiffness and error accumulation, resulting in reduced precision of the end-effector, while parallel mechanisms offer high stability and precision [28,29]. Therefore, a parallel mechanism driven by piezoelectric ceramics can meet the precision requirements for micromanipulation at the cellular level, and researchers have conducted extensive studies on it. Taniguchi designed a flexible 6-SPS (6-spherical-prismatic-spherical) micro-positioning system, which was capable of positioning the wafer with a resolution of less than 0.01 μm [30]. Polit S. developed a high-bandwidth XY nano-positioning stage that utilized two capacitance sensors for closed-loop position control, with a travel range of 15 μm and a precision of 1 nm [31]. Dan Zhang also proposed a 3-DoF parallel mechanism driven by prismatic actuators composed of compliant joints and links and flexure hinges [32]. Traditional hinges primarily achieve rotation through physical contact and mutual friction, but they are prone to backlash, making high-precision operations challenging. The motion or force output from flexible hinges is generated through the elastic deformation of their flexible structures, thereby enhancing the precision and stability of operations.
The improvement of cell manipulation speed is another challenge for clinical applications. One constraint on speed is the success rate of grasping and releasing. Currently, the most common end-effectors are needle-shaped microgrippers, which are prone to misalignment during grasping, leading to grasping failure or target dropping during transportation. Moreover, microscale adhesive forces make cells easily adhere to the end-effector, causing significant difficulty in releasing [33]. Furthermore, the operation efficiency by manual control is low and the operator’s experience will also impact it. There is an urgent need for a robotized approach to manipulate cells automatically. In the past decades, there has been much study on cell recognition, and the iterative upgrade of visual recognition algorithms has improved the speed and accuracy of cell recognition [34,35,36,37,38]. To further increase the manipulation speed, operation path planning is also essential to achieve maximum task efficiency when given a known target position.
To address the lack of micromanipulation methods that balance precision and speed in the field of cell manipulation, this paper presents a high-speed micromanipulation system based on a two-finger microhand. The 3-DoF parallel mechanism driven by piezoelectric actuators endows the microhand with high operation precision and fast response speed. The utilization of a hemispherical end-effector and a vibration-based release method, combined with a gel coating, enhances the success rate of object grasping and releasing. Additionally, optimal grasping and releasing conditions were determined through experiments, aiming to improve operation accuracy while minimizing damage to cells. Finally, this study achieved automated high-speed cell assembly based on microscopic visual feedback and operational path planning. This proposed method holds significant potential for applications in tissue regeneration, organ cultivation, and other related fields in the future.

2. System Setup

2.1. Two-Finger Microhand Structure

The dexterous two-finger microhand is used to grasp and release cells by adjusting the distance between the two needle tips. One of the two fingers is an active finger, which is a 3-PRR (3-prismatic-revolute-revolute) parallel mechanism driven by three piezoelectric actuators, as shown in Figure 1a. The end-effector is installed on the parallel mechanism, which is driven by piezoelectric ceramics (NEC TOKIN AE0203D16, Shiroishi, Japan), as indicated by the green strip in Figure 1a, with a maximum deformation length of 17 μm. The schematic diagram of its joint design is also shown in Figure 1a, including three prismatic joints formed by piezoelectric actuators and eight revolute joints formed by flexible hinges. The motion of every joint is achieved through the designed flexure hinge. The positioning accuracy of the end-effector can reach sub-millimeter levels and, in practical micromanipulations, we have chosen a minimum displacement resolution of 1 μm.
During operation, the mechanism is installed in a way that the plane formed by the X and Y axes coincides with the horizontal operating plane, while the Z-axis aligns with the vertical direction perpendicular to the focal plane of the microscope.
Due to the asymmetry of the 3-PRR structure in this paper, algebraic methods would result in a large computational burden. Therefore, a geometric approach was employed for the inverse kinematic analysis. As shown in Figure 1a, motion chain 1 is composed of P1, R1, R6, and R7. Motion chain 2 is composed of P2, R2, and R4. Motion chain 3 is composed of P3, R3, and R5. As the revolute joint R1 cannot achieve motion when the prismatic joint P1 moves independently, R1 is considered a rigid body when analyzing motion chain 1. As illustrated in Figure 1b, for motion chain 1, with the displacement of the prismatic joint P1, the links FA and AB reach the positions of F′A′ and A′B. The red and blue circles represent the trajectories of points A′ and F′, respectively. The length FF′ can be calculated based on the geometric relationships in this figure, denoted as ∆p1. Motion chains 2 and 3 are symmetric, and the inverse kinetic analysis can be conducted for one, with the joint displacement of the other being opposite. Additionally, the geometric kinematics analysis for motion chain 2 is shown in Figure 1b. With the displacement of the prismatic joint P2, the links AB and BC reach the positions of A′B′ and B′C. The red and blue circles represent the trajectories of points B′ and A′, respectively. The length of AA′ can be calculated based on the geometric relationships in this figure, denoted as ∆p2. The ∆p3 in motion chain 3 is of the same magnitude as ∆p2 but in the opposite direction.
The inverse kinematics analysis for motion chain 1 involves the rotational angle α around the Z-axis, while the inverse kinematics analysis for motion chains 2 and 3 involves the rotational angle β around the Y-axis, as shown in Figure 1a. Translational motion in the X-axis direction is achieved by simultaneous equal displacements of the three prismatic joints. Therefore, when the target pose of the end-effector relative to its original position is represented as [x α β], the displacement changes of the three prismatic joints can be deduced by synthesizing the kinematic relationships of each motion chain.
Based on the inverse kinematic analysis of the three motion chains in Figure 1b, the corresponding relationships between the displacement of each prismatic joint and the pose changes of the microhand around each axis have been determined, as shown in Equations (1)–(4):
Δ p 1 = l 1   cos θ 1 + l 2   sin α − l 1 2 − x a ′ 2
Δ p 2 = l 4   cos θ 4 + l 5   sin   ( β − θ 3 ) − l 4 2 − B ′ H 2
∆p3 = −∆p2
p = [p1, p2, p3]′ = [∆p1 + x, ∆p2 + x, ∆p3 + x]′
where ∆p is the displacement of each prismatic joint; the definitions of l, θ, α, β, etc., are marked in Figure 1b; and x is the displacement of each joint when the three joints move together to achieve translational motion along the X-axis.
Following the mechanism analysis and mathematical modeling, the design parameters of the theoretical 3-PRR parallel mechanism were optimized using methods such as discretization and genetic algorithms. Finally, complementary optimization was performed using ANSYS (Canonsburg, PA, USA) to determine the optimal characteristics of flexible hinge joints.

2.2. Construction of Micromanipulation Robot System

Figure 2 illustrates the configuration diagram of the micromanipulation robot system, mainly divided into the motion control part and the image processing part. The motion control part is utilized to position the cells into an array. The image processing part can send real-time image information to provide visual feedback for automated cell assembly. The two parts are controlled through a PC (Windows).
The motion control part is mainly conducted by a piezo-driven two-finger microhand mounted on a 3-DoF motorized positioning stage, and a 3-DoF stage for cell transportation, as shown in Figure 2. The two-finger microhand is used to grasp the cells and release them at the desired position. The whole microhand is held by an X–Y–Z stage and can be moved in a large 3-D workspace, which enables a substantial travel distance in the cell assembly while maintaining high precision. The 3D stage for cell transportation (sample holder) is utilized to switch between different fields of view to search for cells.
The three piezoelectric actuators of the microhand are driven by a voltage amplifier (Matsusada Precision Inc., PZJ-0.15P, Shiga, Japan). The voltage range outputted to the microhand ranges from 0 to 150 V, and this voltage value will be initialized to 75 V before each use to calibrate the end-effector to the center of the workspace. Two strain gauges are orthogonally attached to the piezoelectric actuator for close-loop control and hysteresis avoidance. The maximum error caused by hysteresis in the piezo actuator is approximately around 0.16 μm. The feedback signal is conditioned by the strain gauge bridge box and amplified by the amplifier (KYOWA, MCD-8A, Tokyo, Japan). The signal is then input into the microhand controller (AD chip ADC78H90, DA chip DAC-AD5363, MCU STM32 F767ZI, STMicroelectronics, Plan-les-Ouates, Switzerland) for AD conversion. The onboard motion control algorithm of both the microhand and stages is executed on this controller. The feedback program in the control board continuously measures the differences between the actual output and the target values and then adjusts the output in real-time based on these differences. This adjustment is facilitated by a proportional-integral (PI) controller, which compares the current and historical errors to calculate both the proportional and integral terms. The X–Y–Z stage (OptoSigma, TAM 655, Tokyo, Japan) is motorized by three linear motors (OptoSigma, SGSP-13ACT-B0) with a maximum travel distance of 13 mm and a maximum speed of 2 mm/s.
The image processing part includes an inverted microscope (OLYMPUS IX73, Tokyo, Japan) for observing the microhand and obtaining the position and status information of the cells. The inverted microscope features a six-slot objective turret that can achieve various magnifications, including 4×, 10×, 20×, 40×, and more. All the experimental processes were recorded using a CCD camera (HIKROBOT, MV-CA050-11UM, Hangzhou, China) that offered a framerate of up to 35 fps under a maximum resolution of 2448 × 2048. The real-time images were processed by computer vision (OpenCV 3.0), providing position and posture feedback to the microhand controller.

3. Methods of Grasping and Releasing Cells

3.1. Hemispherical End-Effector

Current end-effectors are mostly needle-tip shaped, leading to alignment difficulties in the operation plane as well as weak grasping force and susceptibility to vibration interference caused by low end-effector stiffness. To enhance the stability and success rate of grasping, a new hemispherical end-effector was designed, as shown in Figure 3a.
The end-effector fabricated by this method increases the contact area between the cell and the end-effector, transitioning from linear contact to surface contact. Even slight alignment errors during installation do not affect force balance. Therefore, it saves installation time and enhances the stability of grasping.

3.2. Combination of Gel Coating and Vibration-Based Release

Due to the significant impact of adhesive forces at the microscale, addressing the release difficulty caused by adhesion between the target and the end-effector is challenging. In this paper, the active finger connected to the parallel mechanism is capable of releasing cells adhered to it through high-speed vibration. The high-frequency vibration of the end-effector is generated by the piezo driver outputting a high-frequency sinusoidal signal to the piezoelectric ceramics. However, when cells adhere to the other finger, the active release becomes ineffective and release relies on the drag force generated by its vibration-induced microflow. To simplify this process, we transformed the uncertainty of adhesion into certainty by coating the end-effector of the active finger with gel, as shown in Figure 3b. Consequently, the target is consistently adhered to the active finger with stronger adhesion, and release can be achieved through its high-speed vibration in the y-direction.

3.3. Determination of the Optimal Grasping and Releasing Condition

Due to significant variations among individual cells, different cell sizes and shapes can affect grasping and releasing. Therefore, conducting experiments directly on cells to determine the optimal manipulation parameters is challenging and cumbersome owing to large result variances. Therefore, borosilicate glass microspheres with dimensions close to those of cells and more regular shapes serve as better experimental objects. In this study, 50 μm microspheres were used as the operating objects, and, in 200 repetitions of grasping and transporting experiments, the success rates of grasping and transporting with the hemispherical end-effector both reached 100%. As for the releasing effect of the vibration-based release method combined with a gel coating, the amplitude and frequency of vibration still need to be determined. The small amplitude and frequency can lead to failure in overcoming adhesive forces, while the large amplitude and frequency can result in significant deviation from the desired position after release. Table 1 presents the experimental results of the release success rate under different vibration amplitudes and frequencies.
From Table 1, it can be observed that when the frequency is 5 μm and the amplitude is 25 Hz, the release success rate has already reached 98%. When the amplitude is greater than 15 μm or the frequency is higher than 75 Hz, the release success rate could reach 100%.
Next, multiple release experiments were conducted under different frequencies and amplitudes to analyze the release accuracy under different experimental conditions statistically. The results indicate that whether maintaining a constant frequency or amplitude, the post-release position consistently remains in the upper-left direction from the pre-release position. This suggests the presence of a consistent horizontal leftward and vertical upward displacement and the displacement shows systematic variation with changes in amplitude and frequency. The accuracy graphs in the horizontal and vertical directions after release are shown in Figure 4a and Figure 4b, respectively.
The vertical deviation is generated by repeated vertical vibrations. While ensuring a high release success rate, the optimal amplitude and frequency values that minimize the vertical deviation can be determined easily through these repeated experiments. The vertical release deviation graph in Figure 4b shows that the optimal release accuracy, at 2.4 μm, is achieved at a frequency of 25 Hz and an amplitude of 5 μm.
The horizontal deviation is caused by the unique shape of the hemispherical end-effector, resulting in microflow motion towards the left. According to the experimental results in Figure 4a, under different frequency and amplitude conditions, the initial release position has been adjusted to the right by the corresponding deviation to compensate. The compensated horizontal deviation results are shown in Figure 4c. The compensated horizontal release accuracy remains stable within 0.3 μm.
Based on the above experiments, we selected the release condition with an amplitude of 5 μm and a frequency of 25 Hz according to the vertical deviation test as this resulted in the smallest vertical deviation. For the horizontal direction, we chose a compensation of 5.4 μm to the right based on the horizontal deviation test results. Under this condition, we achieved optimal release accuracy and applied it as an ideal universal parameter for cell release. Through experiments, we found that the release accuracy of cells under this condition also reached its optimum. Therefore, this release condition was selected for subsequent cell array experiments.

4. Automated Assembly Strategy for Cells and Experimental Results

4.1. Visual Recognition of the Operational Target and End-Effector

The Hough circle detection method can be employed for microsphere targets. However, this method exhibits a low success rate in recognizing cells due to their irregular circular shape. In this paper, a combination of contour detection and various image processing algorithms is adopted for cell recognition. Firstly, the image undergoes Gaussian filtering and binary conversion. After binary processing, almost every cell in the images has many holes in the center, which is caused by the semi-transparent state of cells in the liquid. To make the region where cells are located a closed black area, morphological operations are employed. Subsequently, the image is eroded, followed by dilation, and then edge detection and contour detection are performed sequentially. After contour detection, it is necessary to exclude interference caused by other backgrounds with closed contours. The area of dust is generally small, while cell fragments often have irregular shapes and exist in large patches. The areas of all contours in the image are calculated, and a filtering process is applied. Contours that match the cell size criteria will be retained, while contours that do not match will be eliminated from the image. The minimum circular contour bounding algorithm is applied to the filtered contours, which can determine the smallest circular region that can encompass the entire cell contour. The obtained center positions and radii of cells are prepared for subsequent automated grasping operations on the cells. Through multiple iterations of image processing, the successful recognition rate for cells reaches 98%.
The template matching method was employed for the recognition of end-effector and the final success rate also reached 98%. Template matching recognizes targets by detecting the position of a given template in the overall image. This method is suitable for scenarios where the main motion of the target is translation in a two-dimensional plane, as applied in this paper. The end-effector of the microhand is primarily involved in translational motion in the operation plane during cell manipulation without significant rotational movements.

4.2. Operational Path Planning

After successfully recognizing the target and end-effector, it is necessary to plan the operation path, including the grasping path and assembly path. Grasping path planning determines the operation sequence for targets and the movement of the end-effector approaching the targets. Assembly path planning includes the sequence of target array positions when forming a specific shape.
Due to the limited imaging range of the microscope, when the number of targets in a single microscopic field of view is insufficient to meet the assembly requirements, it is necessary to supplement targets from adjacent views. We hope to have as abundant targets as possible in the initial field of view, so our first step is the selection of the operational field of view. To avoid complications in target recognition and manipulation caused by an excessive number of cells in the field of view, we controlled the cell density to approximately five cells per average microscopic field. However, precise control of cell density in different microscopic fields is challenging, so this may result in insufficient target numbers in the current field of view during the arraying process.
To address the aforementioned issue, as illustrated in Figure 5a, we assigned numbers to the surrounding operational fields of view. Following the numerical sequence, we added the number of detected targets in the current field of view to the total number of targets detected in all previous fields of view. If the sum was greater than or equal to the required number of targets, we selected the current field of view as the operational field. After completing the manipulation in the current field of view, we sequentially returned to the previous field of view to grasp additional targets for supplementation.
For example, the target number required for assembling the letter ‘T’, as shown in Figure 5b, is six. As shown in Figure 5a, the number of targets in view 1 is four, which is insufficient, so the microscope moves to view 2. The number of targets detected in view 2 is 5, adding the 4 from view 1, for a total of 9 targets, which is sufficient for the array of the letter ‘T.’ Therefore, view 2 is selected to start the array operation, and the assembly order is shown in Figure 5b.
Due to the target array located in the upper right corner of the field of view, to avoid collisions with the targets, the grasping path planning of the end-effector is horizontal first and then vertical. For each array position, find the target with the minimum horizontal and vertical distances from it, as shown in Equation (5). This ensures the shortest path under the premise of the collision-free operation, which is defined as the scanning obstacle-free path planning algorithm.
min (|Tx − Ex| + |Ty − Ey|)
As shown in Figure 5b, after completing the operations on all targets in the current view 2, we return to view 1 to grasp the additional targets, and then return to view 2 to continue the unfinished assembly. The red circles represent the desired array positions, where (Ex, Ey) is the coordinate of the desired array position and the numbers on them represent the order of assembly. The assembly order is from top to bottom and from right to left. The gray–black circles represent the actual positions of the current targets, where (Tx, Ty) denote the actual position coordinate of the target yet to be operated and the numbers on them represent the order of operations. Target 6 is a candidate target from the previous view. Since its horizontal and vertical distance to the sixth desired position is the smallest, it is chosen as the target for supplementation.

4.3. Automated Operation Process

The automated operation flowchart for cell assembly can be implemented based on target recognition and path planning, as shown in Figure 6.
The letters and special shapes, such as triangles, have already been defined in our program. Therefore, when the desired letter or shape for automated assembly is inputted, the program will automatically calculate the required number of operational targets and the desired array positions for these targets.

4.4. Automated Assembly Experimental Results

Many researchers have conducted manipulations similar to micro-grasping. Hui Xie manipulated 20 microspheres with diameters ranging from 3 to 4 μm to construct five two-layered 3D micro-pyramids [39]. The average manipulation time for each microsphere was approximately 48 s. Yu Sun et al. proposed a robotic system based on a MEMS microgripper, using it for high-speed, fully automated pick-and-place operations on borosilicate glass spheres with diameters ranging from 7.5 to 10.9 µm [36]. The average pick-and-place time for each sphere was 6 s. These manipulations are targeted towards ideal spherical targets. Manipulating cells is challenging due to their irregular shapes and adhesive membranes, resulting in a slower manipulation speed. Zhe Lu proposed a micromanipulation system for the automated pick–place of single cells, which could accurately pick up a single cell, transfer the cell, and deposit it at the target location with a speed of 15–30 s per cell [40]. Based on the research by T. Arai et al. [41], this paper proposes a micromanipulation method for tissue engineering, enabling rapid pick-and-place. This method combines high speed and high stability when manipulating cells, addressing a previous research gap.
The cells used in this study are Hela cells, with dimensions of around 10–15 μm, smaller than the microspheres mentioned earlier. If the size of the hemispherical end-effector is too small, it may lead to low stiffness, but if it is too large, it may touch the substance’s bottom and affect grasping. After multiple experimental tests, we selected a 10 μm hemispherical end-effector, achieving a 98% success rate in cell grasping. Another difference between cells and microspheres is their susceptibility to deformation. Experimental tests revealed that when the cell deformation reaches 1 μm after grasping, the success rate is only 50%. However, when the deformation is 2 or 3 μm, the success rate reaches 95%. When the deformation exceeds 3 μm, 100% success in cell grasping can be achieved. However, larger deformations will cause damage to cells, so we chose a grasping deformation of 2 μm to balance successful grasping and minimize the risk of cell damage. Under this grasping deformation condition, we conducted 100 cell grasping experiments, achieving a final success rate of 95%. We applied the optimal manipulation parameters for microspheres to cell release, also achieving a success rate of 97% in 100 cell release experiments.
Finally, the process of automated assembly for the letter ‘T’ is shown in Figure 7. The average precision of cell assembly is 4.5 μm. The average total time of cell grasping and releasing is less than 0.8 s per cell. We conducted 20 experiments of automated assembly of letters, achieving a success rate of 90%. The main reasons for failure were attributed to grasp and release failures during the assembly process. The speed of automatic visual feedback can reach 25 frames per second, meeting our real-time detection needs. Moreover, the algorithm demonstrates robustness against background noise, effectively filtering out cell fragments and impurities in the field of view. By adjusting parameters for specific targets, we can also achieve the assembly of different biological micro-objects, such as cells of different types.

5. Discussion

For any cell manipulation, the extent of damage to the cells is a crucial concern. In this study, cell manipulation was performed using a two-finger microhand, which involves a mechanical contact microoperation. The end-effector is made of capillary glass, exhibiting good biocompatibility, and the cell deformation after grasping is controlled within 2 μm to minimize damage to the cells. Studies related to cell micromanipulations indicated that the damage caused by this single operation on cells was minimal and fell within an acceptable range. However, as the number of operations on the same cell increased, the impact on cell viability became more pronounced [42,43]. This paper only involves a single grasping and releasing of each target cell, so although there is inevitably some damage caused by contact operations, the extent of the damage is relatively low.
Although the micromanipulation robot system can currently accomplish the assembly of cell arrays, it still has several limitations. Firstly, the grasping and releasing of cells cannot achieve a 100% success rate. Grasping cells is much more challenging than microspheres due to variations in cell sizes, more complex morphologies, and unpredictable deformations that occur upon grasping. In addition, the adhesive forces of cells occasionally result in attachment at different positions of the end-effector, making detachment difficult. Secondly, there is an issue of disturbance in micromanipulations within a liquid environment. Manipulating cells in a liquid causes the movement of the end-effector to induce microflows, which is significantly harmful for high-precision operations. While compensation can enhance the placement accuracy for individual cells, continuous operations on multiple cells will cause the previously placed cells to shift due to the microflows induced by the subsequent movement of the end-effector. Thirdly, cell assembly is limited to a two-dimensional plane. Tissue engineering requires the construction of cells into three-dimensional structures. We have not yet attempted 3D cell assembly, which represents an important future research direction.
In the future, we plan to optimize our study from the following directions. Firstly, the microhand structure design can be optimized. Currently, manipulations on micro-objects in the range of tens of micrometers are easily achievable, but manipulations on targets at the level of hundreds of micrometers are challenging due to workspace limitations, which is a major drawback of parallel mechanisms. Therefore, in the future, we consider designing new serial–parallel hybrid structures, such as connecting two 3-PRR modules in series to introduce the advantage of a larger workspace. Secondly, the precision of micromanipulations needs to be further improved. This involves using higher-precision piezoelectric actuators, enhancing the accuracy of sensors, and optimizing controller algorithms. The next goal is to achieve a micromanipulation precision of 0.1 μm. Thirdly, we aim to increase the efficiency of automated operations. Enhancing target detection speed can be achieved through improvements in visual recognition algorithms. For instance, employing convolutional neural network (CNN)-based algorithms can strengthen robustness in identifying targets in complex scenes. Additionally, selecting better hardware configurations can also improve the running speed of algorithms.

6. Conclusions

Current microsurgery techniques in tissue engineering heavily rely on manual operations, which suffer from poor precision, low efficiency, and limited controllability. This paper presents a micromanipulation robot system combined with an automated assembly strategy, achieving robotized high-speed assembly of cells. The 3-DoF parallel micromanipulator, powered by piezoelectric ceramics in this study, offers advantages including a small size, high precision, and fast speed. A hemisphere-shaped end-effector is designed to address the issue of unstable target grasping at the microscale. A release method, combining gel coating with high-speed vibration, is proposed to solve the problem of target release difficulty caused by microscopic adhesion forces. The optimal vibration amplitude and frequency for releasing targets were also obtained through micro-target release experiments. By utilizing piezoelectric actuation, parallel mechanisms, and innovative methods for target grasping and releasing, along with parameter optimization, higher precision micromanipulations on cells have been achieved. Finally, we propose an automated assembly method for micro-targets based on microscopic visual feedback and operation path planning. This method achieves robotized high-speed and high-precision assembly of cell arrays. The average time for grasping and releasing cells is less than 0.8 s. Optimized visual detection algorithms tailored for cells and operation path planning segmented by field of view contribute to a significantly improved efficiency in automated micromanipulations compared to previous studies. In the future, we aim to enhance the precision of micromanipulation by improving the control accuracy of piezoelectric actuators and increasing the sensor resolution. This study holds significance in the fields of biology and medicine, enabling the three-dimensional assembly of biological targets using micromanipulation robots. This technology is expected to find applications in tissue engineering and the assembly of micromachines in the future.

Author Contributions

Y.Z. and Y.D. are co-first authors of the article. Conceptualization, T.A. and X.L.; methodology, Y.Z. and Y.D.; software, J.C.; validation, Y.Z., Y.D. and J.C.; formal analysis, Y.Z. and J.C.; investigation, Y.D.; resources, M.K., Q.H. and T.A.; writing—original draft preparation, Y.Z.; writing—review and editing, Y.D. and X.L.; supervision, M.K., Q.H. and T.A.; project administration, T.A. and X.L.; funding acquisition, Q.H., T.A. and X.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China under Grant 62273052, the Beijing Natural Science Foundation under Grant IS23062, China Postdoctoral Science Foundation under Grant 2022M710378 and the Grant-in-Aid for Scientific Research under Grant 22H01441 from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data is contained within the article.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Tsai, C.H.D.; Kaneko, M.; Sakuma, S.; Arai, F. Observability of cell stiffness in micro-channel method. In Proceedings of the 2013 IEEE International Conference on Robotics and Automation (ICRA), Karlsruhe, Germany, 6–10 May 2013; pp. 2807–2813. [Google Scholar]
  2. Tsai, C.H.D.; Kaneko, M.; Arai, F. Evaluation of cell impedance using a μ-channel. In Proceedings of the 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), San Diego, CA, USA, 28 August–1 September 2012; pp. 5518–5521. [Google Scholar]
  3. Noori, A.; Selvaganapathy, P.R.; Wilson, J. Microinjection in a microfluidic format using flexible and compliant channels and electroosmotic dosage control. Lab Chip 2009, 9, 3202–3211. [Google Scholar] [CrossRef] [PubMed]
  4. Veschgini, M.; Gebert, F.; Khangai, N.; Ito, H.; Suzuki, R.; Holstein, T.W.; Mae, Y.; Arai, T.; Tanaka, M. Tracking mechanical and morphological dynamics of regenerating Hydra tissue fragments using a two fingered micro-robotic hand. Appl. Phys. Lett. 2016, 108, 10. [Google Scholar] [CrossRef]
  5. Won, J.E.; Mateos-Timoneda, M.A.; Castano, O.; Planell, J.A.; Seo, S.J.; Lee, E.J.; Han, C.M.; Kim, H.W. Fibronectin immobilization on to robotic-dispensed nanobioactive glass/polycaprolactone scaffolds for bone tissue engineering. Biotechnol. Lett. 2015, 37, 935–942. [Google Scholar] [CrossRef] [PubMed]
  6. Oktay, K.; Taylan, E.; Kawahara, T.; Cillo, G.M. Robot-assisted orthotopic and heterotopic ovarian tissue transplantation techniques: Surgical advances since our first success in 2000. Fertil. Steril. 2019, 111, 604–606. [Google Scholar] [CrossRef] [PubMed]
  7. Sundaramurthi, D.; Rauf, S.; Hauser, C. 3D bioprinting technology for regenerative medicine applications. Int. J. Bioprinting 2016, 2, 2. [Google Scholar] [CrossRef]
  8. Zhang, H.; Hutmacher, D.W.; Chollet, F.; Poo, A.N.; Burdet, E. Microrobotics and MEMS-based fabrication techniques for scaffold-based tissue engineering. Macromol. Biosci. 2005, 5, 477–489. [Google Scholar] [CrossRef]
  9. Wang, X.; Chen, S.; Kong, M.; Wang, Z.; Costa, K.D.; Li, R.A.; Sun, D. Enhanced cell sorting and manipulation with combined optical tweezer and microfluidic chip technologies. Lab Chip 2011, 11, 3656–3662. [Google Scholar] [CrossRef]
  10. Kirkham, G.R.; Britchford, E.; Upton, T.; Ware, J.; Gibson, G.M.; Devaud, Y.; Ehrbar, M.; Padgett, M.; Allen, S.; Buttery, L.D.; et al. Precision assembly of complex cellular microenvironments using holographic optical tweezers. Sci. Rep. 2015, 5, 8577. [Google Scholar] [CrossRef]
  11. Xu, T.; Hwang, G.; Andreff, N.; Régnier, S. Modeling and swimming property characterizations of scaled-up helical microswimmers. IEEE/ASME Trans. Mechatron. 2013, 19, 1069–1079. [Google Scholar] [CrossRef]
  12. Blümler, P.; Friedrich, R.P.; Pereira, J.; Baun, O.; Alexiou, C.; Mailänder, V. Contactless nanoparticle-based guiding of cells by controllable magnetic fields. Nanotechnol. Sci. Appl. 2021, 14, 91–100. [Google Scholar] [CrossRef]
  13. Cao, H.X.; Jung, D.; Lee, H.S.; Nguyen, V.D.; Choi, E.; Kang, B.; Park, J.O.; Kim, C.S. Holographic Acoustic Tweezers for 5-DoF Manipulation of Nanocarrier Clusters toward Targeted Drug Delivery. Pharmaceutics 2022, 14, 1490. [Google Scholar] [CrossRef] [PubMed]
  14. Wang, Y.; Pan, H.; Mei, D.; Xu, C.; Weng, W. Programmable motion control and trajectory manipulation of microparticles through tri-directional symmetrical acoustic tweezers. Lab Chip 2022, 22, 1149–1161. [Google Scholar] [CrossRef]
  15. Paterson, L.; MacDonald, M.P.; Arlt, J.; Sibbett, W.; Bryant, P.E.; Dholakia, K. Controlled rotation of optically trapped microscopic particles. Science 2001, 292, 912–914. [Google Scholar] [CrossRef] [PubMed]
  16. Yue, T.; Nakajima, M.; Tajima, H.; Kojima, M.; Fukuda, T. High speed cell manipulation by dielectrophoresis and movable microstructure embedding cells fabricated inside microfluidic chips. In Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vilamoura, Algarve, Portugal, 7–12 October 2012; pp. 938–943. [Google Scholar]
  17. Tang, X.; Liu, X.; Li, P.; Liu, F.; Kojima, M.; Huang, Q.; Arai, T. On-Chip cell–cell interaction monitoring at single-cell level by efficient immobilization of multiple cells in adjustable quantities. Anal. Chem. 2020, 92, 11607–11616. [Google Scholar] [CrossRef] [PubMed]
  18. Xie, Y.; Sun, D.; Liu, C. Penetration force measurement and control in robotic cell microinjection. In Proceedings of the 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), St. Louis, MO, USA, 10–15 October 2009; pp. 4701–4706. [Google Scholar]
  19. Liu, J.; Zhang, Z.; Wang, X.; Liu, H.; Zhao, Q.; Zhou, C.; Tan, M.; Pu, H.; Xie, S.; Sun, Y. Automated robotic measurement of 3-D cell morphologies. IEEE Robot. Autom. Lett. 2016, 2, 499–505. [Google Scholar] [CrossRef]
  20. Shen, Y.; Ahmad, M.R.; Nakajima, M.; Kojima, S.; Homma, M.; Fukuda, T. Cell-cell adhesion force measurement using nano picker via nanorobotic manipulators inside ESEM. In Proceedings of the 10th IEEE International Conference on Nanotechnology, Ilsan, Gyeonggi-Do, Republic of Korea, 17–20 August 2010; pp. 870–874. [Google Scholar]
  21. Kolb, I.; Landry, C.R.; Yip, M.C.; Lewallen, C.F.; Stoy, W.A.; Lee, J.; Felouzis, A.; Yang, B.; Boyden, E.S.; Rozell, C.J.; et al. PatcherBot: A single-cell electrophysiology robot for adherent cells and brain slices. J. Neural Eng. 2019, 16, 046003. [Google Scholar] [CrossRef] [PubMed]
  22. Lu, Z.; Moraes, C.; Ye, G.; Simmons, C.A.; Sun, Y. Single cell deposition and patterning with a robotic system. PLoS ONE 2010, 5, e13542. [Google Scholar] [CrossRef] [PubMed]
  23. Shen, Y.; Nakajima, M.; Kojima, S.; Homma, M.; Kojima, M.; Fukuda, T. Single cell adhesion force measurement for cell viability identification using an AFM cantilever-based micro putter. Meas. Sci. Technol. 2011, 22, 115802. [Google Scholar] [CrossRef]
  24. Barnes, C.A.; O’HAGAN, B.M.G.; Howard, C.V.; McKerr, G. Verification of cell viability at progressively higher scanning forces using a hybrid atomic force and fluorescence microscope. J. Microsc. 2007, 228, 185–189. [Google Scholar] [CrossRef]
  25. Kim, B.S.; Park, J.S.; Kang, B.H.; Moon, C. Fabrication and property analysis of a MEMS micro-gripper for robotic micro-manipulation. Robot. Comput. Integr. Manuf. 2012, 28, 50–56. [Google Scholar] [CrossRef]
  26. Gao, X.; Yang, J.; Wu, J.; Xin, X.; Li, Z.; Yuan, X.; Shen, X.; Dong, S. Piezoelectric actuators and motors: Materials, designs, and applications. Adv. Mater. Technol. 2020, 5, 1900716. [Google Scholar] [CrossRef]
  27. Mohith, S.; Upadhya, A.R.; Navin, K.P.; Kulkarni, S.M.; Rao, M. Recent trends in piezoelectric actuators for precision motion and their applications: A review. Smart Mater Struct 2020, 30, 013002. [Google Scholar] [CrossRef]
  28. Deabs, A.; Gomaa, F.R.; Khader, K. Parallel Robot. J. Eng. Sci. Technol. Rev. 2021, 14, 6. [Google Scholar] [CrossRef]
  29. Kumar, S.; Wöhrle, H.; de Gea Fernández, J.; Müller, A.; Kirchner, F. A survey on modularity and distributivity in series-parallel hybrid robots. Mechatronics 2020, 68, 102367. [Google Scholar] [CrossRef]
  30. Taniguchi, M.; Ikeda, M.; Inagaki, A.; Funatsu, R. Ultra precision wafer positioning by six-axis micro-motion mechanism. Int. J. Jpn. Soc. Precis. Eng. 1992, 26, 35–40. [Google Scholar]
  31. Polit, S.; Dong, J. Development of a high-bandwidth XY nanopositioning stage for high-rate micro-/nanomanufacturing. IEEE/ASME Trans. Mechatron. 2010, 16, 724–733. [Google Scholar] [CrossRef]
  32. Zhang, D.; Viegas, K.; Gao, Z.; Ge, Y. June. Modeling and analysis of an enhanced compliant parallel mechanism for high accuracy micro motion. In Proceedings of the 2008 7th World Congress on Intelligent Control and Automation (WCICA), Chongqing, China, 25–27 June 2008; pp. 2289–2294. [Google Scholar]
  33. Diller, E.; Sitti, M. Micro-scale mobile robotics. Found. Trends® Robot. 2013, 2, 143–259. [Google Scholar] [CrossRef]
  34. Xie, H.; Zhang, H.; Song, J.; Meng, X.; Wen, Y.; Sun, L. High-precision automated micromanipulation and adhesive microbonding with cantilevered micropipette probes in the dynamic probing mode. IEEE/ASME Trans. Mechatron. 2018, 23, 1425–1435. [Google Scholar] [CrossRef]
  35. Liu, H.; Wen, J.; Xiao, Y.; Liu, J.; Hopyan, S.; Radisic, M.; Simmons, C.A.; Sun, Y. In situ mechanical characterization of the cell nucleus by atomic force microscopy. ACS Nano 2014, 8, 3821–3828. [Google Scholar] [CrossRef]
  36. Zhang, Y.; Chen, B.K.; Liu, X.; Sun, Y. Autonomous robotic pick-and-place of microobjects. IEEE Trans. Robot. 2009, 26, 200–207. [Google Scholar] [CrossRef]
  37. Wang, F.; Li, P.; Wang, D.; Li, L.; Xie, S.; Liu, L.; Wang, Y.; Li, W.J. Mechanically modulated dewetting by atomic force microscope for micro-and nano-droplet array fabrication. Sci. Rep. 2014, 4, 6524. [Google Scholar] [CrossRef] [PubMed]
  38. Li, G.; Yang, J.; Wang, Y.; Wang, W.; Liu, L. Development of a novel optogenetic indicator based on cellular deformations for mapping optogenetic activities. Nanoscale 2018, 10, 21046–21051. [Google Scholar] [CrossRef] [PubMed]
  39. Xie, H.; Regnier, S. Three-dimensional automated micromanipulation using a nanotip gripper with multi-feedback. J. Micromechan. Microeng. 2009, 19, 075009. [Google Scholar] [CrossRef]
  40. Lu, Z.; Moraes, C.; Zhao, Y.; You, L.; Simmons, C.A.; Sun, Y. A micromanipulation system for single cell deposition. In Proceedings of the 2010 IEEE International Conference on Robotics and Automation (ICRA), Anchorage, AK, USA, 3–7 May 2010; pp. 494–499. [Google Scholar]
  41. Nguyen, C.N.; Ohara, K.; Avci, E.; Takubo, T.; Mae, Y.; Arai, T. Real-time precise 3D measurement of micro transparent objects using All-In-Focus imaging system. J. Micro-Nano Mechatron. 2012, 7, 21–31. [Google Scholar] [CrossRef]
  42. Kang, W.; Robitaille, M.C.; Merrill, M.; Teferra, K.; Kim, C.; Raphael, M.P. Mechanisms of cell damage due to mechanical impact: An in vitro investigation. Sci. Rep. 2020, 10, 12009. [Google Scholar] [CrossRef] [PubMed]
  43. Fulda, S.; Gorman, A.M.; Hori, O.; Samali, A. Cellular stress responses: Cell survival and cell death. J. Cell Biol. 2010, 2010, 214074. [Google Scholar] [CrossRef]
Applsci 14 00617 g001
Figure 1. (a) 3D parallel mechanism model and its joint design. (b) Kinetic chain 1 and kinetic chain 2 or 3.
Figure 1. (a) 3D parallel mechanism model and its joint design. (b) Kinetic chain 1 and kinetic chain 2 or 3.
Applsci 14 00617 g001
Applsci 14 00617 g002
Figure 2. Micromanipulation robot system composition diagram.
Figure 2. Micromanipulation robot system composition diagram.
Applsci 14 00617 g002
Applsci 14 00617 g003
Figure 3. (a) Fabrication processes of hemispherical end-effectors. (b) The vibration-based release method combined with gel coating.
Figure 3. (a) Fabrication processes of hemispherical end-effectors. (b) The vibration-based release method combined with gel coating.
Applsci 14 00617 g003
Applsci 14 00617 g004
Figure 4. (a) Horizontal release displacement deviation (negative values indicate a leftward horizontal direction). (b) Vertical release displacement deviation. (c) Horizontal release displacement deviation after compensation.
Figure 4. (a) Horizontal release displacement deviation (negative values indicate a leftward horizontal direction). (b) Vertical release displacement deviation. (c) Horizontal release displacement deviation after compensation.
Applsci 14 00617 g004
Applsci 14 00617 g005
Figure 5. (a) Illustration of operational field selection. (b) Array strategy illustration.
Figure 5. (a) Illustration of operational field selection. (b) Array strategy illustration.
Applsci 14 00617 g005
Applsci 14 00617 g006
Figure 6. Automated assembly algorithm flowchart.
Figure 6. Automated assembly algorithm flowchart.
Applsci 14 00617 g006
Applsci 14 00617 g007
Figure 7. Assembly process for character ‘T’.
Figure 7. Assembly process for character ‘T’.
Applsci 14 00617 g007
Table 1. Release success rate under different release conditions.
Table 1. Release success rate under different release conditions.
Amplitude5 μm10 μm15 μm20 μm
Frequency
25 Hz98%98%98%100%
50 Hz98%98%98%100%
75 Hz100%100%100%100%
100 Hz100%100%100%100%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhao, Y.; Deng, Y.; Chen, J.; Kojima, M.; Huang, Q.; Arai, T.; Liu, X. High-Speed Cell Assembly with Piezo-Driven Two-Finger Microhand. Appl. Sci. 2024, 14, 617. https://doi.org/10.3390/app14020617

AMA Style

Zhao Y, Deng Y, Chen J, Kojima M, Huang Q, Arai T, Liu X. High-Speed Cell Assembly with Piezo-Driven Two-Finger Microhand. Applied Sciences. 2024; 14(2):617. https://doi.org/10.3390/app14020617

Chicago/Turabian Style

Zhao, Yue, Yan Deng, Junnan Chen, Masaru Kojima, Qiang Huang, Tatsuo Arai, and Xiaoming Liu. 2024. "High-Speed Cell Assembly with Piezo-Driven Two-Finger Microhand" Applied Sciences 14, no. 2: 617. https://doi.org/10.3390/app14020617

APA Style

Zhao, Y., Deng, Y., Chen, J., Kojima, M., Huang, Q., Arai, T., & Liu, X. (2024). High-Speed Cell Assembly with Piezo-Driven Two-Finger Microhand. Applied Sciences, 14(2), 617. https://doi.org/10.3390/app14020617

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop