Cell sorting method based on photoinduced dielectrophoresis virtual channel and dynamic path planning

Abstract

The invention provides a cell sorting method based on light-induced dielectrophoresis virtual channel and dynamic path planning, which comprises the following steps: acquiring an original microscopic image of a cell in real time by using an optical microscope, and acquiring the size and real-time position of the cell through image preprocessing and cell feature extraction operation; determining a respective corresponding target sorting area based on the cell size and the cell position of each cell; on the basis of the current cell position, global path planning is carried out by using a time-space heuristic search algorithm with introduced time dimension, and a global path point sequence containing time sequence information is obtained; projecting the path plan of each cell to a light-induced dielectrophoresis chip by using a digital projector to form a virtual channel so as to guide each cell to move to a respective corresponding target sorting area; and establishing a dynamic potential field environment by utilizing a multi-potential field fusion algorithm so as to adjust parameters of the operation light spots, so that the actual motion trail of the cells tends to the virtual channel. The method is high in anti-interference capability and high in sorting result accuracy.

Description

Technical Field

[0001] This application belongs to the field of cell sorting technology, and in particular relates to a cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning. Background Technology

[0002] Cell sorting technology is a core technology in biology, medicine, and bioengineering, widely used in cancer research, single-cell analysis, and other fields. However, despite strong market demand, the technology itself has encountered significant bottlenecks. Traditional technologies, such as flow cytometry (FACS) and magnetic immunobead sorting (MACS), are widely used but each has its own shortcomings. Specifically, FACS equipment is expensive and bulky, and the shear force from its high-speed flow can easily cause physical damage to fragile cells. It also relies on fluorescent labeling, which may interfere with the natural state of cells. While MACS is gentler, it also suffers from label dependence and has limited sorting purity and flexibility. On the other hand, while microfluidic-based technology platforms offer a gentler environment, their pre-etched, fixed physical channel structures fundamentally limit flexibility, making it difficult to dynamically and adaptively process cell populations of different sizes and types on the same chip, and even more difficult to achieve complex, real-time adjustable cell transport path planning.

[0003] Existing emerging technologies, such as dielectrophoresis (DEP) sorting, have shown great potential for label-free and contactless operation. However, their electrode patterns are usually fixed, leading to rigid sorting patterns and making it difficult to independently and dynamically track and precisely control individual target cells. Their sorting logic is often pre-set and static. A common shortcoming of all these technologies is the lack of an intelligent closed-loop system that integrates "sensing-decision-execution"—a system capable of real-time cell observation, judgment, and dynamic adjustment. Therefore, the market and technology sector are urgently calling for a new, highly integrated technological solution. This solution needs to fundamentally break through the existing framework, with its core being the deep integration of real-time image feedback, dynamic path planning, and independently controllable virtual channels. An ideal next-generation platform should possess the following characteristics: First, it must be label-free and contactless, sorting cells based on their inherent physical properties to maximize cell viability and functional integrity, avoiding bias introduced by labeling. Secondly, it must possess extremely high flexibility and programmability, capable of dynamically generating "virtual channels" through software rather than relying on physical structures, thereby adapting to samples of different cell sizes and changing sorting strategies at any time according to experimental needs. Most importantly, it must have precise manipulation capabilities at the single-cell level, able to identify, lock onto, and independently guide each target cell, delivering it non-damagingly to designated locations such as individual culture chambers or PCR tubes, achieving precise cell sorting. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning, which does not require contact with cells, greatly reduces cell damage, and improves the accuracy of separation; it does not require the preparation of physical virtual channels, is simple to operate, and allows for flexible channel adjustment, making it easy to manipulate; it can sense environmental changes in real time and intelligently plan the optimal obstacle avoidance path for each cell, ensuring that cells can be accurately delivered to the designated location even in complex microenvironments, with strong anti-interference capabilities; it can achieve the precise placement of specific cells or micro / nano materials into designated locations.

[0005] This application provides a cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning, including: The raw microscopic images of cells were acquired in real time using an optical microscope, and the cell size and real-time location were obtained through image preprocessing and cell feature extraction. The target sorting area is determined based on the cell size and cell location of each cell. Based on the current cell location, a spatiotemporal heuristic search algorithm incorporating the time dimension is used for global path planning to obtain a global path point sequence containing temporal information. A digital projector is used to project the path planning of each cell onto the photoinduced dielectrophoresis chip to form a virtual channel, so as to guide each cell to move to its corresponding target sorting area; A dynamic potential field environment is established using a multi-potential field fusion algorithm to adjust the parameters of the operating light spot, thereby making the actual movement trajectory of the cell tend to the virtual channel.

[0006] Furthermore, the process of obtaining cell size and real-time location through image preprocessing and cell feature extraction includes: The original microscopic image is subjected to mean filtering to eliminate random noise interference during the imaging process, and then background subtraction is performed to correct uneven illumination. The corrected image is converted to an 8-bit format, and then an adaptive threshold segmentation algorithm is used to convert the grayscale image into a binary image to highlight the cell outline. Edge detection is performed based on the binarized image to extract the complete contour information of the cell. The area, perimeter and equivalent diameter of the cell are accurately calculated by the pixel statistics algorithm to obtain the cell size. The precise spatial position mapping of the cell on the photoinduced dielectrophoresis chip is established by calculating the centroid coordinates of the cell.

[0007] Furthermore, the global path planning using a spatiotemporal heuristic search algorithm that incorporates a time dimension includes: The surface of the photoinduced dielectrophoresis chip is established as a discrete grid map to obtain the position coordinates of each grid point. Then, a three-dimensional coordinate map is constructed by introducing the time dimension as the search space. Starting with the current cell position as the starting node and the corresponding target sorting area as the ending node, the Manhattan distance is used as the heuristic function to generate the planned path for each cell in batches from the outside to the inside, thereby obtaining a global path point sequence containing the position coordinates and arrival time of each path point.

[0008] Furthermore, the planned path for each cell is calculated based on the actual path cost from the starting node to the current node and the characteristic weight factor of that cell.

[0009] Furthermore, the step of establishing a dynamic potential field environment using a multi-potential field fusion algorithm to adjust the parameters of the operating spot includes: The target sorting area is set as an gravitational potential field, the boundary and obstacles of the photoinduced dielectrophoresis chip are set as repulsive potential fields, and a guiding potential field is generated based on the global path point sequence. The synthetic forces acting on cells were calculated using a multi-field coupling model; The combined force is converted into parameters of the manipulating light spot; wherein the manipulating light spot is a circular pattern generated by a digital projector based on cell size for controlling cell movement, and its parameters include: turning radius and moving speed; The deviation between the actual cell movement trajectory and the virtual channel is calculated, and the light field distribution is adjusted in real time through PID closed-loop control based on the deviation.

[0010] Furthermore, the photoinduced dielectric electrophoresis chip refers to a photoconductive layer composed of a lower layer of ITO glass coated with hydrogenated amorphous silicon and an upper layer of ITO glass. Each cell is guided to move to its corresponding target sorting area using the following method: A signal generator is used to apply an electric field of a certain voltage and frequency to the photoinduced dielectrophoresis chip, so that the cells can follow the manipulated light spot and move along the virtual channel.

[0011] The cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning provided in this application has the following beneficial effects: 1) Photoinduced dielectrophoresis virtual channel technology does not require contact with cells when manipulating them, which greatly reduces damage to cells and improves the accuracy of separation; 2) The optically induced dielectrophoresis virtual channel technology prepares virtual channels using optical technology, eliminating the need to prepare physical virtual channels. It is simple to operate, the channel can be flexibly adjusted, and the system is easy to operate. 3) The light-induced dielectrophoresis virtual channel technology is integrated with path planning and automatic sorting of cells. The dynamic path planning of cell fusion can sense environmental changes in real time and intelligently plan the optimal obstacle avoidance path for each cell, ensuring that cells can be accurately delivered to the designated location even in complex microenvironments, with strong anti-interference ability. 4) It can not only sort cells, but also achieve precise transport and positioning, and can accurately place specific cells or micro / nano materials in designated locations, which is difficult to achieve with many traditional sorting technologies, providing technical support for single-cell analysis. 5) The fusion of path planning and photoinduced virtual channel technology is not only applicable to cells, but can also be extended to the separation, screening and precision assembly of other micro and nanoscale materials, which has great application potential and can become a general technology platform. Attached Figure Description

[0012] Figure 1 This paper shows an overall structural diagram of the photoinduced dielectrophoresis system provided in an embodiment of this application; Figure 2 A structural diagram of the photoinduced dielectrophoresis chip provided in an embodiment of this application is shown; Figure 3 A flowchart of the cell sorting method based on photoinduced dielectrophoresis virtual channel and dynamic path planning provided in this application embodiment is shown; Figure 4 This application provides schematic diagrams illustrating cell sorting of different sizes. Figure labels: 1-Digital projector, 2-Computer screen, 3-Computer screen, 4-Computer host, 5-Light source, 6-CCD microscope system, 7-Microscope objective, 8-Signal generator, 9-Photoinduced dielectrophoresis chip, 10-10× objective, 11-Lens, 12-ITO glass, 13-Manipulating light spot, 14-Virtual channel, 15-Photoconductive layer, 16-ITO glass. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this technical solution clearer, the following detailed description, in conjunction with specific embodiments, further illustrates this technical solution. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of this technical solution.

[0014] Please see as follows Figure 1 The diagram shows the overall structure of the photoinduced dielectrophoresis system. Figure 1As shown, the photoinduced dielectrophoresis system includes: a digital projector 1, a computer screen 2, a computer screen 3, a computer host 4, a light source 5, a CCD microscope system 6, a microscope objective 7, a signal generator 8, a photoinduced dielectrophoresis chip 9, a 10× objective lens 10, and a lens 11. The host 4 is connected to computer screens 3 and 2. Computer screen 2 is connected to the projector 1. The path planning generated on the host 4 is displayed on screen 2, and the projector can project corresponding light patterns. The host 4 is connected to the CCD microscope system 6. The microscopic images continuously acquired by the CCD microscope system 6 are transmitted to the host 4 in real time via a high-speed digital interface and displayed on computer screen 3 in real time. The host 4 briefly saves the initial cell images transmitted and processes them, analyzing the images, creating a constraint environment, and generating path planning. The signal generator 8 is connected to the ITO layers 9 of the upper and lower parts of the photoinduced dielectrophoresis chip. By changing the voltage and frequency applied to both ends of the chip, the force applied to manipulate the cells in the chip is changed. The light pattern projected vertically by projector 1 is focused by two lenses 11 and a 10× microscope objective 10, and finally projected onto the photoinduced dielectrophoresis chip 9. Cell manipulation is achieved by controlling the movement of the manipulating light spot. Above the photoinduced dielectrophoresis chip 9, in sequence, are the microscope objective 7 and the CCD microscope system 6. The function of the microscope objective 7 is to magnify the virtual channel and cell manipulation environment inside the photoinduced chip below. The CCD microscope system 6 stores the magnified image and transmits it to the host computer 4 for display on the computer screen 3. The function of the light source 5 is to provide light for the optical microscope system 6, facilitating real-time observation of the cell manipulation state.

[0015] Please see as follows Figure 2 The diagram shows the structure of a photoinduced dielectrophoresis chip. Figure 2 As shown, the photoinduced dielectrophoresis chip includes: ITO glass 12, a manipulation spot 13, a virtual channel 14, a photoconductive layer 15, and ITO glass 16. The photoconductive layer 15 is composed of 500 nm thick hydrogenated amorphous silicon deposited on the ITO glass 16. ITO has good conductivity and can be used as an electrode. Both the manipulation spot 13 and the virtual channel 14 are composed of... Figure 1 The digital projector 1 projects light onto the photoconductive layer, and the intensity of the manipulated light spot is greater than the intensity of the light in the virtual channel. When illuminated, the conductivity of the photoconductive material changes, generating a non-uniform electric field and creating a potential difference. Different light intensities produce different electric field intensities on the photoconductive material, creating an electric field gradient between them. This results in different photoinduced dielectrophoretic forces on the cells, thus enabling the manipulated light spot 13 to transport cells within the virtual channel 14.

[0016] Please see as follows Figure 3 The flowchart shown is for a cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning. Figure 3 As shown, the method includes: S101. Obtain raw microscopic images of cells in real time using an optical microscope, and obtain cell size and real-time location through image preprocessing and cell feature extraction.

[0017] In practice, cell size and real-time location are obtained through the following methods: Step 1011: Perform mean filtering on the original microscopic image to eliminate random noise interference during the imaging process, and then perform background subtraction to correct uneven illumination.

[0018] Step 1012: Convert the corrected image to 8-bit format, and then use an adaptive threshold segmentation algorithm to convert the grayscale image into a binary image to highlight cell outlines.

[0019] Step 1013: Perform edge detection based on the binarized image to extract the complete outline information of the cell, accurately calculate the area, perimeter and equivalent diameter of the cell through a pixel statistics algorithm to obtain the cell size, and establish the precise spatial position mapping of the cell on the photoinduced dielectrophoresis chip by calculating the cell centroid coordinates.

[0020] As an example, a multi-threaded parallel processing architecture based on the OpenCV computer vision library is used to implement image preprocessing and cell feature extraction. Specifically, firstly, 1392×1040 pixel images are continuously acquired at a frame rate of 1 fps using a CCD camera, and the image data is temporarily stored in a circular buffer. The preprocessing thread performs mean filtering with a 3×3 convolution kernel on each frame image, with the filtering parameters set to kernel = [[1,1,1],[1,1,1],[1,1,1]] / 9, effectively eliminating CCD sensor noise and random interference introduced by the optical system. Subsequently, background subtraction is performed, and then the image is converted to an 8-bit format for subsequent thresholding processing, linearly mapping the original grayscale value range to an 8-bit range of 0-255. Then, an adaptive thresholding segmentation algorithm is used to convert the grayscale image into a binary image, highlighting cell contour features. Binarization employs the Otsu adaptive thresholding algorithm, which automatically determines the optimal segmentation threshold T_optimal = argmax{σ²_between(T)} by analyzing the image pixel grayscale histogram, where σ²_between is the inter-class variance. For cases with blurred cell edges, a supplementary local adaptive threshold T_local(x,y) = mean(neighborhood) - C is used, where neighborhood is an 11×11 pixel window surrounding the (x,y) point, and the constant C=5.

[0021] The cell target and background regions are accurately separated through morphological operations and connected component analysis. The processing employs a multi-threaded parallel computing architecture, simultaneously processing multiple frames of image data to achieve millisecond-level real-time response. Morphological operations utilize opening and closing operations with a 3×3 circle as the structuring element to remove small-area noise spots and fill internal cell cavities. In the feature extraction stage, the Canny operator is used for edge detection, with a high threshold set to twice the standard deviation of pixel grayscale and a low threshold set to 0.4 times the high threshold. Contour extraction uses the Suzuki-Abe algorithm to obtain the complete set of cell boundary points. Geometric parameter calculations include: Area = ∫∫f(x,y)dxdy; Perimeter = ∫| f(s)|ds; Circularity = 4π×Area / Perimeter²; ​​Major and minor axis ratios obtained through ellipse fitting; Centroid coordinates (x_c, y_c) = (M_10 / M_00, M_01 / M_00), where M_ij is the image moment. The processing results are stored in a structured data table in real time, providing accurate cell state information for the path planning algorithm.

[0022] S102. Determine the corresponding target sorting area based on the cell size and cell position of each cell.

[0023] In this step, each cell is divided into two target sorting areas according to its cell size and location.

[0024] S103. Based on the current cell position, a spatiotemporal heuristic search algorithm with time dimension is used to perform global path planning to obtain a global path point sequence containing temporal information.

[0025] In practical implementation, the global pathpoint sequence containing timing information is obtained in the following way: Step 1031: Establish a discrete grid map on the surface of the photoinduced dielectrophoresis chip to obtain the position coordinates of each grid point, and then construct a three-dimensional coordinate map by introducing the time dimension as the search space.

[0026] Step 1032: Using the current cell position as the starting node and the corresponding target sorting area as the ending node, the Manhattan distance is used as the heuristic function to generate the planned path for each cell in batches from the outside to the inside, thereby obtaining a global path point sequence containing the position coordinates and arrival time of each path point.

[0027] The planned path for each cell is calculated based on the actual path cost from the starting node to the current node and the characteristic weight factor of that cell.

[0028] As an example, the global path planning uses the spatiotemporal A_Star algorithm, introducing the time dimension into the search space to construct a three-dimensional coordinate graph (x, y, t). The search space is set as a 1000×800×100 three-dimensional grid, where the xy plane resolution is 1μm / pixel and the time resolution is 0.1s / step. In the heuristic function f(n)=g(n)+h(n)+c(n), the actual cost g(n) is calculated using Euclidean distance g(n) = √[(xn-x0)² + (yn-y0)²], and the heuristic estimate h(n) uses the improved Manhattan distance h(n) = |xgoal-xn| + |ygoal-yn| + α×|tgoal-tn|, where the time weighting factor α=0.5. The cell characteristic weighting factor c(n) = β×(Di / D0)² + γ×(Ki / K0), where β=1.2 and γ=0.8 are empirical coefficients, and D0=10μm and K0=1 are normalized baseline values. The algorithm maintains a priority queue with a capacity of 10000 as an open list, adopts a binary heap data structure, and the node expansion strategy during the search process is 8-neighbor connectivity. The path search termination condition is f(n) < ε or the number of iterations exceeds 50000, where the convergence threshold ε=0.01.

[0029] S104. Using a digital projector, the path planning of each cell is projected onto the photoinduced dielectrophoresis chip to form a virtual channel, so as to guide each cell to move to its corresponding target sorting area.

[0030] The photoinduced dielectric electrophoresis chip refers to a photoconductive layer composed of a lower layer of ITO glass coated with hydrogenated amorphous silicon and an upper layer of ITO glass. In practice, each cell is guided to move to its corresponding target sorting area in the following way: A signal generator is used to apply an electric field of a certain voltage and frequency to the photoinduced dielectrophoresis chip, so that the cells can follow the manipulated light spot and move along the virtual channel.

[0031] Here, the signal generator applies a sinusoidal AC signal to the photoinduced dielectrophoresis chip. The applied voltage is adjustable from 0V to 8V, and the frequency is adjustable from 50Hz to 1MHz. By changing the voltage and frequency applied across the photoinduced chip, the dielectrophoretic force applied to the manipulated particles can be changed.

[0032] As an example, using yeast cells as the research object, the amplitude of the sinusoidal signal applied by the signal generator to the photoinducible chip is 1.5 V, the frequency is 1.5 kHz, the width of the virtual channel is 25 µm, and the moving speed of the manipulated light spot is 15 µm / s.

[0033] S105. A dynamic potential field environment is established using a multi-potential field fusion algorithm to adjust the parameters of the operating light spot, so that the actual movement trajectory of the cell tends to the virtual channel.

[0034] In practice, the parameters of the operating light spot are adjusted in the following ways: Step 1051: Set the target sorting area as an gravitational potential field, set the boundary and obstacles of the photoinduced dielectrophoresis chip as a repulsive potential field, and generate a guiding potential field based on the global path point sequence.

[0035] Step 1052: Calculate the synthetic forces acting on the cells using a multi-field coupling model.

[0036] Step 1053: Convert the combined force into parameters for manipulating the light spot.

[0037] The manipulation spot is a circular pattern generated by a digital projector based on cell size to control cell movement, and its parameters include: turning radius and moving speed.

[0038] Step 1054: Calculate the deviation between the actual cell movement trajectory and the virtual channel, and adjust the light field distribution in real time through PID closed-loop control based on the deviation.

[0039] In this step, the local execution thread employs a multi-potential field fusion algorithm. The potential field computation domain is set to the effective chip area of ​​750×500μm², with a grid resolution of 0.5μm. The gravitational potential field uses a quadratic function model U_att(r) = 0.5×katt×r², with the gravitational coefficient katt adaptively adjusted according to the target area, calculated as katt = 100×(Starget / 1000)^0.5, where Starget is the target area (μm²). The repulsive potential field uses an exponential decay model U_rep(r) = krep×exp(-r / r0), with a repulsive coefficient krep=500, and the influence radius r0 is set to 2.5 times the cell diameter. The chip boundary repulsive field uses a Gaussian function U_boundary(d) = 1000×exp(-d² / σ²), where d is the distance to the boundary, and the standard deviation σ=5μm. The guiding potential field generates a virtual channel based on the global path. The channel width is w = 25 μm, and the potential field strength is U_guide(d) = kguide × d². The guiding coefficient kguide is adaptively adjusted through the path curvature: kguide = 50 × (1 + κ / κmax), where κ is the local curvature and κmax = The resultant force calculation uses the gradient operator F = - U, the numerical differentiation uses a five-point difference scheme, with a computational accuracy of [missing value]. .

[0040] The light field control employs a DMD digital micromirror array with a resolution of 1392×1040. The virtual channel parameter conversion algorithm calculates the spot geometry parameters based on the composite force vector F_total: the initial moving velocity v of the manipulated spot is 15 μm / s, calculated based on the target cell velocity and the current position deviation: v_spot = vtarget + Kv×Δp, where vtarget is the target velocity, and the velocity gain Kv = ... The positional deviation Δp = |pcurrent - pplanned|.

[0041] The PID closed-loop control uses a positional control algorithm with a control period T = 50ms. The deviation is calculated as Δp(k) = ptarget(k) - pcurrent(k). The proportional term is P(k) = Kp × Δp(k), the integral term is I(k) = I(k-1) + Ki × T × Δp(k), and the derivative term is D(k) = Kd × [Δp(k) - Δp(k-1)] / T. The control output is u(k) = P(k) + I(k) + D(k). The PID parameters are tuned using the ZN adjustment method, with Kp = 1.0 and Ki = 1.0. Kd = 0.1s, with an integration limit of ±10μm to prevent integration saturation. The system uses a Kalman filter for cell position state estimation, with process noise variance Q = 0.01μm² and measurement noise variance R = The path tracking performance evaluation metrics include: Mean Absolute Error (MAE) = (1 / N)×Σ|pi -ptarget,i|, Root Mean Square Error (RMSE) = √[(1 / N)×Σ(pi - ptarget,i)²], and Success Rate (η) = Nsuccess / Ntotal×100%, where the success criterion is a final position deviation of less than 5μm.

[0042] Example 2: (1) Cell sample preparation: Yeast cells were selected as the research object. A sucrose solution with an isotonic concentration to that of the target cells was prepared using pure sucrose and deionized water. The target cells were cultured in medium-sized culture dishes. When the cells grew to the ideal state, a certain amount of cells were collected, centrifuged, and the supernatant was discarded. Then, 2 mL of the prepared sucrose solution was added, and the mixture was blown and mixed evenly. Finally, 200 μL of the cell suspension sample was taken into a 1 mL centrifuge tube for use.

[0043] (2) Parameters of the photoinduced dielectrophoresis system: The voltage, frequency, width of the virtual channel, and moving speed of the manipulation spot were determined sequentially. The specific steps are as follows: First, the frequency for manipulating the yeast cells was determined by changing the frequency while keeping the voltage of the signal generator constant. Then, the voltage for manipulating the yeast cells was determined by changing the voltage while keeping the frequency of the signal generator constant. Next, virtual channels of different widths were set, and the manipulation spot was controlled to move at the same speed. The width of the virtual channel was determined by observing the movement of the yeast cells. Finally, the determined virtual channel width was set, and the manipulation spot was controlled to move at different speeds. The moving speed of the manipulation spot was determined by observing the movement of the yeast cells.

[0044] (3) Cell sorting using a light-induced dielectrophoresis system: Step 1: System Initialization and Parameter Setting: Start computer screen 2, computer screen 3, and host 4. Start digital projector 1 and CCD microscope system 6, and adjust the optical focal length to clearly image the virtual channel on the chip surface. Turn on light source 5, and adjust signal generator 8 to output a sine wave signal with a frequency of 1.5 kHz and a voltage of 1.5 V, which is then applied to the photoinduced chip.

[0045] Step 2, Cell Injection: Using a 10 μL pipette, slowly inject 10 μL of the prepared cell suspension into the photoinducible chip 9, being careful to avoid generating air bubbles.

[0046] Step 3: Image Acquisition and Path Planning: Open the CCD Microscope System 6 software and begin continuously acquiring cell images. The image processing algorithm analyzes the cell's position coordinates and size parameters in real time. Cells are classified according to their diameter: small cells (equivalent diameter less than 8 μm) and large cells (equivalent diameter greater than 8 μm). The path planning algorithm calculates the optimal path to different sub-channels for each cell type. The target area for small cells is defined as the upper right corner of the area, with a length of 60 μm and a width of 100 μm, while the target area for large cells is defined as the upper left corner of the area, with a length of 60 μm and a width of 100 μm.

[0047] Step 4: Batch cell manipulation: Please refer to... Figure 4 The diagram shows the sorting of cells of different sizes. Based on the original cell image ( Figure 4 A) First, a circular manipulator spot with a diameter of 25 μm is projected below the centroid of each cell, and the cells are allowed to stabilize for 5 seconds. The first round of sorting then begins, starting with the outermost region. Virtual channels with lower brightness than the circular manipulator spot, designed by a path planning algorithm, are generated below the cells. Figure 4 B), and then the manipulated light spot begins to move along the virtual channel at a speed of 15 μm / s, while the circular manipulated light spot remains unchanged below the cells in other areas. Figure 4C). After the cells in the outermost region reach the target region, the second round of sorting begins. A virtual channel with a lower brightness than the circular manipulation spot, designed by a path planning algorithm, is generated below the cells in the second outermost region. Figure 4 D), and then manipulate the light spot to begin moving along the virtual channel at a speed of 15 μm / s ( Figure 4 E), while the circular manipulation spot remains unchanged below the cells in the inner central region. After the cells in the second outermost region reach the target area, the third round of sorting begins. A virtual channel with a lower brightness than the circular manipulation spot, designed by a path planning algorithm, is generated below the cells in the central region. Figure 4 F), and then manipulate the light spot to begin moving along the virtual channel at a speed of 15 μm / s (F). Figure 4 G), once the cells in the central region reach the target region, the sorting process ends. Figure 4 H). The entire sorting process lasts 5-8 minutes, and the sorting effect is ensured through real-time image monitoring.

[0048] Step 5, Result Verification: After sorting, turn off all manipulating light spots. Cells will disperse somewhat due to interaction forces. Once the cells have stabilized, save the image. ImageJ software is used to count and calculate the size parameters of cells in both large and small cell regions. The calculation results are then used to verify the cell sorting results.

[0049] In summary, the cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning proposed in this application has the following advantages: 1) Simple operation: It combines path planning and light-induced dielectrophoresis virtual channel technology. The virtual channel is projected by a digital projection device, eliminating the need to create a physical channel, making it simple and easy to build; the path planning algorithm is used to design the virtual channel, making operation simple. 2) The method of manipulating cells causes less damage to cells: The sorting is performed using photoinduced dielectrophoresis virtual channel technology, which does not involve direct mechanical contact with cells, thus causing less damage to cells; 3) High precision in sorting and transporting cells of different sizes to designated areas: It can sense environmental changes in real time and intelligently plan obstacle avoidance paths for each cell, ensuring that cells can be accurately delivered to designated locations even in complex microenvironments, with strong anti-interference capabilities.

[0050] The above content is only a preferred embodiment of the present invention. For those skilled in the art, many changes can be made in the specific implementation and application scope based on the ideas of the present invention. As long as these changes do not depart from the concept of the present invention, they all fall within the protection scope of the present invention.

Claims

1. A cell sorting method based on photoinduced dielectrophoresis virtual channels and dynamic path planning, characterized in that, The method includes: The raw microscopic images of cells were acquired in real time using an optical microscope, and the cell size and real-time location were obtained through image preprocessing and cell feature extraction. The target sorting area is determined based on the cell size and cell location of each cell. Based on the current cell location, a spatiotemporal heuristic search algorithm incorporating the time dimension is used for global path planning to obtain a global path point sequence containing temporal information. A digital projector is used to project the path planning of each cell onto the photoinduced dielectrophoresis chip to form a virtual channel, so as to guide each cell to move to its corresponding target sorting area; A dynamic potential field environment is established using a multi-potential field fusion algorithm to adjust the parameters of the operating light spot, thereby making the actual movement trajectory of the cell tend to the virtual channel.

2. The method as described in claim 1, characterized in that, The process of obtaining cell size and real-time location through image preprocessing and cell feature extraction includes: The original microscopic image is subjected to mean filtering to eliminate random noise interference during the imaging process, and then background subtraction is performed to correct uneven illumination. The corrected image is converted to an 8-bit format, and then an adaptive threshold segmentation algorithm is used to convert the grayscale image into a binary image to highlight the cell outline. Edge detection is performed based on the binarized image to extract the complete contour information of the cell. The area, perimeter and equivalent diameter of the cell are accurately calculated by the pixel statistics algorithm to obtain the cell size. The precise spatial position mapping of the cell on the photoinduced dielectrophoresis chip is established by calculating the centroid coordinates of the cell.

3. The method as described in claim 1, characterized in that, The global path planning using a spatiotemporal heuristic search algorithm that incorporates a time dimension includes: The surface of the photoinduced dielectrophoresis chip is established as a discrete grid map to obtain the position coordinates of each grid point. Then, a three-dimensional coordinate map is constructed by introducing the time dimension as the search space. Starting with the current cell position as the starting node and the corresponding target sorting area as the ending node, the Manhattan distance is used as the heuristic function to generate the planned path for each cell in batches from the outside to the inside, thereby obtaining a global path point sequence containing the position coordinates and arrival time of each path point.

4. The method as described in claim 3, characterized in that, The planned path for each cell is calculated based on the actual path cost from the starting node to the current node and the cell's characteristic weighting factor.

5. The method as described in claim 1, characterized in that, The method of establishing a dynamic potential field environment using a multi-potential field fusion algorithm to adjust the parameters of the operating spot includes: The target sorting area is set as an gravitational potential field, the boundary and obstacles of the photoinduced dielectrophoresis chip are set as repulsive potential fields, and a guiding potential field is generated based on the global path point sequence. The synthetic forces acting on cells were calculated using a multi-field coupling model; The combined force is converted into parameters of the manipulating light spot; wherein the manipulating light spot is a circular pattern generated by a digital projector based on cell size for controlling cell movement, and its parameters include: turning radius and moving speed; The deviation between the actual cell movement trajectory and the virtual channel is calculated, and the light field distribution is adjusted in real time through PID closed-loop control based on the deviation.

6. The method as described in claim 1, characterized in that, The photoinduced dielectric electrophoresis chip refers to a photoconductive layer composed of a lower layer of ITO glass coated with hydrogenated amorphous silicon and an upper layer of ITO glass. Each cell is guided to move to its corresponding target sorting area using the following method: A signal generator is used to apply an electric field of a certain voltage and frequency to the photoinduced dielectrophoresis chip, so that the cells can follow the manipulated light spot and move along the virtual channel.

Images