Multi-modal vision-assisted single-cell control and packaging system and multi-modal vision-assisted single-cell control and packaging method

Abstract

The invention discloses a multi-mode vision-assisted single cell control and packaging system and a multi-mode vision-assisted single cell control and packaging method. The system comprises a charge-coupled device (CCD) vision module, a digital micromirror (DMD) light control module, an ultraviolet light source assembly and a digital micro-fluidic chip unit. The system is characterized in that after image enhancement and single cell identification positioning operation, the system can obtain single cell coordinates with extremely high precision; then coordinate mapping and mask generation operation are combined to realize high-precision conversion from a CCD coordinate system (visual system) to a DMD coordinate system (printing system); meanwhile, an operation based on a space-time discrete printing strategy is carried out, and the exposure energy of each micromirror unit is dynamically regulated and controlled successfully by regulating and controlling the switching time of each micromirror unit of the DMD, so that the mechanical hardness gradient construction of the subcellular scale microgel and the precise packaging of the single cell are realized. According to the invention, on-demand identification, positioning and heterogeneous microenvironment packaging of single cells can be realized, and efficient, controllable and biocompatible technical support is provided for cell heterogeneity research.

Description

Technical Field

[0001] This invention belongs to the field of intelligent biomanufacturing and medical-engineering interdisciplinary technology, specifically involving a method for loading and transporting cell-carrying droplets based on digital microfluidic chips, a method for multimodal visual recognition and precise single-cell localization, and a single-cell encapsulation technology based on light field induction. Background Technology

[0002] Single-cell isolation and culture techniques play an irreplaceable fundamental role in cell heterogeneity research. Currently, single-cell encapsulation methods based on micromanipulation and fluid control have become the mainstream technical approach to achieve this goal. Among them, extreme dilution, micromanipulation picking, and fluorescence-activated flow cytometry sorting provide important support for single-cell manipulation; while passive microfluidic droplet technology, through structural design such as fluid focusing, achieves high-throughput encapsulation and independent liquid culture of single cells, significantly improving experimental throughput and controllability. However, the above methods mainly focus on the physical isolation of single cells and two-dimensional liquid culture, and their environment generally lacks extracellular matrix support, making it difficult to reproduce the real three-dimensional microenvironment in vivo. At the same time, significant progress has been made in the research of constructing single-cell microgel systems by combining biomaterials and microfluidic technology. This technology simulates a three-dimensional microenvironment by encapsulating cells in bio-similar hydrogels and preparing structurally regular cell gel microspheres using oil-water two-phase flow. While cell density regulation can promote single-cell microsphere formation to some extent, the highly random nature of the formation process, based on fluid shear and Poisson distribution, leads to low yields of single-cell microspheres, often requiring subsequent separation and purification steps, severely limiting application efficiency. Furthermore, the introduced exogenous oil phase and surfactants are difficult to completely remove, and their potential impact on cell viability and microenvironment biocompatibility cannot be ignored. Therefore, to overcome the dual limitations of existing methods in single-cell encapsulation efficiency and biocompatibility, there is an urgent need to develop an integrated technology solution capable of directly identifying target cells and achieving precise on-demand encapsulation.

[0003] At the forefront of current tissue engineering and cell biology research, the precise construction of single-cell microenvironments has become a key scientific issue in understanding cellular heterogeneity. While traditional homogeneous encapsulation strategies can achieve three-dimensional culture of single cells, they neglect the essential characteristic of mechanical heterogeneity of the extracellular matrix at the subcellular scale. Mechanical stiffness gradients, through differentiated mechanical signal transduction, directly regulate the dynamic response of cell polarity and spatial behavior, and are a core physical element in tissue interface formation and homeostasis maintenance. With the integrated development of photocurable hydrogel technology and digital light processing technology, the construction of multi-regional heterogeneous microgels has achieved preliminary technical feasibility. However, existing technologies have two fundamental limitations: at the physical level, the Gaussian distribution characteristics of the ultraviolet light field result in insufficient precision in regulating the mechanical stiffness within the microgel, making it difficult to simulate a true gradient; at the biological level, the lack of spatial registration capability between single cells and heterogeneous regions makes it difficult to ensure that cells are at key response positions of a preset mechanical gradient. This technological shortcoming means that current research remains focused on observing the statistical behavior of population cells, failing to address the precise mechanical response mechanisms of single cells in heterogeneous microenvironments. Fundamentally, existing methods have not yet bridged the technological gap between "mechanical precision control" and "single-cell spatial registration." This lack of capability not only limits our in-depth understanding of cellular mechanosensing mechanisms but also hinders technological progress in constructing biomimetic complex tissues. Therefore, there is an urgent need to establish a novel heterogeneous encapsulation system capable of synergistically achieving subcellular-level mechanical programming and precise single-cell spatial registration, providing key technological support for revealing the behavioral patterns of cells in real microenvironments. Summary of the Invention

[0004] The purpose of this invention is to propose a multimodal vision-assisted single-cell manipulation and encapsulation system and method for achieving high-precision positioning of single cells and controllable construction of heterogeneous microgels at the subcellular scale.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] This invention proposes a multimodal vision-assisted single-cell manipulation and encapsulation system and method, mainly including the system hardware architecture and the single-cell manipulation and encapsulation process. The single-cell manipulation and encapsulation process includes three key steps: image enhancement and single-cell recognition and localization, coordinate mapping and mask generation, and spatiotemporal discrete printing and heterogeneous microgel construction.

[0007] The present invention is characterized by:

[0008] The system hardware architecture includes a charge-coupled device (CCD) vision module, a digital micromirror (DMD) light control module, an ultraviolet light source assembly, and a microfluidic chip unit, forming a multimodal vision and light-controlled bioprinting platform. In the image enhancement and single-cell recognition and localization section, color compensation and local adaptive contrast enhancement techniques optimize the quality of the microscopic image. Combined with the YOLOv7 deep learning model, single-cell recognition and coarse localization are achieved, and traditional image processing algorithms are integrated to complete pixel-level fine localization, thus enabling single-cell recognition and high-precision coordinate localization in complex microscopic environments. In the coordinate mapping and mask generation section, the optimal mapping relationship between the CCD and DMD coordinate systems is established based on the Levenburg-Marquardt algorithm. Combined with a dynamic digital mask matching the single-cell position information and the target area, seamless integration from visual recognition to the printing system is achieved. In the spatiotemporal discrete printing and heterogeneous microgel construction section, a discrete-time array is constructed to dynamically control the opening time of each micromirror to compensate for the energy unevenness caused by the Gaussian distribution of ultraviolet light intensity, achieving mechanical stiffness control of subcellular-scale microgels and precise encapsulation of single cells.

[0009] The system hardware architecture is characterized by four core components: a visual imaging module, a light control module, a light source system, and a microfluidic execution unit, forming an integrated platform for multimodal perception and light-controlled printing. The visual imaging module uses a high-resolution CCD camera to achieve real-time acquisition of microscopic images; the light control module uses a digital micromirror device as its core, with a micromirror array of 1024×768, achieving dynamic generation and energy distribution control of ultraviolet light patterns through independent micromirror control; the light source system uses a 365nm ultraviolet light source, combined with beam expanding and contracting optical components, to achieve precise control of the beam diameter; the microfluidic execution unit uses a digital microfluidic chip as a carrier to construct a discrete droplet manipulation structure, realizing the controllable transport and positioning of the cell-loaded hydrogel prepolymer solution. The four modules achieve closed-loop collaboration through coordinate mapping and communication protocols, ultimately completing the visual recognition, positioning, and heterogeneous microgel encapsulation of single cells.

[0010] The image enhancement and single-cell recognition and localization steps are characterized by three parts: microscopic image quality improvement, single-cell recognition, and localization accuracy optimization. The original microscopic image suffers from color attenuation and insufficient contrast due to 450nm filtering, affecting the accuracy of single-cell recognition. An adaptive compensation strategy using the RGB color model is employed: first, the average pixel intensity of each channel is calculated; red and blue channel compensation factors are generated using the green channel as a reference; and iterative compensation is performed according to the minimum color loss criterion to ensure the three-channel mean satisfies the gray-scale world assumption. Then, linear stretching is used to expand the dynamic range, and a multi-scale fusion strategy guided by the maximum attenuation map is used to generate a color-corrected image. In the contrast enhancement stage, a local adaptive contrast enhancement algorithm is used. High-frequency components are extracted by calculating the gray-scale mean and variance of image blocks, and an enhancement cutoff factor β=2 is set for controllable enhancement. Finally, guided filtering is introduced to suppress noise, completing the visual quality optimization of the microscopic image. In the single-cell recognition stage, a YOLOv7 deep learning model is used as the detection framework. The labeled single-cell dataset is trained, and the optimal model weights are obtained after approximately 300 training rounds. In the positioning accuracy optimization stage, to address the shortcomings of the YOLO model in positioning, the region of interest within the detection box is first extracted, and an adaptive threshold is used for cell region segmentation. Then, morphological operations are used to remove noise and fill holes, and Canny edge detection is used to extract the precise contour of a single cell. Finally, the geometric center coordinates of the single cell are calculated based on image moments to achieve pixel-level high-precision positioning, providing a reliable spatial coordinate basis for subsequent printing.

[0011] The coordinate mapping and mask generation steps are characterized by achieving a high-precision transformation from the CCD coordinate system to the DMD coordinate system, and the ability to generate a basic mask or a dynamic mask for heterogeneous microgel construction according to printing requirements. During the basic mask generation process, the system uses a CCD camera to capture the spatial coordinates of a single cell, and solves for the optimal mapping transformation matrix (TM) based on the Levenburg-Marquardt algorithm to accurately transform the coordinates to the DMD coordinate system. The positioning error of this cross-coordinate system mapping is less than 1.02 μm, ensuring that the constructed light pattern precisely matches the position of the single cell, placing it at the preset geometric center of the microgel.

[0012] The spatiotemporal discrete printing and heterogeneous microgel construction steps are characterized by dynamic spatial modulation of exposure energy based on a discrete-time array to achieve heterogeneous gradient construction of microgels. The system employs a spatiotemporal discrete printing strategy, the core of which lies in constructing a discrete-time array negatively correlated with the spatial distribution of ultraviolet light intensity, based on a Gaussian distribution model of ultraviolet light intensity. This array uses a dynamic digital mask to control the switching duration of each micromirror unit in a time series. The exposure time is appropriately shortened in the central region with higher light intensity, while the exposure time is extended as needed in the edge region with weaker light intensity, thereby achieving spatial homogenization of the accumulated energy (UED) within the exposure area and effectively compensating for curing deviations caused by inherent uneven light intensity. By flexibly designing the distribution function of the discrete-time array, programmable control of the internal mechanical hardness of the microgel can be further achieved. For example, a radially increasing time distribution can construct a radially gradient microgel with increasing hardness from the center outwards; a linearly gradual time distribution can form a continuously varying hardness zone in a specific direction. This method can construct a variety of complex heterogeneous microgel structures, including bipartite and tripartite structures, providing a highly biomimetic and controllable mechanical microenvironment for studying the response of cells to mechanical stiffness gradients at the subcellular scale. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below.

[0014] Figure 1 Flowchart for mask generation, single-cell recognition and localization, and heterogeneous microgel construction;

[0015] Figure 2 A schematic diagram of a multimodal vision-assisted single-cell manipulation and encapsulation system;

[0016] Figure 3 This is a schematic diagram of a single-cell heterogeneous microgel.

[0017] In the figure: 1-Computer; 2-CCD camera; 3-Objective lens; 4, 5, 8, 14-Prisms; 6-White light source; 7, 9, 10, 12, 13-Convex lens; 11-Ultraviolet light source; 15-DMD digital micromirror; 16-Digital microfluidic chip; 17-Microgel; 18-Encapsulated single cell. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] (1) Fabrication of digital microfluidic chips, the specific process is as follows:

[0020] S1: A two-step ultrasonic cleaning method is used to clean the surface of the ITO conductive glass substrate to remove organic pollutants and particulate impurities.

[0021] S2: A photoresist layer is uniformly coated on the pretreated substrate surface using a spin coating process, followed by pre-baking to enhance the interfacial adhesion between the photoresist and the substrate.

[0022] S3: Use a mask alignment exposure system to perform ultraviolet lithography, and accurately transfer the electrode array pattern from the mask to the photoresist layer;

[0023] S4: Unexposed photoresist is removed through chemical development, followed by deionized water cleaning and hardening treatment to finally form a structurally stable photoresist mask.

[0024] S5: Wet etching is used to selectively remove the ITO conductive layer not protected by photoresist, followed by a photoresist stripping process to completely remove the residual photoresist.

[0025] S6: A uniform and stable insulating barrier is formed on the electrode surface by depositing a SU-8 photoresist dielectric layer using spin coating or by preparing an Al2O3 dielectric layer using atomic layer deposition.

[0026] S7: Teflon AF1600 hydrophobic material is spin-coated onto the surface of the dielectric layer using a spin coating process, and then heat-treated to form a functional interface with low surface energy.

[0027] S8: The prepared upper and lower substrates are bonded together with conductive double-sided adhesive to construct a complete digital microfluidic chip 16 with a parallel plate structure.

[0028] (2) Image enhancement and single-cell recognition and localization, the specific process is as follows:

[0029] Sl: Acquire raw microscopic images using CCD camera 2 and analyze the non-uniform attenuation characteristics of color channels caused by 450nm filtering;

[0030] S2: Construct a color compensation factor based on the minimum color loss criterion, and realize the restoration and balance of the intensity of the red and blue channels through iterative calculation;

[0031] S3: Employs a local adaptive contrast enhancement method to dynamically improve the detail features of microscopic images by calculating the mean and variance of image patch gray levels;

[0032] S4: Apply guided filtering algorithm to suppress noise in the enhanced image, effectively balancing the needs of detail preservation and noise removal;

[0033] S5: Based on the YOLOv7 deep learning framework, single-cell recognition and coarse localization are performed on the enhanced microscopic images, with a recognition accuracy of 94.9%.

[0034] S6: Based on deep learning to identify regions, the integrity of single-cell regions is optimized through adaptive threshold segmentation and morphological operations;

[0035] S7: The Canny edge detection algorithm is used to extract the sub-pixel precision contour of a single cell, and the geometric center coordinates are calculated based on the image moment method.

[0036] (3) Coordinate mapping and mask generation, the specific process is as follows:

[0037] S1: Establish the spatial mapping relationship between the CCD camera coordinate system and the DMD coordinate system, and solve for the parameters of the optimal mapping transformation matrix TM using the Levenburg-Marquardt algorithm;

[0038] S2: Design a calibration mask containing 35 single pixels for exposure experiments, collect corresponding CCD image coordinate data, and substitute it into the loss function to optimize the matrix parameters;

[0039] S3: The mapping accuracy was confirmed through a reverse coordinate verification experiment. The measured coordinate deviation was controlled within 1 pixel, corresponding to an actual spatial error of no more than 1.02μm.

[0040] S4: The single-cell coordinates are accurately transformed from the CCD coordinate system to the DMD coordinate system through the optimal mapping transformation matrix TM, completing the spatial registration from visual positioning to printing execution.

[0041] S5: Perform light field simulation based on the Gaussian distribution model of the ultraviolet light source 11, and establish the discrete light intensity distribution matrix corresponding to the DMD micromirror array 15;

[0042] S6: The required UV exposure energy (UED) is calculated by back-calculating the target mechanical hardness, and a discrete-time control array is constructed through the quantitative relationship between energy and light intensity.

[0043] S7: Merges single-cell positioning coordinates with a discrete-time array to generate a dynamic digital mask with time-space control characteristics.

[0044] (4) Spatiotemporal discrete printing and heterogeneous microgel construction: The specific process is as follows:

[0045] S1: The actual exposure energy of the dynamic digital mask is measured and calibrated using an ultraviolet energy meter to establish a quantitative relationship between exposure energy and the mechanical hardness of the microgel;

[0046] S2: Based on the linear mapping relationship between the target mechanical hardness and the exposure energy, a discrete time control array is calculated using a formula to achieve precise control of the exposure time of each micromirror unit;

[0047] S3: The single-cell positioning coordinates are fused with a discrete-time array to generate a dynamic digital mask with spatiotemporal programming characteristics, and the uniform distribution of exposure energy is achieved through timing control.

[0048] S4: The mechanical hardness of the microgel was characterized by atomic force microscopy and digital holographic imaging to verify the spatiotemporal discrete printing strategy;

[0049] S5: Verification of the successful construction of heterogeneous single-cell microgels 17 with bipartite, tripartite and gradient mechanical hardness distributions using the proposed method.

Claims

1. A system for multimodal vision-assisted single-cell manipulation and encapsulation, characterized in that, include: Charge-coupled device (CCD) vision module, digital micromirror (DMD) light control module, ultraviolet light source assembly and digital microfluidic chip unit; The system uses multimodal visual recognition and coordinate mapping to detect and locate single cells, and uses DMD to dynamically control the ultraviolet beam to achieve high-precision positioning and on-demand encapsulation of single cells.

2. The system as described in claim 1, characterized in that, The microscopic images acquired by the CCD vision module are processed with color compensation and local adaptive contrast enhancement to improve image quality. Combined with the YOLOv7 deep learning model and traditional image processing algorithms, the automatic identification and high-precision localization of single cells are achieved.

3. The system as described in claim 1, characterized in that, The system employs the Levenburg-Marquardt algorithm to establish a mapping transformation matrix between the CCD coordinate system and the DMD coordinate system, thereby achieving high-precision conversion from the visual coordinate system to the printing coordinate system.

4. The system as described in claim 1, characterized in that, The system employs a spatiotemporal discrete printing strategy, which dynamically controls the switching duration of each micromirror in the DMD by constructing a discrete time array. This effectively compensates for curing deviations caused by inherent uneven light intensity, thereby achieving programmable control of the mechanical hardness inside the microgel.

5. The system as described in claim 4, characterized in that, The spatiotemporal discrete printing strategy enables controllable adjustment of the mechanical stiffness of the subcellular scale microgel region.

6. The system as described in claim 5, characterized in that, The system can encapsulate single cells in a hydrogel structure with regionalized mechanical hardness distribution to study the response behavior of single cells to mechanical hardness gradients.

Images