Automatic single-solid-nucleus liquid drop sorting method based on image recognition and dielectrophoretic force

Abstract

The invention discloses a single-solid-core liquid drop automatic sorting method based on image recognition and dielectrophoretic force, which comprises the following steps: establishing a labeling data set containing water-oil liquid drops and single-solid-core solid-water-oil liquid drops, and training a liquid drop classification model; the injection pump is started to generate liquid drop flow, the signal generator is started to generate alternating signals, the high-voltage amplifier is started to amplify the alternating signals, images are collected in real time and recognized through the liquid drop classification model, and when single-solid-core solid-water-oil liquid drops are detected, the computer outputs digital signals to the microcontroller. The microcontroller outputs a high-level electric signal to the direct-current amplifier, the high-voltage switching device is turned on, high voltage is loaded to the dielectrophoresis separation area, the dielectrophoresis electric field is activated, and the single-solid-core solid-water-oil liquid drops are deflected to enter the collection channel. The technical problem that real-time efficient separation is difficult to achieve due to the fact that solid water-oil liquid drops and water-oil liquid drops are mixed in the current micro-flow control preparation process is solved, and the purity and the separation efficiency of the single-solid-core liquid drops are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of microfluidic preparation of new materials, more specifically, the present application relates to a single solid core droplet automatic sorting method based on image recognition and dielectrophoresis force. BACKGROUND

[0002] The single solid core solid-water-oil droplets prepared by microfluidic technology have important application value in inertial confinement fusion (ICF), drug carriers, biological analysis and other fields. In particular, in ICF experiments, single solid core solid-water-oil droplets as precursors of double-layer microspheres directly affect the performance of the final target ball. However, the existing co-flow, cross-flow and flow focusing microfluidic devices inevitably produce multi-solid core droplets, empty droplets and irregular droplets and other defective products during preparation. The existence of these defective droplets seriously affects the utilization efficiency of high-precision hollow microspheres and the preparation quality of subsequent ICF target balls. The traditional droplet screening method mainly relies on manual visual inspection or simple physical screening, which is low in efficiency and limited in accuracy, and is difficult to meet the needs of high-throughput and high-precision industrial production.

[0003] At present, although the existing technology attempts to reduce the generation of defective droplets by optimizing the geometric structure and fluid parameters of the microfluidic device, it is still difficult to completely avoid the formation of defective droplets due to the complexity and randomness of fluid dynamics. At the same time, the existing droplet sorting technology is mainly based on single physical parameters such as size and density, and lacks effective recognition ability for the internal structure of the droplet, which cannot accurately distinguish single solid core droplets from water-oil droplets. As a non-contact manipulation method, dielectrophoresis technology has the advantages of precise control and small damage to samples, but its application in the field of droplet sorting is still in its infancy. Therefore, developing an intelligent single solid core droplet automatic sorting method combining image recognition technology and dielectrophoresis force has important theoretical significance and practical value for improving the overall efficiency and product quality of the microfluidic preparation system. SUMMARY

[0004] An object of the present application is to solve at least the above problems and / or deficiencies, and to provide at least the advantages described below.

[0005] In order to achieve these objects and other advantages according to the present application, the present application provides a single solid core droplet automatic sorting method based on image recognition and dielectrophoresis force, comprising the following steps: Step one, injecting a dispersed phase dispersed with solid microspheres, a continuous phase solution and a sheath flow phase solution into an integrated microfluidic chip, the dispersed phase dispersed with solid microspheres generates single solid core solid-water-oil droplets in the cross-shaped droplet generation area of the integrated microfluidic chip under the action of the continuous phase solution, and the dispersed phase generates water-oil droplets in the cross-shaped droplet generation area under the action of the continuous phase solution; Step two, build an image acquisition system, establish a labeled data set containing water-oil droplets and single solid core solid-water-oil droplets, use the labeled data set to train the droplet classification model using deep learning image recognition algorithm; Step three, build a sorting control system, start the injection pump to generate droplet flow, start the signal generator to generate alternating signal, start the high-voltage amplifier to amplify the alternating signal, adjust the amplification ratio, the industrial camera of the image acquisition system acquires images in real time and identifies whether the collected images are single solid core solid-water-oil droplets through the droplet classification model, when single solid core solid-water-oil droplets are detected, the computer outputs digital signals to the microcontroller, the microcontroller receives the digital signals and outputs high-level electrical signals to the direct current amplifier, which is amplified by the direct current amplifier and output to the high-voltage switching device, at this time the high-voltage switching device is opened, and the high voltage output by the high-voltage amplifier is loaded on the dielectrophoresis sorting area, and the dielectrophoresis electric field is activated to make the single solid core solid-water-oil droplets deflect into the collection channel.

[0006] Preferably, in step one, the dispersed phase is a 2wt% polyvinyl alcohol aqueous solution, the molecular weight of polyvinyl alcohol is 146000, and the degree of hydrolysis is 99%; The solid microspheres are PS microspheres; The continuous phase solution and the sheath flow phase solution are both oil phase solutions mixed with dibutyl phthalate and dioctyl sebacate, and the volume of dibutyl phthalate is 1-5 times that of dioctyl sebacate.

[0007] Preferably, in step one, the integrated microfluidic chip comprises: A cross-shaped droplet generation area comprising a dispersed phase channel and continuous phase channels located on both sides of the dispersed phase channel; A sheath flow adjusting area in communication with the dispersed phase channel, and sheath flow phase channels in communication with both sides of the dispersed phase channel; A dielectrophoresis sorting area corresponding to a pair of parallel liquid metal electrodes, the positive electrode of the parallel liquid metal electrode is connected to the positive electrode of the high-voltage amplifier output through a high-voltage silicone wire, and the negative electrode of the parallel liquid metal electrode is connected to the negative electrode of the microcontroller output through a high-voltage silicone wire, and the high-voltage silicone wire is sealed and fixed by epoxy resin.

[0008] Preferably, the liquid metal used in the parallel liquid metal electrode includes one of gallium-indium-tin alloy, gallium-indium alloy, bismuth-indium alloy, and low-melting-point gallium-based alloy.

[0009] Preferably, in step two, the image acquisition system comprises: An industrial camera provided with an LED light source; A computer connected to the industrial camera.

[0010] Preferably, in step two, the deep learning image recognition algorithm includes one of YOLOv8, YOLO, SSD, Faster R-CNN, and RetinaNet.

[0011] Preferably, in step three, the sorting control system includes: A signal generator is connected to the high-voltage amplifier, which is connected to the parallel liquid metal electrode of the integrated microfluidic chip via a high-voltage silicone wire. A microcontroller, the input of which is connected to the computer, and the output of which is connected to the input of a DC amplifier; the DC amplifier is connected to a DC power supply. A high-voltage switching device, the control terminal of which is connected to the output terminal of the DC amplifier, the positive terminal of the output terminal of the high-voltage switching device is connected to the negative terminal of the output terminal of the high-voltage amplifier, and the negative terminal of the output terminal of the high-voltage switching device is connected to the parallel liquid metal electrode of the integrated microfluidic chip.

[0012] Preferably, in step three, the high-voltage switching device includes one of a high-voltage reed switch, a high-voltage relay, a high-voltage MOSFET, a high-voltage thyristor, and a high-voltage field-effect transistor.

[0013] Preferably, in step three, the microcontroller includes one of Arduino, Raspberry Pi, STM32, ESP32, and PIC microcontroller.

[0014] Preferably, in step three, during dielectrophoretic sorting, the deflection distance of the single solid nucleus solid-water-oil droplet along the y-axis (perpendicular to the flow direction before the single solid nucleus solid-water-oil droplet sorting on a two-dimensional plane) is greater than or equal to 20% of the channel width, and the deflection response time is less than 100 milliseconds.

[0015] The present invention has at least the following beneficial effects: The present invention provides an automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis. For the first time, it realizes high-precision automatic sorting of solid-water-oil (S / W / O) droplets by combining deep learning image recognition algorithm with dielectrophoresis force control technology. It solves the technical problem that solid-water-oil droplets and water-oil (W / O) droplets are mixed in the current microfluidic preparation process, making it difficult to achieve real-time and efficient sorting. It significantly improves the purity and sorting efficiency of single solid-nucleus droplets.

[0016] The application adopts the integrated design of a microfluidic chip, an image recognition system and a dielectrophoresis sorting control system to realize real-time recognition and accurate sorting of S / W / O droplets. The microfluidic chip integrates droplet generation, sheath flow adjustment and dielectrophoresis sorting functions; the image recognition system adopts a YOLOv8 deep learning algorithm to realize real-time classification of droplets; and the dielectrophoresis sorting system generates a non-uniform electric field through a liquid metal electrode to realize directional deflection and sorting of target droplets.

[0017] The application adopts a deep learning algorithm to realize automatic recognition of droplets, and the recognition accuracy is more than 98%, which avoids the subjectivity and inefficiency of manual sorting.

[0018] The application realizes non-contact sorting of droplets through dielectrophoresis force, avoids the damage of mechanical sorting to the structure of droplets, and maintains the integrity of droplets.

[0019] The application integrates droplet generation, recognition and sorting functions in one, realizes fully automatic operation, has high sorting efficiency and simple operation, and is suitable for mass production application.

[0020] The application can control the sorting accuracy and efficiency within a certain range by adjusting the voltage, frequency and flow parameters, and has good system stability and high reproducibility.

[0021] Other advantages, objects and characteristics of the application will be embodied in part through the following description, and will be understood by those skilled in the art through research and practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 It is a real object diagram, a local high-speed camera photo of the microfluidic chip of the application; Figure 2 It is a structure schematic diagram of the microfluidic chip of Example 1 and a sorting process schematic diagram; Figure 3 It is a connection schematic diagram of the single solid core droplet automatic sorting device based on image recognition and dielectrophoresis force in Example 1; Figure 4 It is an optical photo of the droplets collected before and after sorting in Example 2; Figure 5 It is a size distribution diagram of the single solid core solid-water-oil droplets collected in Example 2 of the application; Figure 6 It is a sorting efficiency and collection efficiency diagram of Example 2 of the application. DETAILED DESCRIPTION

[0023] The application will be further described in detail below with reference to the drawings, so that those skilled in the art can implement it according to the description.

[0024] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof. Example 1 like Figure 3 As shown, an automated single-solid-nucleus droplet sorting device based on image recognition and dielectrophoresis includes the following structure: Integrated microfluidic chips, such as Figure 1 and Figure 2 As shown, it includes: The cross-shaped droplet generation region 12 includes a dispersed phase channel 10 and continuous phase channels 11 located on both sides of the dispersed phase channel; Sheath flow regulation zone 14 is connected to the dispersed phase channel 10, and sheath flow phase channels 13 are connected to both sides of the dispersed phase channel 10; Dielectrophoretic sorting area 16 is provided with a pair of parallel liquid metal electrodes 15. The positive electrode of the parallel liquid metal electrode 15 is connected to the positive output terminal of a high-voltage amplifier via a high-voltage silicone wire, and the negative electrode of the parallel liquid metal electrode 15 is connected to the negative output terminal of a microcontroller via a high-voltage silicone wire. The high-voltage silicone wire is sealed and fixed with epoxy resin. The liquid metal used in the parallel liquid metal electrode 15 includes one of gallium indium tin alloy, gallium indium alloy, bismuth indium alloy, and low-melting-point gallium-based alloy, which has good conductivity and fluidity.

[0025] Figure 1 In the middle, the dashed box shows, from left to right, the cross-shaped droplet generation area 12, the sheath flow regulation area 14, and the dielectrophoretic sorting area 16.

[0026] An image acquisition system, comprising: An industrial camera equipped with an LED light source; A computer connected to the industrial camera.

[0027] The sorting control system includes: A signal generator is connected to the high-voltage amplifier, which is connected to the parallel liquid metal electrode of the integrated microfluidic chip via a high-voltage silicone wire. A microcontroller has its input terminal connected to the computer and its output terminal connected to the input terminal of a DC amplifier; the DC amplifier is connected to a DC power supply; the microcontroller is an Arduino; the microcontroller is used to implement the conversion control of computer image recognition results into electrical signal output; A high-voltage switching device, the control end is connected with the output end of the direct current amplifier, the positive output end of the high-voltage switching device is connected with the negative output end of the high-voltage amplifier, and the negative output end of the high-voltage switching device is connected with the parallel liquid metal electrode of the integrated microfluidic chip. The high-voltage switching device is selected from high-voltage dry reed switches; the high-voltage switching device is used to realize the on-off control of the high-voltage alternating current applied to the microfluidic chip.

[0028] Embodiment 2 An automatic sorting method of single solid core droplets based on image recognition and dielectrophoresis force, which uses the automatic sorting device of single solid core droplets based on image recognition and dielectrophoresis force in embodiment 1, comprises the following steps: Step one, an integrated microfluidic chip is prepared by using a mold and a PDMS mold turning process; liquid metal is injected into the electrode channel, and a high-voltage silica gel wire is inserted to establish electrode connection, and then sealed and fixed with epoxy resin; Step two, 2wt% polyvinyl alcohol (PVA) aqueous solution (polyvinyl alcohol molecular weight 146000, hydrolysis degree 99%) is configured as a dispersed phase, and water-containing PS microspheres (diameter 850±10μm) are dispersed therein; 144mL of a mixed oil phase solution of dibutyl phthalate (DBP) and 66mL of dioctyl sebacate (DOS) is configured as a continuous phase solution and a sheath flow phase solution; Step three, an industrial camera is connected with a computer; an injection pump is started to generate a droplet flow, and the dispersed phase containing solid microspheres generates single solid core solid-water-oil droplets in the cross-shaped droplet generation area of the integrated microfluidic chip under the action of the continuous phase solution, and the dispersed phase generates water-oil droplets in the cross-shaped droplet generation area under the action of the continuous phase solution, 300 photos of water-oil droplets and single solid core solid-water-oil droplets are taken in advance, each photo is labeled, a labeled data set containing water-oil droplets (category 0) and single solid core solid-water-oil droplets (category 1) is established, and a droplet classification model is trained by using YOLOv8 algorithm; single solid core solid-water-oil droplet pictures are obtained by taking photos of single solid core solid-water-oil droplets in advance; the labeled droplet pictures are input into YOLOv8 for training to obtain optimal model parameters; the single solid core solid-water-oil droplet pictures are input into the trained YOLOv8 model to verify the accuracy; Step four, a signal generator is connected with a high-voltage amplifier, a computer is connected with an input end of an Arduino controller, a direct current amplifier is connected with an output end of the Arduino controller, a control end of a high-voltage dry reed switch is connected with an output end of the direct current amplifier, a positive output end of the high-voltage amplifier is directly connected with a positive electrode channel, a positive output end of the high-voltage dry reed switch is connected with a negative output end of the high-voltage amplifier, and a negative output end of the high-voltage dry reed switch is connected with a negative electrode channel, so as to realize the conversion of image recognition results to dielectrophoresis sorting signals; Step 5: Start the syringe pumps to generate droplet flow. The flow rates of syringe pump 3 (for injecting the dispersed phase containing PS microspheres into the dispersed phase channel of the integrated microfluidic chip) and syringe pump 1 (for injecting the continuous phase solution into the continuous phase channel of the integrated microfluidic chip) are fixed at 3 mL / h and 20 mL / h, respectively. The flow rate of syringe pump 2 (for injecting the sheath phase solution into the sheath phase channel of the integrated microfluidic chip) is 6 mL / h. The flow rate is mL / h. An alternating signal is generated by a signal generator with an amplitude of 5V and a frequency of 500Hz. A high-voltage amplifier is then activated to amplify the alternating signal, adjusting the amplification factor to achieve an output voltage of 1600V. The high-voltage voltage is connected to a high-voltage silicone wire via a high-voltage reed switch. An industrial camera is positioned to capture images in real-time from the integrated microfluidic chip's recognition area and analyzed using a droplet classification model. Specifically, the industrial camera inputs the captured images into a trained YOLOv8 model, which can identify whether the droplets in the image are the target droplets (single-core solid-water-oil droplets). When a single-core solid-water-oil droplet is detected, the computer outputs a digital signal to the Arduino controller. Upon receiving the digital signal 1, the Arduino controller outputs a high-level electrical signal to a DC amplifier. After DC amplification, a 12V voltage is output to the high-voltage reed switch. This activates the high voltage output from the high-voltage amplifier, which is then applied to the dielectric electrophoresis sorting area. The dielectric electrophoresis electric field activates, causing the single-core solid-water-oil droplet to deflect into the collection channel and be collected using a petri dish.

[0029] like Figure 4 As shown in the photos before and after sorting, the droplets before sorting were a mixture of single-nucleus solid-water-oil droplets and water-oil droplets, while the droplets collected after sorting contained only single-nucleus solid-water-oil droplets and water-oil droplets, indicating that the single-nucleus solid-water-oil droplets and water-oil droplets were successfully sorted.

[0030] like Figure 5 As shown, the size distribution of the single solid-water-oil droplets obtained ranges from 853.8 to 887.0 μm, with a mean of 870 μm and a coefficient of variation of 0.92%.

[0031] Sorting efficiency is defined as the ratio of the number of single-nucleus solid-water-oil droplets that successfully enter the collection channel to the total number of single-nucleus solid-water-oil droplets. Figure 6 As shown, the results indicate that the sorting efficiency reached 99.6%. The collection efficiency is defined as the ratio of the number of single solid-nucleus solid-water-oil droplets in the target collection channel to the total number of droplets collected in that channel, and the collection efficiency also reached 99.6%.

[0032] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0033] While embodiments of the application have been disclosed in connection with the above specification and drawings this description is not intended to limit the scope of the application and many modifications, enhancements, alternatives, and variations will become apparent to those skilled in the art from this disclosure. Accordingly, it is intended that the application not be limited to the described embodiments, but that it include all variations falling within the scope of the claims, and their equivalents.

Claims

1. An automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis, characterized in that, Includes the following steps: Step 1: Inject a dispersed phase containing solid microspheres, a continuous phase solution, and a sheath flow phase solution into the integrated microfluidic chip. Under the action of the continuous phase solution, the dispersed phase containing solid microspheres generates single-core solid-water-oil droplets in the cross-shaped droplet generation region of the integrated microfluidic chip. Under the action of the continuous phase solution, the dispersed phase generates water-oil droplets in the cross-shaped droplet generation region. Step 2: Build an image acquisition system and create a labeled dataset containing water-oil droplets and single-solid-nucleus solid-water-oil droplets. Using the labeled dataset, train a droplet classification model using a deep learning image recognition algorithm. Step 3: Construct a sorting control system. Start the injection pump to generate droplet flow, start the signal generator to generate an alternating signal, start the high-voltage amplifier to amplify the alternating signal, adjust the amplification factor, and use the industrial camera of the image acquisition system to acquire images in real time. The system then uses a droplet classification model to identify whether the acquired image is a single-core solid-water-oil droplet. When a single-core solid-water-oil droplet is detected, the computer outputs a digital signal to the microcontroller. When the microcontroller receives the digital signal, it outputs a high-level electrical signal to the DC amplifier. After being amplified by the DC amplifier, the signal is output to the high-voltage switching device. At this time, the high-voltage switching device is turned on, and the high voltage output by the high-voltage amplifier is applied to the dielectric electrophoresis sorting area. The dielectric electrophoresis electric field is activated, causing the single-core solid-water-oil droplet to deflect into the collection channel.

2. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step one, the dispersed phase is a 2 wt% aqueous solution of polyvinyl alcohol, the molecular weight of polyvinyl alcohol is 146,000, and the degree of hydrolysis is 99%. The solid microspheres are PS microspheres; Both the continuous phase solution and the sheath flow phase solution are oil phase solutions of a mixture of dibutyl phthalate and dioctyl sebacate, with the volume of dibutyl phthalate being 1 to 5 times that of dioctyl sebacate.

3. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step one, the integrated microfluidic chip includes: The cross-shaped droplet generation region includes a dispersed phase channel and continuous phase channels located on both sides of the dispersed phase channel; A sheath flow regulation zone, which is connected to the dispersed phase channel, and sheath flow phase channels are connected to both sides of the dispersed phase channel; The dielectric electrophoretic sorting area is provided with a pair of parallel liquid metal electrodes. The positive electrode of the parallel liquid metal electrode is connected to the positive output terminal of the high-voltage amplifier through a high-voltage silicone wire, and the negative electrode of the parallel liquid metal electrode is connected to the negative output terminal of the microcontroller through a high-voltage silicone wire. The high-voltage silicone wire is sealed and fixed with epoxy resin.

4. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 3, characterized in that, The liquid metal used in the parallel liquid metal electrode includes one of gallium indium tin alloy, gallium indium alloy, bismuth indium alloy, and low-melting-point gallium-based alloy.

5. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step two, the image acquisition system includes: An industrial camera equipped with an LED light source; A computer connected to the industrial camera.

6. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step two, the deep learning image recognition algorithm includes one of YOLOv8, YOLO, SSD, Faster R-CNN, and RetinaNet.

7. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step three, the sorting control system includes: A signal generator is connected to the high-voltage amplifier, which is connected to the parallel liquid metal electrode of the integrated microfluidic chip via a high-voltage silicone wire. A microcontroller, the input of which is connected to the computer, and the output of which is connected to the input of a DC amplifier; the DC amplifier is connected to a DC power supply. A high-voltage switching device, the control terminal of which is connected to the output terminal of the DC amplifier, the positive terminal of the output terminal of the high-voltage switching device is connected to the negative terminal of the output terminal of the high-voltage amplifier, and the negative terminal of the output terminal of the high-voltage switching device is connected to the parallel liquid metal electrode of the integrated microfluidic chip.

8. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step three, the high-voltage switching device includes one of the following: high-voltage reed switch, high-voltage relay, high-voltage MOSFET, high-voltage thyristor, and high-voltage field-effect transistor.

9. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step three, the microcontroller includes one of Arduino, Raspberry Pi, STM32, ESP32, and PIC microcontroller.

10. The automatic sorting method for single solid-nucleus droplets based on image recognition and dielectrophoresis as described in claim 1, characterized in that, In step three, during dielectrophoretic sorting, the deflection distance of the single solid-water-oil droplet along the y-axis is greater than or equal to 20% of the channel width, and the deflection response time is less than 100 milliseconds.

Images