Micro-fluidic chip manufacturing method and application method based on 3D printing technology

Abstract

The invention relates to the technical field of biomedical detection, in particular to a micro-fluidic chip manufacturing method and an application method based on a 3D printing technology, and the method comprises the following steps: manufacturing a micro-fluidic chip through combination of 3D printing and soft lithography, optimizing a liquid drop generation and detection process, and developing a YOLO-Drop deep learning model to realize automatic identification and counting of liquid drop targets. According to the invention, the manufacturing cost and time of the micro-fluidic chip can be obviously reduced, the droplet generation precision and detection efficiency are improved, and meanwhile, ultra-sensitive protein detection is realized through intelligent analysis. Experimental verification shows that the detection limit of the system reaches 0.015 fM, the recovery rate range is 94.0%-116.96%, the coefficient of variation is lower than 10%, and an innovative solution is provided for biomedical detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical detection, and in particular to a microfluidic chip manufacturing method and application method based on 3D printing technology. BACKGROUND

[0002] In the field of biomedical detection, high sensitivity and high specificity detection of trace protein markers has always been an important direction of biological analysis technology research, and is also a key link to realize early disease diagnosis and precision treatment. As a new generation of protein quantitative analysis technology, digital enzyme-linked immunosorbent assay (ddELISA) has been widely used in the ultra-sensitive detection of tumor markers, inflammatory factors and neurodegenerative disease-related proteins due to its single molecule detection capability and extremely low detection limit. The core principle is to divide the immune reaction system into a large number of small droplets, and each droplet independently completes immune recognition and signal amplification, thereby realizing accurate counting of a single target molecule. However, the preparation of traditional microfluidic chips is highly dependent on photolithography process, which is tedious, expensive and time-consuming, and is difficult to quickly iterate and customize, which seriously restricts its promotion in scientific research and clinical application. In addition, traditional photolithography process has shortcomings in microchannel size precision, surface roughness and material biocompatibility, which may have a negative impact on droplet generation efficiency and immune analysis accuracy, limiting the further development of ddELISA technology.

[0003] In recent years, 3D printing technology provides a new idea for rapid prototyping of microfluidic chips due to its significant advantages such as flexible design, rapid forming and low cost. In theory, 3D printing can realize free construction and rapid iteration of microfluidic chip structure, greatly improving the development efficiency of microfluidic system. However, the application of 3D printing in microfluidic chip manufacturing still faces a series of challenges, including insufficient microchannel size precision due to limited printing resolution, uneven biocompatibility and chemical stability of printing materials, and high microstructure surface roughness, which may affect droplet formation efficiency and immune analysis accuracy. Therefore, how to combine 3D printing technology with microfluidic chip manufacturing process to overcome the above technical bottlenecks has become a key problem to improve the performance of ddELISA platform. SUMMARY

[0004] The purpose of the present application is to provide a microfluidic chip manufacturing method and application method based on 3D printing technology, which combines 3D printing and soft lithography technology, optimizes the design and manufacturing process of microfluidic chips, significantly reduces manufacturing cost and time, and improves the precision and efficiency of droplet generation and detection.

[0005] According to an aspect of the present application, a microfluidic chip manufacturing method based on 3D printing technology is provided, comprising:

[0006] Designing a microfluidic chip schematic diagram using computer-aided design software, including microchannel structure and droplet generation area;

[0007] Exporting the design file into STL format and slicing it through slicing software, then printing the microfluidic chip template using a high-precision 3D printer;

[0008] Cleaning, drying and surface treating the printed template to ensure the template surface is smooth and free of residual resin;

[0009] Using the photocured resin template as a mold, pouring polydimethylsiloxane precursor liquid and curing to form a PDMS soft template with microfluidic structure;

[0010] Peeling off the cured PDMS layer from the template and completing the preliminary formation of the microfluidic chip through cutting and punching operations;

[0011] Surface activating the PDMS layer and glass substrate using a plasma treatment instrument, then pressing and heating them to achieve firm bonding;

[0012] Checking the bonded chip under an optical microscope to ensure the channel structure is complete and the connectivity is good, and finally obtaining a microfluidic chip with micron-level structure.

[0013] In some embodiments, the height of the microchannel is 10-60 μm, and the pitch is 6-60 μm.

[0014] In some embodiments, during the PDMS soft template forming process, the mold surface is subjected to fluorination treatment to enhance the hydrophobicity of the microchannels.

[0015] In some embodiments, the polydimethylsiloxane precursor liquid is mixed from PDMS matrix and crosslinking agent in a mass ratio of 10:1.

[0016] In some embodiments, the surface activation treatment time of the plasma treatment instrument is 1 min.

[0017] According to another aspect of the present application, a microfluidic chip application method based on 3D printing technology is provided, comprising:

[0018] Injecting the oil phase and the water phase into the continuous phase inlet and the dispersed phase inlet of the microfluidic chip, respectively;

[0019] Controlling the flow rate through a multi-channel syringe pump to stabilize the generation of uniform microdroplets;

[0020] Encapsulating an immune complex in the droplet and generating a fluorescent signal through an enzymatic reaction;

[0021] After the droplet generation, the droplets are incubated and the droplet images are collected by fluorescence microscopy.

[0022] In some embodiments, the oil phase flow rate is set to 20 μL / min and the water phase flow rate is set to 2 μL / min.

[0023] In some embodiments, the droplet image is subjected to target detection by using a YOLO-Drop model, which is based on a YOLOv8 architecture and introduces a patch detection strategy and a bidirectional feature pyramid network.

[0024] In some embodiments, the Focal Loss loss function and the multi-scale data augmentation strategy are used in the training process of the YOLO-Drop model.

[0025] In some embodiments, the YOLO-Drop model weight file is exported in the ONNX format and is optimized by the TensorRT tool into an inference engine file suitable for embedded platforms.

[0026] Compared with the prior art, the present application has the following beneficial effects: through the combination of 3D printing technology and soft lithography technology, low-cost and high-efficiency manufacturing of microfluidic chips is realized, solving the problem of complex and expensive traditional process; through the optimization of microfluidic chip design and droplet generation process, high precision and high stability of droplet generation are realized; through the development and deployment of the YOLO-Drop model, the automation degree and accuracy of droplet target detection are significantly improved; experimental verification shows that the system constructed by the present application has ultra-high sensitivity in IL-6 antigen detection, with a detection limit of 0.015 fM, a recovery rate range of 94.0%-116.96%, and a coefficient of variation less than 10%, fully verifying its high sensitivity and reliability in complex biological matrix. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0028] Fig. 1 A schematic diagram of the microfluidic chip manufacturing process based on 3D printing;

[0029] Fig. 2 A flow chart of the microfluidic chip manufacturing method;

[0030] Fig. 3 A flow chart of the application method based on the microfluidic chip. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described below in conjunction with accompanying drawings of the embodiments of the present application. Figs. 1-3 The technical solutions in the embodiments of the present application will be described below in conjunction with accompanying drawings of the embodiments of the present application.

[0032] The present application provides a microfluidic chip manufacturing method based on 3D printing technology and its application in digital enzyme-linked immunoassay.

[0033] The present application provides a microfluidic chip manufacturing method based on 3D printing technology and its application in digital enzyme-linked immunoassay.

[0034] First, in the design and manufacture of the microfluidic chip, the AutoCAD 2020 computer-aided design software is used to draw the schematic diagram of the microfluidic chip, including the microchannel structure and the droplet generation area. According to the experimental requirements, the height and spacing of the microchannel can be flexibly adjusted, the height of the microchannel is 20 μm, and the total height of the template base plate is 3 mm. After the design is completed, the schematic diagram is exported as an STL format file, and the BMF_Slicer_1.6.4 slicing software is used for slicing processing. Subsequently, a high-precision 3D printer is used to print the microfluidic chip template. After printing is completed, the platform screw is removed and the platform is taken out, the surface is cleaned with anhydrous ethanol to ensure that the template surface is smooth and free of residual resin. After the printed part is cleaned by ultrasonic and ultraviolet curing, its integrity and accuracy are checked. If there are slight defects, they can be polished by fine sandpaper to ensure that the template quality meets the subsequent experimental requirements.

[0035] In the PDMS soft template forming process, the treated 3D printed template is used as a mold, and a polydimethylsiloxane precursor liquid is poured into the mold, and vacuum is applied in a vacuum drying box to remove air bubbles. The precursor liquid is mixed with PDMS matrix and crosslinking agent in a mass ratio of 10:1, and after stirring, it is poured into the mold. The curing condition is set to 65℃ for 4h, so that the PDMS precursor liquid becomes a solid PDMS layer, and the microfluidic structure is copied onto the PDMS layer. After curing is completed, the PDMS layer is peeled off from the template, and a cutting machine and a puncher are used to complete the preliminary forming. The punching operation uses a 0.6mm needle to punch through the inlet and outlet channel ports of the chip to ensure good channel connectivity. Subsequently, the plasma treatment instrument 5 is used to activate the surface of the PDMS layer and the glass substrate to form hydroxyl structures on the surface. The activated PDMS layer and the glass substrate are quickly aligned and pressed tightly, and are heated to 400℃ for 40s to achieve firm bonding. After bonding is completed, the chip is placed under a conventional optical microscope to check the channel structure integrity and connectivity, to ensure that the microfluidic chip has micron-level structure and no channel blockage.

[0036] In the droplet generation and digital enzyme-linked immunoassay, a microfluidic chip was used for the experiment. The oil phase and the water phase were injected into the continuous phase inlet and the dispersed phase inlet of the microfluidic chip, respectively, where the oil phase was a mixture of HFE7500 and 5% fluorosurfactant, and the water phase was a solution containing immune complex magnetic beads. The flow rate was controlled by a multi-channel syringe pump, and the oil phase flow rate was set to 20 μL / min, and the water phase flow rate was set to 2 μL / min, so that uniform microdroplets were stably generated. The droplet size can be precisely controlled by adjusting the microchannel height and spacing, for example, when the channel height is 10 μm and the spacing is 6 μm, the generated droplet diameter is about 8 μm, and when the channel height is 60 μm and the spacing is 60 μm, the generated droplet diameter can reach more than 80 μm. After the droplets were generated, they were incubated for about 2 min to ensure that the enzymatic reaction was fully carried out, and then the droplet images were collected by a fluorescence microscope. The positive fluorescence signal in the droplets increased significantly, verifying the excellent performance of the platform in detecting fM-level ultra-low concentration proteins.

[0037] To improve the automation and accuracy of droplet target detection, a deep learning-based target detection model YOLO-Drop was developed. When constructing the dataset, based on the droplet images generated by the above microfluidic chip, 5574 high-resolution fluorescence images were collected, covering five types of targets: single magnetic bead signal droplet, single magnetic bead non-signal droplet, non-magnetic bead but signal droplet, multiple magnetic bead signal droplet, and multiple magnetic bead non-signal droplet. The dataset was labeled using the X-AnyLabeling tool, and the labeling process included manually labeling 1000 images and training the classic target detection algorithm YOLOv8n, and then using its automatic target detection function to automatically label the remaining images on a large scale, and manually correcting them one by one. After labeling, the data was converted to YOLO format to adapt to the training requirements of YOLOv8 and YOLO-Drop models.

[0038] The YOLO-Drop model is optimized on the basis of YOLOv8, including the introduction of block detection strategy, bidirectional feature pyramid network BiFPN and convolutional attention module CBAM. The backbone network part uses EfficientNetV2-S architecture instead of the original Backbone of YOLOv8, and embeds CBAM attention mechanism at multiple stages to strengthen the model's attention to significant droplet regions. The feature fusion module part uses an improved BiFPN structure, which effectively alleviates the missed detection problem of traditional detection networks when dealing with targets with dramatic size changes or dense overlaps through a bottom-up and top-down bidirectional information fusion mechanism, combined with channel weighting strategy and lightweight SEBlock attention module. The detection head part uses a decoupled structure, with independent classification branch and bounding box regression branch, reducing the mutual interference between tasks and improving the stability and accuracy of the detection results.

[0039] During training, multi-scale data augmentation and Focal Loss loss function are adopted to significantly improve the model's detection ability for small targets and dense targets. Data augmentation strategies include random cropping, flipping, color transformation, and Gaussian noise, among which Gaussian noise simulates the possible changes in lighting or image blur during the experiment, thereby improving the robustness of the model. Focal Loss reduces the weight of easily classified samples and enhances the impact of difficult-to-classify samples, thereby reducing background false positives and improving small target recall. After training, the YOLO-Drop model weight file is exported in ONNX format and further converted to a TensorRT engine file to accelerate inference performance on embedded platforms. The embedded platform uses NVIDIA Jetson Orin Nano, and the ONNX model is optimized for layer fusion and tensor memory optimization using the TensorRT optimization tool to generate an engine file suitable for Orin Nano GPU acceleration. During deployment, the TensorRT runtime environment is initialized, GPU memory buffer is allocated, and the optimized model weight is loaded through the TensorRT API.

[0040] To support real-time image loading, model calling for inference, and visualization of detection results, a graphical user interface named DropleSense is developed. The interface uses modular design and includes four core functional areas: raw image frame, detection result frame, function button area, and result statistics frame. Users can load model weight files by clicking the "Load Model" button in the function button area, select images to be detected by clicking the "Load Image" button, call the YOLO-Drop model for inference by clicking the "Start Detect" button, and visualize the output in the detection result frame. After detection, the result statistics frame synchronously displays the counting results of each type of target and the key parameters of the model during runtime, such as inference time and confidence distribution. DropleSense uses multi-thread asynchronous inference, which does not affect the smoothness of the interface while performing detection tasks, ensuring a smooth user experience.

[0041] Experimental verification shows that the system constructed by the present application has ultra-high sensitivity in IL-6 antigen detection, with a detection limit of 0.015 fM, a recovery rate range of 94.0%-116.96%, and a coefficient of variation less than 10%, fully verifying its high sensitivity and reliability in complex biological matrices.

[0042] In summary, the present application realizes low-cost and high-efficiency manufacturing of microfluidic chips by combining 3D printing technology and soft lithography technology, solves the problem of complex and expensive traditional process, and significantly improves the automation and accuracy of droplet target detection through the development and deployment of the YOLO-Drop model, providing an innovative solution for the biomedical detection field.

[0043] The above shows and describes the basic principles and main features of the present application and the advantages of the present application, and it is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting in any respect, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended to encompass all variations falling within the meaning and scope of the equivalent elements of the claims. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0044] In addition, it should be understood that although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the description manner of the specification is only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that those skilled in the art can understand.

Claims

1. A method for manufacturing microfluidic chips based on 3D printing technology, characterized in that, include: Design the schematic diagram of a microfluidic chip using computer-aided design software, including the microchannel structure and droplet generation region; The design file is exported as an STL format and sliced ​​using slicing software. Then, a high-precision 3D printer is used to print the microfluidic chip template. The printed templates are cleaned, dried, and surface-treated to ensure that the template surface is smooth and free of residual resin. A photocurable resin template is used as a mold, into which polydimethylsiloxane precursor liquid is poured and cured to form a PDMS soft template with a microfluidic structure. The cured PDMS layer is peeled off from the template, and the initial forming of the microfluidic chip is completed through cutting and drilling operations; The PDMS layer and the glass substrate were surface activated using a plasma treatment device, and then the two were pressed together and heated to achieve a strong bond. After bonding, the chip is inspected under an optical microscope to ensure that the channel structure is intact and the connectivity is good, ultimately resulting in a microfluidic chip with a micron-scale structure.

2. The manufacturing method according to claim 1, characterized in that, The height of the microchannels is 10 μm to 60 μm, and the spacing is 6 μm to 60 μm.

3. The manufacturing method according to claim 2, characterized in that, During the PDMS soft template molding process, the mold surface is subjected to fluorination treatment to enhance the hydrophobicity of the microchannels.

4. The manufacturing method according to claim 1, characterized in that, The polydimethylsiloxane precursor liquid is composed of a PDMS matrix and a crosslinking agent mixed in a mass ratio of 10:

1.

5. The manufacturing method according to claim 1, characterized in that, The surface activation treatment time of the plasma processor is 1 minute.

6. A method for applying microfluidic chips based on 3D printing technology, characterized in that, include: The oil phase and the aqueous phase are injected into the continuous phase inlet and the dispersed phase inlet of the microfluidic chip, respectively; The flow rate is controlled by a multi-channel syringe pump to stably generate uniform microdroplets; Immunocomplexes are encapsulated in droplets and fluorescent signals are generated via enzymatic reactions. After the droplets are generated, they are incubated and images of the droplets are acquired using a fluorescence microscope.

7. The application method according to claim 6, characterized in that, The oil phase flow rate was set to 20 μL / min, and the water phase flow rate was set to 2 μL / min.

8. The application method according to claim 6, characterized in that, The YOLO-Drop model is used to detect targets in droplet images. The YOLO-Drop model is based on the YOLOv8 architecture and introduces a block detection strategy and a bidirectional feature pyramid network.

9. The method of application according to claim 6, characterized in that, The YOLO-Drop model employs Focal Loss and multi-scale data augmentation strategies during training.

10. The application method according to claim 8, characterized in that, The YOLO-Drop model weight file was exported in ONNX format and optimized using the TensorRT tool into an inference engine file suitable for embedded platforms.

Images