Bimodal immunochromatography detection system based on smart phone deep learning

Abstract

The invention discloses a dual-mode immunochromatography detection system based on smart phone deep learning, and the system comprises a smart phone which is provided with a camera and a detection APP, and is disposed on an adaptive hardware platform; the self-adaptive hardware platform is provided with a stretchable support, an optical path module, an imaging macro lens and other structures. The stretchable bracket utilizes a micro slide rail to adjust the position of the mobile phone; the light path module is used for exciting an imaging signal of the immunochromatography chip and guiding the imaging signal to the imaging lens; the imaging macro lens is used for amplifying a to-be-detected area of the chip and guiding the to-be-detected area to the smart phone camera; and the bimodal immunochromatography chip is respectively configured to be a distance type fluorescence chip and a traditional colorimetric chip. The detection system does not need an extra large detection device and complex image processing software, the detection process is remarkably simplified, the data size needed by deep learning model training is reduced, computing resources are optimized, convenience and intelligence of detection are achieved, and expandability and maintainability of the system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of immunochromatographic detection, and in particular to a dual-modal immunochromatographic detection system based on deep learning of a smart phone. Background Art

[0002] At present, instrumental analysis and immunoassay are the main methods for detecting biomarkers. Instrumental analysis has high sensitivity and accuracy, but requires expensive equipment, complex pretreatment, professional operation and long detection time, which limits the application of such methods in large-scale and on-site detection. Rapid immunochromatographic detection technology (LFIA, lateral flow immunoassay) is a chromatographic analysis technology that combines membrane chromatography technology with immunolabeling. Through the luminescence or color development of markers (such as quantum dot microspheres QBs, colloidal gold AuNPs, etc.), it can be determined whether the sample contains the analyte and its content. Immunochromatography has become one of the most advanced point of care testing (POCT) platforms due to its simple use, short time consumption, good portability, low cost, high reliability, and no need for complex instruments. It has played a vital role in food safety testing, environmental testing and clinical medicine.

[0003] As the demand for POCT diagnosis increases, improving the accuracy and sensitivity of immunochromatographic testing has become a research focus. Machine learning, especially deep learning technology, plays a key role by combining mathematical calculations, signal and image processing methods with biological analysis to optimize test results. Deep learning uses more intelligent algorithms to automatically and multi-level learn complex features, analyze images and data, and thus quickly and accurately identify, classify and grade concentrations of targets, providing a basis for determining risk levels. However, the implementation of deep learning is usually accompanied by a large amount of computing resources and excellent hardware configuration requirements, including high-performance processors, large memory, GPU acceleration, fast storage and a stable operating system.

[0004] When exploring cutting-edge areas to improve the performance of immunochromatographic detection systems, we should not only focus on the refinement and efficiency of signal processing algorithms, but also on the breakthrough design of chips and detection methods. For example, immunochromatographic chips based on material innovation and structural optimization have emerged. And innovative chip signal reading strategies will become the key force to promote the performance leap of detection systems. Unlike traditional readout strategies based on intensity changes, converting signals into results related to visual distance is largely independent of operator intervention and easy to handle. However, the current microfluidic detection methods based on distance readout still require the additional introduction of enzyme substrates or colorimetric reagents, and cannot complete the reading of distance signals in one step, which increases the complexity of the operation. Given that the design of distance-based immunochromatographic chips is still in its infancy, efforts should be made to develop a one-step distance-based immunochromatographic detection system to simplify the operating process, improve detection efficiency, and make it a competitive new rapid analysis tool.

[0005] Therefore, technicians in this field are committed to developing a dual-modal immunochromatographic detection system based on smartphone deep learning. Summary of the invention

[0006] In view of the above-mentioned defects of the prior art, the present invention provides a dual-modal immunochromatography detection system based on deep learning of a smartphone, the detection system comprising:

[0007] A smartphone, equipped with a camera and a detection app, is placed on an adaptive hardware platform;

[0008] The detection APP has built-in deep learning models and quantitative detection algorithms;

[0009] Adaptive hardware platform with stretchable bracket, optical path module, imaging macro lens and other structures, no battery space required, integrated long-term wireless charging function;

[0010] The stretchable bracket is adapted to smartphones of different sizes, and uses a micro-slide structure to adjust the relative position of the smartphone camera and the imaging lens;

[0011] The optical path module is used to excite the imaging signal of the immunochromatographic chip and guide it to the imaging lens;

[0012] The imaging macro lens is used to magnify the area to be tested on the immunochromatographic chip and guide it to the camera of the smartphone;

[0013] The dual-modal immunochromatographic chip is configured as a distance-type fluorescent chip and a traditional colorimetric chip respectively; the distance-type fluorescent chip has a T zone, the spray pattern is dot-shaped, and the T zone of each chip contains 10 capture points.

[0014] Furthermore, the mobile phone is configured with a detection APP, which encapsulates the trained deep learning model YOLO v5 through the Flask framework. The deep learning model YOLO v5 can be used to process the images of the immunochromatography chip imaging to obtain image analysis results related to visual distance or signal strength.

[0015] Furthermore, the mobile phone remotely communicates with another terminal to transmit the captured picture to the terminal, and the terminal is configured with a trained deep learning model YOLO v5, and the deep learning model YOLO v5 can be used to process the pictures imaged by the immunochromatography chip to obtain image analysis results related to visual distance or signal strength.

[0016] Furthermore, the deep learning model YOLO v5 uses a lightweight network, and fuses shallow features and deep semantic features of the image obtained by the convolutional neural network of the deep learning model YOLO v5 to achieve detection of targets of different scales.

[0017] Furthermore, the deep learning model YOLO v5 configured on the mobile phone is pruned, distilled and compressed.

[0018] Furthermore, the smartphone detection APP also includes a quantitative detection module for calling a quantitative detection algorithm to complete the acquisition of OD values ​​from RGB values ​​in the imaging image analysis of the immunochromatographic chip, and to obtain concentration information by fitting the relationship between the OD value and the concentration standard curve.

[0019] Furthermore, the hardware platform has a closed black housing with dimensions of 9.8×7.3×3.4 cm. 3 , the material is polystyrene, and the inner and outer walls are treated with black matte.

[0020] Furthermore, the hardware platform also includes a microcircuit module that integrates functions such as wireless power supply, low-power Bluetooth communication, and control of the light source irradiated on the immunochromatographic chip.

[0021] Furthermore, the imaging lens is embedded in the closed black housing.

[0022] Furthermore, the dual mode is a detection mode of fluorescence or colorimetry. When the detection mode is colorimetry, the immunochromatographic chip further includes a colorimetric immunochromatographic chip, and each chip contains a T line and a C line.

[0023] The detection system of the present invention integrates a smart phone platform, fully exploits the functional development of smart phones, does not require additional large-scale detection devices and complex image processing software, significantly simplifies the detection process, reduces the amount of data required for deep learning model training, and optimizes computing resources; achieves significant sensitivity improvement; realizes system integration and portability, convenient and intelligent detection, and has better scalability and maintainability;

[0024] The detection system of the present invention has excellent reliability and excellence, and also has excellent specificity, repeatability and sensitivity;

[0025] The detection system of the present invention uses a smartphone APP as a detection platform, which not only greatly reduces production costs, but also significantly improves the deployment efficiency and popularity of the system through convenient installation. In addition, the APP-based detection system can continuously optimize functions and performance through software updates to meet the growing needs of users.

[0026] The detection system of the present invention realizes fully intelligent and highly stable immunochromatographic rapid detection, showing great potential and value in POCT applications, and providing strong technical support and ample application possibilities for the community, family and personalized applications of POCT. Its convenient design not only allows users to easily operate it without professional skills, but also promotes the comprehensive application of AI technology in medical, environmental monitoring and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] FIG. 1( a ) is a structural diagram of the overall appearance of the hardware platform of the dual-modal immunochromatographic detection system based on deep learning of a smartphone of the present invention;

[0028] FIG1( b ) is a partial perspective view of the hardware platform of the dual-modal immunochromatographic detection system based on smartphone deep learning of the present invention;

[0029] Figure 2 It is a test result diagram of a dual-modal immunochromatographic chip of a dual-modal immunochromatographic detection system based on deep learning of a smartphone of the present invention;

[0030] Figure 3 It is a graph showing the concentration classification results of standard solutions in two detection modes according to an embodiment of the dual-modal immunochromatographic detection system based on deep learning of a smartphone of the present invention;

[0031] Figure 4 It is a standard fitting curve diagram of fluorescence detection in an embodiment of a dual-modal immunochromatographic detection system based on deep learning of a smartphone of the present invention;

[0032] Figure 5 It is a standard fitting curve diagram of colorimetric detection in an embodiment of the dual-modal immunochromatographic detection system based on deep learning of a smartphone of the present invention.

[0033] 1-Adaptive mobile phone holder; 2-Imaging macro lens; 3-Chip card slot; 4-Optical path module; 5-Micro circuit module. DETAILED DESCRIPTION

[0034] The following describes several preferred embodiments of the present invention with reference to the drawings in the specification, so that the technical content is clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.

[0035] In the drawings, components with the same structure are indicated by the same numerical reference numerals, and components with similar structures or functions are indicated by similar numerical reference numerals. The size and thickness of each component shown in the drawings are arbitrarily shown, and the present invention does not limit the size and thickness of each component. In order to make the illustration clearer, the thickness of the components is appropriately exaggerated in some places in the drawings.

[0036] The present invention takes aflatoxin B1 (AFB1) biomarker as an example to illustrate an implementation method of a dual-modal immunochromatographic detection system based on smartphone deep learning based on the present invention, thereby realizing intelligent graded diagnosis and highly sensitive quantitative detection of AFB1. At the same time, the present invention can also be directly expanded to immunochromatographic intelligent detection of other sample biomarkers.

[0037] The dual-modal immunochromatography detection system based on deep learning of smart phones of the present invention mainly includes three parts: a smart phone APP based on YOLOv5, a dual-modal immunochromatography chip, and an adaptive hardware platform based on wireless power supply of smart phones. After the immunochromatography reaction is completed, for the fluorescence and colorimetric detection modes, the distance and intensity signal of the chip can be visually observed under 365nm excitation light and natural light, respectively, and the concentration classification or qualitative results can be preliminarily judged. Further, the chip and the smart phone are placed in the corresponding positions of the hardware platform, the smart phone APP is operated, and after the detection mode and image source are selected, the concentration level distinction and quantitative results of the object to be tested are obtained.

[0038] As shown in FIG. 1( a ) and ( b ), the smartphone-based adaptive hardware platform of the present invention has a closed black housing with a size of 9.8×7.3×3.4 cm. 3, the material is polystyrene, and the inner and outer walls are treated with black matte. It is equipped with an adaptive mobile phone holder, a microcircuit module, an optical path module, an imaging macro lens (preferably embedded in a black shell) and a chip card slot. The adaptive mobile phone holder uses a micro-slide structure to flexibly adjust the mobile phone camera to the best position to fit the macro lens, so as to accurately capture and enlarge the area to be tested. In addition, the adaptive mobile phone holder adopts a stretchable form to improve the wide adaptability to smartphones of different sizes. The microcircuit module integrates functions such as wireless power supply, low-power Bluetooth communication, and light source control. Among them, the wireless power supply module uses the principle of electromagnetic induction and the reverse charging function of the smartphone to convert the electromagnetic signal captured by the coil group into electrical energy for the normal operation of the hardware platform, saving battery space for the system. For the fluorescence and colorimetric detection modes, the optical path module uses symmetrical 395nm and 520nm LED long light plates as signal excitation light sources, respectively, and optical adjustment is performed to ensure that it can evenly and fully cover the area to be tested, thereby ensuring the portability, adaptability and stability of the detection system of the present invention.

[0039] In order to improve the convenience of reading the immunochromatographic detection signal and verify the feasibility of the system for multimodal detection, two types of immunochromatographic chips were prepared. The preparation process is as follows:

[0040] 1) For the distance-type fluorescent immunochromatographic chip, AFB1-BSA antigen (0.5 mg / ml) was sprayed onto the NC membrane (length 2.5 cm) as the T zone, and the spray pattern was set to dots with a longitudinal spacing of 2 mm to ensure that the T zone of each chip contained 10 capture points, and the spraying volume of each point was 0.10 μl.

[0041] 2) For the conventional colorimetric immunochromatographic chip, the spraying mode was set to linear, and AFB1-BSA antigen (1 mg / ml) and IgG (1 mg / ml) were sprayed onto the NC membrane at a speed of 1 μl / cm as T line and C line, respectively, with a spacing of 4 mm.

[0042] 3) After the corresponding antigen or antibody is sprayed, the strip is dried at 37°C for 3 hours. The chip's sample pad, NC membrane, and absorbent pad are assembled on the PVC base plate in a way that the adjacent parts overlap by 1-2 mm. Finally, the assembled strip is cut into 3 mm width to complete the chip preparation and stored in a dry environment for later use.

[0043] The QBs-mAb for fluorescence detection and the AuNPs-mAb immunoprobe for colorimetric detection were synthesized by the commonly used carbodiimide method and electrostatic binding method, respectively. In order to ensure the detection sensitivity, a series of optimization experiments were carried out on key factors such as NC membrane pore size, probe usage conditions, and probe synthesis coupling ratio to obtain the best signal intensity and gradient. Finally, for the fluorescence chip, the CN95 NC membrane was selected, the probe concentration and sample volume were 0.25 mg / ml and 3 μl, respectively, and the QBs and antibody coupling ratio was 1 mg: 25 μg. For the colloidal gold chip, the NC membrane selected CN140, the probe sample volume was 8 μl, and the AuNPs and antibody coupling ratio was 1 mg: 80 μg antibody.

[0044] Six concentrations of AFB1 standard solution samples were prepared, and according to industry standards, 0 and 0.05 ng / ml were defined as low concentrations, 0.2 and 1 ng / ml were defined as medium concentrations, and 10 and 40 ng / ml were classified as high concentrations.

[0045] like Figure 2 As shown in the figure, for the traditional colorimetric immunochromatographic chip, the visual detection limit was determined to be 0.2ng / ml, and the cutoff value was 20ng / ml, which provided a clear standard for the visual determination of positive samples. For the distance-type fluorescent immunochromatographic chip, 51 sets of experimental data were analyzed in detail, and obvious rules were obtained under naked eye observation: in the low concentration range, at least 8 points in the T zone were clearly lit, and if the brightness of 10 points could be observed to decrease successively, it was regarded as the reaction result of the positive sample. If the brightness difference was not large, it was judged as negative; when the bright points in the T zone were reduced to 4 to 7, this provided a basis for detection at medium concentrations; and in the high concentration range, the clear and bright points in the T zone were reduced to 3 or less. Therefore, the determination method based on the number of T zone lights further subdivided the concentration range and realized semi-quantitative detection with naked eyes. On the other hand, the readout strategy converts the signal into a result directly related to the visual distance, which is largely independent of the operator's intervention, making the test results more objective and easy to handle, and improving the detection efficiency.

[0046] In the present invention, the YOLO v5 model is used to train and identify the detection data. First, the network structure is optimized and customized for high performance. On the input side, considering the trade-off between network performance and accuracy, computational efficiency and memory usage, the total number of pixels of the input image is defined as 640×640×3 to ensure that the image size is moderate and the color information is rich. The Mosaic image enhancement method is adopted, and the data set is expanded by adjusting the saturation, brightness, contrast, noise degree of the image, and the color and geometric transformation methods of flipping the image, thereby significantly improving the robustness and generalization ability of the model. Backbone uses the lightweight network CSPDarknet53 to efficiently extract features through 53 carefully designed convolutional layers, 1 fully connected layer, and 3×3 convolutional kernels. The Neck of the model fuses the shallow features and deep semantic features of the image obtained by the convolutional neural network, thereby realizing the detection of targets of different scales. The 255×20×20, 255×40×40, and 255×80×80 images output by the head are further responsible for detecting large, medium, and small-scale targets, respectively. During the training process, the head uses the bounding box loss function and NMS technology to ensure the precise adjustment of the bounding box, thereby providing accurate and stable recognition results. In addition, the model is further compressed through pruning, distillation and other technical means to ensure that it can run smoothly on smartphones.

[0047] For the six concentrations of AFB1 samples, three smartphones were used to capture eight images of the chip, and each image was expanded to five. Therefore, for the two detection modes, 3×8×6×5=720 images were generated to construct the dataset. After the network construction and dataset expansion, in order to balance the accuracy and training time of the model, the key training parameters batchsize and epochs were adjusted according to the actual situation. Finally, the batchsize was selected as 16, and the epochs were determined to be 150 and 125 for fluorescence and colorimetric data, respectively. The training effect and test quality of the model were evaluated by exploring multiple key indicators. After the model was trained on fluorescence data, the overall accuracy, Precision, Recall and mAP@0.5 were as high as 99.3%, 99.3%, 99.1% and 99.6%, respectively. The YOLO v5 model trained on colorimetric data has an accuracy, precision, recall, and mAP@0.5 of 99.3%, 99.1%, 99.7%, and 99.6%, respectively. The Precision, Recall, and mAP@0.5 of the deep learning model on each category are all higher than 98%, laying a solid foundation for its embedding in smartphones.

[0048] Based on the above-mentioned deep learning model after optimization training, the model transplantation and quantitative detection algorithm are studied, and a smart phone APP is developed. Through Flask, a RESTful API is quickly built to encapsulate the interaction between the deep learning model (.pt format) in Python and the JAVA environment as a Web service. During the deployment process, the routing and processing functions in Flask are freely defined according to the needs based on Python, and the API endpoint is written to process the POST request, which is used to receive the reasoning request sent by the JAVA environment through the HTTP client library OKHTTP to call the API, and use the model for reasoning. The reasoning result is then returned to the JAVA environment through HTTP. Once JAVA receives the response of Flask, it will parse the result at a speed close to real time. Based on the above technical foundation, the present invention realizes the intelligent classification determination of the concentration of the object to be measured. Further, the image processing algorithm is integrated in the mobile phone APP to obtain more accurate quantitative detection results. The clarity and details of the area to be measured are ensured by performing preprocessing such as bilinear interpolation scaling, median filtering denoising, and threshold segmentation background light on the image. Traverse each pixel of the original image and accurately locate the key fluorescent points (or T and C lines) in the T zone to analyze the optical signal. The weighted average algorithm and the nonlinear model obtained by FIJI software analysis were used to complete the accurate conversion of RGB value-grayscale value-OD value. Finally, the concentration of the analyte in the sample was determined according to the OD-concentration standard curve fitting equation.

[0049] In addition, in order to optimize the convenience and personalized experience of the detection system, the present invention has comprehensively expanded the functions of the smartphone APP. After completing the login and Bluetooth device connection, the user will select the detection mode (fluorescence / colorimetry) and data source (camera image / album pre-stored image) according to the needs. Afterwards, the APP immediately and automatically calls the deep learning model, performs a preliminary classification of the concentration of the target substance, and further obtains quantitative results based on the built-in algorithm. In addition, the present invention specially accesses the Wenxin Yiyan intelligent consulting assistant with powerful natural language processing capabilities and knowledge base support in the detection program to provide more comprehensive and convenient user support. The program also optimizes the monitoring and control capabilities of food safety issues through big data analysis technology by integrating the geographic location positioning function of the detector.

[0050] System test plan 1: Standard sample test. Prepare 6 concentrations (0, 0.05, 0.2, 1, 10, 40 ng / ml) of AFB1 standard solution samples for testing. For the concentration classification effect under the two detection modes, the present invention uses a total of 48 standard solution samples of on-site detection data or pre-stored detection images for strict verification.

[0051] like Figure 3As shown in the figure, the established detection system can process images and obtain grading results at a speed close to real time, showing a classification capability with an overall accuracy of 98%, which fully proves the successful transplantation of the YOLO v5 model to smartphones through the Flask framework.

[0052] For the quantitative detection effect in the two modes, each concentration of solution was tested 10 times and verified using its on-site detection data or pre-stored detection images.

[0053] like Figure 4-5 As shown in the figure, the signal of the key fluorescent point (or T line) in the T zone of the dual-modal immunochromatographic detection chip is strongly correlated with the concentration of the target. The R 2 The LOD of colorimetric detection is 0.015 ng / ml, and the LOD of fluorescence detection can be as low as 2.2 pg / ml.

[0054] Test plan 2: Detection of actual samples. Corn and soybean were selected, negative samples were extracted and added with AFB1 standards of different masses, and the established system was used for detection.

[0055] The results showed that the detection system had 100% accuracy in determining the grade of fluorescence and colorimetric responses in real grain samples, and the average recovery of quantitative detection results was in the range of 90.00%-112.00%, with CV less than 6.15%. The above results fully demonstrated the excellent reliability of the detection system in actual field analysis.

[0056] The detection system of the present invention conforms to the urgent need for the cross-integration of POCT and AI technology. It can realize the prediction, precise classification and accurate measurement of AFB1 concentration only through a smartphone, without the need for additional large-scale detection devices and complex image processing software, significantly simplifying the detection process. In terms of graded diagnosis, compared with other studies that apply AI models to smartphones for immunochromatography detection, the present invention further reduces the amount of data required for deep learning model training and optimizes computing resources. In terms of quantitative detection, compared with other AFB1 immunochromatography detection studies based on smartphones, the present invention achieves a significant improvement in sensitivity. In conjunction with the development of a new immunochromatography chip and a hardware platform based on wireless power supply for smartphones, the system is truly integrated and portable, and the detection is convenient and intelligent, and the scalability and maintainability of the entire system are significantly improved.

[0057] The detection system of the present invention can complete all processes within 20 minutes, achieving an accuracy rate of up to 98% for the grading of AFB1 concentration, which fully verifies the reliability and excellence of the deep learning strategy of smartphones in practical applications. The system also has ultra-sensitive quantitative detection capabilities, and the R 2 As high as 0.993, the linear range is as wide as 0.05-40ng / ml. The detection chip has excellent specificity, repeatability and sensitivity. The standard deviation at each concentration is less than 0.03, and the LOD can be as low as 2.2pg / ml, which is much lower than the industry standard. For the detection of actual grain samples, the detection system showed a grade judgment with an accuracy of 100%, and the average recovery rate of quantitative detection results was in the range of 90.00%-112.00%, and the CV was less than 6.15%. The above results fully demonstrate the excellent reliability of the detection system in actual on-site analysis.

[0058] In addition, the detection system of the present invention uses a smartphone APP as a detection platform, which not only greatly reduces production costs, but also significantly improves the deployment efficiency and popularity of the system through convenient installation. And with the continuous advancement of technology and changes in user needs, the APP-based detection system can continuously optimize functions and performance through software updates, ensuring that the detection system can keep up with the times and meet the growing needs of users.

[0059] In summary, the detection system of the present invention realizes full-intelligent, highly stable rapid immunochromatography detection, showing great potential and value in POCT applications. Under the background of artificial intelligence, the development of this system has pushed immunochromatography technology to a new height of intelligence, convenience and digitization, providing strong technical support and ample application possibilities for the community, family and personalized applications of POCT.

[0060] In summary, the present invention proposes an immunochromatographic detection system based on deep learning and smartphones, providing a faster, more convenient, user-friendly and cost-effective detection method. With the help of smartphones with high computing power, network connection function and high-resolution camera, they become an ideal carrier for integrating image processing and data analysis technology to generate a client for POCT mobile diagnosis and monitoring system, and the deep embedding of artificial intelligence algorithms into smartphones can effectively eliminate the interference caused by different light, camera position and even operator stability, further improving the accuracy and efficiency of detection.

[0061] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A dual-modal immunochromatographic detection system based on deep learning of smart phones, characterized in that: The detection system comprises: A smartphone, equipped with a camera and a detection app, is placed on an adaptive hardware platform; The detection APP has built-in deep learning models and quantitative detection algorithms; Adaptive hardware platform with stretchable bracket, optical path module, imaging macro lens and other structures; The stretchable bracket is adapted to smartphones of different sizes, and uses a micro-slide structure to adjust the relative position of the smartphone camera and the imaging lens; The optical path module is used to excite the imaging signal of the immunochromatographic chip and guide it to the imaging lens; The imaging macro lens is used to magnify the area to be tested on the immunochromatographic chip and guide it to the camera of the smartphone; The dual-modal immunochromatographic chip is configured as a distance-type fluorescent chip and a traditional colorimetric chip respectively; the distance-type fluorescent chip has a T zone, the spray pattern is dot-shaped, and the T zone of each chip contains 10 capture points.

2. The dual-modal immunochromatography detection system based on deep learning of smart phones according to claim 1, characterized in that: The mobile phone is configured with a detection APP, and the trained deep learning model YOLO v5 is encapsulated through the Flask framework. The deep learning model YOLO v5 can be used to process the images of the immunochromatography chip imaging to obtain image analysis results related to visual distance or signal strength.

3. The dual-modal immunochromatography detection system based on deep learning of smart phones according to claim 1, characterized in that: The mobile phone remotely communicates with another terminal and transmits the captured picture to the terminal. The terminal is configured with a trained deep learning model YOLO v5. The deep learning model YOLO v5 can be used to process the pictures imaged by the immunochromatography chip to obtain image analysis results related to visual distance or signal strength.

4. The dual-modal immunochromatography detection system based on deep learning of smart phones as claimed in claim 2, characterized in that: The deep learning model YOLO v5 uses a lightweight network and fuses shallow features and deep semantic features of the image obtained by the convolutional neural network of the deep learning model YOLO v5 to achieve detection of objects of different scales.

5. The dual-modal immunochromatography detection system based on deep learning of smart phones as claimed in claim 2, characterized in that: The deep learning model YOLO v5 configured on the mobile phone is pruned, distilled and compressed.

6. The dual-modal immunochromatographic detection system based on deep learning of smart phones according to claim 1, characterized in that: The smartphone detection APP also includes a quantitative detection module for calling a quantitative detection algorithm to complete the acquisition of OD values ​​from RGB values ​​in the imaging image analysis of the immunochromatographic chip, and to obtain concentration information by fitting the relationship between the OD values ​​and the concentration standard curve.

7. The dual-modal immunochromatographic detection system based on deep learning of smart phones according to claim 1, characterized in that: The hardware platform has a closed black housing with dimensions of 9.8×7.3×3.4 cm 3 , the material is polystyrene, and the inner and outer walls are treated with black matte.

8. The dual-modal immunochromatographic detection system based on deep learning of smart phones according to claim 7, characterized in that: The hardware platform also includes a microcircuit module that integrates functions such as wireless power supply, low-power Bluetooth communication, and control of the light source irradiated on the immunochromatographic chip.

9. The dual-modal immunochromatographic detection system based on deep learning of smart phones according to claim 7, characterized in that: The imaging lens is embedded in a closed black housing.

10. The dual-modal immunochromatographic detection system based on deep learning of smart phones according to claim 1, characterized in that: The dual mode is that the detection mode is fluorescence or colorimetry. When the detection mode is colorimetry, the immunochromatographic chip further includes a colorimetric immunochromatographic chip, and each chip contains a T line and a C line.

Images