Virus antigen detection method and device, electronic equipment and storage medium

Abstract

The invention provides a virus antigen detection method and device, electronic equipment and a storage medium, and relates to the technical field of detection, any one virus antigen fluorescence image is acquired as a current fluorescence image, the current fluorescence image is subjected to noise reduction processing on RGB channels by adopting a time-space domain combination method, and the current fluorescence image is obtained; and inputting the denoised fluorescence image into the trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image, and finally obtaining a virus antigen detection result based on the positive cell detection result. The method effectively solves the problems of low detection efficiency and low detection result accuracy of a traditional antigen detection method, and can effectively avoid the influence of human factors, so that a more accurate detection result is obtained, and the sensitivity, accuracy and credibility of virus antigen detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of immunological detection of viral antigens, and in particular to a method, device, electronic equipment and storage medium for detecting viral antigens. Background Art

[0002] Most viruses are transmitted mainly through blood, respiratory droplets and contact. At present, the more mature immunoassay methods mainly include enzyme-linked immunosorbent assay (ELISA), chemiluminescence, immunofluorescence, etc. The main principle of immunofluorescence (IF) is to use the specific binding between antigen and antibody to mark the target protein, and fluorescence can be observed through a fluorescence microscope, and then the target protein can be observed.

[0003] In practical applications, most of the existing interpretation of fluorescent images is based on manual interpretation. However, manual interpretation is highly subjective, labor-intensive, and easily causes visual fatigue, resulting in inaccurate interpretation results. With the application of artificial intelligence technology, machine learning, especially deep learning, plays an increasingly important role in the field of medical image analysis. Therefore, a deep learning-based viral antigen detection solution is urgently needed to improve the accuracy of detection. Summary of the invention

[0004] The technical purpose to be achieved by the embodiments of the present invention is to provide a virus antigen detection method to solve the problem that the efficiency and accuracy of manual interpretation in the prior art are low and accurate antigen detection cannot be achieved. The virus antigen detection results are used to assist medical staff in diagnosis.

[0005] To achieve the above-mentioned purpose, the embodiments of the present invention provide a virus antigen detection method, device, electronic device and storage medium, including: obtaining any virus antigen fluorescence image as the current fluorescence image, using the spatiotemporal domain joint method to perform noise reduction processing on the RGB three channels, inputting the noise-reduced fluorescence image into the trained convolutional neural network model, obtaining the positive cell detection result of the fluorescence image, and finally obtaining the virus antigen detection result based on the positive cell detection result. This can effectively solve the problems of low detection efficiency and low accuracy of detection results in traditional antigen detection methods, effectively avoid the influence of human factors, thereby obtaining more accurate detection results, and improving the sensitivity, accuracy and credibility of virus antigen detection.

[0006] In a first aspect, an embodiment of the present invention provides a method for detecting viral antigens, the method comprising: Obtain any virus antigen fluorescence image as the current fluorescence image; Performing image denoising on the current fluorescence image using a spatiotemporal joint method on the RGB three channels to obtain a denoised fluorescence image; Inputting the de-noised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image; The viral antigen detection result is obtained based on the positive cell detection result.

[0007] Preferably, the image denoising process of the current fluorescent image using a spatiotemporal joint method on the RGB three channels includes: Perform threshold segmentation on the three channels of RGB to obtain noise images of the current fluorescence image on the three channels of RGB; Performing morphological processing and connected domain marking on the current fluorescent image to obtain a connected domain image of the current fluorescent image; Image denoising is performed according to the noise map and the connected domain image.

[0008] Preferably, performing image denoising processing according to the noise map and the connected domain image includes: For any pixel point of the current fluorescence image, if it exists in the noise map on the three RGB channels and simultaneously exists in the connected domain of the connected domain image corresponding to the viral antigen fluorescence image, the pixel point is considered to be a suspected noise pixel point; If the above judgment conditions are not met, it is considered that the pixel is not a noise pixel.

[0009] Preferably, if the pixel points in three consecutive viral antigen fluorescence images are all suspected noise pixels, the pixel points are considered to be true noise pixels, and image noise reduction processing is performed; If not satisfied, no image noise reduction processing is performed.

[0010] Preferably, the image noise reduction process is to add or subtract pixel values ​​of current noise pixels.

[0011] Preferably, the convolutional neural network model includes a first convolutional neural network model and a second convolutional neural network model; wherein the first convolutional neural network model includes 8 convolutional layers, 4 maximum pooling layers and 3 residual modules; the second convolutional neural network model includes 8 convolutional layers, 4 average pooling layers and 3 residual modules; the output of the first convolutional neural network model and the output of the second convolutional neural network model are weightedly fused and input into the fully connected layer.

[0012] Preferably, obtaining the viral antigen detection result based on the positive cell detection result comprises: The number of positive cells in the fluorescent image is determined based on the positive cell detection result. When the number of positive cells exceeds a preset first threshold, the viral antigen detection result of the current fluorescent image is determined to be positive.

[0013] In a second aspect, an embodiment of the present invention further provides a viral antigen detection device, the detection device comprising: An acquisition module, used for acquiring any one of the virus antigen fluorescence images as the current fluorescence image; A denoising module is used to perform image denoising on the current fluorescence image using a spatiotemporal joint method on the RGB three channels to obtain a denoised fluorescence image; A detection module, used for inputting the de-noised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image; A determination module is used to obtain a viral antigen detection result based on the positive cell detection result.

[0014] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps in the viral antigen detection method as described in any one of the embodiments of the present invention are implemented.

[0015] The present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the viral antigen detection method as described in any one of the embodiments of the present invention are implemented.

[0016] The virus antigen detection method of the present invention can be used to obtain any virus antigen fluorescence image as the current fluorescence image, and the current fluorescence image can be subjected to noise reduction processing using a spatiotemporal joint method on the RGB three channels, and the noise-reduced fluorescence image can be input into a trained convolutional neural network model to obtain the positive cell detection result of the fluorescence image, and finally the virus antigen detection result can be obtained based on the positive cell detection result. The present invention effectively solves the problems of low detection efficiency and low accuracy of detection results in traditional antigen detection methods, can effectively avoid the influence of human factors, thereby obtaining more accurate detection results, and improving the sensitivity, accuracy and credibility of virus antigen detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a schematic diagram of a process of viral antigen detection provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a convolutional neural network model provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of a one-dimensional residual convolutional neural network structure provided by an embodiment of the present invention; Figure 4 It is a structural schematic diagram of a viral antigen detection device provided by an embodiment of the present invention; Figure 5 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, Figure 1 This is a flow chart of a virus antigen detection method provided by an embodiment of the present invention. It can be used to determine whether the results of antigen detection such as HBsAg (hepatitis B surface antigen) and respiratory virus are positive. Specifically, it is to detect positive cells in the specimen, and finally determine whether the specimen to be tested is positive based on the detection results of positive cells. The virus antigen detection method includes: 101. Obtain any one viral antigen fluorescence image as the current fluorescence image.

[0021] Continuously acquire N virus antigen fluorescence images, where N ≥ 3 and is a positive integer; and acquire any one of the N virus antigen fluorescence images as a current fluorescence image.

[0022] In an embodiment of the present invention, it can be used for respiratory virus antigen detection. The acquisition of viral antigen fluorescence images includes two basic steps: specimen preparation and photographing fluorescent images, wherein specimen preparation generally includes specimen collection, specimen processing, specimen preparation, specimen reaction, and specimen slide preparation. The step of photographing fluorescent images mainly includes placing the prepared viral antigen specimen slide under a fluorescent microscope, taking pictures continuously to obtain N viral antigen fluorescent images, where N ≥ 3 and is a positive integer, and taking any one of them as the current fluorescent image to be processed. As a preferred value, N = 10 can also be selected according to the actual application scenario. The method of photographing to obtain fluorescent images can be timed continuous shooting, or manual continuous shooting, and the implementation method may not be specifically limited.

[0023] 102. The current fluorescence image is subjected to image denoising by adopting a spatiotemporal domain joint method on the RGB three channels to obtain a denoised fluorescence image.

[0024] Fluorescence images collected by fluorescence microscopy may be affected by uneven light illumination, thermoelectric effect, current noise, etc., resulting in poor quality of fluorescence images, background noise or salt and pepper noise; or ripple noise or speckle noise may appear in the image due to fluorescence interference, which may easily cause artificial misjudgment; and conventional image processing methods cannot effectively remove such noise. In this embodiment, since the fluorescence image is a color RGB image, a joint method of time and space domains can be used on three channels for image denoising, which effectively reduces noise interference and improves the quality of the fluorescence image, thereby helping to effectively improve the detection accuracy. Specifically including: 1021. Perform threshold segmentation on the three channels of RGB respectively on the current fluorescence image to obtain noise maps of the three channels of RGB on the current fluorescence image.

[0025] The existing noise image segmentation mostly adopts the threshold segmentation method, which usually sets a fixed threshold to segment the effective area and the noise area in the image. However, the fixed threshold is easy to mistake some useful information for noise, resulting in the loss of the features of the effective area and affecting the accuracy of subsequent detection. In this embodiment, the adaptive threshold segmentation method is used on the RGB three channels to determine the maximum gray threshold t max And the minimum gray threshold t min, the area between the maximum grayscale threshold and the minimum grayscale threshold is considered to be the effective area, and the other areas are noise areas. For the fluorescent image of the specimen slide, the entire field of view is the sample to be detected, and there is no additional background area, so the minimum grayscale threshold can be taken as 0. In this embodiment, the noise map on the three RGB channels is obtained based on the threshold segmentation. In order to clearly characterize the noise points in the noise map and facilitate the subsequent binary operation, the pixel points with grayscale values ​​greater than the maximum grayscale threshold are considered as noise points in the threshold segmentation, and the binarization is set to 1.

[0026] The maximum grayscale threshold can be determined based on the grayscale histogram. First, the acquired current fluorescence image is recorded as P0, and P0 is separated into grayscale images on three channels, namely R, G, and B, which are recorded as P 0R , P 0G , P 0B For the grayscale image P 0R , count the number of pixels corresponding to each gray value (i=0-255) to construct a gray histogram, and calculate the gray mean G ave ; Determine whether each gray value satisfies: G(i)-G(i-1)>G ave / 5; if it satisfies, G(i) is taken as the maximum grayscale threshold, that is, t max =G(i). Based on the maximum grayscale threshold G(i), the grayscale image P 0R Binarization segmentation is performed, and the binarization values ​​of pixels whose grayscale values ​​are greater than or equal to the maximum grayscale threshold are set to 1, and the binarization values ​​of pixels whose grayscale values ​​are less than the maximum grayscale threshold are set to 0, thereby obtaining the grayscale image P of the fluorescence image P0 on the R channel. 0R The noise map P 0RN Repeat the above steps in sequence, and the corresponding noise maps of the current fluorescence image P0 on the G and B channels are P 0GN , P 0BN .

[0027] Grayscale histograms usually present a single peak distribution. In the area near the peak, grayscale changes have little effect on the number of pixels. In the area far from the peak, the number of pixels is small and the overall change is small. In addition, different segmentation thresholds are required for images of different brightness. When the brightness is high, the maximum grayscale threshold is larger, and the difference between adjacent grayscale values ​​is also larger, so G ave / 5 is used as the benchmark for the adaptive threshold to cope with fluorescent images of different brightness.

[0028] As another preferred implementation, the binary segmentation threshold can be determined based on the mean and standard deviation of the distribution function: T=µ+3σ, where µ and σ are the mean and standard deviation of the pixel distribution function, respectively. If the gray value of a pixel is less than T, it is considered to be noise, and the binary value of the pixel is set to 1. On the contrary, if it is greater than or equal to T, it is considered to belong to the valid area, and the binary value of the pixel is set to 0. For the gray image P 0R , count the number of pixels corresponding to each gray value (i=0-255) to construct a gray histogram, which represents the distribution of pixels with different gray values, denoted as P 0RD , represents the pixel distribution of different gray values ​​of the gray image P0 on the R channel; similarly, P 0GD , P 0BD . Calculate the difference in the grayscale pixel distribution between the current fluorescent image and the adjacent fluorescent image, and calculate it for each channel separately, so as to obtain: U 01R =P 0RD -P 1RD , U 01G =P 0GD -P 1GD , U 01B =P 0BD -P 1BD Among them, U 01R It represents the difference between the distribution of pixels of different gray values ​​in the two adjacent fluorescence images P0 and P1 on the R channel. The mean and standard deviation of the difference distribution function on each channel are calculated, and µ+3σ is used as the binarization threshold of the channel for segmentation. Specifically, the segmentation thresholds on the three RGB channels are obtained: T 01R 、T 01G 、T 01B ; According to the segmentation threshold T 01R 、T 01G 、T 01B Binarization segmentation is performed on the three channels of RGB respectively, and the pixels greater than or equal to the segmentation threshold are set to 1, and the pixels less than the segmentation threshold are set to 0, so that the noise map of the current fluorescence image P0 on the three channels of R, G, and B can be obtained: 0RN , P 0GN , P 0BN In this embodiment, the adaptive threshold of image segmentation is determined by the mean and standard deviation of the difference between two adjacent fluorescence images to determine the noise caused by uncertain factors such as motion. Pixels greater than or equal to the segmentation threshold are considered to be noise pixels, so that noise maps on three channels are obtained based on the determined adaptive threshold.

[0029] 1023. Perform morphological processing and connected domain labeling on the current fluorescent image to obtain a connected domain image of the current fluorescent image.

[0030] In this embodiment, the basic steps of obtaining the connected domain of the fluorescence image include: converting the fluorescence color image into a grayscale image; performing threshold segmentation on the grayscale image to obtain a binary image; and marking the connected domain of the binary image to obtain a connected domain image of positive cells in the fluorescence image.

[0031] For the current fluorescent image P0, the grayscale values ​​on the RGB three channels are weighted and summed to obtain the grayscale pixel value, and the calculation formula is: Gray=0.3R+0.6G+0.1B; the grayscale image is segmented based on the global threshold to obtain a binary image, and the pixels with grayscale values ​​greater than or equal to the global threshold are set to 0, and the pixels with grayscale values ​​less than the global threshold are set to 1; the global threshold segmentation method only needs to segment the obvious noise points as the background, and then the noise pixels are processed by pixel value processing. The connected domain of the binary image is marked to obtain the connected domain image corresponding to the positive cells in the fluorescent image. There are only foreground pixels in the connected domain area, and in the same connected domain, any two pixels are directly connected. Based on the steps in this embodiment, the connected domain image P of the current fluorescent image P0 can be obtained. 0CC .

[0032] 1024. Perform image denoising processing according to the noise map and the connected domain image.

[0033] For any pixel of the current fluorescence image, if it exists in the noise map on the three RGB channels and also exists in the connected domain of the connected domain image corresponding to the virus antigen fluorescence image, the pixel is considered to be a suspected noise pixel; if the above judgment conditions are not met, the pixel is considered not to be a noise pixel. Furthermore, if the pixel in three consecutive virus antigen fluorescence images are suspected noise pixels, the pixel is considered to be a true noise pixel, and image denoising is performed; if not, image denoising is not performed.

[0034] In this embodiment, the binary calculation is performed based on the noise map of the three channels of the current fluorescent image and the connected domain image of the current fluorescent image, that is, the binary value in the connected domain is 1, the binary value in the non-connected domain is 0, the binary value of the noise pixel in the noise map is 1, and the binary value of other pixels is 0, so that the two binary images are directly ANDed, and if the result is 1, it can be considered as a suspected noise pixel. It can be understood that for any fluorescent image, each pixel therein has a corresponding pixel in the noise map of the RGB three channels and the connected domain image, and can be considered to belong to the same pixel at the same position. The judgment of this step is mainly based on whether the noise pixel in the noise map exists in the corresponding foreground image connected domain, which is used as the preliminary judgment basis for determining whether it is a suspected noise pixel, and then based on whether the pixel at the same position in the three consecutive fluorescent images is also a suspected noise pixel. If it is satisfied, it is finally considered that the pixel belongs to a real noise pixel, and image denoising is performed.

[0035] It can be understood that for the fluorescence image P0, its noise maps on the RGB three channels are P 0RN , P 0GN , P 0BN , the corresponding connected domain image is P 0CC . 0RN , P 0GN , P 0BN , P 0CC The pixels at the same position are binary ANDed to obtain the binary image P' 0CC , the pixel points whose operation result is binary value 1 in the image are suspected noise pixels; in order to improve the accuracy of noise detection, it is necessary to judge three consecutive fluorescence images to eliminate the misjudgment of noise points that may be affected by external uncertain factors, so that the pixels that meet the judgment conditions here are called suspected noise pixels.

[0036] Repeat the above calculation steps for the fluorescent images P1 and P2 that are consecutively adjacent to the current fluorescent image P0. 1RN , P 1GN , P 1BN , P 1CC The pixels at the same position are binary ANDed to obtain the binary image P' 1CC , P 2RN , P 2GN , P 2BN , P 2CC The pixels at the same position are binary ANDed to obtain the binary image P' 2CC .

[0037] Finally, the binary image P' 0CC , P'1CC , P' 2CC Perform binary AND operation again to get the final noise map P Noise . Noise map P Noise The pixels with binary value 1 are the real noise pixels.

[0038] At this point, the process of determining noise pixels in the spatiotemporal domain is completed.

[0039] After the judgment process is completed, the noise pixels need to be processed for image noise reduction. Noise The pixel point p(x,y) with a pixel value of 1 in the fluorescence images P0, P1, and P2 corresponds to p0(x,y), p1(x,y), and p2(x,y) respectively; in this embodiment, for each pixel point, the pixel value is added or subtracted on the three channels of RGB to achieve noise reduction processing, and it is necessary to satisfy that the sum of the grayscale values ​​of the pixel point on the three channels of RGB before and after the noise reduction processing remains unchanged. Preferably, for the pixel point p0(x,y), the grayscale values ​​corresponding to the three channels of RGB are: G 0R (x,y),G 0G (x,y),G 0B (x,y), the noise reduction process needs to meet the following conditions: (G 0R (x,y)+a)+(G 0G (x,y)+b)+(G 0B (x,y)+c)=G 0R (x,y)+G 0G (x,y)+G 0B (x, y), that is, a+b+c=0, where a, b, and c are the noise addition and subtraction values ​​of the pixel point p0(x, y) on the RGB channels respectively, and the absolute values ​​of a, b, and c are all ≤3. For example, a=-1, b=-2, and c=3, that is, for G 0R (x, y) minus 1 gray value, for G 0G (x, y) minus 2 gray values, for G 0B (x, y) plus 3 grayscale values ​​ensures that the grayscale of the current fluorescence image in the spatial dimension will not change, thereby effectively avoiding the distortion of the fluorescence image after denoising. For the fluorescence images P1 and P2, the noise addition and subtraction values ​​of the corresponding pixel points p1(x, y) and p2(x, y) on the RGB three channels do not need to be exactly the same as the current fluorescence image P0. It is only necessary that each pixel point satisfies the condition a+b+c=0, that is, the sum of the grayscale values ​​of the pixel point on the RGB three channels remains unchanged before and after denoising. Furthermore, the denoised RGB three-channel images are synthesized to obtain the denoised fluorescence image, that is, the denoised fluorescence image P 0denoise , P1denoise , P 2denoise .

[0040] 103. Input the denoised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image.

[0041] Preferably, the convolutional neural network model includes a first convolutional neural network model and a second convolutional neural network model; wherein the first convolutional neural network model includes 8 convolutional layers, 4 maximum pooling layers and 3 residual modules; the second convolutional neural network model includes 8 convolutional layers, 4 average pooling layers and 3 residual modules; the output of the first convolutional neural network model and the output of the second convolutional neural network model are weightedly fused and input into the fully connected layer.

[0042] In this embodiment, based on the conventional convolutional neural network structure, a convolutional neural network model for virus antigen detection is constructed, and the model is composed of two-way one-dimensional residual convolutional neural networks; common convolutional neural networks are used to recognize two-dimensional or three-dimensional image data. The one-dimensional residual convolutional neural network in the present invention is improved based on the VGG network, and the convolution layer and the pooling layer are replaced with one-dimensional convolution and one-dimensional pooling, and the residual network is introduced to obtain a one-dimensional residual convolutional neural network model. The basic framework structure of the convolutional neural network model can be found in Figure 2 .

[0043] In this embodiment, the convolutional neural network model includes a first convolutional neural network model and a second convolutional neural network model. The first convolutional neural network model and the second convolutional neural network model are both one-dimensional residual convolutional neural networks. The one-dimensional residual convolutional neural network structure is shown in Figure 3 ; Among them, the first convolutional neural network contains 8 convolutional layers, 4 maximum pooling layers and 3 residual modules; the second convolutional neural network contains 8 convolutional layers, 4 average pooling layers and 3 residual modules; the residual modules are connected in a jump manner, and are additively connected to the input after two convolutional layers. This connection method can avoid the gradient descent problem in network training and improve network performance. At the same time, a batch normalization layer is set after each convolutional layer to improve the training efficiency of the model. In order to improve the performance of the model, the weights of all network layers of the first convolutional neural network and the second convolutional neural network come from the pre-trained network, where the weights of the first convolutional neural network are set to be fixed during the training process, and the weights of the second convolutional neural network are fine-tuned during the training period, where the pre-trained network can be a transfer learning network. It should be noted that the weights of the fully connected layer and the Softmax layer need to be fine-tuned during the model training process.

[0044] The fluorescence image P after noise reduction is 0denoise , P 1denoise , P2denoise After constructing feature matrices for any two images in the image, they are used as inputs of the first convolutional neural network and the second convolutional neural network respectively. The maximum pooling features output by the first convolutional neural network and the average pooling features output by the second convolutional neural network are weightedly fused to obtain fused features. The fused features are input into the fully connected layer, and Softmax is introduced to output the probability that the detection target is a positive cell. Preferably, the denoised fluorescence image is an RGB color image, and the size of the input image can be 640 × 640 × 3.

[0045] As a preferred embodiment, after the convolutional neural network model is built, a model training process is also included. The fluorescence image data set contains 500 images, of which 80% of the images are randomly selected as training images, a total of 400 images, and the remaining 20% ​​are used as test images, a total of 100 images; the training images are manually annotated using image marking software, and only viral antigen-positive cells after a fluorescent reaction are annotated. The training images are input into the convolutional neural network model for training, the number of iterations or convergence conditions are set, and the ratio of cells recognized by the model is used as an evaluation index of the model detection effect. Training and testing are performed on the experimental data set, and finally a convolutional neural network model with a recognition accuracy of 90% is obtained. The training process belongs to the conventional technical knowledge in this field and is not described here.

[0046] 104. The viral antigen test results are obtained based on the positive cell test results.

[0047] The de-noised fluorescence image is input into the convolutional neural network model to obtain the positive cell detection result. In this embodiment, by counting the positive cells in the detection result, if the number of positive cells is greater than a pre-set first threshold, the sample is considered positive. Since positive cells will exist once infected with the virus, preferably, the first threshold can be 2.

[0048] As another preferred embodiment, the number of cells to be tested and the number of positive cells detected in the fluorescent image can also be calculated respectively, and if the number ratio of positive cells exceeds a set second threshold, the sample is considered positive. The second threshold can be 10%, 30%, etc., and the specific value can be set according to the actual application scenario.

[0049] As other preferred implementations, the binary AND operation can be replaced by other binary operations, such as a binary OR operation. In the present invention, the noise pixel needs to be output as a binary value of 1. It can be understood by those skilled in the art that the binary value of the corresponding threshold segmentation output needs to be adjusted to ensure that the binary value of the noise pixel outputted in the final output is 1. This can be obtained by those skilled in the art through conventional simple adjustments, and such a variation is also included in the protection scope of the present invention.

[0050] like Figure 4 As shown, Figure 4 : is a schematic diagram of the structure of a virus antigen detection device provided by an embodiment of the present invention. The virus antigen detection device comprises: An acquisition module 301 is used to continuously acquire N viral antigen fluorescence images, where N ≥ 3 and is a positive integer; A noise reduction module 302 is used to perform image noise reduction processing on the fluorescence image using a spatiotemporal domain joint method on the RGB three channels to obtain a noise-reduced fluorescence image; A detection module 303 is used to input the denoised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image; The determination module 304 is used to obtain the viral antigen detection result based on the positive cell detection result.

[0051] The viral antigen detection device provided in the embodiment of the present invention can implement each process implemented by the viral antigen detection method in the above method embodiment, and can achieve the same beneficial effects. To avoid repetition, it will not be described here.

[0052] See also Figure 5 , Figure 5 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 5 As shown, it includes: a memory 402, a processor 401, and a computer program of a virus antigen detection method stored in the memory 402 and executable on the processor 401, wherein: The processor 401 is used to call the computer program stored in the memory 402 to perform the following steps: Continuously acquire N viral antigen fluorescence images, where N ≥ 3 and is a positive integer; The fluorescence image is subjected to image denoising by using a spatiotemporal joint method on the RGB three channels to obtain a denoised fluorescence image. The denoised fluorescence image is input into the trained convolutional neural network model to obtain the positive cell detection result of the fluorescence image; The viral antigen test results are obtained based on the positive cell test results.

[0053] The electronic device provided in the embodiment of the present invention can implement each process implemented by the virus antigen detection method in the above method embodiment, and can achieve the same beneficial effects. To avoid repetition, it will not be described here.

[0054] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the viral antigen detection method provided in the embodiment of the present invention are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0055] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the computer-readable storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0056] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for detecting viral antigens, characterized in that: The detection method comprises: Obtain any virus antigen fluorescence image as the current fluorescence image; Performing image denoising on the current fluorescence image using a spatiotemporal joint method on the RGB three channels to obtain a denoised fluorescence image; Inputting the de-noised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the fluorescence image; The viral antigen detection result is obtained based on the positive cell detection result.

2. The viral antigen detection method according to claim 1, characterized in that: The method of performing image noise reduction processing on the current fluorescent image using a spatiotemporal domain joint method on the RGB three channels includes: Perform threshold segmentation on the three channels of RGB to obtain noise images of the current fluorescence image on the three channels of RGB; Performing morphological processing and connected domain marking on the current fluorescent image to obtain a connected domain image of the current fluorescent image; Image denoising is performed according to the noise map and the connected domain image.

3. The viral antigen detection method according to claim 2, characterized in that: The performing image denoising processing according to the noise map and the connected domain image comprises: For any pixel point of the current fluorescence image, if it exists in the noise map on the three RGB channels and simultaneously exists in the connected domain of the connected domain image corresponding to the viral antigen fluorescence image, the pixel point is considered to be a suspected noise pixel point; If the above judgment conditions are not met, it is considered that the pixel is not a noise pixel.

4. The viral antigen detection method according to claim 3, characterized in that: The performing image denoising processing according to the noise map and the connected domain image further includes: If the pixel points in three consecutive viral antigen fluorescence images are all suspected noise pixels, the pixel points are considered to be real noise pixels, and image noise reduction processing is performed; If not satisfied, no image noise reduction processing is performed.

5. The viral antigen detection method according to claim 4, characterized in that: The image noise reduction process is to add or subtract pixel values ​​of the real noise pixels.

6. The method for detecting viral antigens according to claim 1, wherein: The convolutional neural network model includes a first convolutional neural network model and a second convolutional neural network model; wherein, the first convolutional neural network model and the second convolutional neural network model are both one-dimensional residual convolutional neural networks; the first convolutional neural network model includes 8 convolutional layers, 4 maximum pooling layers and 3 residual modules; the second convolutional neural network model includes 8 convolutional layers, 4 average pooling layers and 3 residual modules; the output of the first convolutional neural network model and the output of the second convolutional neural network model are weightedly fused and input into the fully connected layer.

7. The viral antigen detection method according to claim 1, characterized in that: The method of obtaining a viral antigen detection result based on the positive cell detection result comprises: The number of positive cells in the fluorescent image is determined based on the positive cell detection result. When the number of positive cells exceeds a preset first threshold, the viral antigen detection result of the current fluorescent image is determined to be positive.

8. A viral antigen detection device, characterized in that: The detection device comprises: An acquisition module, used for acquiring any one of the virus antigen fluorescence images as the current fluorescence image; A denoising module is used to perform image denoising on the current fluorescence image using a spatiotemporal joint method on the RGB three channels to obtain a denoised fluorescence image; A detection module, used to input the denoised fluorescence image into a trained convolutional neural network model to obtain a positive cell detection result of the current fluorescence image; A determination module is used to obtain a viral antigen detection result based on the positive cell detection result.

9. An electronic device, characterized in that: comprising a processor and a memory, wherein: The memory has computer-readable instructions stored thereon; The computer-readable instructions are executed by one or more processors, so that the electronic device implements the viral antigen detection method as described in any one of claims 1-7.

10. A storage medium having computer-readable instructions stored thereon, characterized in that: The computer-readable instructions are executed by one or more processors to implement the viral antigen detection method described in any one of claims 1-7.

Images