An immune test weak positive sample intelligent interpretation method and system

Abstract

The application provides an immune test weak positive sample intelligent interpretation method and system, a weak positive sample image containing a detection line and a quality control line region in an immune chromatography test strip is divided into multiple local region blocks; structural similarities between each group of adjacent local region blocks are determined, candidate line regions and background regions of the detection line are screened from the weak positive sample image based on all the structural similarities; a probability heat map of the detection line is generated according to the gray gradient distribution of the candidate line region, the contribution characteristics of the background region to the weak positive sample image are determined, the background correction characteristic region of the detection line in the weak positive sample image is determined through the contribution characteristics; the contrast index of the signal-to-noise ratio corresponding to the detection line in the immune chromatography test strip is determined by the signal-to-noise ratio of the detection line in the background correction characteristic region, and then a grading report of the weak positive sample is generated. By adopting the scheme, the influence of the continuity of the detection line on the weak positive sample test can be reduced.

Description

Technical Field

[0001] This application relates to the field of sample interpretation technology, and more specifically, to an intelligent interpretation method and system for weakly positive samples in immunoassay. Background Technology

[0002] Intelligent sample interpretation refers to the process of automatically analyzing and hierarchically determining the images or data corresponding to various test samples (such as immunochromatographic test strips, medical images, biological samples, etc.) by using technologies such as image sensors, computer vision, and machine learning.

[0003] Immunochromatographic test strips are widely used in disease diagnosis, health screening, and many other fields due to their ease of operation and low cost. Their working principle is based on the specific binding of antigens and antibodies, reflecting the presence or absence of the target analyte in the sample through the color development of the test line and control line. In practical applications, to obtain test results, image sensors are often used to acquire images of the immunochromatographic test strip after the reaction, which are then analyzed and processed. With technological advancements, the requirements for the accuracy and precision of test results are constantly increasing, especially for the interpretation of weakly positive samples. Traditional simple visual interpretation or basic image analysis methods are no longer sufficient. Currently, some studies have attempted to use image processing techniques and machine learning algorithms to improve interpretation accuracy. However, in areas such as precise screening of the test line area, signal continuity analysis, and comprehensive multi-dimensional information for graded interpretation, the analysis of the continuity of the test line is not perfect, and the impact of minor breaks on the results is easily overlooked. Therefore, how to reduce the impact of test line continuity on the testing of weakly positive samples has become a problem facing the industry. Summary of the Invention

[0004] This application provides an intelligent interpretation method and system for weak positive samples in immunoassay, which can reduce the impact of the continuity of the test line on the testing of weak positive samples.

[0005] Firstly, this application provides an intelligent interpretation method for weakly positive samples in immunochromatographic testing, applied to immunochromatographic test strips. The immunochromatographic test strip includes a test line and a control line. Image acquisition is performed after the immunochromatographic test strip has reacted with the weakly positive sample. The method includes the following steps:

[0006] The image sensor is used to acquire images of weakly positive samples in the regions containing the detection line and control line of the immunochromatographic test strip;

[0007] The weakly positive sample image is divided into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip;

[0008] Determine the structural similarity between adjacent local regions in each group, and based on all structural similarities, select candidate line regions and background regions for detection lines from the weak positive sample images;

[0009] A probability heatmap of the detection line is generated based on the gray-level gradient distribution of the candidate line region. The contribution features of the background region to the weak positive sample image are determined. The continuity of the detection line in the probability heatmap is constrained by the contribution features to obtain the background correction feature region of the detection line in the weak positive sample image.

[0010] A contrast index for the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image is constructed. The contrast index of the signal-to-noise ratio is compared with the preset weak positive interpretation interval of the weak positive sample to generate a grading report for the weak positive sample.

[0011] In some embodiments, dividing the weakly positive sample image into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip specifically includes:

[0012] Determine the chromatographic flow direction of the immunochromatographic test strip;

[0013] Determine the smoothing window for segmenting the weak positive sample image;

[0014] Based on the smooth window, the weak positive sample image is smoothly divided according to the chromatographic flow direction to obtain multiple local region blocks.

[0015] In some embodiments, determining the structural similarity between adjacent local region blocks specifically includes:

[0016] Select a set of adjacent local region blocks as the selected adjacent local region blocks, and determine the structural features of each local region block in the selected adjacent local region blocks;

[0017] The structural similarity between the selected adjacent local regions is determined based on the structural features of each local region block in the selected adjacent local region blocks.

[0018] Continue to determine the structural similarity between the remaining adjacent local region blocks.

[0019] In some embodiments, selecting candidate line regions and background regions for detection lines from the weakly positive sample images based on all structural similarities specifically includes:

[0020] The weak positive sample image is enhanced to obtain an enhanced image of the weak positive sample;

[0021] Determine the similarity mapping of the detection lines based on all structural similarities;

[0022] Candidate line regions and background regions for detection lines are determined based on the similarity map and the enhanced image.

[0023] In some embodiments, generating a probability heatmap of the detection line based on the grayscale gradient distribution of the candidate line region specifically includes:

[0024] The horizontal and vertical gray-level gradient values ​​of each pixel within the candidate line region are determined, thereby obtaining the gray-level gradient distribution of the candidate line region;

[0025] Multiple probability values ​​for the detection line are determined based on the gray-scale gradient distribution;

[0026] The probability heatmap of the detection line is determined by all probability values.

[0027] In some embodiments, determining the contribution features of the background region to the weakly positive sample image specifically includes:

[0028] Determine the brightness and fluctuation characteristics of the background region;

[0029] Extract the texture detail features within the background area;

[0030] The contribution features of the background region to the weak positive sample image are determined based on the brightness features, the fluctuation features, and the detail features.

[0031] In some embodiments, the background correction feature region of the detection line in the probability heatmap is obtained by applying background constraints to the continuity of the detection line in the weak positive sample image using the contribution features, specifically including:

[0032] The suspected breakage region and candidate continuity region of the detection line are extracted from the probability heatmap;

[0033] The suspected fracture region of the detection line is smoothed using the contribution characteristics to obtain the fracture smoothing state of the detection line.

[0034] The background correction feature region of the detection line in the weak positive sample image is determined based on the fracture smoothing state and the candidate continuous region.

[0035] Secondly, this application provides an intelligent interpretation system for weakly positive samples in immunoassay testing, comprising:

[0036] The acquisition module is used to acquire images of weakly positive samples containing the detection line and control line areas of the immunochromatographic test strip based on an image sensor;

[0037] The processing module is used to divide the weakly positive sample image into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip;

[0038] The processing module is also used to determine the structural similarity between adjacent local region blocks in each group, and to filter out candidate line regions and background regions for detection lines from the weak positive sample images based on all structural similarities.

[0039] The processing module is further configured to generate a probability heatmap of the detection line based on the gray-level gradient distribution of the candidate line region, determine the contribution features of the background region to the weak positive sample image, and apply background constraints to the continuity of the detection line in the probability heatmap through the contribution features to obtain the background correction feature region of the detection line in the weak positive sample image.

[0040] The execution module is used to construct a contrast index of the signal-to-noise ratio of the detection line in the immunochromatographic test strip based on the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image. Based on the contrast index of the signal-to-noise ratio, it is compared with the preset weak positive interpretation interval of the weak positive sample, and then a grading report of the weak positive sample is generated.

[0041] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described intelligent interpretation method for weak positive samples in immune testing.

[0042] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent interpretation method for weakly positive samples in immune testing.

[0043] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0044] The intelligent interpretation method and system for weak positive samples in immunochromatographic testing provided in this application first acquires images of weak positive samples containing detection lines and control lines on the immunochromatographic test strip using an image sensor; the weak positive sample images are divided into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip; the structural similarity between adjacent local region blocks is determined, and candidate line regions and background regions for detection lines are screened from the weak positive sample images based on all structural similarities; a probability heatmap of the detection lines is generated based on the gray-level gradient distribution of the candidate line regions, and the contribution characteristics of the background regions to the weak positive sample images are determined. The continuity of the detection lines in the probability heatmap is constrained by the contribution characteristics to obtain the background correction feature regions of the detection lines in the weak positive sample images; a contrast index of the signal-to-noise ratio of the detection lines in the background correction feature regions and the signal-to-noise ratio of the control lines in the weak positive sample images is constructed using the signal-to-noise ratio of the detection lines in the background correction feature regions and the signal-to-noise ratio of the control lines in the weak positive sample images; the contrast index of the signal-to-noise ratio is compared with a preset weak positive interpretation interval for the weak positive sample, thereby generating a grading report for the weak positive sample.

[0045] Therefore, this application, in the intelligent interpretation of weak positive samples, divides the image into local region blocks along the chromatographic flow direction, simulating the physical diffusion process of the test strip, making the analysis more closely resemble the characteristics of actual samples. This refined partitioning helps isolate the influence of continuity, for example, decomposing discontinuous test lines into independent units, facilitating subsequent identification of local anomalies (such as background noise or artifacts), thereby reducing the interference of continuity on the overall interpretation; by calculating the structural similarity of adjacent region blocks, the difference in local continuity is quantitatively evaluated, effectively distinguishing candidate line regions (which may be affected by continuity) from background regions. This directly addresses the technical problem; the screening process can suppress continuity artifacts (such as stripes or gradients), avoiding the misjudgment of discontinuous features of test lines in weak positive samples as background noise, and improving the accuracy of region segmentation; a probability heatmap is generated using gray-level gradient distribution to highlight the potential location of the test line; simultaneously, the extraction and reduction of background contribution features... Bundle operations (such as denoising or model correction) apply background suppression to the probability heatmap, directly correcting for continuity effects (such as pseudo-continuity caused by diffusion or gradation). This enhances the robustness of detection line identification, ensuring accurate reconstruction of low-contrast or discontinuous detection lines in weak positive samples and reducing misjudgments caused by background interference. A quantitative indicator is provided to evaluate the significance of the detection lines by constructing a signal-to-noise ratio contrast index between the detection lines and control lines. A graded report is generated after comparison with a preset interpretation interval. This process mitigates the bias of continuity in signal-to-noise ratio calculation (such as the contribution of background noise), ensuring objective and reliable interpretation. The final report generation reduces subjective dependence and solves the sensitivity fluctuation problem caused by continuity effects in weak positive samples. Using this approach, the impact of detection line continuity on the testing of weak positive samples can be reduced. Attached Figure Description

[0046] Figure 1 This is an exemplary flowchart of an intelligent interpretation method for weakly positive immunoassay samples according to some embodiments of this application;

[0047] Figure 2 This is a diagram illustrating the reaction process between an immunochromatographic test strip and a weakly positive sample, as shown in some embodiments of this application.

[0048] Figure 3 This is an exemplary flowchart illustrating the determination of structural similarity according to some embodiments of this application;

[0049] Figure 4 This is a schematic diagram of the structure of an intelligent interpretation system for weakly positive immunoassay samples, as shown in some embodiments of this application.

[0050] Figure 5 This is a schematic diagram of the structure of a computer device for implementing an intelligent interpretation method for weakly positive samples in immune testing, according to some embodiments of this application. Detailed Implementation

[0051] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] refer to Figure 1 The figure is an exemplary flowchart of an intelligent interpretation method for weakly positive immunoassay samples according to some embodiments of this application. The intelligent interpretation method for weakly positive immunoassay samples mainly includes the following steps:

[0053] In step 101, an image of a weakly positive sample containing the detection line and control line area of ​​the immunochromatographic test strip is acquired using an image sensor.

[0054] In specific implementation, a high-resolution image sensor (≥5 million pixels) is selected, along with a ring-shaped white light source, to ensure that the uniformity deviation of the illumination intensity is ≤5%. The completed immunochromatographic test strip is fixed on the stage, and the image sensor is vertically aligned with the area in the middle of the immunochromatographic test strip that covers the detection line and the control line to collect the image. The collected image is used as the weak positive sample image of the area containing the detection line and the control line on the immunochromatographic test strip. In other embodiments, other methods can also be used to determine the image, which is not limited here.

[0055] It should be noted that the weak positive sample image in this application refers to the image of the weak positive sample after the reaction of the immunochromatographic test strip. It can be used to analyze the weak positive status of the weak positive sample. During the acquisition process, the acquisition frame rate is set to 10fps and the exposure time is 50-200ms. Simultaneously, the image is preprocessed by median filtering to remove noise and stretching the grayscale range to [50,200] to obtain a clear and interference-free weak positive sample image.

[0056] In some embodiments, reference Figure 2 As shown, this figure illustrates the reaction process between the immunochromatographic test strip and a weakly positive sample in some embodiments of this application. Figure 2As shown, the sample addition and initial infiltration (0-1 minute) are as follows: The weakly positive sample (e.g., serum, urine) is added to the sample pad. The physical process involves the sample rapidly infiltrating the pad through capillary action. The buffer solution in the pad dilutes the sample, and the pH is adjusted to a suitable reaction range (e.g., pH 7.2-7.4). Chemical pretreatment occurs when the proteins in the sample are dissolved by the buffer solution, leaving the target antigen (weakly positive) in a free state. The sample then undergoes chromatography and label binding (1-5 minutes). Chromatographic kinetics are formed when the absorbent pad creates a siphon effect, propelling the sample along the test strip towards the absorbent pad. The sample flows through the binding pad, triggering label release. Labeled antibody-antigen binding occurs when the labeled antibody (e.g., colloidal gold-labeled antibody) in the binding pad specifically binds to the target antigen in the sample, forming a "labeled antibody-antigen" complex. In weakly positive samples, the antigen concentration is low, resulting in a smaller amount of complex formation. The chromatographic migration of the immune complex (5-10 minutes) involves the "labeled antibody-antigen" complex continuing to migrate with the sample towards the NC. During membrane migration, unbound labeled antibodies (free state) also migrate simultaneously. Key characteristics: The complex has a larger molecular weight, and its migration speed is slightly slower than that of the free labeled antibody. During chromatography, the diffusion of the complex on the NC membrane is limited by the pore size, maintaining a linear migration path. The immunoreaction between the detection line and the control line (10-15 minutes): Detection line (T line) reaction: When the "labeled antibody-antigen" complex reaches the T line, it is captured by the coated secondary antibody, forming a "secondary antibody-antigen-labeled antibody" sandwich structure. Due to the low antigen content in weakly positive samples, less label accumulates at the T line, resulting in weak color intensity. Control line (C line) reaction: When the free labeled antibody migrates to the C line, it is captured by the coated secondary antibody, forming a "secondary antibody-labeled antibody" complex. The C line color development serves as a control for reaction effectiveness and should be performed regardless of whether the sample is positive or not. Color development principle: Colloidal gold labels appear as red bands when aggregated; fluorescent labels require excitation light detection. The T line color development of weakly positive samples... The line color is lighter than that of a strong positive, possibly close to the visual recognition threshold; the reaction terminates and the signal stabilizes (15-20 minutes), and the chromatography ends: the sample reaches the absorbent pad, the liquid flow stops, and the signal is fixed: the immune complex is fixed on the NC membrane and no longer migrates, and the color intensity of the marker remains stable for a period of time (usually within 30 minutes). Note: if the effective interpretation time is exceeded (e.g., after 30 minutes), the background may deepen due to non-specific adsorption, affecting the interpretation of the results.

[0057] In step 102, the weakly positive sample image is divided into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip.

[0058] In some embodiments, dividing the weakly positive sample image into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip can be achieved by the following steps:

[0059] Determine the chromatographic flow direction of the immunochromatographic test strip;

[0060] Determine the smoothing window for segmenting the weak positive sample image;

[0061] Based on the smooth window, the weak positive sample image is smoothly divided according to the chromatographic flow direction to obtain multiple local region blocks.

[0062] In specific implementation, firstly, the gray-level accumulation curves of pixels in the horizontal / vertical directions are calculated using the gray-level projection method for the weak positive sample image. The direction corresponding to the peak value in the pixel gray-level accumulation curve on the immunochromatographic test strip is taken as the chromatographic flow direction of the immunochromatographic test strip, where the chromatographic flow direction represents the direction in which the weak positive sample flows on the immunochromatographic test strip. Secondly, the effective region size on the immunochromatographic test strip is obtained through contour detection methods (such as edge detection and threshold segmentation). A smoothing window is calculated using an adaptive window algorithm (such as S=min(width, height) / 10). Finally, the smoothing window is combined with the image gray-level variance (set to 30% when σ²>50). The window overlap rate is set (20% otherwise). The window slides in steps along the flow direction of the chromatography with the overlap rate. At the same time, the overlapping areas of adjacent windows are fused by weighted average (α=0.5). The window position is fine-tuned to avoid strong edges by Sobel gradient detection method. Non-overlapping windows are processed in parallel by open multiprocessing and multithreading. Finally, the corresponding areas of each window obtained by sliding are all taken as local region blocks. Other methods can be used for division in other embodiments, which are not limited here.

[0063] It should be noted that the local region block in this application represents the region block corresponding to a local window in a weak positive sample image, which can be used to analyze the features in the weak positive sample image, thereby improving the analysis efficiency and accuracy.

[0064] In step 103, the structural similarity between adjacent local region blocks in each group is determined, and candidate line regions and background regions for detection lines are selected from the weak positive sample images based on all structural similarities.

[0065] In some embodiments, reference Figure 3 As shown, this figure is an exemplary flowchart of determining structural similarity in some embodiments of this application. In this embodiment, determining the structural similarity between adjacent local region blocks can be achieved by the following steps:

[0066] First, in step 1031, a group of adjacent local region blocks are selected as selected adjacent local region blocks, and the structural features of each local region block in the selected adjacent local region blocks are determined.

[0067] Secondly, in step 1032, the structural similarity between the selected adjacent local regions is determined based on the structural features of each local region in the selected adjacent local region blocks;

[0068] Finally, in step 1033, the structural similarity between the remaining adjacent local region blocks is determined.

[0069] In specific implementation, firstly, the texture features (such as contrast, correlation, energy, and entropy) of each local region block in the selected adjacent local region blocks can be calculated using the gray-level co-occurrence matrix. The local keypoint features of each local region block in the selected adjacent local region blocks are extracted using the scale-invariant feature transform operator. The edge intensity and orientation distribution of each local region block in the selected adjacent local region blocks are quantified using the Sobel gradient detection method. The set of the above texture features, local keypoint features, edge intensity, and orientation distribution is taken as the structural features of the corresponding local region block, where the structural features represent the structural characteristics of the local region block. Secondly, the degree of difference between the structural features of two local region blocks in the selected adjacent local region blocks is calculated based on similarity algorithms (such as cosine similarity, Euclidean distance, and structural similarity index), and this degree of difference is taken as the structural similarity between the structural features of each local region block in the selected adjacent local region blocks. In some embodiments, other methods can also be used, which are not limited here.

[0070] It should be noted that the structural similarity parameter in the text represents the degree of similarity between the structural features of each local region block in adjacent local region blocks. It can be used to analyze the degree of difference in weak positive sample images, thereby obtaining more accurate image information.

[0071] In some embodiments, the process of selecting candidate line regions and background regions for detection lines from the weakly positive sample images based on all structural similarities can be achieved using the following steps:

[0072] The weak positive sample image is enhanced to obtain an enhanced image of the weak positive sample;

[0073] Determine the similarity mapping of the detection lines based on all structural similarities;

[0074] Candidate line regions and background regions for detection lines are determined based on the similarity map and the enhanced image.

[0075] In specific implementation, firstly, adaptive histogram equalization is used to enhance the local contrast in the weak positive sample image, combined with Gaussian filtering to suppress noise in the weak positive sample image, thereby generating an enhanced image of the weak positive sample image, where the enhanced image represents the image after enhancement of the weak positive sample image; secondly, a similarity mapping map is constructed based on the structural similarity (such as SSIM value) of adjacent local regions, and the structural similarity values ​​are normalized to [0,1] and then mapped to grayscale values ​​(the higher the value, the more likely it is to belong to the detection line region), and the normalized structural similarity values ​​are linearly mapped to grayscale images (0 corresponds to black, 1 corresponds to white), and then this grayscale image is used as the similarity mapping map of the detection lines, where the similarity mapping map represents the mapping map of the degree of structural similarity between the detection lines; finally, a pixel-level fusion method based on masking is used to fuse the enhanced image and the similarity mapping map, and morphological closing operation (5×5) is used to achieve this. The small holes in the candidate region corresponding to the detection line in the image filled with structural elements are selected from the filled image by combining connected component analysis (area threshold set to 100 pixels²) to select the region that meets the morphological characteristics of the detection line (aspect ratio > 3:1) as the candidate line region, and the remaining corresponding region is divided into the background region; other methods can be used to determine the region in other embodiments, which are not limited here.

[0076] It should be noted that in this application, the candidate line region refers to the region corresponding to the candidate line of the detection line, and the background region refers to the background region excluding the candidate region corresponding to the detection line, which can be used to distinguish weak signal detection lines from background noise in weak positive sample images.

[0077] In step 104, a probability heatmap of the detection line is generated based on the gray-level gradient distribution of the candidate line region. The contribution features of the background region to the weak positive sample image are determined. The continuity of the detection line in the probability heatmap is constrained by the contribution features to obtain the background correction feature region of the detection line in the weak positive sample image.

[0078] In some embodiments, generating a probability heatmap of the detection line based on the gray-level gradient distribution of the candidate line region can be achieved using the following steps:

[0079] The horizontal and vertical gray-level gradient values ​​of each pixel within the candidate line region are determined, thereby obtaining the gray-level gradient distribution of the candidate line region;

[0080] Multiple probability values ​​for the detection line are determined based on the gray-scale gradient distribution;

[0081] The probability heatmap of the detection line is determined by all probability values.

[0082] In specific implementation, firstly, for each pixel within the candidate line region, the Sobel operator (3×3 convolution kernel) is used to calculate the horizontal and vertical gradient values ​​of each pixel within the candidate line region. Each horizontal and vertical gradient value is then mapped to a specific location within the candidate line region to form a grayscale gradient distribution. This grayscale gradient distribution refers to the statistical distribution characteristics of the rate of change (gradient) of pixel grayscale values ​​in space, used to describe the local details and structural information of the image. Secondly, the relationship between the detection line and the flow direction of the immunochromatographic test strip is obtained from the database. Based on this relationship, a gradient direction threshold (e.g., ±30°) is set, and gradient values ​​smaller than the gradient direction threshold are selected from the grayscale gradient distribution. All selected gradient values ​​are normalized, and the normalized values ​​are used as the probability values ​​of the corresponding pixel belonging to the detection line. The probability value represents the parameter value indicating the probability that a pixel belongs to the detection line. Finally, Gaussian smoothing (σ=1.5) is applied to eliminate isolated high probability values ​​among all probability values, followed by an 8... The neighborhood voting mechanism iterates through each pixel. If the probability value of a pixel is lower than the majority probability value of the 8 pixels in its neighborhood, the probability value of that pixel is corrected to the majority probability value in the neighborhood; otherwise, it remains unchanged. This enhances the spatial continuity of the detection line region. The corrected probability value is mapped to different pseudo-colors (e.g., probability value 0 corresponds to blue, and 1 corresponds to red), thereby generating a probability heatmap of the detection line. Other methods can also be used in other embodiments, which are not limited here.

[0083] It should be noted that the probability heatmap in this application represents the hotspot distribution of the probability of a detection line in a weakly positive sample image, and can be used to visually represent the probability distribution density of a detection line in a weakly positive sample image.

[0084] In some embodiments, determining the contribution features of the background region to the weakly positive sample image can be achieved by the following steps:

[0085] Determine the brightness and fluctuation characteristics of the background region;

[0086] Extract the texture detail features within the background area;

[0087] The contribution features of the background region to the weak positive sample image are determined based on the brightness features, the fluctuation features, and the detail features.

[0088] In specific implementation, firstly, the mean and standard deviation of pixel grayscale values ​​in the background region are calculated. The mean is used as a brightness feature, and the standard deviation is used as a fluctuation feature. The brightness feature reflects the background brightness, and the fluctuation feature reflects the background noise level. Secondly, a texture histogram is generated for the background region using a local binary mode to quantify the local grayscale variation pattern. Simultaneously, a Gabor filter (5 scales × 8 directions) is used to extract multi-scale texture information from the texture histogram of the background region. Principal component analysis is then used to reduce the dimensionality of the multi-scale texture information to 16. A 3D feature vector is generated and used as the texture detail feature within the background region. This texture detail feature represents the detailed characteristics of the texture within the background region, including its grayscale distribution, spatial arrangement, periodicity, and directionality. Finally, the brightness, fluctuation, and texture detail features of the background region are standardized to eliminate dimensional differences between different features (e.g., unifying the numerical range of brightness and texture features). The standardized three types of features are then combined into a comprehensive feature vector. A Gaussian mixture model is used to quantify the matching degree between each pixel and the background region. Global optimization fusion based on a graph model, combined with preset weight ratios, fuses the matching degree of the three types of features (e.g., brightness features account for 30%, fluctuation features for 20%, and detail features for 50%) to generate a contribution feature map of the background region to the weak positive sample image. This contribution feature map is used as the contribution feature of the background region to the weak positive sample image. Other methods can be used in other embodiments, which are not limited here.

[0089] It should be noted that the contribution feature map in this application represents the degree of contribution of the background region to the weak positive sample image. It can be used to analyze the distribution of the background region. The higher the value in the contribution feature map, the more the region conforms to the background characteristics, thereby effectively identifying the interference area of ​​background noise on the weak positive signal.

[0090] In some embodiments, the background correction feature region of the detection line in the weak positive sample image can be obtained by applying background constraints to the continuity of the detection line in the probability heatmap using the contribution features, which can be achieved through the following steps:

[0091] The suspected breakage region and candidate continuity region of the detection line are extracted from the probability heatmap;

[0092] The suspected fracture region of the detection line is smoothed using the contribution characteristics to obtain the fracture smoothing state of the detection line.

[0093] The background correction feature region of the detection line in the weak positive sample image is determined based on the fracture smoothing state and the candidate continuous region.

[0094] In specific implementation, the extraction of suspected break regions and candidate continuous regions of the detection line from the probability heatmap can be achieved as follows: A continuous judgment threshold for the detection line is determined by combining cross-validation and index optimization with historical weak positive sample image data. Here, the break-continuity judgment threshold represents the probability judgment threshold for a continuous detection line. Regions in the probability heatmap with a probability value ≥ the break-continuity judgment threshold are binarized into foreground, and the region corresponding to this foreground is designated as a candidate continuous region, with the remainder as background. Here, the candidate continuous region represents the region in the weak positive sample image where the detection line is continuous. Then, the opening operation in morphological opening and closing operations is applied to the binary image to eliminate isolated noise points in the foreground, and the closing operation is performed to fill internal small holes. The detection line is then connected to adjacent foreground regions to enhance its continuity. Next, the connected components of the detection line in the closed region are extracted. The boundaries of each connected component are traversed, and background gaps between adjacent connected components are detected using the scan line method. For each background gap, the width of the gap perpendicular to the detection line direction is calculated. A width judgment threshold for the detection line is determined by combining cross-validation and index optimization with historical weak positive sample image data. This width judgment threshold represents the minimum width threshold required for the detection line to be present during training. If the width is greater than or equal to the width judgment threshold, the region is determined to be a suspected break region. A suspected break region refers to a region in a weak positive sample image where the detection line is broken. Other methods can be used in other embodiments, which are not limited here.

[0095] In specific implementation, the suspected breakage area of ​​the detection line is smoothed using the contribution features to obtain the breakage smoothing state of the detection line. This can be achieved in the following way: First, for each pixel in the suspected breakage area, a 3×3 neighborhood is defined centered on it, and the weighted average of the gray values ​​of neighboring pixels is calculated. The weights decrease inversely with distance (e.g., the weight of distance d is 1 / d), so that the contribution of neighboring pixels to the repair value is greater. Random noise conforming to the Gaussian distribution N(0, σ²) is generated according to the fluctuation characteristics of the background area (i.e., standard deviation σ), and superimposed on the weighted average to simulate the natural gray-level fluctuation of the background. Finally, a Gabor filter (preset with 5 scales and 8 directions) matching the background texture is used to convolve the repair area to extract the frequency and direction features of the background texture, generate details consistent with the surrounding background texture, and fuse the texture details into the suspected breakage area through linear superposition. The fused area is used as the breakage smoothing state of the detection line. The breakage smoothing state represents the state of the area after the smooth transition between the breakage area and the surrounding background in terms of gray value, fluctuation characteristics, and texture mode, so as to eliminate artificial repair traces.

[0096] In specific implementation, the background correction feature region of the detection line in the weak positive sample image can be determined based on the broken smooth state and the candidate continuous region in the following way: pixel merging based on binary logic merges the broken smooth state and the candidate continuous region, introduces a conditional random field (CFD) model to optimize the boundary of the merged region, uses the probability heatmap value, grayscale features and neighborhood spatial relationship of the pixel as observation variables, and uses whether the pixel belongs to the detection line as a hidden variable, and achieves fine adjustment of the boundary by minimizing the energy function, eliminating the over-extension or blurring caused by smoothing processing, and uses the optimized region as the background correction feature region of the detection line in the weak positive sample image; other methods can also be used in other embodiments, which are not limited here.

[0097] It should be noted that the background correction feature region in this application represents the region of physical characteristics and signal characteristics of the detection line in the weak positive sample image, which can be used to analyze the properties of the weak positive sample and facilitate obtaining accurate properties of the weak positive sample.

[0098] In step 105, a contrast index of the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image are constructed to determine the signal-to-noise ratio of the detection line in the immunochromatographic test strip. The contrast index of the signal-to-noise ratio is compared with the preset weak positive interpretation interval of the weak positive sample to generate a grading report of the weak positive sample.

[0099] In some embodiments, the contrast index of the signal-to-noise ratio corresponding to the detection line in the immunochromatographic test strip, constructed using the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weakly positive sample image, can be achieved by the following steps:

[0100] Determine the signal-to-noise ratio of the detection line in the background correction feature region;

[0101] Determine the signal-to-noise ratio of the control lines in the weak positive sample image;

[0102] Construct a contrast index for the signal-to-noise ratio of the detection line in the immunochromatographic test strip based on the ratio of the signal-to-noise ratio of the detection line to that of the control line;

[0103] In specific implementation, within the background correction feature area of ​​the detection line, its average gray value is calculated, and the gray variance of adjacent background areas is used as the noise value. The two are divided to obtain the signal-to-noise ratio of the detection line. In the weak positive sample image, the quality control line area is located, and the ratio of the average gray value of this area to the background noise value is calculated to obtain the signal-to-noise ratio of the quality control line. The signal-to-noise ratio of the detection line is divided by the signal-to-noise ratio of the quality control line, and the value obtained by division is used as the contrast index of the signal-to-noise ratio corresponding to the detection line in the immunochromatographic test strip. In other embodiments, other methods can also be used to determine it, which are not limited here.

[0104] It should be noted that the contrast index in this application represents a parameter value indicating the degree of contrast between the signal-to-noise ratio corresponding to the test line and the signal-to-noise ratio corresponding to the control line in the immunochromatographic test strip, and can be used to analyze the properties of weakly positive samples.

[0105] In some embodiments, the comparison between the contrast index based on the signal-to-noise ratio and a preset weak positive interpretation interval for the weak positive sample, thereby generating a classification report for the weak positive sample, can be achieved through the following steps:

[0106] Determine the preset weak positive interpretation interval for the weak positive sample;

[0107] When the contrast index of the signal-to-noise ratio is greater than or equal to the upper limit of the weak positive interpretation interval, the weak positive sample is determined to be positive.

[0108] When the contrast index of the signal-to-noise ratio is within the weak positive interpretation interval, the weak positive sample is determined to be weak positive.

[0109] When the contrast index of the signal-to-noise ratio is less than or equal to the lower limit of the weak positive interpretation interval, the weak positive sample is determined to be negative.

[0110] A grading report for the weakly positive sample is generated based on whether the weakly positive sample is positive, weakly positive, or negative.

[0111] In specific implementation, the preset weak positive interpretation interval for weak positive samples is determined by combining the contrast index distribution of historical clinical samples with the quantile method. The weak positive interpretation interval represents the interval for interpreting weak positive samples. After the judgment is completed, the report generation engine is called to integrate the weak positive sample as positive, the weak positive sample as weak positive, and the weak positive sample as negative with multidimensional features (CI value, detection line contour map, texture matching degree). Through template design, the integrated results are combined to generate a graded report containing confidence score (such as threshold matching probability based on cross-validation) and retesting recommendations. Other methods can also be used in other embodiments, which are not limited here.

[0112] It should be noted that the grading report described in this application is a systematic document generated after the assessment objects are classified into levels based on standards. It presents the difference levels of assessment results through a combination of quantitative and qualitative methods, realizing full-process automation from numerical comparison to clinical decision-making.

[0113] Furthermore, in another aspect of this application, in some embodiments, this application provides an intelligent interpretation system for weakly positive samples in immune testing, with reference to... Figure 4The figure is a schematic diagram of the structure of an intelligent interpretation system for weakly positive immunoassay samples according to some embodiments of this application. The intelligent interpretation system 400 for weakly positive immunoassay samples includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below:

[0114] Acquisition module 401, in this application, is mainly used to acquire images of weakly positive samples containing the detection line and control line areas of the immunochromatographic test strip based on an image sensor;

[0115] Processing module 402, in this application, is used to divide the weak positive sample image into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip;

[0116] It should be noted that the processing module 402 in this application is also used to determine the structural similarity between adjacent local region blocks in each group, and to filter out candidate line regions and background regions of the detection line from the weak positive sample image based on all structural similarities.

[0117] Additionally, it should be noted that the processing module 402 in this application is also used to generate a probability heatmap of the detection line based on the gray-scale gradient distribution of the candidate line region, determine the contribution features of the background region to the weak positive sample image, and use the contribution features to impose background constraints on the continuity of the detection line in the probability heatmap to obtain the background correction feature region of the detection line in the weak positive sample image.

[0118] The execution module 403 in this application is mainly used to construct the contrast index of the signal-to-noise ratio of the detection line in the immunochromatographic test strip based on the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image. Based on the contrast index of the signal-to-noise ratio, it is compared with the preset weak positive interpretation interval of the weak positive sample, and then a grading report of the weak positive sample is generated.

[0119] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described intelligent interpretation method for weak positive samples in immune testing.

[0120] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing an intelligent interpretation method for weakly positive immunoassay samples according to some embodiments of this application. The intelligent interpretation method for weakly positive immunoassay samples in the above embodiments can... Figure 5The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0121] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0122] The communication bus 502 can be used to transmit information between the aforementioned components.

[0123] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0124] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The method used in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0125] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0126] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0127] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0128] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent interpretation method for weakly positive samples in immune testing.

[0129] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0130] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart method for interpreting weakly positive samples in immunochromatographic testing, applied to immunochromatographic test strips, wherein, The immunochromatographic test strip includes a detection line and a control line. Image acquisition is performed after the immunochromatographic test strip reacts with a weakly positive sample. The method is characterized by the following steps: The image sensor is used to acquire images of weakly positive samples in the regions containing the detection line and control line of the immunochromatographic test strip; The weakly positive sample image is divided into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip; Determine the structural similarity between adjacent local regions in each group, and based on all structural similarities, select candidate line regions and background regions for detection lines from the weak positive sample images; A probability heatmap of the detection line is generated based on the gray-level gradient distribution of the candidate line region. The contribution features of the background region to the weak positive sample image are determined. The continuity of the detection line in the probability heatmap is constrained by the contribution features to obtain the background correction feature region of the detection line in the weak positive sample image. A contrast index for the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image is constructed. The contrast index of the signal-to-noise ratio is compared with the preset weak positive interpretation interval of the weak positive sample to generate a grading report for the weak positive sample.

2. The method as described in claim 1, characterized in that, The weakly positive sample image is divided into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip, specifically including: Determine the chromatographic flow direction of the immunochromatographic test strip; Determine the smoothing window for segmenting the weak positive sample image; Based on the smooth window, the weak positive sample image is smoothly divided according to the chromatographic flow direction to obtain multiple local region blocks.

3. The method as described in claim 1, characterized in that, Determining the structural similarity between adjacent local regions in each group specifically includes: Select a set of adjacent local region blocks as the selected adjacent local region blocks, and determine the structural features of each local region block in the selected adjacent local region blocks; The structural similarity between the selected adjacent local region blocks is determined based on the structural features of each local region block in the selected adjacent local region blocks. Continue to determine the structural similarity between the remaining adjacent local region blocks.

4. The method as described in claim 1, characterized in that, Based on all structural similarities, candidate line regions and background regions for detection lines are selected from the weakly positive sample images, specifically including: The weak positive sample image is enhanced to obtain an enhanced image of the weak positive sample; Determine the similarity mapping of the detection lines based on all structural similarities; Candidate line regions and background regions for detection lines are determined based on the similarity map and the enhanced image.

5. The method as described in claim 1, characterized in that, Generating a probability heatmap of the detection line based on the gray-level gradient distribution of the candidate line region specifically includes: The horizontal and vertical gray-level gradient values ​​of each pixel within the candidate line region are determined, thereby obtaining the gray-level gradient distribution of the candidate line region; Multiple probability values ​​for the detection line are determined based on the gray-scale gradient distribution; The probability heatmap of the detection line is determined by all probability values.

6. The method as described in claim 1, characterized in that, Determining the contribution features of the background region to the weakly positive sample image specifically includes: Determine the brightness and fluctuation characteristics of the background region; Extract the texture detail features within the background area; The contribution features of the background region to the weak positive sample image are determined based on the brightness features, the fluctuation features, and the detail features.

7. The method as described in claim 1, characterized in that, By applying background constraints to the continuity of the detection lines in the probability heatmap using the contribution features, the background correction feature region of the detection lines in the weakly positive sample image is obtained, specifically including: The suspected breakage region and candidate continuity region of the detection line are extracted from the probability heatmap; The suspected fracture region of the detection line is smoothed using the contribution characteristics to obtain the fracture smoothing state of the detection line. The background correction feature region of the detection line in the weak positive sample image is determined based on the fracture smoothing state and the candidate continuous region.

8. An intelligent interpretation system for weakly positive samples in immunochromatographic testing, applied to immunochromatographic test strip detection, wherein, The immunochromatographic test strip includes a detection line and a control line. Image acquisition is performed after the immunochromatographic test strip reacts with a weakly positive sample. The system is characterized by comprising: The acquisition module is used to acquire images of weakly positive samples containing the detection line and control line areas of the immunochromatographic test strip based on an image sensor; The processing module is used to divide the weakly positive sample image into multiple local region blocks according to the chromatographic flow direction of the immunochromatographic test strip; The processing module is also used to determine the structural similarity between adjacent local region blocks in each group, and to filter out candidate line regions and background regions for detection lines from the weak positive sample images based on all structural similarities. The processing module is further configured to generate a probability heatmap of the detection line based on the gray-level gradient distribution of the candidate line region, determine the contribution features of the background region to the weak positive sample image, and apply background constraints to the continuity of the detection line in the probability heatmap through the contribution features to obtain the background correction feature region of the detection line in the weak positive sample image. The execution module is used to construct a contrast index of the signal-to-noise ratio of the detection line in the immunochromatographic test strip based on the signal-to-noise ratio of the detection line in the background correction feature region and the signal-to-noise ratio of the control line in the weak positive sample image. Based on the contrast index of the signal-to-noise ratio, it is compared with the preset weak positive interpretation interval of the weak positive sample, and then a grading report of the weak positive sample is generated.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the intelligent interpretation method for weakly positive samples in immunoassay as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent interpretation method for weakly positive samples in immune testing as described in any one of claims 1 to 7.

Images