Test paper detection method and device of IVD, electronic equipment and storage medium
By performing a series of processing on the original image of the IVD test strip detection results, extracting and fusing the accurate detection features, the problem of insufficient reading accuracy of the IVD test strip detection results in the prior art is solved, and higher detection accuracy and efficiency are achieved.
Patent Information
- Application Number
- CN202510072889.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-17
- Publication Date
- 2025-06-13
AI Technical Summary
The existing methods for reading IVD test strip test results rely on visual comparison methods, and there is a problem of insufficient accuracy, especially in the case of environmental impact and differences in human eye color recognition.
By acquiring the original image of the detected IVD test strip, a series of processes are performed to extract the foreground area images, obtain the positive detection image, standard detection feature map, and layered and fused along the long side direction to obtain the accurate detection features, and finally match the detection results based on the accurate detection features.
It reduces the impact of external environmental factors on the detection results, reduces the workload of subsequent feature extraction, improves the detection accuracy, and can quickly and effectively obtain the test strip test results.
Smart Images
Figure CN120147668A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of test strip detection, and specifically to a test strip detection method, device, electronic device and storage medium for IVD. Background Art
[0002] IVD test strips are a type of in vitro diagnostic reagent, mainly used for detecting and diagnosing disease markers or biomolecules in human samples. IVD reagents usually include components such as reagent kits, test strips, instruments and related software.
[0003] IVD test strips have a very wide range of applications, covering multiple fields, including clinical medicine, preventive health care, drug research and development, etc. They can help doctors judge the type, severity of diseases and the selection of treatment plans, provide accurate and timely diagnostic results for patients, and are of great significance for the early screening and monitoring of diseases.
[0004] Currently, the method for reading the results of IVD test strips is the visual comparison method. When using the visual comparison method to identify the reaction color gamut of the test strip, it will be affected by the environment and the differences in human eye color recognition. Moreover, the saliva test strip is relatively small, and the area of its effective test strip foreground region is 15mm x 15mm. And due to the unstable concentration of the detection source, some detection results may have unclear colors due to too low concentration of the detection source. Just relying on visual inspection, the detection results lack accuracy.
[0005] With the popularization of mobile devices and the development of image recognition technology, it has become possible to form a more efficient and convenient IVD test strip recognition method by combining imaging devices such as mobile phones and cameras with machine learning and other methods. However, due to the influence of imaging devices, environmental light, color temperature, etc., problems such as low contrast between the image target and the background, high noise, shadow interference and blurred edges are caused, which affect the recognition and detection accuracy of IVD test strip images.
[0006] Based on this, it is necessary to propose a test strip detection method, device, electronic device and storage medium for IVD to improve the accuracy of obtaining IVD test strip detection results, which has become an important technical problem to be solved urgently. Summary of the Invention
[0007] The purpose of the present invention is to provide a test strip detection method, device, electronic device and storage medium for IVD to solve the problem that the existing method for reading the results of IVD test strips is the visual comparison method and the detection results lack accuracy.
[0008] To achieve the above object, the present invention provides the following technical solution: A test strip detection method for IVD, the detection method comprising the following steps: S1, obtaining an original image of the test strip of the detected IVD; S2, processing the original image to extract a foreground region image of the test strip of the IVD; S3, processing the foreground region image to obtain a positive detection image; S4, processing the positive detection image to obtain a standard detection feature map; S5, stratifying the standard detection feature map along the long side direction, and then fusing according to position and color information to obtain a precise detection feature; S6, matching the detection result according to the precise detection feature.
[0009] Preferably, the above S2 further comprises the following steps: S21, obtaining the boundary of the detection region through a U-net network; S22, obtaining the contour of the detection region through a Canny operator; S23, obtaining the foreground region image according to the contour and the original image.
[0010] Preferably, the above S3 specifically comprises the following steps: obtaining a straight boundary of the positive detection image through Hough transform, and obtaining the positive detection image according to the straight boundary and the foreground region image.
[0011] Preferably, the above Hough transform specifically comprises the following steps: mapping the points in the foreground image space to the parameter space to form a series of straight lines; in the parameter space, voting for each straight line and counting the number of times it appears; finding local maxima in the parameter space, and these maxima correspond to the straight lines in the image; extracting the parameters of the straight lines in the image according to the positions of the peaks; the straight lines form the boundary of the positive detection image;
[0012] Preferably, the above S4 specifically comprises the following steps:
[0013] S41, numerically processing the positive detection image to obtain a digitalized feature map;
[0014] S42, adjusting the size of the digitalized feature map to obtain an intermediate feature map of a standard size;
[0015] S42, normalizing the intermediate feature map to form a standard detection feature map, and the normalization method is as follows:
[0016]
[0017] wherein, P is the pixel value of the intermediate feature map, and P 1 is the pixel value of the standard detection feature map.
[0018] Preferably, the method for adjusting the size of the digitalized feature map is as follows:
[0019]
[0020] Among them, x 1 and y 1 are the coordinates in the intermediate feature map, x i and y i are the coordinates of the four pixels closest to x 1 and y 1 in the digital feature map. I(x i , y i ) is the pixel value of the digital feature map at the point (x i , y i ), and I 1 (x 1 , y 1 ) is the pixel value of the intermediate feature map at the point (x 1 , y 1 ).
[0021] Preferably, the above S5 specifically includes the following steps: Traverse the standard detection feature map to obtain several layers of small feature maps: P 1 , P 2 , …, P n , and then use the (R, G, B) information of the point mapping in the small feature map to construct the plane - gray - scale similarity quantization value M of different - level small feature maps:
[0022]
[0023] In the formula, t represents the number of points in the top - most layer of small feature maps, o represents the number of points in the small feature map adjacent to the top - most layer of small feature maps, p(r, g, b) represents the rgb value of the point in the top - most layer of small feature maps, q(r, g, b) represents the rgb value of the point in the small feature map adjacent to the top - most layer of small feature maps, p(x, y) represents the plane position of the point in the top - most layer of small feature maps, q(x, y) represents the plane position of the point in the small feature map adjacent to the top - most layer of small feature maps, and k is the weighting factor of the normalized image distance and the set distance;
[0024] When M <= Dm, merge two layers of small feature maps, and use the merged small feature map as the new outermost - layer small feature map;
[0025] When M > Dm, save the outermost - layer small feature map as the precise feature map, and use the small feature map adjacent to the outermost - layer small feature map as the new outermost - layer small feature map;
[0026] Traverse P 1 , P 2 , …, P n , and obtain multiple layers of precise feature maps, and use each precise feature map as the precise detection feature.
[0027] Based on another object of the present invention, the present invention provides the following technical solution: An IVD test strip detection device, the detection device uses the detection method as described above, and the detection device includes: a detector, the detector is provided with a display area; a camera unit, the camera unit is used to obtain the original image of the IVD test strip after detection; a foreground unit, the foreground unit is used to process the original image to extract the foreground area image of the IVD test strip; a positive detection unit, the positive detection unit is used to process the foreground area image to obtain a positive detection image; a standard unit, the standard unit is used to process the positive detection image to obtain a standard detection feature map; a precision unit, the precision unit is used to layer the standard detection feature map along the long side direction, and then fuse according to the position and color information to obtain a precise detection feature; a matching unit, the matching unit is used to match the detection result according to the precise detection feature.
[0028] Based on another object of the present invention, the present invention provides the following technical solution: An electronic device, the electronic device runs the detection method as described above.
[0029] Based on another object of the present invention, the present invention provides the following technical solution: A storage medium, the storage medium stores the detection method as described above.
[0030] Compared with the prior art, the beneficial effect of the test strip detection method of the present invention is: By performing a series of processes on the original image, the influence of external environmental factors on the detection result is minimized as much as possible, and the workload of subsequent feature extraction can be greatly reduced. The test strip is segmented according to the position and color information, which greatly reduces the workload required for matching and can greatly improve the detection accuracy. In summary, through the IVD test strip detection method, the test strip detection result can be obtained quickly and effectively. Description of the Drawings
[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.
[0032] Figure 1 It is a technical roadmap of an IVD test strip detection method;
[0033] Figure 2 It is a system structure diagram of an IVD test strip detection device;
[0034] Figure 3 It is a structural schematic diagram of the detector.
[0035] In the figure: a detection device 10 , a camera unit 110 , a foreground unit 120 , a positive detection unit 130 , a standard unit 140 , a precision unit 150 , a matching unit 160 , a detector 2 , and a detection area 21 . DETAILED DESCRIPTION
[0036] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.
[0037] In this article, the terms "upper, lower, inside, outside" and the like are established based on the positional relationships shown in the drawings. The corresponding positional relationships may also change depending on the drawings, and therefore cannot be understood as absolute limitations on the scope of protection; moreover, relational terms such as "first" and "second" and the like are merely used to distinguish one component from another with the same name, and do not necessarily require or imply any actual relationship or order between these components.
[0038] Embodiment 1
[0039] like Figure 1 As shown, this embodiment provides an IVD test paper detection method, and the detection method includes the following steps: S1, obtaining the original image of the IVD test paper after detection; the relevant detection personnel need to perform the detection according to the IVD test paper detection operation instructions, and after waiting for a preset time, they need to take a photo of the test paper that has completed the detection, and the original image can be obtained by taking a photo, and after obtaining the original image, it needs to be uploaded to the cloud. Preferably, when obtaining the original image, the conditions such as the shooting height and the shooting angle should be as consistent as possible with the angle and height specified in the IVD test paper detection operation instructions to avoid incomplete test strips and minimize the workload of subsequent processing.
[0040] S2. Processing the original image to extract the foreground area image of the IVD test strip; processing the original image includes extracting the boundary by the U-net network, obtaining the contour of the detection area by the Canny operator, and intercepting the foreground area image on the original image by the contour and the original image. The above operation can quickly obtain the contour of the detection area, and the accuracy of the contour of the detection area is high.
[0041] S3. Process the foreground region image to obtain the positive detection image. The processing of the foreground region image includes Hough transform. Since the detection region of the IVD test strip is rectangular and the boundaries of its detection region are all straight lines, the straight boundaries of the detection region can be accurately obtained through Hough transform. Then, the original image can be further cropped based on the contour formed by the straight boundaries to further remove the background image, enhance the subsequent detection accuracy, and reduce the subsequent detection workload. After cropping the foreground region image, its angle needs to be corrected.
[0042] S4. Process the positive detection image to obtain the standard detection feature map. The processing of the positive detection image includes digitization processing, size adjustment, and normalization processing. Digitization processing facilitates subsequent feature matching. Size adjustment makes all input images of standard size, which is convenient for reducing the workload required for subsequent matching. Normalization processing is also for reducing the workload required for subsequent matching.
[0043] S5. Layer the standard detection feature map along the long side direction, and then fuse it according to the position and color information to obtain the accurate detection feature. Based on the detection result reading method of the IVD test strip, it mainly relies on color information and position information. The test strip is accurately segmented according to the position information and color information of different regions of the test strip. Feature matching is performed based on the segmented test strip, which can greatly reduce the workload required for matching and can significantly improve the detection accuracy.
[0044] S6. Match the detection result according to the accurate detection feature. Match the accurate detection feature according to the SIFT algorithm, SURF algorithm, and ORB algorithm. After obtaining the matching result, send it back to the relevant detection personnel to complete a detection process.
[0045] Specifically, through a series of processes on the original image, the influence of external environmental factors on the detection result is minimized as much as possible, and the workload of subsequent feature extraction can be significantly reduced. The test strip is segmented according to the position and color information, which can greatly reduce the workload required for matching and can significantly improve the detection accuracy. In summary, through the IVD test strip detection method, the test strip detection result can be obtained quickly and effectively.
[0046] Furthermore, the above S2 also includes the following steps: S21. Obtain the boundary of the detection region through the U-net network; the boundary extraction model is obtained after training the U-NET network on the dataset, and the model can be used for boundary extraction. The dataset includes 80% of the pictures of IVD test strips and 20% of negative samples.
[0047] S22. Obtain the contour of the detection area through the Canny operator; since the boundaries obtained by u-net generally have small holes and the edges are not smooth enough, after extracting the boundaries, use the Canny operator for edge extraction and encode the contour to obtain a more accurate contour of the detection area.
[0048] S23. Obtain the foreground area image according to the contour and the original image. Using the contour as a mask, the foreground area image can be obtained on the original image.
[0049] Further, the above S3 specifically includes the following steps: Obtain the straight line boundary of the positive detection image through the Hough transform, and obtain the positive detection image according to the straight line boundary and the foreground area image. The above Hough transform specifically includes the following steps: Map the points in the foreground image space to the parameter space to form a series of straight lines; in the parameter space, vote for each straight line and count the number of times it appears; find the local maximum in the parameter space, and these maximum values correspond to the straight lines in the image; according to the position of the peak, extract the parameters of the straight lines in the image; the straight lines form the boundary of the positive detection image. In this embodiment, the Hough transform is used to detect straight lines in the image according to the characteristic that the edge of the test strip is a straight line. The parametric equation of the Hough transform is as follows:
[0050] ρ = xcosθ + ysinθ
[0051] Use the θ and ρ parameters to describe the straight line. ρ represents the perpendicular distance from the origin to the straight line, and θ represents the angle between the straight line and the x-axis. Preferably, an angle transformation is also performed on the obtained positive detection image to make the two short boundaries of the positive detection image parallel to the X-axis.
[0052] Further, the above S4 specifically includes the following steps:
[0053] S41. Numerically process the positive detection image to obtain a digital feature map; after digitization, store the digitized feature map through a tensor matrix; S42. Adjust the size of the digital feature map to obtain an intermediate feature map of the standard size. The method for adjusting the size of the digital feature map is as follows:
[0054]
[0055] where x 1 and y 1 are the coordinates in the intermediate feature map, x i and y i are the coordinates of the four pixels closest to x 1 and y 1 in the digital feature map, and I(x i , y i ) is the digital feature map at (xi , y i ), the pixel value of the point, I 1 (x 1 , y 1 ) is the pixel value of the intermediate feature map at the point (x 1 , y 1 ); S42. Normalize the intermediate feature map to form a standard detection feature map. The normalization method is as follows:
[0056]
[0057] where P is the pixel value of the intermediate feature map, and P 1 is the pixel value of the standard detection feature map. The size adjustment makes all input images of a standard size, facilitating reducing the workload required for subsequent matching. The normalization process is also for reducing the workload required for subsequent matching.
[0058] Further, the above S5 specifically includes the following steps: Traverse the standard detection feature map to obtain several layers of small feature maps: P 1 , P 2 , …, P n . The chunking method here is: Obtain the maximum and minimum y - coordinates y min 、y max of the points in the standard detection feature map. Establish n - 1 equally - spaced and x - axis - parallel dividing lines between the maximum and minimum y - coordinates. Divide the standard detection feature map according to the division to obtain several layers of small feature maps. Then, use the (R, G, B) information of the points mapped in the small feature maps to construct the plane - gray - scale similarity quantization value M of different - level small feature maps:
[0059]
[0060] In the formula, t represents the number of points in the top - most layer of small feature maps, o represents the number of points in the small feature map adjacent to the top - most layer of small feature maps, p(r, g, b) represents the rgb value of the point in the top - most layer of small feature maps, q(r, g, b) represents the rgb value of the point in the small feature map adjacent to the top - most layer of small feature maps, p(x, y) represents the plane position of the point in the top - most layer of small feature maps, q(x, y) represents the plane position of the point in the small feature map adjacent to the top - most layer of small feature maps, and k is the weighting factor of the normalized image distance and the set distance;
[0061] The planar-gray similarity quantization value M is used to evaluate the similarity between two layers of small feature maps. When the similarity is large enough, they can be merged. When M <= Dm, the two layers of small feature maps are merged, and the merged small feature map becomes the new outermost small feature map; Dm is a threshold. Preferably, Dm = 0.1 mm. When M > Dm, the outermost small feature map is saved as an accurate feature map, and the small feature map adjacent to the outermost small feature map becomes the new outermost small feature map; traverse P 1 ,P 2 ,…,P n , and obtain multiple accurate feature maps, using each accurate feature map as the precise detection feature. Through the above steps, relying on color information and position information, the test strip can be accurately segmented according to the position information and color information of different regions of the test strip, and feature matching can be performed based on the segmented test strip, which can greatly reduce the workload required for matching and significantly improve the detection accuracy.
[0062] Embodiment 2
[0063] For the parts in this embodiment that are the same as those in Embodiment 1, the same reference numerals are given and the same textual descriptions are omitted.
[0064] As Figure 2 and Figure 3 shown, this embodiment discloses a test strip detection device for IVD. The detection device 10 uses the detection method in Embodiment 1. The detection device 10 includes: a detector 2, with a detection area 21 set on the detector; an imaging unit 110, which is used to obtain the original image of the test strip of the IVD after detection; a foreground unit 120, which is used to process the original image to extract the foreground area image of the test strip of the IVD; a positive detection unit 130, which is used to process the foreground area image to obtain a positive detection image; a standard unit 140, which is used to process the positive detection image to obtain a standard detection feature map; a precise unit 150, which is used to layer the standard detection feature map along the long side direction and then perform fusion according to position and color information to obtain a precise detection feature; a matching unit 160, which is used to match the detection result according to the precise detection feature.
[0065] Embodiment 3
[0066] In this embodiment, for the parts that are the same as those in Embodiment 1 and Embodiment 2, the same reference numerals are given and the same textual descriptions are omitted.
[0067] This embodiment discloses an electronic device that runs the detection method in Embodiment 1.
[0068] Embodiment 4
[0069] In this embodiment, the parts that are the same as those in Embodiment 1 and Embodiment 2 are given the same reference numerals, and the same textual descriptions are omitted.
[0070] This embodiment discloses a storage medium that stores the detection method in Embodiment 1.
[0071] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.
Claims
1. A test paper detection method for IVD, characterized in that: The detection method comprises the following steps: S1, obtaining the original image of the IVD test strip after detection; S2, processing the original image to extract a foreground area image of the IVD test strip; S3, processing the foreground area image to obtain a positive detection image; S4, processing the positive detection image to obtain a standard detection feature map; S5, layering the standard detection feature map along the long side direction, and then fusing it according to the position and color information to obtain accurate detection features; S6. Matching detection results according to the precise detection features.
2. The IVD test paper detection method according to claim 1, characterized in that: The above S2 further includes the following steps: S21, obtaining the boundary of the detection area through the U-net network; S22, obtaining the contour of the detection area through a Canny operator; S23. Acquire the foreground area image according to the outline and the original image.
3. The IVD test paper detection method according to claim 1, characterized in that: The above S3 specifically includes the following steps: obtaining the straight line boundary of the positive detection image through Hough transform, and obtaining the positive detection image according to the straight line boundary and the foreground area image.
4. The IVD test paper detection method according to claim 3, characterized in that: The above-mentioned Hough transform specifically includes the following steps: mapping the points in the foreground image space into the parameter space to form a series of straight lines; in the parameter space, voting for each straight line and counting the number of times it appears; finding local maxima in the parameter space, which correspond to the straight lines in the image; extracting the parameters of the straight lines in the image according to the positions of the peak values; the straight lines form the boundaries of the positive detection image.
5. The IVD test paper detection method according to claim 1, characterized in that: The above S4 specifically includes the following steps: S41, performing digital processing on the positive detection image to obtain a digital feature map; S42, adjusting the size of the digitized feature map to obtain an intermediate feature map of a standard size; S42, normalizing the intermediate feature map to form the standard detection feature map, wherein the normalization method is as follows: Among them, P is the pixel value of the intermediate feature map, and P1 is the pixel value of the standard detection feature map.
6. A test paper detection method for IVD according to claim 5, characterized in that, The method for adjusting the size of the digitized feature map is as follows: Among them, x1 and y1 are the coordinates in the intermediate feature map, x i and i are the coordinates of the four nearest pixels corresponding to x1 and y1 in the digitized feature map, I(x i ,y i ) is the digital feature map in (x i ,y i ), I1(x1,y1) is the pixel value of the intermediate feature map at the point (x1,y1).
7. The IVD test paper detection method according to claim 5, characterized in that: The above S5 specifically includes the following steps: traverse the standard detection feature map to obtain several layers of small feature maps: P1, P2, ..., P n , and then use the (R, G, B) information of the point mapping in the small feature map to construct the plane-grayscale similarity quantization value M of the small feature map at different levels: Wherein, t represents the number of points in the small feature map of the top layer, o represents the number of points in the small feature map adjacent to the small feature map of the top layer, p(r, g, b) represents the RGB value of the point in the small feature map of the top layer, q(r, g, b) represents the RGB value of the point in the small feature map adjacent to the small feature map of the top layer, p(x, y) represents the plane position of the point in the small feature map of the top layer, q(x, y) represents the plane position of the point in the small feature map adjacent to the small feature map of the top layer, and k is the weighting factor of the normalized image distance and the set distance; When M<=Dm, the two layers of small feature maps are merged, and the merged small feature maps are used as the new outermost small feature maps; When M>Dm, the small feature map in the outermost layer is saved as a precise feature map, and the small feature map adjacent to the small feature map in the outermost layer is used as the new small feature map in the outermost layer; Traverse P1, P2, ..., P n , obtain multiple layers of the precise feature map, and use each of the precise feature maps as the precise detection feature.
8. A test paper detection device for IVD, characterized in that: The detection device uses the detection method according to any one of claims 1 to 7, and the detection device comprises: A detector, wherein the detector is provided with a display area; A camera unit, the camera unit is used to obtain an original image of the IVD test paper after detection; A foreground unit, the foreground unit is used to process the original image to extract a foreground area image of the IVD test strip; A positive detection unit, the positive detection unit is used to process the foreground area image to obtain a positive detection image; A standard unit, the standard unit is used to process the positive detection image to obtain a standard detection feature map; A precision unit, the precision unit is used to layer the standard detection feature map along the long side direction, and then fuse them according to the position and color information to obtain the precision detection feature; A matching unit is used to match the detection result according to the precise detection feature.
9. An electronic device, characterized in that: The electronic device runs the detection method described in any one of claims 1-7.
10. A storage medium, characterized in that: The storage medium stores the detection method described in any one of claims 1-7.