A digital image target detection method for medical testing

By employing a local self-normalized color field and a space-color coupled diffusion structure measurement algorithm, the problem of unstable color and spatial structure detection in medical examination images is solved, achieving stable and accurate detection of target structures in medical examination images.

CN123118569APending Publication Date: 2026-11-17FOURTH MILITARY MEDICAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611375014.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-07
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing medical laboratory image target detection methods rely on single color features or simple edge information, which makes it difficult to take into account both color change features and spatial structural continuity, resulting in unstable detection results and a high risk of false positives or false negatives.

Method used

A local self-normalized color field construction algorithm and a spatial-color coupled diffusion structure metric calculation algorithm are adopted. The local self-normalized color field construction algorithm reduces the influence of differences in illumination and staining intensity. The structure propagation enhancement response value is calculated by combining the spatial-color coupled diffusion structure metric value, and an adaptive detection threshold is generated to generate a mask for the target area.

Benefits of technology

It improves the consistency and stability of color features in medical examination images, enhances the color contrast features of target structures, reduces background noise interference, improves the ability to distinguish target structures from background areas, and enhances the accuracy and stability of detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118569A_ABST
    Figure CN123118569A_ABST
Patent Text Reader

Abstract

This invention relates to the field of image processing technology, and more particularly to a digital image target detection method for medical examination. First, pixel color vectors are obtained based on the acquired medical examination image. A normalized color field vector is then obtained using a local self-normalized color field construction algorithm. Next, a spatial-color coupling diffusion structure metric is calculated using a spatial-color coupling diffusion structure metric calculation algorithm. The maximum spatial-color coupling diffusion structure metric is searched among all pixels in the neighborhood of each pixel, and coupled with the metric of the current pixel to obtain a structure propagation enhancement response value. Finally, an adaptive detection threshold is obtained based on the structure propagation enhancement response value and compared with the structure propagation enhancement response value to generate a target region mask. This method solves the technical problems of easily misidentifying random noise or colored particles as target structures and missing detections when the structure boundaries are weak, leading to insufficient detection accuracy and stability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a digital image target detection method for medical testing. Background Technology

[0002] With the development of digital imaging and computer vision technologies, the field of medical testing is gradually shifting from traditional manual microscopic observation to digital and automated image analysis. In medical testing processes such as blood cell testing, pathological slide analysis, and microbial detection, it is typically necessary to acquire sample images using microscopic imaging equipment and then identify and statistically analyze cell morphology, bacterial distribution, or tissue structure using image analysis techniques. In actual medical testing, samples often require staining to enhance the distinguishability between different structures; for example, cell nuclei, cytoplasm, or microbial structures often exhibit specific color characteristics through different staining methods. However, during microscopic imaging, factors such as uneven staining intensity, changes in lighting conditions, and differences in imaging equipment parameters can easily lead to significant color differences in the same structure across different regions or between different samples. This color instability affects the extraction of structural features by image analysis algorithms, making target detection methods based on color or grayscale information prone to false positives or false negatives, thereby reducing the reliability of automated medical testing image analysis. On the other hand, target structures in medical examination images often possess complex spatial morphological features. For example, cell boundaries, bacterial outlines, or abnormal tissue regions often exhibit continuous spatial structures, while background regions may contain interfering information such as staining noise, impurities, or imaging noise. Existing image detection methods mainly rely on single color features or simple edge information for detection, making them highly sensitive to local noise and unable to simultaneously consider color variation features and spatial structural continuity. When color fluctuations or background interference exist in the image, isolated noise is easily misidentified as a target structure, affecting the accuracy of the detection results. Therefore, there is an urgent need to propose a digital image target detection method for medical examination to address the aforementioned problems. Summary of the Invention

[0003] This invention provides a digital image target detection method for medical testing, which addresses the problems of existing methods that often analyze directly based on changes in original color or local pixel intensity values, lack comprehensive measurement of color differences and spatial structural continuity, and are prone to misidentifying random noise or stained particles as target structures; and the technical problems of missed detection when structural boundaries are weak, resulting in insufficient accuracy and stability in the detection of target areas such as cells, bacteria or tissue structures in medical testing images.

[0004] The present invention provides a digital image target detection method for medical testing, comprising the following steps:

[0005] S1. Acquire medical examination images; Based on the light intensity of the acquired medical examination images in the red, green and blue bands, obtain the intensity values ​​of the pixels in the red, green and blue color channels. The intensity values ​​of the pixels in the three color channels together constitute the pixel color vector, and the normalized color field vector is obtained through the local self-normalization color field construction algorithm.

[0006] S2. Based on the normalized color field vector, the spatial-color coupled diffusion structure metric is calculated using a spatial-color coupled diffusion structure metric calculation algorithm. After searching for the maximum spatial-color coupled diffusion structure metric in the neighborhood of each pixel, it is coupled with the spatial-color coupled diffusion structure metric of the current pixel to obtain the structure propagation enhancement response value. Based on the structure propagation enhancement response value, an adaptive detection threshold is obtained. The structure propagation enhancement response value is compared with the adaptive detection threshold to generate a target region mask for application to medical examination images.

[0007] Preferably, S1 specifically includes:

[0008] In the local self-normalized color field construction algorithm, a neighborhood is established with the current pixel as the center, and the intensity values ​​of all pixels in the neighborhood on the three color channels are statistically averaged to obtain the neighborhood color mean vector.

[0009] Preferably, S1 specifically includes:

[0010] Based on the intensity values ​​of neighboring pixels in the three color channels and the mean vector of neighboring colors, the difference between neighboring pixels in the three color channels is calculated, and the difference is statistically analyzed to obtain the degree of color change in the neighborhood.

[0011] Preferably, S1 specifically includes:

[0012] Based on the pixel color vector and the neighborhood color mean vector, calculate the offset and color offset magnitude of the current pixel's color vector relative to the neighborhood color mean vector; add 1 to the sum of the neighborhood color change and the color offset magnitude to obtain the normalization factor; based on the normalization factor, normalize the offset to obtain the normalized color field vector.

[0013] Preferably, S2 specifically includes:

[0014] In the spatial-color coupled diffusion structure metric calculation algorithm, the color difference between the current pixel and its neighboring pixels is calculated based on the normalized color field vector; the spatial distance factor is obtained by calculating the square of the geometric distance based on the spatial coordinates of the current pixel and the spatial coordinates of its neighboring pixels in the medical examination image.

[0015] Preferably, S2 specifically includes:

[0016] Based on the spatial distance factor, the color differences are weighted, accumulated, and normalized to obtain the spatial-color coupling diffusion structure metric of the pixel.

[0017] Preferably, S2 specifically includes:

[0018] The structure propagation enhancement response value of a pixel is obtained by dividing the product of the spatial-color coupled diffusion structure metric of a pixel and the largest spatial-color coupled diffusion structure metric of any other pixel in the pixel's neighborhood by 1 and adding the largest spatial-color coupled diffusion structure metric of any other pixel in the pixel's neighborhood.

[0019] Preferably, S2 specifically includes:

[0020] The adaptive detection threshold is obtained by averaging the structure propagation enhancement response values ​​of all pixels in the medical examination image.

[0021] The beneficial effects of the technical solution of the present invention are:

[0022] 1. By constructing a local self-normalized color field, the color of medical examination images is stabilized, and pixel color normalization is achieved. This reduces the impact of changes in illumination, staining intensity, and imaging equipment parameters on the color distribution of medical examination images during microscopic imaging. It enables the color information of different regions to be expressed at a uniform scale, improving the consistency and stability of color features in medical examination images and providing a stable color information foundation for subsequent target structure detection.

[0023] 2. By jointly analyzing the mean color vector of the neighborhood and the degree of color change in the neighborhood, and normalizing the pixel color vector, the suppression of local abnormal color brightness and random noise is achieved. At the same time, the regional features with obvious color changes, such as cell boundaries, bacterial outlines and tissue structures, are preserved, thereby enhancing the color contrast features of the target structure, reducing the interference of the background area on the detection results, and improving the ability to express the structural information of medical test images.

[0024] 3. By constructing a spatial-color coupled diffusion structure metric, a synergistic measurement of pixel color difference and spatial distance factor is achieved, enabling a unified expression of color change features and spatial structure features in medical examination images. This effectively characterizes spatially continuous structural information such as cell boundaries, microbial outlines, and tissue structure edges in medical examination images, improving the ability to distinguish between target structures and background regions.

[0025] 4. By searching for the maximum spatial-color coupling diffusion structure metric value among pixels other than the current pixel in the pixel's neighborhood and coupling it with the current pixel's spatial-color coupling diffusion structure metric value, the propagation enhancement of structural information in the local region is achieved. This allows pixels in the real structural region with spatially continuous features to obtain higher response intensity compared to isolated noise pixels with the same spatial-color coupling diffusion structure metric value but lower maximum spatial-color coupling diffusion structure metric values ​​among pixels other than the current pixel in the neighborhood. Isolated noise pixels are effectively suppressed because their neighbors lack structural support and have lower corresponding maximum spatial-color coupling diffusion structure metric values. This reduces the probability of isolated noise pixels being misidentified as target structures and improves the stability of target detection results. Attached Figure Description

[0026] Figure 1 This is a flowchart of a digital image target detection method for medical testing according to the present invention. Detailed Implementation

[0027] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments; based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] The following description, in conjunction with the accompanying drawings, details a specific scheme for a digital image target detection method for medical testing provided by the present invention.

[0030] See attached document Figure 1 The diagram illustrates a flowchart of a digital image target detection method for medical examination provided by an embodiment of the present invention. The method includes the following steps:

[0031] S1. Acquire medical examination images; Based on the light intensity of the acquired medical examination images in the red, green and blue bands, obtain the intensity values ​​of the pixels in the red, green and blue color channels. The intensity values ​​of the pixels in the three color channels together constitute the pixel color vector, and the normalized color field vector is obtained through the local self-normalized color field construction algorithm.

[0032] Medical examination images are acquired from medical imaging equipment, such as pathology slide scanners and microbial detection microscopes. These images are desensitized and authorized sample images, or images taken from publicly available medical image datasets, and their acquisition and use comply with relevant regulations on personal information protection. In the acquired medical examination images, such as blood cell images, pathology slide images, or microbial detection images, different structures, such as cell boundaries, bacterial outlines, and tissue structures, exhibit different color characteristics through different staining methods such as Wright staining, HE staining, and fluorescence staining. Changes in staining intensity and microscopic lighting conditions often lead to different color brightness in different areas of the same structure. A local self-normalized color field construction algorithm is used for color stabilization processing of medical examination images. Local statistical analysis is used to normalize the color information of the medical examination images, resulting in a normalized color field vector. This reduces the impact of changes in lighting conditions, staining intensity differences, and imaging equipment parameters during microscopic imaging on the image analysis results, thereby providing a stable color feature representation for subsequent structural detection.

[0033] The acquired medical examination images are three-channel color images. Indicates the spatial location of medical examination images The pixel color vector at that location. This represents the spatial coordinates of pixels in a medical examination image. The color vector of each pixel is represented by the intensity values ​​of the pixel in the red, green, and blue channels. Indicates spatial location The intensity value of the pixel at that location in the red channel. Indicates spatial location The intensity value of the pixel at that location in the green channel. Indicates spatial location The intensity value of a pixel in the blue channel is represented by the intensity values ​​of the pixel in all three color channels. These values ​​are obtained by medical imaging equipment, such as microscopes or scanners, through optical sensors that collect light intensities in the red, green, and blue bands of the acquired medical image and convert the data between analog and digital. The combination of the three color channels can describe the comprehensive color characteristics of different structures in the medical image after staining.

[0034] The entire medical examination image is traversed pixel by pixel, and a local neighborhood region is established at each pixel location. The neighborhood is a fixed-size window centered on the current pixel, such as... or The pixel region; for pixels located at the edge of the medical examination image, a portion of their window falls outside the range of the medical examination image. The neighborhood of this pixel is determined by the effective pixels within the window that fall within the range of the medical examination image, and the number of neighborhood pixels is the number of these effective pixels. Alternatively, mirror filling, edge copying, or other equivalent methods can be used to fill in the portion of the window that falls outside the range of the medical examination image before calculation. In this case, the number of neighborhood pixels is the total number of pixels within the window. The above edge processing methods are all well-known in the art and will not be elaborated here. After the neighborhood is established, the intensity values ​​of all pixels in the neighborhood on the three color channels are statistically averaged to obtain the neighborhood color mean vector.

[0035] After obtaining the neighborhood color mean vector, the difference between each pixel in the neighborhood and the three color channels is calculated, and the differences are statistically analyzed to obtain the degree of color change in the neighborhood, which reflects the degree of dispersion of color distribution within the neighborhood. If the degree of color change in a certain area is small, it indicates that the area may belong to the background or a uniformly stained area. If the degree of color change is large, it indicates that the area may contain target structures with obvious color changes, such as cell boundaries, bacterial outlines, or tissue structures.

[0036] After obtaining the neighborhood color mean vector and the degree of color change, the color vector of the current pixel is normalized to obtain the normalized color field vector. Specifically, the offset of the current pixel's color vector relative to the neighborhood color mean vector is first calculated. The sum of the neighborhood color change degree and the color offset amplitude is then added by 1 to form the normalization factor. Normalizing the offset can suppress the influence of abnormal color brightness while maintaining color change information.

[0037] The formula for calculating the normalized color field vector is:

[0038] ;

[0039] in, Indicates spatial location The normalized color field vector of the pixel at that location, including its spatial position. The normalized value of the pixel in the red channel Normalized value in the green channel and the normalized value in the blue channel ; Indicates spatial location The offset of the color vector of a pixel at a given location relative to the mean color vector of its neighborhood; Indicates spatial location The color mean vector of the pixel neighborhood at that location. , Indicates spatial location The average intensity of the red channel in the pixel neighborhood at a given location is calculated using the following formula: , Indicates spatial location The average intensity of the green channel in the pixel neighborhood at a given location is calculated using the following formula: , Indicates spatial location The average intensity of the blue channel in the pixel neighborhood at a given location is calculated using the following formula: , Indicates the number of neighboring pixels. Indicates spatial location The neighborhood of the pixel at that location; for pixels located at the edge of a medical examination image, The value is determined by the effective pixels within its neighborhood window that fall within the range of the medical examination image. The corresponding number of valid pixels is taken; Represents the spatial coordinates of neighboring pixels. Indicates spatial location The intensity value of the neighboring pixels in the red channel. Indicates spatial location The intensity value of the neighboring pixels in the green channel. Indicates spatial location The intensity value of the neighboring pixels at that location in the blue channel; Indicates spatial location The formula for calculating the degree of color change in the neighborhood of a pixel is:

[0040] , Indicates spatial location The difference in the red channel between the neighboring pixels at a given location. Indicates spatial location The difference between neighboring pixels in the green channel. Indicates spatial location The difference in the blue channel between the neighboring pixels at a given location; Indicates spatial location The formula for calculating the pixel color offset at a given location is: ; This represents the normalization factor.

[0041] By calculating the normalized color field vector, a stable expression of color information in medical examination color images can be achieved. This can effectively reduce the impact of illumination changes and differences in staining intensity on image analysis results, enhance the color contrast features of target structures, and suppress background noise, thereby improving the stability and detection accuracy of subsequent image structure analysis and target detection processes.

[0042] S2. Based on the normalized color field vector, the spatial-color coupled diffusion structure metric is calculated using a spatial-color coupled diffusion structure metric calculation algorithm. After searching for the maximum spatial-color coupled diffusion structure metric in the neighborhood of each pixel, it is coupled with the spatial-color coupled diffusion structure metric of the current pixel to obtain the structure propagation enhancement response value. Based on the structure propagation enhancement response value, an adaptive detection threshold is obtained. The structure propagation enhancement response value is compared with the adaptive detection threshold to generate a target region mask for application to medical examination images.

[0043] In medical examination images, cell boundaries, bacterial outlines, or abnormal tissue regions often exhibit obvious color changes and continuous structures. In order to simultaneously characterize color change features and spatial structure features, a spatial-color coupled diffusion structure metric calculation algorithm is used to calculate the spatial-color coupled diffusion structure metric value.

[0044] The color difference between the current pixel and every neighboring pixel is calculated. Since color information is composed of three components: red, green, and blue, the color difference actually reflects the degree of comprehensive change among the three color components. By uniformly measuring the changes of the three color components, a value representing the intensity of the color difference between two pixels can be obtained. The greater the color difference, the more obvious the color structure change between the two pixels, corresponding to locations such as boundary regions, cell outlines, or tissue structure edges.

[0045] After obtaining the color difference, a spatial distance factor is further introduced. This spatial distance factor describes the square of the geometric distance between the current pixel and its neighboring pixels on the medical examination image plane. It is used to weight the color difference; neighboring pixels with closer geometric distances have a greater impact on the structural features of the current pixel, while neighboring pixels with farther geometric distances have a relatively smaller impact on structural features. Through this spatial distance attenuation mechanism, structural measurements can better highlight local structural changes while avoiding excessive influence from distant pixels on the results.

[0046] For all pixels in the neighborhood, calculate the color difference and spatial distance factor between the current pixel and all pixels in the neighborhood, and then accumulate all the calculation results. To ensure comparability between different neighborhood sizes, the accumulated results are also normalized. The calculation formula is as follows:

[0047] ;

[0048] in, Indicates spatial location The spatial-color coupling diffusion structure metric of the pixel at that location; Indicates spatial location The normalized color field vector of the neighboring pixels at that location; Indicates spatial location Pixel and spatial position The color difference of neighboring pixels at a given location is calculated using the following formula: , Indicates spatial location The normalized value of the pixel at that location in the red channel. Indicates spatial location The normalized value of the pixel at that location in the green channel. Indicates spatial location The normalized value of the pixel in the blue channel at that location. Indicates spatial location The normalized value of the neighboring pixels in the red channel. Indicates spatial location The normalized value of the neighboring pixels in the green channel. Indicates spatial location The normalized value of the neighboring pixels at that location in the blue channel; This represents the spatial distance factor, which is the square of the geometric distance between the current pixel and its neighboring pixels; The cumulative result is represented by the denominator, which consists of the sum of the spatial distance factor and the constant term 1. The constant term 1 is used to avoid the denominator being zero when the neighboring pixel is the current pixel itself and the spatial distance factor is zero, and to limit the weight determined by the spatial distance factor to a limited range.

[0049] In medical examination images, true structures exhibit spatial continuity, while random noise manifests as isolated pixel variations. For each pixel, a search is performed on the maximum spatial-color coupled diffusion structure metric (SCD) among all pixels in its neighborhood. The structure propagation enhancement response value is obtained by coupling the current pixel's SCD with this maximum value. The formula for calculating the structure propagation enhancement response value shows that, given a single SCD for the current pixel, the enhancement value increases with the increase of the maximum SCD among all pixels in its neighborhood. Therefore, when a pixel has a high SCD and significant SCD values ​​also exist among its neighbors, its structure propagation enhancement response value is significantly enhanced compared to pixels with the same SCD but lower maximum SCD values ​​among their neighbors, effectively highlighting the true structural regions. Furthermore,... Although an isolated noise pixel has a color difference from every pixel in its neighborhood and has a high spatial-color coupling diffusion structure metric, all other pixels in its neighborhood are located in a uniformly colored background region. The color difference in each pixel's neighborhood comes only from the isolated noise pixel. The spatial-color coupling diffusion structure metric obtained by averaging the number of neighboring pixels is low. Therefore, the maximum spatial-color coupling diffusion structure metric of the isolated noise pixel's neighborhood is low, and its structure propagation enhancement response value is the same as its own spatial-color coupling diffusion structure metric. Pixels in continuous structure regions are suppressed, thereby reducing the probability of isolated noise being misidentified as the target structure.

[0050] The formula for calculating the structure propagation enhanced response value is:

[0051] ;

[0052] in, Indicates spatial location Structure propagation enhancement response value at the pixel location; Indicates spatial location The maximum spatial-color coupling diffusion structure metric within the pixel neighborhood of a given pixel, excluding that pixel, is calculated using the following formula: , Indicates spatial location neighborhood of the pixel at that location The set consisting of all pixels except the current pixel, meaning that the search range for the maximum value does not include the current pixel itself; Indicates spatial location The spatial-color coupling diffusion structure metric of the neighboring pixels at a given location; the constant term 1 in the denominator of the above formula for calculating the structure propagation enhancement response value is used to avoid... When zero is taken, the denominator is zero, and will be determined by... The determined weights are limited to a finite range.

[0053] By statistically analyzing the entire medical examination image, the overall average level of the structure propagation enhancement response value of all pixels is calculated and used as the adaptive detection threshold. The formula for calculating the adaptive detection threshold is:

[0054] ;

[0055] in, Indicates the adaptive detection threshold; This represents the set of spatial coordinates of all pixels in a medical examination image. This indicates the total number of pixels in a medical examination image; This represents the sum of the structure propagation enhancement response values ​​of all pixels in a medical examination image.

[0056] The structure propagation enhancement response value of each pixel is compared with an adaptive detection threshold to generate a target region mask. The calculation formula for the target region mask is as follows:

[0057] ;

[0058] in, Indicates spatial location The target region mask value is set at a pixel, and the target region mask is a binary image with the same size as the medical examination image; when the structure propagation enhancement response value of a certain pixel is... Greater than or equal to the adaptive detection threshold When the target region mask value is 1, the pixel is identified as part of the target region; that is, pixels whose structure propagation enhancement response value equals the adaptive detection threshold are classified as part of the target region. Less than the adaptive detection threshold When the target area mask value is 0, the pixel is considered to belong to the background area or a non-target area.

[0059] The target region mask is then applied to the medical examination image, which involves multiplying the target region mask by the medical examination image pixel by pixel. The calculation formula is as follows:

[0060] .

[0061] in, Indicates spatial location Target region detection results at the pixel location; target region mask values A pixel with a value of 1 retains its original pixel color vector. Target area mask value Pixels with an intensity of 0 have their intensity values ​​set to 0 in all three color channels. This means that pixels outside the target area mask are not retained, thus obtaining a target area detection result that retains the original color information. Multiplying the target area mask by the medical examination image by pixels is a well-known technique in the field and will not be described in detail here.

[0062] As an illustrative embodiment, a blood cell image acquired through microscopic imaging is used as the processing object. The blood cell image is a desensitized sample image with appropriate authorization, or it is taken from a publicly available medical image dataset. The blood cell image is a three-channel color image, and the intensity value of each color channel is represented by 8-bit quantization after analog-to-digital conversion, with a value range of 0 to 255. The sample is treated with Wright's staining, and there are differences in microscopic illumination conditions and staining intensity between different imaging fields. In this embodiment, the neighborhood... Take the pixel centered at the current pixel Pixel window, number of neighboring pixels The value is set to 25. For pixels located at the edge of a medical examination image, their neighborhood is determined by the number of valid pixels falling within the image's range. The search range for the maximum value during the calculation of the structure propagation enhancement response value is also considered. For this The pixel window contains 24 pixels excluding the current pixel; adaptive detection threshold. The value is determined by the arithmetic mean of the structural propagation enhancement response values ​​of the entire medical examination image.

[0063] To verify the detection effect of this embodiment, the proposed method was compared with a fixed threshold segmentation method and a single color feature detection method on the same set of medical examination images. The fixed threshold segmentation method directly sets a fixed threshold for the intensity value of pixels in each color channel and determines the target region accordingly; the single color feature detection method determines the target region only based on the difference between the pixel color vector and a preset target color; the proposed method determines the target region based on the comparison result of the structure propagation enhancement response value and the adaptive detection threshold. All three methods output target region masks on the same set of medical examination images. Using manually labeled target regions as a benchmark, the detection rate and false detection rate were statistically analyzed using the same criteria. The detection rate is the ratio of the number of pixels correctly identified as target regions to the total number of pixels manually labeled as target regions, and the false detection rate is the ratio of the number of pixels incorrectly identified as target regions to the total number of pixels identified as target regions. In the above comparison, the proposed method has a higher detection rate than the fixed threshold segmentation method and the single color feature detection method, and a lower false detection rate than the two methods mentioned above.

[0064] As can be seen from the calculation process of the aforementioned steps, both the fixed threshold segmentation method and the single color feature detection method make judgments based solely on the color information of the pixel itself. When the staining intensity is uneven or the microscopic illumination conditions change, the color intensity values ​​of the same structure in different regions drift accordingly, and the judgment results also change. Furthermore, isolated staining particles and imaging noise are judged as target regions because their colors differ from the background. This method first eliminates the overall drift of color intensity at the local neighborhood scale through a local self-normalized color field construction algorithm. Then, it couples the spatial-color coupling diffusion structure metric of the current pixel with the largest spatial-color coupling diffusion structure metric of its neighbors, excluding the current pixel. This makes the judgment result depend simultaneously on the color difference of the pixel itself and the structural support within its neighbors. Isolated noise pixels lack structural support from pixels other than the current pixel, and their corresponding structure propagation enhancement response values ​​are lower than those of pixels with the same spatial-color coupling diffusion structure metric that are located within a continuous structural region. This reduces the probability that isolated noise pixels are judged as target regions under the same adaptive detection threshold.

[0065] In summary, a digital image target detection method for medical testing has been developed.

[0066] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments; the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results; in some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0068] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A digital image target detection method for medical examination, characterized in that, Includes the following steps: S1. Acquire medical examination images; Based on the light intensity of the acquired medical examination images in the red, green and blue bands, obtain the intensity values ​​of the pixels in the red, green and blue color channels. The intensity values ​​of the pixels in the three color channels together constitute the pixel color vector, and the normalized color field vector is obtained through the local self-normalization color field construction algorithm. S2. Based on the normalized color field vector, the spatial-color coupled diffusion structure metric is calculated using the spatial-color coupled diffusion structure metric calculation algorithm. After searching for the maximum spatial-color coupled diffusion structure metric in the neighborhood of each pixel, it is coupled with the spatial-color coupled diffusion structure metric of the current pixel to obtain the structure propagation enhancement response value. An adaptive detection threshold is obtained based on the structure propagation enhanced response value. The structure propagation enhanced response value is compared with the adaptive detection threshold to generate a target region mask for application to medical examination images.

2. The digital image target detection method for medical examination according to claim 1, characterized in that, S1 specifically includes: In the local self-normalized color field construction algorithm, a neighborhood is established with the current pixel as the center, and the intensity values ​​of all pixels in the neighborhood on the three color channels are statistically averaged to obtain the neighborhood color mean vector.

3. The method for digital image target detection for medical examination according to claim 2, characterized in that, S1 specifically includes: Based on the intensity values ​​of neighboring pixels in the three color channels and the mean vector of neighboring colors, the difference between neighboring pixels in the three color channels is calculated, and the difference is statistically analyzed to obtain the degree of color change in the neighborhood.

4. The digital image target detection method for medical examination according to claim 3, characterized in that, S1 specifically includes: Based on the pixel color vector and the neighborhood color mean vector, calculate the offset and color offset magnitude of the current pixel's color vector relative to the neighborhood color mean vector; add 1 to the sum of the neighborhood color change and the color offset magnitude to obtain the normalization factor; based on the normalization factor, normalize the offset to obtain the normalized color field vector.

5. The digital image target detection method for medical examination according to claim 1, characterized in that, S2 specifically includes: In the spatial-color coupled diffusion structure metric calculation algorithm, the color difference between the current pixel and its neighboring pixels is calculated based on the normalized color field vector; the spatial distance factor is obtained by calculating the square of the geometric distance based on the spatial coordinates of the current pixel and the spatial coordinates of its neighboring pixels in the medical examination image.

6. The digital image target detection method for medical examination according to claim 5, characterized in that, S2 specifically includes: Based on the spatial distance factor, the color differences are weighted, accumulated, and normalized to obtain the spatial-color coupling diffusion structure metric of the pixel.

7. The digital image target detection method for medical examination according to claim 6, characterized in that, S2 specifically includes: The structure propagation enhancement response value of a pixel is obtained by dividing the product of the spatial-color coupled diffusion structure metric of a pixel and the largest spatial-color coupled diffusion structure metric of any other pixel in the pixel's neighborhood by 1 and adding the largest spatial-color coupled diffusion structure metric of any other pixel in the pixel's neighborhood.

8. A digital image target detection method for medical examination according to claim 7, characterized in that, S2 specifically includes: The adaptive detection threshold is obtained by averaging the structure propagation enhancement response values ​​of all pixels in the medical examination image.