Method and device for high-throughput quantitative analysis of cell and histological staining images

Abstract

This invention provides a high-throughput quantitative analysis method and apparatus for biomedical cell and histological staining images. The method includes: image preprocessing, identification of tissue / cell regions, identification of stained regions, and calculation of pixel values ​​of the identified regions to output corresponding quantitative values. This invention can meet the needs of high-throughput, batch processing and analysis of stained image results. It can flexibly adjust the image recognition threshold and selection range through various parameters based on differences in staining background color effects and different background colors generated during image acquisition from different batches of experimental samples, to achieve optimal batch recognition results. Furthermore, this invention can effectively identify and ignore background noise in the image, such as specks similar in color to the target analyte. This invention can automatically process large batches of sample images, saving time and effort, and has greater flexibility in parameter adjustment, generating the required quantitative results based on parameter adjustments.

Description

Technical Field

[0001] This invention relates to the quantitative analysis of cellular and histological features and specific staining results in the biomedical field, and particularly to a high-throughput quantitative analysis method and apparatus for cell and histological staining images based on image recognition. Background Technology

[0002] In biomedical research and diagnosis, various staining methods are commonly used to label and identify tissues and cells. Quantitative processing of positively stained areas is a routine and fundamental procedure. These results are widely applied in biological experiments, disease diagnosis, and drug development. For example, among various cell and histological staining methods, Alizarin Red S staining and Von Kossa staining are two commonly used methods for identifying calcium components in tissues and cells. They are widely used in studies of osteoblast differentiation, osteoblast or tissue pathophysiology, such as identification and diagnosis (calcification detection) in orthopedics, dentistry, and cardiovascular disease pathology research. Alizarin Red S is an anthraquinone derivative that can chelate with various metal ions to form stable orange-red or dark red complexes. It is an excellent metal indicator and can also be used as a colorimetric reagent for the determination of metal ions by absorbance, as an acid-base indicator, or for the determination of metal ions by polarography. Von Kossa staining utilizes the metal displacement method to reduce calcium deposited by silver nitrate, allowing observation of calcium or calcium salt deposition. Von Kossa staining reagent is an authoritative and classic technique for analyzing calcium deposition in fixed tissue and cell samples. It is mainly suitable for detecting calcium deposition and calcified nodules in tissues, and is also widely used in the study of osteocytes or other tissue pathophysiology.

[0003] Quantitative analysis of the staining results images is expected to yield effective data including the positive rate, positive area, and staining / intensity of tissue or cell staining (such as Alizarin Red S staining and Von Kossa staining for calcification). The most widely used analytical tool is ImageJ, developed by the National Institutes of Health (NIH). This tool is based on manually defining the stained area to find the optimal threshold for identifying the stained area and the extent of the entire tissue or cell region, thereby calculating quantitative values ​​such as area and staining intensity. The tool suffers from several limitations that restrict its application in high-throughput batch processing and the accuracy of its results: First, its batch image processing capabilities are weak, requiring preprocessing for most data analysis. Preprocessing is limited to single imported images and manual adjustments, such as grayscale conversion and pixel value flipping. Second, the staining range of the samples must be manually selected. This manual selection is time-consuming and labor-intensive, and the threshold values ​​set for each operation can vary due to visual observation, making it impossible to apply the same selection criteria to image recognition for the same batch of samples. Finally, the quantitative data values ​​in the calculation results only cover basic data such as the selected area and grayscale difference; subsequent data still require manual processing, and the quantitative data that can be included in subsequent calculations is limited to the output content and options set by the software.

[0004] It should be noted that the information disclosed in the background section above is only for understanding the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The main objective of this invention is to overcome the deficiencies of the aforementioned background technology and provide a method and apparatus for high-throughput quantitative analysis of cell and histological staining images.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A high-throughput quantitative analysis method for cell and histological staining images includes the following steps:

[0008] (1) Convert the original color image to a grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0 to 1 to 0 to 255;

[0009] (2) Identify tissue and / or cellular regions:

[0010] Based on the set parameter value ROI.thr, the image is converted into a black and white image with a pixel value of 0 or 1;

[0011] Based on the set parameter value fill.thr, the white areas in the black and white image are deepened;

[0012] Completely fill connected targets in the image;

[0013] Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion;

[0014] Based on the set parameter value obj.thr, extract the target whose proportion is within the parameter value range;

[0015] (3) Identify the stained areas:

[0016] Based on the set parameter value stain.thr, within the range of the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1.

[0017] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0018] Furthermore, in step (2), the range of ROI.thr is [0,1], the range of fill.thr is (0,+∞), and the range of obj.thr is [0,1]; in step (3), the range of stain.thr is [0,1].

[0019] Furthermore, in step (2), the range of obj.thr is 0.05 to 0.1, and in step (3), the parameter value of stain.thr for the red staining region of Alizarin Red S is 0.5, and the parameter value for the black staining region of Von Kossa is 0.6 to 0.7.

[0020] Furthermore, the quantitative values ​​in step (4) include one or more of the following: cell / tissue area and intensity, staining area and intensity, and the percentage of staining area and intensity.

[0021] A high-throughput quantitative analysis method for cell and histological staining images includes the following steps:

[0022] (1) Convert the original color image to a grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0 to 1 to 0 to 255;

[0023] (2) Identify tissue and / or cellular regions:

[0024] A three-class iterative method is used for region identification in images;

[0025] Based on the set parameter value fill.thr, the white areas in the black and white image are deepened;

[0026] Completely fill connected targets in the image;

[0027] Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion;

[0028] Based on the set parameter value obj.thr, extract the target whose proportion is within the parameter value range;

[0029] (3) Identify the stained areas:

[0030] Based on the set parameter value stain.thr, within the range of the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1.

[0031] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0032] Furthermore, in the three-class iterative method, the region is divided into three groups according to the within-group variance and the within-group mean is calculated. These mean values ​​are set as the thresholds for the background and foreground, respectively, and the middle group is the region to be determined. The above actions are repeated in the region to be determined until the background and foreground are completely divided.

[0033] A high-throughput quantitative analysis method for cell and histological staining images includes the following steps:

[0034] (1) Convert the original color image to a grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0 to 1 to 0 to 255;

[0035] (2) Identify tissue and / or cellular regions:

[0036] The foreground and background of the existing image are manually selected and labeled as px.fg and px.bg respectively; the data structure of the training set and the validation set are adjusted, where the px.fg and px.bg of the training set are converted into three-channel color pixel matrices fgMat and bgMat respectively, and the foreground is labeled as 1 and the background as 0, and the pixel values ​​of the entire image are selected for the validation set.

[0037] The K-nearest neighbor method was used to fit the data, and the foreground was marked as 1, i.e., the tissue / cell region;

[0038] (3) Identify the stained areas:

[0039] Based on the set parameter value stain.thr, within the range of the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1.

[0040] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0041] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the high-throughput quantitative analysis method for cell and histological staining images.

[0042] A high-throughput quantitative analysis device for cell and histological staining images includes a processor that executes a computer program to implement the high-throughput quantitative analysis method for cell and histological staining images.

[0043] The present invention has the following beneficial effects:

[0044] This invention provides a high-throughput quantitative analysis method and apparatus for histological and cytological staining, especially calcium staining changes, based on image recognition. This invention can meet the needs of high-throughput, batch processing and analysis of staining image results. It can flexibly adjust the threshold and selection range of image recognition through various parameters according to the differences in staining background color effect and different background colors generated during image acquisition of different batches of experimental samples, so as to achieve the optimal batch recognition effect. Moreover, this invention can effectively identify and ignore background noise in the image, such as specks with colors similar to the target object.

[0045] By importing stained images of tissues or cells in batches, machine learning algorithms can be used to perform intelligent image recognition and result analysis on the overall tissue or cell structure and its positively stained areas. Compared with existing methods, this approach can automate the processing of large batches of sample images, saving time and effort, and offers greater flexibility in parameter adjustment, allowing for independent parameter adjustment and generation of the required quantitative results as needed. Attached Figure Description

[0046] Figure 1 This is a flowchart of a high-throughput quantitative analysis method for cell and histological staining images according to an embodiment of the present invention.

[0047] Figure 2 This is a flowchart of a high-throughput quantitative analysis method for cell and histological staining images according to another embodiment of the present invention.

[0048] Figure 3 This is a flowchart of a high-throughput quantitative analysis method for cell and histological staining images according to another embodiment of the present invention.

[0049] Figure 4 This is a system architecture diagram of the software tool for applying the high-throughput quantitative analysis method for cell and histological staining images of the present invention.

[0050] Figure 5 This is a front-end service display page (browser version) for the software tool that applies the high-throughput quantitative analysis method for cell and histological staining images of the present invention.

[0051] Figure 6This is a page for adjusting some parameters of the software tool used in the high-throughput quantitative analysis method for cell and histological staining images of the present invention.

[0052] Figure 7 These are partial analysis results images from the software tool that utilizes the high-throughput quantitative analysis method for cell and histological staining images of this invention.

[0053] Figure 8 This is a comparison of some quantitative values ​​of the high-throughput quantitative analysis method for cell and histological staining images in this embodiment of the invention with the results of manual processing using ImageJ. Detailed Implementation

[0054] The embodiments of the present invention will be described in detail below. It should be emphasized that the following description is merely exemplary and not intended to limit the scope and application of the present invention.

[0055] It should be noted that when a component is referred to as "fixed to" or "set on" another component, it can be directly on or indirectly on that other component. When a component is referred to as "connected to" another component, it can be directly connected to or indirectly connected to that other component. Furthermore, a connection can be used for fixing, coupling, or communication.

[0056] It should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention.

[0057] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0058] See Figure 1 In one embodiment, a method for high-throughput quantitative analysis of biomedical staining based on image recognition is provided, the method comprising the following steps ( Figure 1 Steps S1 to S4 shown:

[0059] (1) Image preprocessing:

[0060] 1.1 Import the original image.

[0061] 1.2 Convert color images to grayscale images.

[0062] 1.3 Flip pixel values,

[0063] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0064] (2) Identify tissue / cell regions:

[0065] 2.1 Set the parameter value ROI.thr to convert the image into a black and white image with a pixel value of 0 or 1.

[0066] 2.2 Set the parameter value fill.thr to darken the white areas in a black and white image.

[0067] 2.3 Completely fill connected targets in the image.

[0068] 2.4 Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion.

[0069] 2.5 Set the parameter value obj.thr to extract targets whose proportion falls within the parameter value range;

[0070] (3) Identify the stained areas:

[0071] 3.1 Set the parameter value stain.thr to threshold the stained area or point within the tissue / cell region identified in the second step, and convert the image into a black and white image with a pixel value of 0 or 1.

[0072] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0073] In a preferred embodiment, in step (2) of the method, the range of parameter ROI.thr is [0,1], the range of fill.thr is (0,+∞), and the range of obj.thr is [0,1]; in step (3), the range of parameter stain.thr is [0,1].

[0074] In a preferred embodiment, the preferred parameter of obj.thr in step (2) of the method is 0.05 to 0.1, and the preferred parameter of stain.thr in step (3) is 0.5 for Alizarin Red S staining and 0.6 to 0.7 for Von Kossa staining.

[0075] In a specific embodiment, steps (1) to (3) of the method can be used to calculate corresponding quantitative values ​​based on the pixel values ​​of the identified area, such as cell / tissue area and intensity, staining area and intensity, staining area and intensity ratio, etc.

[0076] The algorithm in the example can employ unsupervised machine learning.

[0077] This embodiment works best for sample images with clean backgrounds, free of clutter or noise. Users can refer to examples such as... Figure 5 The interface shown allows for batch import of images, and you can also... Figure 6 As shown, first select one or more images to adjust the parameters. Figure 6 The parameters shown are for illustrative purposes only and do not represent all adjustable parameters of the method. By selecting the optimal parameters and threshold, batch image recognition is performed to obtain results such as... Figure 7 and Figure 8 The identification results are saved and the relevant quantitative values ​​are exported.

[0078] See Figure 2 In another embodiment, a method for high-throughput quantitative analysis of biomedical staining based on image recognition, compared with the first embodiment, adopts a three-class thresholding technique for region identification in step (2) 2.1 of the method in this embodiment, while steps 2.2 to 2.5 remain unchanged, and steps (1), (3) and (4) remain unchanged.

[0079] The method specifically includes the following steps ( Figure 2 Steps S1 to S4 shown:

[0080] (1) Image preprocessing:

[0081] 1.1 Import the original image.

[0082] 1.2 Convert color images to grayscale images.

[0083] 1.3 Flip pixel values,

[0084] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0085] (2) Identify tissue / cell regions:

[0086] 2.1 A three-class thresholding technique is used for region identification.

[0087] 2.2 Set the parameter value fill.thr to darken the white areas in a black and white image.

[0088] 2.3 Completely fill connected targets in the image.

[0089] 2.4 Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion.

[0090] 2.5 Set the parameter value obj.thr to extract targets whose proportion falls within the parameter value range;

[0091] (3) Identify the stained areas:

[0092] 3.1 Set the parameter value stain.thr to threshold the stained area or point within the tissue / cell region identified in the second step, and convert the image into a black and white image with a pixel value of 0 or 1.

[0093] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0094] The three-class iterative method, based on the Otsu method, divides the region into three groups according to the within-group variance and calculates the within-group mean, which is then set as the threshold for the background and foreground, respectively. The middle group is the undetermined region. The above process is repeated in the undetermined region until the background and foreground are completely separated. The three-class iterative method performs significantly better than the traditional Otsu method or other algorithms that perform a one-time binary classification of the entire image for images with large overall pixel variance.

[0095] In a preferred embodiment, in step (2) of the method, the range of parameter ROI.thr is [0,1], the range of fill.thr is (0,+∞), and the range of obj.thr is [0,1]; in step (3), the range of parameter stain.thr is [0,1].

[0096] In a preferred embodiment, the preferred parameter of obj.thr in step (2) of the method is 0.05 to 0.1, and the preferred parameter of stain.thr in step (3) is 0.5 for Alizarin Red S staining and 0.6 to 0.7 for Von Kossa staining.

[0097] In a specific embodiment, through steps (1) to (3) of the method, the corresponding quantitative values, such as cell / tissue area and intensity, staining area and intensity, staining area and intensity ratio, etc., can be calculated based on the pixel values ​​of the identification area.

[0098] The algorithm in the example can employ unsupervised machine learning.

[0099] Because the three-class iterative method is less effective for images with large pixel variance (specifically biological images, i.e., cluttered backgrounds or other interfering factors outside the target area), this embodiment performs best for sample images with cluttered backgrounds, inclusions, or noise. Users can also use methods such as... Figure 5 The interface shown allows for batch import of images, and you can also... Figure 6 As shown, first select one or more images to adjust the parameters. Figure 6 The parameters shown are for illustrative purposes only and do not represent all adjustable parameters of the method. By selecting the optimal parameters and threshold, batch image recognition is performed to obtain results such as... Figure 7 and Figure 8 The identification results are saved and the relevant quantitative values ​​are exported.

[0100] See Figure 3 In another embodiment, a method for high-throughput quantitative analysis of biomedical staining based on image recognition, compared with the first embodiment, uses the K-nearest neighbor method for region identification in step (2) of the method in this embodiment, while steps (1), (3), and (4) remain unchanged. The method specifically includes the following steps ( Figure 3 Steps S1 to S4 shown:

[0101] (1) Image preprocessing:

[0102] 1.1 Import the original image.

[0103] 1.2 Convert color images to grayscale images.

[0104] 1.3 Flip pixel values,

[0105] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0106] (2) Identify tissue / cell regions:

[0107] 2.1 Manually select the foreground and background of the existing image, labeling them as px.fg and px.bg respectively.

[0108] 2.2 Adjust the data structure of the training and validation sets. The px.fg and px.bg of the training set are converted into three-channel color pixel matrices fgMat and bgMat, respectively, with foreground values ​​set to 1 and background values ​​to 0. The validation set selects the pixel values ​​of the entire image.

[0109] 2.3 The K-nearest neighbor method was used to fit the data, and the foreground was marked as 1, i.e., the tissue / cell region;

[0110] (3) Identify the stained areas:

[0111] 3.1 Set the parameter value stain.thr to threshold the stained area or point within the tissue / cell region identified in the second step, and convert the image into a black and white image with a pixel value of 0 or 1.

[0112] (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values.

[0113] In a preferred embodiment, the parameter stain.thr in step (3) of the method is preferably 0.5 for Alizarin Red S staining and 0.6 to 0.7 for Von Kossa staining;

[0114] In a specific embodiment, through steps (1) to (3) of the method, the corresponding quantitative values, such as cell / tissue area and intensity, staining area and intensity, staining area and intensity ratio, etc., can be calculated based on the pixel values ​​of the identification area.

[0115] The algorithm in the example can employ unsupervised machine learning.

[0116] The method in this embodiment is more complex in implementation than the previous two embodiments, involving additional model training and fitting processes. However, it can simultaneously identify tissue / cell regions with or without interfering factors. For users, this reduces the need for manual parameter tuning and increases the types of acceptable image samples for batch image processing. Similarly, it can yield results such as... Figure 7 and Figure 8 The identification results are saved and the relevant quantitative values ​​are exported.

[0117] Depend on Figure 8 It can be seen that the results of high-throughput quantitative analysis of biomedical staining based on image recognition provided in this manual and the results of manual processing with ImageJ for the same batch of images are very close (error range not exceeding 5%). This indicates that the method described in this manual can replicate the image analysis capabilities of ImageJ. Furthermore, the tools and methods described can process over a thousand images simultaneously, with each image processing time not exceeding 1 second, while manual processing takes approximately 5-6 minutes per image. Therefore, the tools and methods described in this manual significantly shorten image analysis time and improve work efficiency. Moreover, because manual image processing with ImageJ involves variations in parameter settings due to human factors, it is impossible to apply the same standard to the same batch of images. The method described in this manual, however, can standardize the processing of images within the same batch, resulting in consistent and more reliable quantitative values.

[0118] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the high-throughput quantitative analysis method for cell and histological staining images.

[0119] This invention also provides a high-throughput quantitative analysis device for cell and histological staining images, including a processor, which executes the high-throughput quantitative analysis method for cell and histological staining images when executing a computer program.

[0120] Using machine learning algorithms, it is possible to perform batch intelligent identification of tissues / cells and their positively stained areas. Multiple algorithms are applied to different types of images, and multi-parameter adjustment is used to adapt images of different batches and staining depths to meet the needs of image recognition in different scenarios.

[0121] The following describes specific examples.

[0122] Example 1

[0123] See Figure 1 A method for high-throughput quantitative analysis of biomedical staining based on image recognition, the method comprising the following steps:

[0124] (1) Image preprocessing:

[0125] 1.1 Import the original image.

[0126] 1.2 Convert color images to grayscale images.

[0127] 1.3 Flip pixel values,

[0128] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0129] (2) Identify tissue / cell regions:

[0130] 2.1 Set the initial value of the parameter ROI.thr to 0.012 to convert the image into a black and white image with a pixel value of 0 or 1.

[0131] 2.2 Set the initial value of the parameter fill.thr to 10 to darken the white areas in the black and white image.

[0132] 2.3 Completely fill connected targets in the image.

[0133] 2.4 Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion.

[0134] 2.5 Set the initial value of the parameter value obj.thr to 0.1, and extract targets whose proportion is within the range of the parameter value;

[0135] (3) Identify the stained areas:

[0136] 3.1 Set the parameter value of stain.thr. Within the tissue / cell region identified in the second step, threshold the stained area or point and convert the image into a black and white image with a pixel value of 0 or 1. For Alizarin Red S staining, the initial value of stain.thr is 0.5, and for Von Kossa staining, the initial value of stain.thr is 0.6.

[0137] (4) By calculating the pixel values ​​of the recognition area, output the corresponding quantitative values ​​such as ROI.area, ROI.intensity, stained.area, stained.intensity.

[0138] Example 2

[0139] See Figure 2 A method for high-throughput quantitative analysis of biomedical staining based on image recognition, the method comprising the following steps:

[0140] (1) Image preprocessing:

[0141] 1.1 Import the original image.

[0142] 1.2 Convert color images to grayscale images.

[0143] 1.3 Flip pixel values,

[0144] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0145] (2) Identify tissue / cell regions:

[0146] 2.1 A three-class thresholding technique is adopted.

[0147] Perform region identification.

[0148] 2.2 Set the initial value of the parameter fill.thr to 10 to darken the white areas in the black and white image.

[0149] 2.3 Completely fill connected targets in the image.

[0150] 2.4 Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion.

[0151] 2.5 Set the initial value of the parameter value obj.thr to 0.1, and extract targets whose proportion is within the range of the parameter value;

[0152] (3) Identify the stained areas:

[0153] 3.1 Set the parameter value of stain.thr. Within the tissue / cell region identified in the second step, threshold the stained area or point and convert the image into a black and white image with a pixel value of 0 or 1. For Alizarin Red S staining, the initial value of stain.thr is 0.5, and for Von Kossa staining, the initial value of stain.thr is 0.6.

[0154] (4) By calculating the pixel values ​​of the recognition area, output the corresponding quantitative values ​​such as ROI.area, ROI.intensity, stained.area, stained.intensity.

[0155] Example 3

[0156] See Figure 3 A method for high-throughput quantitative analysis of biomedical staining based on image recognition, the method comprising the following steps:

[0157] (1) Image preprocessing:

[0158] 1.1 Import the original image.

[0159] 1.2 Convert color images to grayscale images.

[0160] 1.3 Flip pixel values,

[0161] 1.4 Generate a new two-dimensional pixel matrix and convert its pixel values ​​from 0 to 1 to 0 to 255;

[0162] (2) Identify tissue / cell regions:

[0163] 2.1 Manually select the foreground and background of the existing image, labeling them as px.fg and px.bg respectively.

[0164] 2.2 Adjust the data structure of the training and validation sets. The px.fg and px.bg of the training set are converted into three-channel color pixel matrices fgMat and bgMat, respectively, with foreground values ​​set to 1 and background values ​​to 0. The validation set selects the pixel values ​​of the entire image.

[0165] 2.3 The K-nearest neighbor method was used to fit the data, and the foreground was marked as 1, i.e., the tissue / cell region;

[0166] (3) Identify the stained areas:

[0167] 3.1 Set the parameter value of stain.thr. Within the tissue / cell region identified in the second step, threshold the stained area or point and convert the image into a black and white image with a pixel value of 0 or 1. For Alizarin Red S staining, the initial value of stain.thr is 0.5, and for Von Kossa staining, the initial value of stain.thr is 0.6.

[0168] (4) By calculating the pixel values ​​of the recognition area, output the corresponding quantitative values ​​such as ROI.area, ROI.intensity, stained.area, stained.intensity.

[0169] To apply the high-throughput quantitative analysis method for cell and histological staining images according to embodiments of the present invention, the following methods can be used: Figure 4 The software tools shown.

[0170] like Figure 4 As shown, the tool is an integrated software system architecture, consisting of a three-layer architecture: infrastructure, backend services, and frontend services. The backend services utilize the Spring Boot framework, and the frontend services utilize the Vue framework. The infrastructure includes three modules: database, disk, and image analysis service. The image analysis service reads images from the disk and saves the analysis results. The database is MySQL, and the algorithms in the image analysis service can be written in R. The backend services include three modules: parameter management service, image management service, and execution analysis service. The parameter management service is used to configure storage in the database. The image management service uses the database to manage image locations and store or delete images on the disk. The execution analysis service is used to call the algorithms in the image analysis service. The frontend services include three modules: a parameter configuration management page, an image management page, and an execution analysis page. The parameter configuration management page uses the parameter management service to add, modify, and delete parameter configurations. The image management page uses the image management service to batch upload images, view analysis results, and download images. The execution analysis page uses the execution analysis service to select images and perform analysis.

[0171] The software tool can be divided into three main modules: configuration management, image management, and execution of analysis commands. Configuration management includes two functions: manual parameter configuration and automatic parameter configuration. Image management includes two functions: source image and analysis result image management. Execution of analysis commands includes two functions: triggering algorithms and saving results.

[0172] Known image recognition or tissue / cell and staining region identification for biomedical samples is typically performed manually, which is subject to significant human influence and cannot effectively handle large batches of such images. This paper addresses this issue by providing a tool and method for high-throughput quantitative analysis of biomedical staining based on image recognition. This tool offers both manual and automatic parameter tuning options, providing users with multiple choices to meet their personalized needs. Output results include a comparison of the original and recognized images, as well as quantitative values ​​such as the area and intensity of relevant regions.

[0173] The system architecture of this software tool can be deployed in the cloud, which can solve the problem of the complex operating environment required for image analysis languages. It also supports multi-user use, makes full use of computing resources, and facilitates the future optimization and updating of algorithms.

[0174] The background section of this invention may include background information about the problems or environment in which the invention is being developed, and is not necessarily a description of prior art. Therefore, the content included in the background section does not constitute an admission of prior art by the applicant.

[0175] The above description provides a further detailed explanation of the present invention in conjunction with specific / preferred embodiments, and it should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various substitutions or modifications can be made to these described embodiments without departing from the concept of the present invention, and all such substitutions or modifications should be considered within the scope of protection of the present invention. In the description of this specification, the reference to terms such as "an embodiment," "some embodiments," "preferred embodiment," "example," "specific example," or "some examples," etc., indicates that the specific features, structures, materials, or characteristics described in connection with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of different embodiments or examples. Although the embodiments of the present invention and their advantages have been described in detail, it should be understood that various changes, substitutions, and modifications can be made herein without departing from the scope of protection of the patent application.

Claims

1. A high-throughput quantitative analysis method for cell and histological staining images, characterized in that, Includes the following steps: (1) Convert the original color image to grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0~1 to 0~255; (2) Identify tissue and / or cellular regions: Based on the set parameter value ROI.thr, the image is converted into a black and white image with a pixel value of 0 or 1; Based on the set parameter value fill.thr, the white areas in the black and white image are deepened; Completely fill connected targets in the image; Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion; Based on the set parameter value obj.thr, extract the target whose proportion is within the parameter value range; Where ROI.thr has a range of [0, 1], and fill.thr has a range of (0, +). The range of obj.thr is [0, 1]. (3) Identify the stained areas: Based on the set parameter value stain.thr, within the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1; where the range of stain.thr is [0,1]; (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values; Therefore, based on the differences in the staining effect and background color generated during the image acquisition process of different batches of experimental samples, the optimal batch recognition effect is achieved, and the background noise of the image is identified and ignored.

2. The high-throughput quantitative analysis method for cell and histological staining images according to claim 1, characterized in that, In step (2), the range of obj.thr is 0.05~0.

1. In step (3), the parameter value of stain.thr for the red staining region of Alizarin Red S is 0.5, and the parameter value for the black staining region of Von Kossa is 0.6~0.

7.

3. The high-throughput quantitative analysis method for cell and histological staining images according to any one of claims 1 to 2, characterized in that, The quantitative values ​​in step (4) include one or more of the following: cell / tissue area and intensity, staining area and intensity, and the percentage of staining area and intensity.

4. A high-throughput quantitative analysis method for cell and histological staining images, characterized in that, Includes the following steps: (1) Convert the original color image to grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0~1 to 0~255; (2) Identify tissue and / or cellular regions: A three-class iterative method is used for region identification in images; Based on the set parameter value fill.thr, the white areas in the black and white image are deepened; Completely fill connected targets in the image; Label the unconnected targets in the image, list the area of ​​each target, and calculate its proportion; Based on the set parameter value obj.thr, extract the target whose proportion is within the parameter value range; The range of fill.thr is (0, +). The range of obj.thr is [0, 1]. (3) Identify the stained areas: Based on the set parameter value stain.thr, within the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1; where the range of stain.thr is [0,1]; (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values; Therefore, based on the differences in the staining effect and background color generated during the image acquisition process of different batches of experimental samples, the optimal batch recognition effect is achieved, and the background noise of the image is identified and ignored.

5. The high-throughput quantitative analysis method for cell and histological staining images according to claim 4, characterized in that, In the three-class iterative method, the region is divided into three groups according to the within-group variance and the within-group mean is calculated. These mean values ​​are set as the thresholds for the background and foreground, respectively, and the middle group is the region to be determined. Repeat the above steps in the undefined area until the background and foreground are completely separated.

6. A high-throughput quantitative analysis method for cell and histological staining images, characterized in that, Includes the following steps: (1) Convert the original color image to grayscale image, flip the pixel values, and generate a new two-dimensional pixel matrix so that the pixel values ​​are changed from 0~1 to 0~255; (2) Identify tissue and / or cellular regions: Manually select the foreground and background of the existing image and label them as px.fg and px.bg respectively; Adjust the data structure of the training set and the validation set. In the training set, px.fg and px.bg are converted into three-channel color pixel matrices fgMat and bgMat, respectively, and the foreground is marked as 1 and the background as 0. The validation set selects the pixel values ​​of the entire image. The K-nearest neighbor method was used to fit the data, and the foreground was marked as 1, i.e., the tissue / cell region; (3) Identify the stained areas: Based on the set parameter value stain.thr, within the tissue / cell region identified in step (2), the stained region or point is thresholded and the image is converted into a black and white image with a pixel value of 0 or 1; where the range of stain.thr is [0,1]; (4) Calculate the pixel values ​​of the recognition area and output the corresponding quantitative values; Therefore, based on the differences in the staining effect and background color generated during the image acquisition process of different batches of experimental samples, the optimal batch recognition effect is achieved, and the background noise of the image is identified and ignored.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the high-throughput quantitative analysis method for cell and histological staining images as described in any one of claims 1 to 6.

8. A high-throughput quantitative analysis device for cell and histological staining images, characterized in that, It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the high-throughput quantitative analysis method for cell and histological staining images according to any one of claims 1 to 6.

Images