Classification method, system and computer-readable storage medium for tumor tissue slice medical images

Abstract

The present invention provides a classification method, system and computer-readable storage medium for tumor tissue slice medical images, and relates to the field of image processing technology. The present invention obtains tumor tissue slice images, grays the images, then performs Fourier transform, filtering and inverse Fourier transform to obtain a second classification image, extracts texture features of the second classification image, performs threshold processing on the second classification image to obtain a third classification image, extracts cell features of the third classification image; analyzes cell features and texture features to obtain cell reproduction coefficients and cell differentiation coefficients, and generates tumor tissue recognition coefficients based on them, compares the tumor tissue recognition coefficients with preset thresholds, and divides tumor cells into #imgabs0#, #imgabs1# and #imgabs2# levels according to the comparison results. The present invention quantifies the pathological severity of tumor cells, improves the accuracy and efficiency of tumor tissue slice medical image classification, and provides a scientific basis for clinical diagnosis and treatment.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a classification method and system for tumor tissue slice medical images and a computer-readable storage medium. Background Art

[0002] The degree of differentiation and reproductive capacity of tumor cells are important characteristics for identifying tumor cells. These two aspects not only affect the biological behavior of tumors, but are also directly related to the prognosis and treatment strategies of patients. First, the degree of differentiation of tumor cells refers to the degree of similarity between tumor cells and normal cells in morphology and function. Highly differentiated tumor cells usually show morphology and function that are closer to normal cells, grow relatively slowly, and respond better to treatment. In contrast, poorly differentiated tumor cells show obvious abnormal characteristics, often grow rapidly and are more invasive, and a decrease in the degree of differentiation is often associated with an increase in malignancy. Secondly, tumor cells usually have abnormal proliferation characteristics that can break through the normal cell proliferation control mechanism, leading to rapid growth of tumors. The assessment of this reproductive capacity involves not only the rate of cell division, but also the regulatory mechanism of the cell cycle and the interaction between cells.

[0003] Traditionally, the inference of tumor cells is generally made by professionals through manual inference and judgment. However, with the development of image recognition technology, images can be analyzed and understood by computers, and objects and features in images can be automatically identified, classified and understood. Therefore, the differentiation and reproduction characteristics of tumor cells can also be analyzed and diagnosed through image recognition technology. This can quickly process large amounts of image data and provide rapid results. In contrast, manual inference requires more time and human resources. Therefore, the use of image recognition to infer tumor cells can effectively improve the efficiency of tumor cell identification and inference and shorten the time.

[0004] In the prior art, the publication number CN112861826A discloses a cancer cell recognition method, which belongs to the field of medical devices, including collecting images, analyzing images and outputting detection results, wherein the image analysis includes noise reduction processing, color space transformation, color adjustment, image segmentation using the K-means method, feature extraction of images by convolution analysis, geometric analysis of images by calculating rotation and calculating size transformation, and recognition of images using vector machines; the prior art provides a cancer cell recognition system and method, which is reasonably priced and has good detection effect, and can realize automatic analysis of pathological images to achieve the purpose of early detection of tumor cells. However, the prior art still has shortcomings, using convolution for feature extraction and vector machines for pathological recognition. This feature extraction and pathological recognition use a black box model. Although the combination of CNN and SVM has achieved good results in feature extraction and classification, it is difficult to understand how the model makes specific decisions due to its complex structure and parameter settings. Especially in the medical field, doctors usually need to understand the decision-making basis of the model in order to provide more reasonable clinical advice.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a classification method, system and computer-readable storage medium for tumor tissue slice medical images to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A classification method for tumor tissue slice medical images, the specific steps comprising:

[0009] Step 1: Obtain a tumor tissue slice image, first perform grayscale processing on the tumor tissue slice image to obtain a first classification image, perform Fourier transform on the first classification image to filter out the low-frequency part, and then use inverse Fourier transform to convert the image frequency domain into spatial domain to obtain a second classification image;

[0010] Step 2: Use the gray level co-occurrence matrix to extract texture features from the second classification image to obtain tumor cell texture features; perform threshold processing on the second classification image to obtain a third classification image, perform feature extraction on the third classification image to obtain cell features, wherein the texture features include energy and contrast, and the cell features include cell nuclear density, cell nuclear spacing, and cell nuclear circularity.

[0011] Step 3: Perform mathematical analysis on the texture features to obtain the cell regularity coefficient, build a cell differentiation model based on the cell regularity coefficient and the circularity of the cell nucleus, and generate the cell differentiation coefficient;

[0012] Step 4: Dedimensionalize the cell nuclear density and the cell nuclear distance to obtain the cell proliferation coefficient, and construct a tumor cell recognition model based on the cell proliferation coefficient and the cell differentiation coefficient to generate a tumor tissue recognition coefficient;

[0013] Step 5: Compare the tumor tissue recognition coefficient with the preset threshold value, and divide the tumor cells into grade 1, grade 21, and grade 311 according to the comparison result.

[0014] Further,

[0015] The specific logic for obtaining the second classification image is as follows: the coordinates of the pixel points of the image are marked, the bottom pixel row is the first row, the leftmost pixel column is the first column, the first classification image is Fourier transformed, the low-frequency part is filtered out, and then the inverse Fourier transform is used to convert the image frequency domain into the spatial domain to obtain the second classification image; the specific formula for Fourier transform is:

[0016]

[0017] Among them, F(u,v) is the complex number representation with the coordinates in the frequency domain (u,v), f(x,y) is the gray value with the coordinates in the spatial domain (x,y), M is the total number of horizontal pixels, and N is the total number of vertical pixels; x and y are the pixel positions in the spatial domain, and the leftmost and bottommost pixels are zero points; u is the frequency in the frequency domain; v is the intensity in the frequency domain;

[0018] The specific formula for filtering out the low-frequency part is:

[0019]

[0020] Where H(u,v) is the complex number representation after filtering with the frequency domain coordinates (u,v), and D0 is the cutoff frequency;

[0021] The specific formula on which the inverse Fourier transform is based is:

[0022]

[0023] Furthermore, the specific logic for generating the cell features is: counting the number of cell nuclei and the number of pixels of the third recognition image, and dividing the number of cell nuclei by the number of pixels to obtain the cell nucleus density, and the specific formula is:

[0024]

[0025] Among them, ρ is the cell nucleus density, a is the number of cell nuclei, and A is the number of pixels;

[0026] The average distance between all adjacent cell nuclei is calculated to obtain the cell and distance; the specific formula is:

[0027]

[0028] Where L is the distance between nuclei, l i is the shortest distance between adjacent cells in the i-th group, Ax is the number of adjacent cell groups; the shortest distance between the cells is the number of pixel points in the shortest path between two cell nuclei;

[0029] Count the number of pixels occupied by the cell nucleus area and the cell nucleus edge area, and divide the cell nucleus edge area into

[0030] The specific formula for obtaining the circularity of the nucleus is:

[0031]

[0032] Wherein, P is the circularity of the cell nucleus, O is the area of ​​the cell nucleus, and K is the perimeter of the cell nucleus. The cell nucleus area is the number of pixels occupied by the cell nucleus area, and the cell nucleus perimeter is the number of pixels occupied by the edge area of ​​the cell nucleus.

[0033] Furthermore, the specific logic for generating the cell differentiation coefficient is as follows: mathematically analyzing the texture features to obtain the cell regularity coefficient, constructing a cell differentiation model based on the cell regularity coefficient and the circularity of the cell nucleus, and generating the cell differentiation coefficient; the specific formula for obtaining the cell regularity coefficient is as follows:

[0034] H=Co*AS

[0035] H is the cell regularity coefficient, Co is the texture contrast, and AS is the texture energy;

[0036] The specific logic for generating cell rule coefficients is:

[0037]

[0038] Among them, Fh is the cell differentiation coefficient and P is the roundness of the cell nucleus.

[0039] Furthermore, the specific logic for generating the tumor tissue recognition coefficient is as follows: the cell nuclear density and the cell nuclear distance are dedimensionalized to obtain the cell reproduction coefficient, a tumor tissue recognition model is constructed according to the cell reproduction coefficient and the cell differentiation coefficient, and the tumor tissue recognition coefficient is generated; the specific formula for generating the cell reproduction coefficient is as follows:

[0040]

[0041] Wherein, Fz is the cell proliferation coefficient, ρ is the nucleus density, and L is the nucleus spacing;

[0042] The specific formula for generating the tumor tissue recognition coefficient is:

[0043]

[0044] Wherein, Sv is the tumor tissue recognition coefficient, Fz is the cell proliferation coefficient, and Fh is the cell differentiation coefficient.

[0045] Further, the tumor tissue is divided into grade Ι, grade ΙΙ, and grade ΙΙΙ, and the specific logic is as follows: if Sv < 0.4Sv0, the tumor tissue is grade Ι tumor tissue; if 0.4Sv0 ≤ Sv < Sv0, the tumor tissue is grade ΙΙ tumor tissue; if Sv ≥ Sv0, the tumor tissue is grade ΙΙΙ tumor tissue;

[0046] Wherein, Sv is the tumor tissue recognition coefficient, and Sv0 is the tumor tissue recognition threshold.

[0047] The present invention further provides a classification system for medical images of tumor tissue sections. The system is used to execute the classification method for medical images of tumor tissue sections, including:

[0048] An acquisition and processing module, configured to obtain a tumor tissue section image, first perform grayscale processing on the tumor tissue section image to obtain a first classification image, perform Fourier transform on the first classification image, filter out the low-frequency part therein, and then use inverse Fourier transform to convert the image frequency domain into the spatial domain to obtain a second classification image;

[0049] A feature extraction module, configured to use a gray-level co-occurrence matrix to extract texture features from the second classification image to obtain tumor cell texture features; perform threshold processing on the second classification image to obtain a third classification image, and extract features from the third classification image to obtain cell features. The texture features include energy and contrast, and the cell features include nucleus density, nucleus spacing, and nucleus roundness.

[0050] An analysis and operation module, configured to perform mathematical analysis on the texture features to obtain a cell regularity coefficient, construct a cell differentiation model based on the cell regularity coefficient and the nucleus roundness, and generate a cell differentiation coefficient.

[0051] A comprehensive processing module, configured to perform dimensionless processing on the nucleus density and nucleus spacing to obtain a cell proliferation coefficient, construct a tumor cell recognition model based on the cell proliferation coefficient and the cell differentiation coefficient, and generate a tumor tissue recognition coefficient.

[0052] The recognition and classification module is used to compare the tumor tissue recognition coefficient with a preset threshold value, and classify the tumor cells into grade 1, grade 21 and grade 311 according to the comparison result.

[0053] The present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program executable by a processor, and when the computer program is executed by the processor, the above-mentioned classification method of tumor tissue slice medical images can be implemented.

[0054] Compared with the prior art, the present invention has the following beneficial effects: the present invention extracts the texture features and cell features of tumor cells, which can reflect the differentiation and reproduction characteristics of cells, and generates cell differentiation coefficients and cell reproduction coefficients that can reflect the degree of cell differentiation and reproduction ability through these features, providing an important basis for the identification of tumor cells;

[0055] The present invention obtains a tumor tissue recognition coefficient that can comprehensively reflect the state of tumor cells based on the above-mentioned cell differentiation coefficient and cell reproduction coefficient, quantifies the pathological severity of tumor cells, and classifies tumor cells by analyzing the size of the tumor tissue recognition coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0057] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0058] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0059] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0060] Example:

[0061] See also Figure 1 , the present invention provides a technical solution:

[0062] A classification method for tumor tissue slice medical images, the specific steps comprising:

[0063] Step 1: Obtain a tumor tissue slice image, first perform grayscale processing on the tumor tissue slice image to obtain a first classification image, perform Fourier transform on the first classification image to filter out the low-frequency part, and then use inverse Fourier transform to convert the image frequency domain into spatial domain to obtain a second classification image;

[0064] Grayscale processing is the process of converting a color image into a grayscale image, making the values ​​of the three components R (red), G (green), and B (blue) equal. The specific formula is:

[0065] Y=0.298*R+0.587*G+0.114*B

[0066] Wherein, Y is the gray value of the pixel in the first recognition image, and R, G, and B are the red component value, green component value, and blue component value of the pixel of the leaf image after image scaling, respectively.

[0067] The specific logic for obtaining the second classification image is as follows: the coordinates of the pixel points of the image are marked, the bottom pixel row is the first row, the leftmost pixel column is the first column, the first classification image is Fourier transformed, the low-frequency part is filtered out, and then the inverse Fourier transform is used to convert the image frequency domain into the spatial domain to obtain the second classification image; the specific formula for Fourier transform is:

[0068]

[0069] Among them, F(u,v) is the complex number representation with the coordinates in the frequency domain (u,v), f(x,y) is the gray value with the coordinates in the spatial domain (x,y), M is the total number of horizontal pixels, and N is the total number of vertical pixels; x and y are the pixel positions in the spatial domain, and the leftmost and bottommost pixels are zero points; u is the frequency in the frequency domain; v is the intensity in the frequency domain;

[0070] The specific formula for filtering out the low-frequency part is:

[0071]

[0072] Where H(u,v) is the complex number representation after filtering with the frequency domain coordinates (u,v), and D0 is the cutoff frequency;

[0073] The specific formula on which the inverse Fourier transform is based is:

[0074]

[0075] After Fourier transform, the high-frequency part of the image can reflect the contour and edge information of the image. The smooth part of the image is filtered using a high-pass filter, and only the high-frequency part that symbolizes the contour and edge of the image is retained. The image is converted from the frequency domain to the spatial domain using an inverse Fourier transform. At this time, the image only retains the edge and contour information that is useful for subsequent analysis, reducing unnecessary interference for subsequent calculations and improving the accuracy of subsequent feature extraction.

[0076] Step 2: Use the gray level co-occurrence matrix to extract texture features from the second classification image to obtain tumor cell texture features; perform threshold processing on the second classification image to obtain a third classification image, perform feature extraction on the third classification image to obtain cell features, wherein the texture features include energy and contrast, and the cell features include cell nuclear density, cell nuclear spacing, and cell nuclear circularity.

[0077] Create the gray level co-occurrence matrix of the second classification image, and calculate the energy and contrast of the gray level co-occurrence matrix. The specific formula is as follows:

[0078]

[0079] AS is the grayscale co-occurrence matrix energy, Co is the grayscale co-occurrence matrix contrast; the grayscale co-occurrence matrix energy AS reflects the repeatability of the texture pattern of tumor tissue, and the larger its value, the higher the frequency of repeated texture pattern; the grayscale co-occurrence matrix contrast Co reflects the uniformity of tumor tissue texture, and the larger its value, the more uniform the distribution of tumor tissue texture; usually the texture pattern of cells is highly repeatable and evenly distributed, while the texture pattern and texture distribution of tumor cells tend to be more chaotic, so studying the texture characteristics of tumor cells can provide an important basis for the classification of tumor cells.

[0080] Compared with cell texture, the grayscale values ​​of cell nuclei and cell boundaries are larger. By performing threshold segmentation on the grayscale values, the cell nucleus boundary and the cell boundary can be partially segmented. The specific logic is to set a grayscale threshold, traverse all pixel points of the second classification image, retain the grayscale values ​​of pixels with grayscale values ​​greater than the grayscale threshold, and set the grayscale values ​​of pixels with grayscale values ​​less than the grayscale threshold to zero.

[0081] The specific logic for generating the cell features is: counting the number of cell nuclei and the number of pixels in the third recognition image and dividing the number of cell nuclei by the number of pixels can obtain the cell nucleus density. The specific formula is:

[0082]

[0083] Among them, ρ is the cell nucleus density, a is the number of cell nuclei, and A is the number of pixels; the cell nucleus density ρ reflects the number of cell nuclei in an area. The larger the value, the more cell nuclei there are, and the more cell nuclei there are, the stronger the cell reproduction ability. Extracting this feature can provide an important basis for evaluating the reproduction ability of tumor cells.

[0084] The average distance between all adjacent cell nuclei is calculated to obtain the cell and distance; the specific formula is:

[0085]

[0086] Where L is the distance between nuclei, l i is the shortest distance between adjacent cells in the i-th group, and Ax is the number of adjacent cell groups; the shortest distance between the cells is the number of pixel points in the shortest path between two cell nuclei; the cell nucleus distance L reflects the distance between adjacent cells, and the larger its value, the larger the distance between adjacent cells; tumor cells have a strong reproductive ability, the stronger the reproductive ability, the more tumor cells there are, and the more tumor cells there are, the smaller the distance between cells; extracting this feature can provide an important basis for evaluating the reproductive ability of tumor cells.

[0087] Count the number of pixels occupied by the cell nucleus area and the cell nucleus edge area, and divide the cell nucleus edge area into

[0088] The specific formula for obtaining the circularity of the nucleus is:

[0089]

[0090] Among them, P is the circularity of the cell nucleus, O is the area of ​​the cell nucleus, and K is the circumference of the cell nucleus. The cell nucleus area is the number of pixels occupied by the cell nucleus area, and the cell nucleus circumference is the number of pixels occupied by the edge area of ​​the cell nucleus. The circularity of the cell nucleus P reflects the regularity of the cell nucleus. The larger its value is, the more regular the shape of the cell nucleus is, and the closer it is to a perfect circle. Generally, the higher the severity of tumor cell pathology, the higher the degree of irregularity of the cell nucleus, and the lower the circularity of the cell nucleus.

[0091] Step 3: Perform mathematical analysis on the texture features to obtain the cell regularity coefficient, build a cell differentiation model based on the cell regularity coefficient and the circularity of the cell nucleus, and generate the cell differentiation coefficient;

[0092] The specific logic for generating the cell differentiation coefficient is as follows: mathematically analyze the texture features to obtain the cell regularity coefficient, construct a cell differentiation model based on the cell regularity coefficient and the circularity of the cell nucleus, and generate the cell differentiation coefficient; the specific formula for obtaining the cell regularity coefficient is as follows:

[0093] H=Co*AS

[0094] H is the cell regularity coefficient, Co is the texture contrast, and AS is the texture energy; the cell regularity coefficient H reflects the regularity of the texture distribution on the cell surface. The larger its value, the more regular the texture distribution on the cell surface; the grayscale co-occurrence matrix energy AS reflects the repeatability of the texture pattern of tumor tissue. The larger its value, the higher the frequency of repeated texture pattern; the grayscale co-occurrence matrix contrast Co reflects the uniformity of the texture of tumor tissue. The larger its value, the more uniform the texture distribution of tumor tissue.

[0095] The specific logic for generating cell rule coefficients is:

[0096]

[0097] Among them, Fh is the cell differentiation coefficient, and P is the nucleus circularity. The cell differentiation coefficient Fh reflects the degree of cell differentiation. The larger its value, the higher the degree of cell differentiation. The nucleus circularity P reflects the regularity of the nucleus. The larger its value, the more regular the shape of the nucleus, and the closer it is to a perfect circle. The cell regularity coefficient H reflects the regularity of the texture distribution on the cell surface. The larger its value, the more regular the texture distribution on the cell surface. This formula constructs the cell differentiation coefficient Fh that reflects the degree of cell differentiation. Cell differentiation is an important feature of tumor cells. The generation of the cell differentiation coefficient Fh can provide an important basis for the subsequent classification of tumor cells.

[0098] Step 4: Dedimensionalize the cell nuclear density and the cell nuclear distance to obtain the cell proliferation coefficient, and construct a tumor cell recognition model based on the cell proliferation coefficient and the cell differentiation coefficient to generate a tumor tissue recognition coefficient;

[0099] The specific logic for generating the tumor tissue recognition coefficient is as follows: the cell nuclear density and the cell nuclear distance are dedimensionalized to obtain the cell reproduction coefficient, and a tumor tissue recognition model is constructed based on the cell reproduction coefficient and the cell differentiation coefficient to generate the tumor tissue recognition coefficient. The specific formula for generating the cell reproduction coefficient is as follows:

[0100]

[0101] Among them, Fz is the cell reproduction coefficient, ρ is the cell nuclear density, and L is the cell nuclear distance; the cell reproduction coefficient Fz reflects the cell's reproductive ability. The larger its value, the stronger the cell reproduction ability. The cell nuclear density ρ reflects the number of cell nuclei in an area. The larger its value, the more cell nuclei there are, and the more the number, the stronger the cell reproduction ability. The cell nuclear distance L reflects the distance between adjacent cells. The larger its value, the larger the distance between adjacent cells.

[0102] The specific formula for generating the tumor tissue recognition coefficient is:

[0103]

[0104] Among them, Sv is the tumor tissue recognition coefficient, Fz is the cell proliferation coefficient, and Fh is the cell differentiation coefficient. The tumor tissue recognition coefficient Sv reflects the pathological severity of tumor cells; the larger its value, the higher the pathological severity of tumor cells; the cell proliferation coefficient Fz reflects the proliferation ability of cells, and the larger its value, the stronger the cell proliferation ability. The cell differentiation coefficient Fh reflects the degree of cell differentiation, and the larger its value, the higher the degree of cell differentiation. The differentiation and proliferation abilities of tumor cells are important bases for judging the pathological degree of tumor cells. This formula generates a tumor tissue recognition coefficient that can reflect the pathological severity of tumor cells through the cell proliferation system and the cell differentiation coefficient, providing an important basis for the subsequent classification of tumor cells.

[0105] Step 5: Compare the tumor tissue recognition coefficient with a preset threshold, and according to the comparison result, classify the tumor cells into grade Ι, grade ΙΙ, and grade ΙΙΙ.

[0106] The specific logic for classifying the tumor tissue into grade Ι, grade ΙΙ, and grade ΙΙΙ is as follows: if Sv < 0.4Sv0, the tumor tissue is grade Ι tumor tissue; if 0.4Sv0 ≤ Sv < Sv0, the tumor tissue is grade ΙΙ tumor tissue; if Sv ≥ Sv0, the tumor tissue is grade ΙΙΙ tumor tissue.

[0107] Please refer to Figure 2 , the present invention further provides a classification system for medical images of tumor tissue sections. The system is used to implement the classification method for medical images of tumor tissue sections, and includes:

[0108] An acquisition and processing module, configured to obtain a tumor tissue section image, first perform grayscale processing on the tumor tissue section image to obtain a first classification image, perform Fourier transform on the first classification image, filter out the low-frequency part therein, and then use inverse Fourier transform to convert the image frequency domain into the spatial domain to obtain a second classification image;

[0109] A feature extraction module, configured to extract texture features of tumor cells by using a gray-level co-occurrence matrix for the second recognition image; perform threshold processing on the second classification image to obtain a third classification image, and extract features from the third classification image to obtain cell features. The texture features include energy and contrast, and the cell features include nucleus density, nucleus spacing, and nucleus roundness.

[0110] An analysis and operation module, configured to perform mathematical analysis on the texture features to obtain a cell regularity coefficient, construct a cell differentiation model based on the cell regularity coefficient and the nucleus roundness, and generate a cell differentiation coefficient.

[0111] A comprehensive processing module is used to perform dimensionless processing on the cell nuclear density and the cell nuclear distance to obtain the cell proliferation coefficient, build a tumor cell recognition model based on the cell proliferation coefficient and the cell differentiation coefficient, and generate a tumor tissue recognition coefficient;

[0112] The recognition and classification module is used to compare the tumor tissue recognition coefficient with a preset threshold value, and classify the tumor cells into grade 1, grade 21 and grade 311 according to the comparison result.

[0113] The present invention further provides a computer-readable storage medium, characterized in that: the storage medium internally stores a computer program that can be executed by a processor, and when the computer program is executed by the processor, it can implement the classification method of tumor tissue slice medical images.

[0114] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0115] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0116] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for classifying medical images of tumor tissue slices, characterized in that: The specific steps include: Step 1: Obtain a tumor tissue section image. First, perform grayscale processing on the tumor tissue section image to obtain a first classified image. Then, perform Fourier transform on the first classified image, filter out the low-frequency part, and use inverse Fourier transform to convert the image frequency domain into the spatial domain to obtain a second classified image. Step 2: Use the gray-level co-occurrence matrix to extract texture features from the second classified image to obtain tumor cell texture features. Perform threshold processing on the second classified image to obtain a third classified image, and extract features from the third classified image to obtain cell features. The texture features include energy and contrast, and the cell features include nuclear density, nuclear spacing, and nuclear roundness. Step 3: Perform mathematical analysis on the texture features to obtain a cell regularity coefficient. Construct a cell differentiation model based on the cell regularity coefficient and nuclear roundness, and generate a cell differentiation coefficient. Step 4: Perform dimensionless processing on the nuclear density and nuclear spacing to obtain a cell proliferation coefficient. Construct a tumor cell recognition model based on the cell proliferation coefficient and cell differentiation coefficient, and generate a tumor tissue recognition coefficient. Step 5: Compare the tumor tissue recognition coefficient with a preset threshold. According to the comparison result, classify the tumor cells into grade Ι, grade ΙΙ, and grade ΙΙΙ. Perform mathematical analysis on the texture features to obtain a cell regularity coefficient. Construct a cell differentiation model based on the cell regularity coefficient and nuclear roundness, and generate a cell differentiation coefficient. The specific formula for obtaining the cell regularity coefficient is: H = Co * AS H is the cell regularity coefficient, Co is the texture contrast, and AS is the texture energy. The specific logic for generating the cell regularity coefficient is: where Fh is the cell differentiation coefficient and P is the nuclear roundness. The specific logic for generating the tumor tissue recognition coefficient is: Perform dimensionless processing on the nuclear density and nuclear spacing to obtain a cell proliferation coefficient. Construct a tumor tissue recognition model based on the cell proliferation coefficient and cell differentiation coefficient, and generate a tumor tissue recognition coefficient. The specific formula for obtaining the cell proliferation coefficient is: where Fz is the cell proliferation coefficient, ρ is the nuclear density, and L is the nuclear spacing. The specific formula for generating the tumor tissue recognition coefficient is: where Sv is the tumor tissue recognition coefficient, Fz is the cell proliferation coefficient, and Fh is the cell differentiation coefficient. The specific logic for classifying the tumor tissue into grade Ι, grade ΙΙ, and grade ΙΙΙ is: If Sv < 0.4Sv0, the tumor tissue is grade Ι tumor tissue. If 0.4Sv0 ≤ Sv < Sv0, the tumor tissue is grade ΙΙ tumor tissue. If Sv ≥ Sv0, the tumor tissue is grade ΙΙΙ tumor tissue. where Sv is the tumor tissue recognition coefficient and Sv0 is the tumor tissue recognition threshold.

2. The method for classifying tumor tissue slice medical images according to claim 1, characterized in that: The specific logic for obtaining the second classified image is: Label the coordinates of the pixel points of the image. The bottommost pixel row is the first row, and the leftmost pixel column is the first column. Perform Fourier transform on the first classified image, filter out the low-frequency part, and use inverse Fourier transform to convert the image frequency domain into the spatial domain to obtain the second classified image. The specific formula for performing Fourier transform is: Among them, F(u,v) is the complex number representation with the coordinates in the frequency domain (u,v), f(x,y) is the gray value with the coordinates in the spatial domain (x,y), M is the total number of horizontal pixels, and N is the total number of vertical pixels; x and y are the pixel positions in the spatial domain, and the leftmost and bottommost pixels are zero points; u is the frequency in the frequency domain; v is the intensity in the frequency domain; The specific formula for filtering out the low-frequency part is: Where H(u,v) is the complex number representation after filtering with the frequency domain coordinates (u,v), and D0 is the cutoff frequency; The specific formula on which the inverse Fourier transform is based is:

3. The method for classifying tumor tissue slice medical images according to claim 1, characterized in that: The specific logic for generating the cell features is: counting the number of cell nuclei and the number of pixels in the third recognition image and dividing the number of cell nuclei by the number of pixels can obtain the cell nucleus density. The specific formula is: Among them, ρ is the cell nucleus density, a is the number of cell nuclei, and A is the number of pixels; The average distance between all adjacent cell nuclei is calculated to obtain the cell nucleus distance; the specific formula is: Where L is the distance between nuclei, l i is the shortest distance between adjacent cells in the i-th group, Ax is the number of adjacent cell groups; the shortest distance between the cells is the number of pixel points in the shortest path between two cell nuclei; Count the number of pixels occupied by the cell nucleus area and the cell nucleus edge area, and divide the cell nucleus edge area; The specific formula for obtaining the circularity of the nucleus is: Wherein, P is the circularity of the cell nucleus, O is the area of ​​the cell nucleus, and K is the circumference of the cell nucleus. The cell nucleus area is the number of pixels occupied by the cell nucleus area, and the cell nucleus circumference is the number of pixels occupied by the edge area of ​​the cell nucleus.

4. A classification system for tumor tissue slice medical images, characterized by: The system is used to implement the classification method of tumor tissue slice medical images according to any one of claims 1 to 3, comprising: An acquisition and processing module is used to obtain tumor tissue slice images, first grayscale the tumor tissue slice images to obtain a first classification image, perform Fourier transform on the first classification image to filter out the low-frequency part, and then use inverse Fourier transform to convert the image frequency domain into spatial domain to obtain a second classification image; The feature extraction module is used to extract texture features from the second classification image using a gray level co-occurrence matrix to obtain texture features of tumor cells; perform threshold processing on the second classification image to obtain a third classification image, and perform feature extraction on the third classification image to obtain cell features, wherein the texture features include energy and contrast, and the cell features include cell nuclear density, cell nuclear spacing, and cell nuclear circularity. The analysis and operation module is used to perform mathematical analysis on the texture features to obtain the cell regularity coefficient, construct a cell differentiation model based on the cell regularity coefficient and the circularity of the cell nucleus, and generate the cell differentiation coefficient; A comprehensive processing module is used to perform dimensionless processing on the cell nuclear density and the cell nuclear distance to obtain the cell proliferation coefficient, build a tumor cell recognition model based on the cell proliferation coefficient and the cell differentiation coefficient, and generate a tumor tissue recognition coefficient; The recognition and classification module is used to compare the tumor tissue recognition coefficient with a preset threshold value, and classify the tumor cells into grade 1, grade 21 and grade 311 according to the comparison result.

5. A computer-readable storage medium, characterized in that: The storage medium stores a computer program executable by a processor, and when the computer program is executed by the processor, the method for classifying tumor tissue slice medical images according to any one of claims 1 to 3 can be implemented.

Images