Method, System and Storage Medium for Identifying Breast Tumor Region Based on Ultrasonic Image

Abstract

The present invention discloses a method, system and storage medium for identifying breast tumor regions based on ultrasonic images, which relates to the technical field of tumor region identification, and includes the following steps: obtaining a breast ultrasonic image, preprocessing the breast ultrasonic image to obtain a preprocessed image; dividing the preprocessed image into regions through an image segmentation algorithm to obtain breast regions; integrating the breast regions to obtain homologous regions, extracting features of the homologous regions to obtain the regional features of the homologous regions; analyzing the regional features to determine whether the homologous regions are tumor regions; The present invention is used to solve the problem that the existing tumor region identification technology still has inaccurate division of regions in the image, resulting in easy errors in the identification results of tumor regions.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor region recognition, and specifically to a method, system, and storage medium for breast tumor region recognition based on ultrasonic images. Background Art

[0002] Tumor region recognition technology refers to a technology in the field of medical image analysis for automatically or semi-automatically identifying and demarcating tumor tissue regions in a patient's body in medical images (such as MRI, CT, ultrasound, etc.). This technology usually involves methods in fields such as image processing, pattern recognition, and machine learning, aiming to assist doctors in diagnosing, locating, and tracking tumors.

[0003] Existing tumor region recognition technologies usually analyze the entire breast image, identify the tumor region based on the entire breast analysis, and the extraction of the tumor region usually involves contour extraction. The larger the image, the more blurred the contour division by the contour extraction technology, which will lead to inaccurate region division. Subsequently, incorrect features will be extracted during feature extraction, further resulting in incorrect tumor region recognition results. For example, in the patent application with the publication number: CN111311553A, a method, device, and storage medium for breast tumor recognition based on a region of interest are disclosed. This solution performs edge detection on the region of interest and then identifies it, but the origin of the region of interest is not clear. Therefore, the demarcation of the tumor region also lacks a data basis. Existing tumor region recognition technologies also have the problem of inaccurate division of regions within the image, resulting in easy occurrence of incorrect tumor region recognition results. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems in the existing technology to some extent. By obtaining a breast ultrasound image, preprocessing the breast ultrasound image to obtain a preprocessed image, then dividing the region of the preprocessed image through an image segmentation algorithm to obtain a breast region, integrating the breast regions to obtain a homologous region, extracting region features of the homologous region, and finally analyzing the region features to determine whether the homologous region is a tumor region, so as to solve the problem that existing tumor region recognition technologies have inaccurate division of regions within the image, resulting in easy occurrence of incorrect tumor region recognition results.

[0005] To achieve the above object, in a first aspect, the present application provides a method for breast tumor region recognition based on ultrasonic images, including the following steps:

[0006] Obtain a breast ultrasound image, preprocess the breast ultrasound image to obtain a preprocessed image;

[0007] Divide the region of the preprocessed image through an image segmentation algorithm to obtain a breast region;

[0008] Integrate the breast region to obtain a homologous region, extract features from the homologous region to obtain the regional features of the homologous region;

[0009] Analyze the regional features to determine whether the homologous region is a tumor region.

[0010] Further, obtain a breast ultrasound image, and preprocess the breast ultrasound image to obtain a preprocessed image, including the following sub-steps:

[0011] Obtain a breast ultrasound image;

[0012] Perform grayscale processing on the breast ultrasound image to obtain a breast grayscale image;

[0013] Perform contrast enhancement on the breast grayscale image to obtain a preprocessed image.

[0014] Further, perform region division on the preprocessed image through an image segmentation algorithm to obtain breast regions, including the following sub-steps:

[0015] Perform grid division on the preprocessed image, and perform preliminary processing on the divided image regions to obtain partial breast regions;

[0016] Perform grayscale analysis on the remaining image regions to obtain the remaining breast regions.

[0017] Further, perform grid division on the preprocessed image, and perform preliminary processing on the divided image regions to obtain partial breast regions, including the following sub-steps:

[0018] Obtain the pixel resolution of the preprocessed image, where the pixel resolution includes the number of horizontal pixel points and the number of vertical pixel points in the preprocessed image, represented by symbols n and m respectively;

[0019] Divide n by a first quantity, mark the calculation result as the first grid number, divide m by the first quantity, and mark the calculation result as the second grid number;

[0020] Multiply the first grid number by the second grid number to obtain the number of regions;

[0021] Divide the preprocessed image into the number of image regions, where the image regions are square and each image region contains the number of pixel points obtained by multiplying the first quantity by the first quantity;

[0022] Mark the pixel points in the image region as regional pixel points, and mark the grayscale value of the regional pixel points as the regional pixel value;

[0023] Statistically analyze the regional pixel values in the image region, and mark the proportion of the regional pixel values with the same value among all the regional pixel values in the image region as the grayscale proportion;

[0024] Obtain the maximum value in the gray-scale proportion, mark it as the maximum proportion value, compare the maximum proportion value with the first proportion threshold. If the maximum proportion value is greater than or equal to the first proportion threshold, output a gray-scale filling signal; if the maximum proportion value is less than the first proportion threshold, output a gray-scale analysis signal;

[0025] If a gray-scale filling signal is output, change the region pixel values of all region pixel points in the image region to the region pixel values corresponding to the maximum proportion value, and name the filled image region as the breast region; if a gray-scale analysis signal is output, perform gray-scale analysis on the image region to obtain the breast region.

[0026] Further, performing gray-scale analysis on the remaining image region to obtain the remaining breast region includes the following sub-steps:

[0027] Count the number of different region pixel values, mark it as the same-color number;

[0028] Sort and number the numerical values of the region pixel values existing in the image region in ascending order, represented by the symbol P i where i is a non-zero natural number and i is the serial number of P;

[0029] Using i in P i as the X-axis and the same-color number as the Y-axis to establish a plane rectangular coordinate system, named the color value distribution line graph. Enter P i and the corresponding same-color number into the color value distribution line graph, and mark the coordinate points formed by P i and the corresponding same-color number in the color value distribution line graph as color value distribution points;

[0030] Connect adjacent color value distribution points with a straight line, name the straight line as the inter-point straight line, and name the included angle formed between two adjacent inter-point straight lines as the trend angle, and the trend angle is less than or equal to 180°;

[0031] Calculate the average value of P i mark the calculation result as the average color value, divide the color value distribution line graph into two left and right regions based on the average color value, and name them the first region and the second region respectively. Calculate the sum of the same-color numbers of all color value distribution points in the first region, and name the calculation result the first color number. Calculate the sum of the same-color numbers of all color value distribution points in the second region, and name the calculation result the second color number;

[0032] Compare the first color number with the second color number. If the first color number is less than the second color number, output a first processing signal; if the first color number is equal to the second color number, output a second processing signal; if the first color number is greater than the second color number, output a third processing signal;

[0033] Process the color values of the image region based on the output signal to obtain the breast region.

[0034] Further, processing the color values of the image region based on the output signal to obtain the breast region includes the following sub-steps:

[0035] If the first processing signal is output, mark the second region as the color value analysis area; if the third processing signal is output, mark the first region as the color value analysis area;

[0036] Obtain the trend angle adjacent to the color value analysis area. If the opening of the trend angle is downward, output a color value valid signal; if the opening of the trend angle is upward, output a color value invalid signal;

[0037] If a color value valid signal is output, include the color value distribution points on the trend angle into the color value analysis area and continue to search for the next trend angle for analysis; if a color value invalid signal is output, stop searching;

[0038] Sum the region pixel values of all regions within the color value analysis area and calculate the average value, and mark the calculation result as the filled color value;

[0039] If the second processing signal is output, calculate the average value of the region pixel values of all region pixel points within the image region, and mark the calculation result as the filled color value;

[0040] Change the region pixel values of all region pixel points within the image region to the filled color value, and mark the image region as the breast region.

[0041] Further, integrate the breast regions to obtain homologous regions, and extract features of the homologous regions. Obtaining the region features of the homologous regions includes the following sub-steps:

[0042] Name the figure composed of breast regions as the region figure, and extract the contour of the region figure to obtain different contour regions;

[0043] Mark the regions corresponding to the contour regions in the preprocessed figure as homologous regions, and extract the contour of the homologous regions to obtain the contour to be analyzed;

[0044] Obtain the shape features and texture features of the contour to be analyzed, and the shape features and texture features constitute the region features.

[0045] Further, analyze the region features to determine whether the homologous region is a tumor region, including the following sub-steps:

[0046] Establish a tumor region recognition model, and train the tumor region recognition model through the region features of big data;

[0047] After training is completed, the extracted regional features are input into the tumor region recognition model. If the output result is "yes", the contour to be analyzed within the homologous region is marked as the tumor region; if the output result is "no", the contour to be analyzed within the homologous region is marked as the normal region.

[0048] In a second aspect, the present application provides a breast tumor region recognition system based on ultrasonic images, including a preprocessing module, a region division module, a feature extraction module, and a tumor recognition module; the preprocessing module, the feature extraction module, and the tumor recognition module are respectively connected to the region division module for data connection;

[0049] The preprocessing module is used to obtain breast ultrasonic images, preprocess the breast ultrasonic images, and obtain preprocessed images;

[0050] The region division module is used to divide the preprocessed image through an image segmentation algorithm to obtain breast regions;

[0051] The feature extraction module is used to integrate the breast regions to obtain homologous regions, extract features from the homologous regions, and obtain the regional features of the homologous regions;

[0052] The tumor recognition module is used to analyze the regional features and determine whether the homologous region is a tumor region.

[0053] In a third aspect, the present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, it runs the steps in the above method.

[0054] Advantages of the present invention: By obtaining breast ultrasonic images, preprocessing the breast ultrasonic images to obtain preprocessed images, then performing grid division on the preprocessed images, and performing preliminary processing on the divided image regions to obtain some breast regions, and then performing gray-scale analysis on the remaining image regions to obtain the remaining breast regions. The advantage is that, different from directly performing edge detection on the entire preprocessed image to obtain the contours of multiple regions, first perform grid division and then perform certain processing on the image regions to unify the gray-scale values of the image regions to obtain breast regions. Several breast regions represent the approximate distribution of gray-scale values in the preprocessed image, providing a reliable data basis for subsequent analysis and improving the reliability and effectiveness of tumor region recognition;

[0055] By integrating the breast region, the present invention obtains a homologous region, extracts features from the homologous region to obtain the regional features of the homologous region, and finally analyzes the regional features to determine whether the homologous region is a tumor region. The advantage is that by extracting the contour of the regional image composed of the breast region, since edge detection is performed on several regions rather than pixel points, the number of color values to be processed is greatly reduced, so the extraction of the contour becomes more accurate, and the subsequent feature extraction of the homologous region can extract the regional features closest to the real ones, improving the accuracy and reliability of the tumor region. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a schematic block diagram of the system of the present invention;

[0057] Figure 2 is the preprocessed image of the present invention;

[0058] Figure 3 is the line graph of the color value distribution of the present invention;

[0059] Figure 4 is the contour to be analyzed of the present invention;

[0060] Figure 5 is the flowchart of the steps of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0062] Example 1, please refer to Figure 1 As shown, the present application provides a breast tumor region recognition system based on ultrasonic images, including a preprocessing module, a region division module, a feature extraction module, and a tumor recognition module; the preprocessing module, the feature extraction module, and the tumor recognition module are respectively connected to the region division module for data connection;

[0063] Please refer to Figure 2 As shown, the preprocessing module is used to obtain a breast ultrasonic image and preprocess the breast ultrasonic image to obtain a preprocessed image;

[0064] The preprocessing module is configured with a preprocessing strategy, and the preprocessing strategy includes:

[0065] Obtain a breast ultrasonic image;

[0066] Perform gray-scale processing on the breast ultrasonic image to obtain a breast gray-scale image;

[0067] Perform contrast enhancement on the breast grayscale image to obtain a preprocessed image;

[0068] In practical applications, obtain breast ultrasound images, perform grayscale processing on the breast ultrasound images to obtain breast grayscale images, and then perform contrast enhancement on the breast grayscale images to obtain preprocessed images as Figure 2 shown.

[0069] The region division module is used to divide the preprocessed image into regions through an image segmentation algorithm to obtain breast regions; the region division module includes a preliminary division unit, a secondary division unit, and a final processing unit;

[0070] The preliminary division unit is used to perform grid division on the preprocessed image and perform preliminary processing on the divided image regions to obtain partial breast regions;

[0071] The preliminary division unit is configured with a preliminary division strategy, and the preliminary division strategy includes:

[0072] Obtain the pixel resolution of the preprocessed image, where the pixel resolution includes the number of horizontal pixel points and the number of vertical pixel points in the preprocessed image, represented by the symbols n and m respectively;

[0073] Divide n by the first quantity, mark the calculation result as the first number of grids, divide m by the first quantity, and mark the calculation result as the second number of grids;

[0074] Multiply the first number of grids by the second number of grids to obtain the number of regions;

[0075] Divide the preprocessed image into the number of image regions, where the image regions are square and each image region contains the number of pixels equal to the product of the first quantity multiplied by the first quantity;

[0076] In practical applications, the obtained pixel resolution is 600×350, that is, n = 600, m = 350; the setting of the first quantity is to divide the preprocessed image into several grid regions to enhance the accuracy of subsequent edge detection. The range of the first quantity is usually from 10 to 40. In this embodiment, the first quantity is set to 20. The calculated first number of grids is 30, and the second number of grids is 18. The calculation result is rounded up to an integer; since 350 is not divisible by 20, the size of the image region in the 18th column is 20×10 instead of 20×20. The image regions except the 18th column are all 20×20; the calculated number of regions is 540, and the preprocessed image is divided into 540 grid image regions, where the size of the image region in the 18th column is 20×10, and the sizes of the remaining image regions are all 20×20;

[0077] Mark the pixel points in the image region as region pixel points, and mark the grayscale value of the region pixel points as the region pixel value;

[0078] Statistically analyze the regional pixel values in the image region, and calculate the proportion of the regional pixel values with the same value among all the regional pixel values in the image region, which is marked as the gray-scale proportion.

[0079] Obtain the maximum value in the gray-scale proportion, which is marked as the maximum proportion value. Compare the maximum proportion value with the first proportion threshold. If the maximum proportion value is greater than or equal to the first proportion threshold, output a gray-scale filling signal; if the maximum proportion value is less than the first proportion threshold, output a gray-scale analysis signal.

[0080] If a gray-scale filling signal is output, change all the regional pixel values of all the regional pixel points in the image region to the regional pixel value corresponding to the maximum proportion value, and name the filled image region as the breast region; if a gray-scale analysis signal is output, perform gray-scale analysis on the image region to obtain the breast region.

[0081] In practical applications, take the image region in the first row and first column and the image region in the first row and second column as examples, which are referred to as the first image region and the second image region in this embodiment respectively; in the first image region, there are multiple values for the regional pixel values, such as 255, 238, 216, 168, 242, 253, 155, and 58. Among them, the gray-scale proportion of 255 is the largest proportion and is 0.8. The first proportion threshold is set to 0.5, that is, the regional pixel values of half of the pixel points in the image region are the same. Then, directly fill the image region with the corresponding regional pixel value; by comparison, it is obtained that the maximum proportion is greater than the first proportion threshold, and a gray-scale filling signal is output. Change the regional pixel values of all the pixel points in the first image region to 255 to obtain the first breast region; for the second image region, the regional pixel value corresponding to the largest proportion is still 255, but the largest proportion is 0.27. By comparison, it is obtained that the largest proportion is less than the first proportion threshold, and a gray-scale analysis signal is output, and gray-scale analysis is performed on the second image region.

[0082] The secondary partitioning unit is used to perform gray-scale analysis on the remaining image region to obtain the remaining breast region.

[0083] The secondary partitioning unit is configured with a secondary partitioning strategy, and the secondary partitioning strategy includes:

[0084] Statistically analyze the number of different regional pixel values, which is marked as the number of same colors.

[0085] Sort and number the values of the regional pixel values existing in the image region in ascending order, and represent them by the symbol P i where i is a non-zero natural number and i is the serial number of P.

[0086] Please refer to Figure 3 as shown, take P iIn it, i is the X-axis, and the number of same-color pixels is the Y-axis to establish a plane rectangular coordinate system, named the color value distribution line graph. P i and the corresponding number of same-color pixels are entered into the color value distribution line graph. P i and the corresponding number of same-color pixels are marked as color value distribution points in the color value distribution line graph;

[0087] The adjacent color value distribution points are connected by a straight line. The straight line is named the inter-point straight line, and the included angle formed between two adjacent inter-point straight lines is named the trend included angle. The trend included angle is less than or equal to 180°;

[0088] In practical applications, for the second image region, the number of same-color pixels with a region pixel value of 255 is 108, and the number of same-color pixels with a region pixel value of 0 is 9. The remaining numbers of same-color pixels are omitted in this embodiment. For details, please refer to Figure 3 as shown;

[0089] Calculate the average value of P i . Mark the calculation result as the average color value. Based on the average color value, divide the color value distribution line graph into two regions on the left and right, named the first region and the second region respectively. Calculate the sum of the numbers of same-color pixels of all color value distribution points in the first region, and name the calculation result the first color number. Calculate the sum of the numbers of same-color pixels of all color value distribution points in the second region, and name the calculation result the second color number;

[0090] Compare the first color number with the second color number. If the first color number is less than the second color number, output the first processing signal; if the first color number is equal to the second color number, output the second processing signal; if the first color number is greater than the second color number, output the third processing signal;

[0091] Process the color value of the image region based on the output signal to obtain the breast region;

[0092] In practical applications, as Figure 3 shown, in P i , 1 ≤ i ≤ 20, where P1 to P 20 are 0, 5, 9, 18, 34, 56, 73, 88, 96, 105, 115, 124, 153, 169, 185, 196, 225, 242, 253, and 255 in sequence. Calculate the average value to obtain the average color value of 120 (the calculation result is reserved as an integer); the average color value of 120 is between 115 and 124, that is, between P 11 and P 12 . Based on the average color value, divide the color value distribution line graph into two regions on the left and right, that is, the first region includes P1 to P 11 , and the second region includes P 12 to P 20, the sum is obtained, the first color value is 182, and the second color value is 218. By comparison, the first color value is less than the second color value, and the first processing signal is output;

[0093] The final processing unit is configured with a final processing strategy, and the final processing strategy includes:

[0094] If the first processing signal is output, the second area is marked as the color value analysis area; if the third processing signal is output, the first area is marked as the color value analysis area;

[0095] Obtain the trend angle adjacent to the color value analysis area. If the opening of the trend angle is downward, output the color value valid signal; if the opening of the trend angle is upward, output the color value invalid signal;

[0096] If the color value valid signal is output, include the color value distribution points on the trend angle in the color value analysis area and continue to search for the next trend angle for analysis; if the color value invalid signal is output, stop searching;

[0097] Sum the pixel values of all areas in the color value analysis area and calculate the average value, and mark the calculation result as the filling color value;

[0098] If the second processing signal is output, calculate the average value of the regional pixel values of all regional pixel points in the image area, and mark the calculation result as the filling color value;

[0099] Change the regional pixel values of all regional pixel points in the image area to the filling color value, and mark the image area as the breast area;

[0100] In practical applications, the first processing signal is output, and the second area is marked as the color value analysis area. At this time, the color value analysis area includes P 12 to P 20 . At this time, the trend angle adjacent to the color value analysis area is the trend angle at the color value distribution point corresponding to P 11 . The opening of this trend angle is upward, and the color value invalid signal is output, and the search stops; if the opening is downward, it means that P 11 to P 12 shows a downward trend, P 11 is greater than P 12 . In this case, P 11 and P 12The pixel values of the regions are usually relatively close and have a certain correlation. Therefore, they need to be included in the color value analysis area. The color value analysis area is only used to analyze the part with a strong correlation with the image pixel values within the image area, that is, the overall trend of the color. The part with a weak correlation is ignored to better match the color presented by this image area to the naked eye. Sum up all the region pixel values within the color value analysis area and calculate the average value. The filled color value is obtained as 213. Then, change the region pixel values of all region pixel points within the second image area to 213 to obtain the second breast region.

[0101] The feature extraction module is used to integrate the breast regions to obtain homologous regions, and extract the regional features of the homologous regions.

[0102] The feature extraction module is configured with a feature extraction strategy, and the feature extraction strategy includes:

[0103] Name the figure composed of breast regions as the regional figure, and extract the contour of the regional figure to obtain different contour regions.

[0104] Please refer to Figure 4 As shown, mark the region corresponding to the contour region in the preprocessed figure as the homologous region, and extract the contour of the homologous region to obtain the contour to be analyzed.

[0105] Obtain the shape features and texture features of the contour to be analyzed. The shape features and texture features constitute the regional features.

[0106] In practical applications, the regional image is the preprocessed image after color value filling. At this time, each grid can be regarded as a pixel point. At this time, performing edge detection on the regional image to extract the contour can narrow the detection range; by extracting the contour region, and the homologous region corresponding to the contour region in the preprocessed image is as Figure 4 As shown, extract the contour of the homologous region to obtain the contour to be analyzed as Figure 4 As shown, and then extract the Figure 4 regional features through existing tumor feature extraction technologies. Among them, the shape features include the contour and the area. The contour is Figure 4 , and the area is 7.24mm 2 , and extract the Figure 4 texture features through the gray level co-occurrence matrix technology, and obtain a contrast of 2.1. The texture features are not only the contrast, but there are also other features. However, the focus of this embodiment is on the division of the image area and the extraction of the contour. Since the extraction and analysis of image features both adopt existing technologies, no specific description is given in this embodiment.

[0107] The tumor recognition module is used to analyze the regional features and determine whether the homologous region is a tumor region.

[0108] The tumor recognition module is configured with a tumor recognition strategy, and the tumor recognition strategy includes:

[0109] Establish a tumor region recognition model, and train the tumor region recognition model through the regional features of big data;

[0110] After the training is completed, input the extracted regional features into the tumor region recognition model. If the output result is yes, mark the contour to be analyzed within the homologous region as the tumor region; if the output result is no, mark the contour to be analyzed within the homologous region as the normal region;

[0111] In practical applications, the tumor region recognition model adopts convolutional neural network technology, and trains the tumor region recognition model through the regional features of big data. That is, obtain the breast images with existing diagnostic results through big data, and train the model based on the actual determination results. For the training of the contour, calculate the similarity to analyze whether the input contour is the tumor region, and use the probability calculation method for the area and texture features. For the tumor recognition model, the technologies used are all existing technologies. This embodiment aims to optimize the result of contour extraction, so that the subsequent extracted image features are more accurate, thereby improving the accuracy of tumor recognition. Therefore, the specific training process of the tumor recognition model will not be described in detail; after the training is completed, input the extracted regional features into the tumor region recognition model. If the output result is yes, mark the contour to be analyzed within the homologous region as the tumor region; if the output result is no, mark the contour to be analyzed within the homologous region as the normal region.

[0112] Example 2, please refer to Figure 5 As shown, the present application provides a method for identifying breast tumor regions based on ultrasound images, including the following steps:

[0113] Step S1, obtain a breast ultrasound image, preprocess the breast ultrasound image to obtain a preprocessed image; Step S1 includes the following sub-steps:

[0114] Step S101, obtain a breast ultrasound image;

[0115] Step S102, perform grayscale processing on the breast ultrasound image to obtain a breast grayscale image;

[0116] Step S103, perform contrast enhancement on the breast grayscale image to obtain a preprocessed image;

[0117] Step S2, perform regional division on the preprocessed image through an image segmentation algorithm to obtain breast regions; Step S2 includes the following sub-steps:

[0118] Step S201, perform grid division on the preprocessed image, and perform preliminary processing on the divided image regions to obtain partial breast regions;

[0119] Step S201 includes the following sub-steps:

[0120] Step S2011: Obtain the pixel resolution of the preprocessed image. The pixel resolution includes the number of horizontal pixel points and the number of vertical pixel points in the preprocessed image, which are represented by symbols n and m respectively;

[0121] Step S2012: Divide n by the first quantity, mark the calculation result as the first number of grids, divide m by the first quantity, and mark the calculation result as the second number of grids;

[0122] Step S2013: Multiply the first number of grids by the second number of grids to obtain the number of regions;

[0123] Step S2014: Divide the preprocessed image into the number of image regions obtained. The image regions are square, and each image region contains the number of pixels equal to the product of the first quantity by the first quantity;

[0124] Step S2015: Mark the pixel points in the image region as region pixel points, and mark the gray value of the region pixel points as the region pixel value;

[0125] Step S2016: Statistically analyze the region pixel values in the image region, and calculate the proportion of the region pixel values with the same value among all the region pixel values in the image region, which is marked as the gray proportion;

[0126] Step S2017: Obtain the maximum value in the gray proportion, which is marked as the maximum proportion value. Compare the maximum proportion value with the first proportion threshold. If the maximum proportion value is greater than or equal to the first proportion threshold, output a gray filling signal; if the maximum proportion value is less than the first proportion threshold, output a gray analysis signal;

[0127] Step S2018: If a gray filling signal is output, change the region pixel values of all the region pixel points in the image region to the region pixel value corresponding to the maximum proportion value, and name the filled image region as the breast region; if a gray analysis signal is output, perform gray analysis on the image region to obtain the breast region;

[0128] Step S202: Perform gray analysis on the remaining image regions to obtain the remaining breast regions;

[0129] Step S202 includes the following sub-steps:

[0130] Step S2021: Statistically analyze the number of different region pixel values, which is marked as the number of same colors;

[0131] Step S2022: Sort and number the values of the region pixel values existing in the image region in ascending order, using symbol P iIt is represented that, where i is a non-zero natural number and i is the serial number of P;

[0132] Step S2023: Take i in P i as the X-axis and the same-color number as the Y-axis to establish a plane rectangular coordinate system, named the color value distribution broken line graph. Enter P i and the corresponding same-color number into the color value distribution broken line graph, and mark the coordinate points formed by P i and the corresponding same-color number in the color value distribution broken line graph as color value distribution points;

[0133] Step S2024: Connect adjacent color value distribution points with straight lines, name the straight lines as inter-point straight lines, and name the included angle formed between two adjacent inter-point straight lines as the trend angle, where the trend angle is less than or equal to 180°;

[0134] Step S2025: Calculate the average value of P i , mark the calculation result as the average color value, divide the color value distribution broken line graph into two regions on the left and right based on the average color value, name them the first region and the second region respectively, calculate the sum of the same-color numbers of all color value distribution points in the first region, name the calculation result the first color number, calculate the sum of the same-color numbers of all color value distribution points in the second region, and name the calculation result the second color number;

[0135] Step S2026: Compare the first color number with the second color number. If the first color number is less than the second color number, output the first processing signal; if the first color number is equal to the second color number, output the second processing signal; if the first color number is greater than the second color number, output the third processing signal;

[0136] Step S2027: Process the color value of the image region based on the output signal to obtain the breast region;

[0137] Step S2027 includes the following sub-steps:

[0138] Step S2027.a: If the first processing signal is output, mark the second region as the color value analysis area; if the third processing signal is output, mark the first region as the color value analysis area;

[0139] Step S2027.b: Obtain the trend angle adjacent to the color value analysis area. If the trend angle opens downward, output the color value valid signal; if the trend angle opens upward, output the color value invalid signal;

[0140] Step S2027.c: If the color value valid signal is output, include the color value distribution points on the trend angle into the color value analysis area and continue to find the next trend angle for analysis; if the color value invalid signal is output, stop searching;

[0141] Step S2027.d: Sum the pixel values of all regions within the color value analysis area and calculate the average value, and mark the calculation result as the filling color value;

[0142] Step S2027.e: If the second processing signal is output, calculate the average value of the regional pixel values of all regional pixel points within the image area, and mark the calculation result as the filling color value;

[0143] Step S2027.f: Change the regional pixel values of all regional pixel points within the image area to the filling color value, and mark the image area as the breast area;

[0144] Step S3: Integrate the breast areas to obtain homologous areas, and extract features of the homologous areas to obtain the regional features of the homologous areas; Step S3 includes the following sub-steps:

[0145] Step S301: Name the figure composed of breast areas as the regional figure, extract the contour of the regional figure to obtain different contour areas;

[0146] Step S302: Mark the areas corresponding to the contour areas in the preprocessed figure as homologous areas, extract the contours of the homologous areas to obtain the contours to be analyzed;

[0147] Step S303: Obtain the shape features and texture features of the contours to be analyzed, and the shape features and texture features constitute the regional features;

[0148] Step S4: Analyze the regional features to determine whether the homologous area is a tumor area; Step S4 includes the following sub-steps:

[0149] Step S401: Establish a tumor area recognition model, and train the tumor area recognition model through the regional features of big data;

[0150] Step S402: After training is completed, input the extracted regional features into the tumor area recognition model. If the output result is yes, mark the contour to be analyzed within the homologous area as the tumor area; if the output result is no, mark the contour to be analyzed within the homologous area as the normal area.

[0151] Embodiment 3 provides an electronic device, which may include: a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus. The memory stores computer-readable instructions, and the processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps in the method for identifying breast tumor regions based on ultrasonic images are run to achieve the following functions: obtaining breast ultrasonic images, preprocessing the breast ultrasonic images to obtain preprocessed images; dividing the preprocessed images into regions through an image segmentation algorithm to obtain breast regions; integrating the breast regions to obtain homologous regions, extracting features of the homologous regions to obtain the regional features of the homologous regions; analyzing the regional features to determine whether the homologous regions are tumor regions.

[0152] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical disks and other various media that can store program codes.

[0153] Embodiment 4: This application also provides a computer program product. The computer program product includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the method for identifying breast tumor regions based on ultrasonic images provided by the above-mentioned various methods. The method includes: obtaining breast ultrasonic images, preprocessing the breast ultrasonic images to obtain preprocessed images; dividing the preprocessed images into regions through an image segmentation algorithm to obtain breast regions; integrating the breast regions to obtain homologous regions, extracting features of the homologous regions to obtain the regional features of the homologous regions; analyzing the regional features to determine whether the homologous regions are tumor regions.

[0154] Embodiment 5. The present application further provides a computer-readable storage medium. The present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method for identifying a breast tumor region based on ultrasonic images are run. Through the above technical solution, when the computer program is executed by the processor, the method in any optional implementation manner of the above-mentioned embodiment is executed to achieve the following functions: obtaining a breast ultrasonic image, preprocessing the breast ultrasonic image to obtain a preprocessed image; dividing the preprocessed image into regions through an image segmentation algorithm to obtain a breast region; integrating the breast regions to obtain a homologous region, extracting features of the homologous region to obtain the regional features of the homologous region; analyzing the regional features to determine whether the homologous region is a tumor region.

[0155] Through the description of the above embodiments, the embodiments of the present invention can be provided as a method, a system or a computer program product. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0156] In the embodiments provided in the present application, it should be understood that the disclosed system or method can be implemented in other ways. The above-described embodiments are merely illustrative. For example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some communication interfaces. The indirect coupling or communication connection of the system, module and unit can be in an electrical, mechanical or other form.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying breast tumor regions based on ultrasound images, characterized in that: The steps include: Acquire breast ultrasound images, preprocess the breast ultrasound images, and obtain preprocessed images; The pre-processed image is divided into regions by an image segmentation algorithm to obtain the breast region; Integrate the mammary gland regions to obtain homologous regions, extract features from the homologous regions, and obtain regional features of the homologous regions; Analyze regional features to determine whether the homologous region is a tumor region; The pre-processed image is divided into regions by an image segmentation algorithm to obtain the breast region, which includes the following sub-steps: Divide the preprocessed image into grids, and perform preliminary processing on the divided image areas to obtain partial breast areas; Perform grayscale analysis on the remaining image area to obtain the remaining breast area; Grayscale analysis of the remaining image area to obtain the remaining breast area includes the following sub-steps: Count the number of different area pixel values ​​and mark them as the same color number; The pixel values ​​in the image area are sorted and numbered in ascending order, and the symbol P is used to represent the pixel values ​​in the image area. i Represents, where i is a non-zero natural number and i is the serial number of P; P i The i in the figure is the X-axis, and the color number is the Y-axis to establish a plane rectangular coordinate system, named color value distribution line chart, and P i Enter the corresponding number of the same color into the color value distribution line chart, and put P i The coordinate points formed by the corresponding same color numbers in the color value distribution line graph are marked as color value distribution points; Connect adjacent color value distribution points through straight lines, name the straight lines as point-to-point straight lines, name the angle formed between two adjacent point-to-point straight lines as trend angle, and the trend angle is less than or equal to 180°; Calculate P i The average value of the color value distribution is marked as the average color value. Based on the average color value, the color value distribution line graph is divided into two areas on the left and right, named the first area and the second area respectively. The sum of the same color numbers of all color value distribution points in the first area is calculated, and the calculation result is named the first color number. The sum of the same color numbers of all color value distribution points in the second area is calculated, and the calculation result is named the second color number. Compare the first color number with the second color number, if the first color number is less than the second color number, output a first processing signal; if the first color number is equal to the second color number, output a second processing signal; if the first color number is greater than the second color number, output a third processing signal; The color value of the image area is processed based on the output signal to obtain the breast area.

2. The method for identifying breast tumor regions based on ultrasound images according to claim 1, characterized in that: Acquiring a breast ultrasound image and preprocessing the breast ultrasound image to obtain a preprocessed image includes the following sub-steps: Obtain breast ultrasound images; Grayscale processing is performed on the breast ultrasound image to obtain a breast grayscale image; The mammary gland grayscale image is contrast enhanced to obtain a preprocessed image.

3. The method for identifying breast tumor regions based on ultrasound images according to claim 2, characterized in that: The preprocessed image is gridded and the image area obtained by the division is preliminarily processed to obtain a partial breast area, which includes the following sub-steps: Obtaining a pixel resolution of the preprocessed image, where the pixel resolution includes the number of horizontal pixels and the number of vertical pixels in the preprocessed image, represented by symbols n and m respectively; Divide n by the first number and mark the result as the first grid number; divide m by the first number and mark the result as the second grid number; Multiply the first grid number by the second grid number to get the number of areas; Divide the preprocessed image into a number of image regions, where the image regions are in a positive direction and include a first number of pixels times a first number of pixels in the image regions; Mark the pixels in the image area as regional pixels, and mark the grayscale values ​​of the regional pixels as regional pixel values; Count the regional pixel values ​​in the image area, and count the proportion of regional pixel values ​​with the same value in all regional pixel values ​​in the image area, which is marked as grayscale proportion; Obtain the maximum value of the grayscale proportion, mark it as the maximum proportion value, compare the maximum proportion value with the first proportion threshold, if the maximum proportion value is greater than or equal to the first proportion threshold, output the grayscale filling signal; if the maximum proportion value is less than the first proportion threshold, output the grayscale analysis signal; If a grayscale filling signal is output, the regional pixel values ​​of all regional pixels in the image area are changed to the regional pixel values ​​corresponding to the maximum proportion value, and the filled image area is named the breast area; if a grayscale analysis signal is output, the image area is grayscale analyzed to obtain the breast area.

4. The method for identifying breast tumor regions based on ultrasound images according to claim 3, characterized in that: Processing the color value of the image area based on the output signal to obtain the breast area includes the following sub-steps: If the first processed signal is output, the second area is marked as a color value analysis area; if the third processed signal is output, the first area is marked as a color value analysis area; Obtain the trend angle adjacent to the color value analysis area, and if the trend angle opens downward, output a color value valid signal; If the trend angle opens upward, an invalid color value signal is output; If a valid color value signal is output, the color value distribution points on the trend angle are included in the color value analysis area, and the next trend angle is searched for analysis; If the output color value is invalid, the search stops; Sum up all pixel values ​​in the color value analysis area and calculate the average value, and mark the calculation result as the filling color value; If the second processed signal is output, the average value of the regional pixel values ​​of all regional pixel points in the image area is calculated, and the calculation result is marked as the filling color value; The pixel values ​​of all the pixels in the image area are changed to the fill color value, and the image area is marked as the breast area.

5. The method for identifying breast tumor regions based on ultrasound images according to claim 4, characterized in that: Integrating the mammary gland regions to obtain homologous regions, extracting features from the homologous regions, and obtaining regional features of the homologous regions include the following sub-steps: The graph composed of breast regions is named as a regional graph, and contour extraction is performed on the regional graph to obtain different contour regions; Mark the area corresponding to the contour area in the preprocessed graph as a homologous area, extract the contour of the homologous area, and obtain the contour to be analyzed; The shape features and texture features of the contour to be analyzed are obtained, wherein the shape features and texture features constitute regional features.

6. The method for identifying breast tumor regions based on ultrasound images according to claim 5, characterized in that: Analyzing the regional features and determining whether the homologous region is a tumor region includes the following sub-steps: Establish a tumor region recognition model and train the tumor region recognition model through the regional features of big data; After the training is completed, the extracted regional features are input into the tumor region recognition model. If the output result is yes, the contour to be analyzed in the homologous region is marked as a tumor region; if the output result is no, the contour to be analyzed in the homologous region is marked as a normal region.

7. A breast tumor region recognition system based on ultrasound imaging, used to implement the breast tumor region recognition method based on ultrasound imaging according to any one of claims 1 to 6, characterized in that: It includes a preprocessing module, a region division module, a feature extraction module and a tumor recognition module; the preprocessing module, the feature extraction module and the tumor recognition module are respectively connected with the region division module data; The preprocessing module is used to obtain breast ultrasound images, preprocess the breast ultrasound images, and obtain preprocessed images; The region division module is used to divide the pre-processed image into regions by using an image segmentation algorithm to obtain breast regions; The feature extraction module is used to integrate the mammary gland regions to obtain homologous regions, extract features from the homologous regions, and obtain regional features of the homologous regions; The tumor recognition module is used to analyze regional features and determine whether a homologous region is a tumor region.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 6 are executed.

Images