Breast cancer pathological image segmentation method

Abstract

The invention discloses a breast cancer pathological image segmentation method, and relates to the technical field of medical image processing. Comprising the steps of inputting a to-be-segmented breast cancer medical image and performing initialization; according to the breast cancer image to be segmented, obtaining a gray level image of the breast cancer image; performing non-local mean filtering operation on the grayscale image, and forming a two-dimensional histogram by the non-local mean filtering image and the grayscale image; based on the constructed two-dimensional histogram, taking a Renyi entropy as a target function, and using an improved backtracking search optimization algorithm as a search method to search an optimal segmentation threshold combination; performing multi-threshold image segmentation on the breast cancer image to be segmented by using the searched optimal threshold combination, and outputting a segmentation result; according to the improved backtracking search optimization algorithm, a random reselection strategy is introduced to help the algorithm to avoid local optimum and effectively improve the exploration capability of the algorithm, and meanwhile, a quality enhancement mechanism is adopted to enhance the quality of candidate solutions and keep certain randomness at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a breast cancer pathology image segmentation method. Background Art

[0002] At present, multi-threshold image segmentation is an important technology in medical image processing, which is widely used in the fields of tumor detection, pathological analysis, etc. The core of multi-threshold image segmentation is to segment the image into different regions by setting multiple thresholds, so as to highlight the characteristics of the lesion area. However, traditional multi-threshold image segmentation methods often rely on manually setting thresholds or using histogram methods for segmentation. These methods often fail to achieve ideal segmentation results when facing complex pathological images.

[0003] In the existing technology, metaheuristic algorithm has gradually become an important technology in multi-threshold image segmentation due to its powerful global search ability and adaptability to various optimization problems. The core of metaheuristic algorithm lies in the balance between exploration and utilization, that is, finding new areas in the search space and optimizing solutions in known areas. Due to its high efficiency and global convergence, it is very effective in practical applications.

[0004] However, existing meta-heuristic algorithms still have some problems in practical applications, such as being prone to falling into local optimal solutions and being sensitive to initial parameters, which leads to insufficient segmentation accuracy and stability. Especially in the segmentation of breast cancer pathology images, the complex tissue structure and diverse lesion morphology further increase the difficulty of segmentation, and traditional algorithms are difficult to meet the requirements of high accuracy and high robustness.

[0005] Therefore, improving the multi-threshold image segmentation method to improve the accuracy and efficiency of segmentation and adapt to the complexity of breast cancer pathology images has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] Based on this, it is necessary to provide a breast cancer pathology image segmentation method to address the above technical issues.

[0007] The present invention adopts the following technical solutions: The present invention provides a breast cancer pathology image segmentation method, comprising: Acquire a breast cancer pathology image to be segmented; and initialize the segmentation threshold; Graying the breast cancer pathology image to obtain a grayscale image; and performing a non-local mean filtering operation based on the grayscale image to obtain a non-local mean image; Through non-local mean filtering of images and grayscale images, a two-dimensional histogram is constructed and Renyi entropy is selected as the fitness objective function; An improved backtracking search optimization algorithm is used to find the optimal segmentation threshold combination, and the optimal threshold combination is used to segment the breast cancer pathology image to be segmented to obtain the final segmentation result; wherein the improved backtracking search optimization algorithm is obtained by introducing a quality enhancement mechanism for improving the quality of candidate solutions and a random reselection strategy for maintaining population diversity into the backtracking search algorithm.

[0008] Preferably, the method of finding the optimal segmentation threshold combination by using an improved backtracking search optimization algorithm comprises the following steps: Initialize the parameters of the improved backtracking search optimization algorithm, including: population size, maximum number of evaluations, and dimension; Calculate the fitness values ​​of all individuals in the population, and obtain the optimal individual and the optimal fitness value; Generate an initial historical population and perform a mutation process to obtain a mutant population; Determine the crossover population and select the population with better fitness by calculating the binary integer mapping value whose size is the product of the population size and the dimension; Introduce a quality enhancement mechanism to determine whether the random number used to generate the initial historical population is less than 0.5. If the random number is less than 0.5, the quality enhancement mechanism is executed; otherwise, the quality enhancement mechanism is skipped. A random reselection strategy is introduced to randomly select the position of new individuals and obtain newly generated solutions; Compare the newly generated solution with the current solution. If it is better than the current solution, update it, and update the optimal individual and optimal fitness value at the same time; Determine whether the current evaluation times are less than the maximum evaluation times. If so, redefine the historical population; otherwise, output the individual in the optimal position as the optimal threshold combination, and output the optimal fitness value as the maximum Renyi entropy; wherein the individual in the optimal position is a multidimensional vector, which includes multiple values, each value corresponds to an optimal threshold, and the multidimensional vector composed of the optimal thresholds corresponding to all values ​​is the optimal threshold combination.

[0009] Preferably, the mutant population is obtained by: Generate initial historical population : ; ; in, is the individual of the initial historical population; , , and Respectively represent Maximum and minimum bounds in each dimension; Used to generate random numbers between 0 and 1; At the beginning of each iteration, we redefine : ; Among them, a and b are Randomly generated internally; Indicates that in the interval Uniform distribution on In the definition After that, randomly change The order of individuals in is as follows: ; in, Indicates an update operation. represents a random shuffle function; By performing mutation operations on the initial historical population, the mutant population is obtained. The formula is as follows: ; in, is the mutation population, search direction matrix Depend on 1 control, , , is a standard normal distribution.

[0010] Preferably, determining the crossover population and selecting a population with better fitness by calculating a binary integer mapping value whose size is the product of the population size and the dimension comprises: The calculated size is Binary integer value mapping : ; ; in, is the dimension after random shuffling; Determine the crossover population individuals: ; in, is a crossover population individual, is the mutant population individual, is the initial historical population individual; Select a population with better fitness. If the fitness value of the updated individual is better than the current global minimum, it will be updated as the optimal individual and the corresponding optimal fitness value will be updated. The formula is as follows: ; in, Represents the fitness function of the population, which is used to calculate the fitness value of the population.

[0011] Preferably, the quality enhancement mechanism comprises: Create a solution , the formula is as follows: ; ; ; ; in, is an integer between [-1,1]; is the optimal individual, are three random individuals ( )’s average value; is a random number in the range of 0 to 1; To control the parameters, () represents the exponential function, is a random number in the range [0,5]. is the current evaluation number, is the maximum number of evaluations; When the solution is met The fitness value is greater than the current solution The fitness value of The fitness value of the solution is not better than the fitness value of the current solution, and the random number Less than the control parameter Create a new solution , the formula is as follows: ; ; ; ; in, is a random number in the range [0,2], used to emphasize the importance of the best solution; is an adaptability factor, To control the parameters, and is a constant with values ​​of 20 and 12 respectively. Generated by the Runge-Kutta method, where Calculated by the following formula: ; ; ; ; ; ; In the formula, All are random numbers between [0,1]; and Represent the optimal solution and the worst solution respectively; , represents the optimal solution the importance of Indicates the increment of position; is the scaling factor, which decreases exponentially.

[0012] Preferably, the control parameter The value of decreases with the increase of the number of evaluations. In the early stage of the improved backtracking search optimization algorithm, the control parameters When the improved backtracking search optimization algorithm performs a global search, the improved backtracking search optimization algorithm controls the parameters in the later stage. When, the improved backtracking search optimization algorithm is locally developed.

[0013] Preferably, the adaptability factor It is used to balance the relationship between exploration and development. In the early iterations of the improved backtracking search optimization algorithm, a larger fitness value is obtained to increase diversity and enhance the exploration search capability. In the later stage of the improved backtracking search optimization algorithm, its value is reduced by increasing the number of evaluations to improve the development search capability of the algorithm.

[0014] Preferably, the random reselection strategy includes: Generate new candidate solutions based on Levy flight search strategy , the formula is as follows: ; ; in, represents the current candidate solution, Indicates the flight step size; , represents normal distribution; is the Levy index; is the Gamma function; Individuals reproduce in a random manner and by probability The new candidate solution found is as follows: ; in, and is a random number that follows a uniform distribution in the range [0,1]. and are two randomly selected candidate solutions.

[0015] At least one of the above technical solutions adopted by the present invention can achieve the following beneficial effects: The present invention introduces a quality enhancement mechanism and a random reselection strategy into the backtracking search algorithm, thereby enhancing the capabilities of the backtracking search algorithm in terms of improving the quality of candidate solutions and maintaining population diversity.

[0016] Specifically, in the segmentation of breast cancer pathology images, an improved backtracking search optimization algorithm is used as the search method, and Renyi entropy is used as the fitness objective function. The optimal segmentation threshold combination is obtained for multi-threshold segmentation, which overcomes the shortcomings of traditional methods in insufficient segmentation accuracy when dealing with complex tissue structures, improves the effect of multi-threshold image segmentation, and has the advantages of high segmentation accuracy, strong algorithm robustness and wide adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 A schematic diagram of a process for segmenting a breast cancer pathology image provided by the present invention; Figure 2 A plan view of a two-dimensional histogram in a method for segmenting a breast cancer pathology image provided by an embodiment of the present invention; Figure 3 The present invention provides a specific execution flow chart of an improved backtracking search optimization algorithm for a breast cancer pathology image segmentation method. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] The technical solutions provided by various embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0020] Figure 1 The present invention is a schematic flow chart of a method for segmenting a breast cancer pathology image, which specifically includes the following steps: S101: Input a breast cancer pathology image to be segmented and initialize the threshold segmentation level.

[0021] S102: graying the input breast cancer image to obtain its grayscale image, and performing a non-local mean filtering operation based on the grayscale image to obtain a non-local mean image, specifically including: The non-local mean filtering method is a technology used for image denoising. Its main purpose is to retain the details of the image as much as possible while removing noise. Its basic idea is to calculate the similarity between the current pixel and other pixels with similar neighborhood structures based on the redundant information in the image, and perform weighted averaging on these similar pixels to obtain the estimated value of the current pixel, including: set up and is the pixel in image I e and f The gray value of image I is as follows: ; ; ; ; ; in, is the pixel in the image through The gray value after denoising, Refers to pixels and The corresponding weight of represents the normalization coefficient, Refers to the control parameters, and is the local mean, Pixel The surrounding area is The pixel block, Pixel The surrounding area is Pixel blocks; If the image Grayscale image of The size is , then the NM image obtained by NM filtering The size is also .

[0022] S103: Based on the non-local mean filter image and the grayscale image, construct a corresponding two-dimensional histogram, see Figure 2 , and select Renyi entropy as the objective function, including: point It is a grayscale image Pixels in Gray value and NM image Pixels in The gray value is generated by Indicate point The number of occurrences, its joint probability density The final two-dimensional histogram is given by the following formula; ; At the same time, let the grayscale image The grayscale levels of the NM image G are Pixel gray level and NM grayscale Constitute a The binary group Then Renyi entropy The expression is: ; ; By maximizing the Renyi total entropy, the optimal threshold is found, and finally the binary The setting of makes the Renyi total entropy maximum. is the required optimal segmentation threshold.

[0023] S104: using the improved backtracking search optimization algorithm to find the optimal segmentation threshold, and using the optimal segmentation threshold to segment the breast cancer pathology image to obtain the final segmentation result; wherein, the specific execution process of the improved backtracking search optimization algorithm is shown in Figure 3 , specifically including: Initialize the parameters of the improved backtracking search optimization algorithm, including: population size, maximum number of evaluations, and dimension; Calculate the fitness values ​​of all individuals in the population, and obtain the optimal individual and the optimal fitness value; Generate an initial historical population and perform the mutation process to obtain a mutant population; Generate the initial historical population: ; ; in, is the individual of the initial historical population; , , and Respectively represent Maximum and minimum bounds in each dimension; Used to generate random numbers between 0 and 1; At the beginning of each iteration, we redefine : ; Among them, a and b are Randomly generated internally; Indicates that in the interval Uniform distribution on .

[0024] In the definition After that, randomly change The order of individuals in is as follows: ; in, Indicates an update operation. represents a random shuffle function; By performing mutation operations on the initial historical population, the mutant population is obtained. The formula is as follows: ; in, is the mutation population, search direction matrix Depend on 1 control, , , is a standard normal distribution.

[0025] Determine the crossover population and select the population with better fitness by calculating the binary integer mapping value whose size is the product of the population size and the dimension, including: The calculated size is Binary integer value mapping : ; ; in, is the dimension after random shuffling; Determine the crossover population individuals, the formula is as follows: ; in, is a crossover population individual, is the mutant population individual, is the initial historical population individual.

[0026] Select a population with better fitness. If the fitness value of the updated individual is better than the current global minimum, it will be updated as the best individual and the corresponding optimal fitness value will be updated. The formula is as follows: ; in, Represents the fitness value of individuals in the population; Determine whether the random number when generating the initial historical population is less than 0.5. If the random number is less than 0.5, the quality enhancement mechanism is executed; otherwise, the quality enhancement mechanism is skipped. Among them, the quality enhancement mechanisms include: Create a solution , the formula is as follows: ; ; ; ; in, is an integer between [-1,1]; is the optimal individual, are three random individuals ( )’s average value; is a random number in the range of 0 to 1; To control the parameters, () represents the exponential function, is a random number in the range [0,5]. is the current evaluation number, is the maximum number of evaluations; When the solution is met The fitness value is greater than the current solution The fitness value of The fitness value of the solution is not better than the fitness value of the current solution, and the random number Less than the control parameter Create a new solution , the formula is as follows: ; ; ; ; in, is a random number in the range [0,2], used to emphasize the importance of the best solution; is an adaptability factor, To control the parameters, and is a constant with values ​​of 20 and 12 respectively. Generated by the Runge-Kutta method, where Calculated by the following formula: ; ; ; ; ; ; in, All are random numbers between [0,1]; and Represent the optimal solution and the worst solution respectively; , represents the optimal solution the importance of Indicates the increment of position; is the scaling factor, which decreases exponentially.

[0027] The random reselection (RR) strategy is introduced to randomly select the position of new individuals, so that the algorithm can avoid the problem of local optimality, including: Generate new candidate solutions based on Levy flight search strategy , the formula is as follows: ; ; in, represents the current candidate solution, Indicates the flight step size; , represents normal distribution; is the Levy index; is the Gamma function; Individuals reproduce in a random manner, with probability The new candidate solution found is as follows: ; in, and is a random number that follows a uniform distribution in the range [0,1]. and are two randomly selected candidate solutions.

[0028] Compare the newly generated solution with the original solution. If it is better than the original solution, update the solution and update the optimal individual and optimal fitness value at the same time. Determine whether the current evaluation times are less than the maximum evaluation times. If so, redefine the historical population; otherwise, output the optimal position and optimal fitness value as the optimal threshold combination and the maximum Renyi entropy. The optimal threshold combination is used to segment breast cancer pathology images.

[0029] The following table shows the use of the Wilcoxon signed rank test at different threshold levels to evaluate the quality of the segmented image using three image quality evaluation indicators, namely Peak Signal to Noise Ratio (PSNR), Structural Similarity Index (SSIM) and Feature Similarity Index (FSIM). These three evaluation indicators are used to evaluate the difference between the segmented image and the original image, measure the structural integrity of the image, and reflect the feature similarity between the original image and the segmented image. In the table, Thresholds represents the threshold level, Mean represents the average of the overall ranking, Rank represents the ranking level, "+" indicates that the performance of REBSA is better than the comparison algorithm, "=" indicates that the performance of REBSA is the same as the comparison algorithm, and "-" indicates that the performance of REBSA is worse than the comparison algorithm. According to the results of the Wilcoxon signed rank test, it can be seen that among all threshold levels, the image segmented by the threshold set obtained using REBSA shows superiority in multiple angles, and the overall average ranking is the first, and it has certain stability and robustness, which can effectively maintain the details and visual quality of the segmented image.

[0030] Among them, the comparison results of PSNR are shown in the following table:

[0031] The comparison results of SSIM are shown in the following table:

[0032] The comparison results of FSIM are shown in the following table:

[0033] In summary, this embodiment introduces a quality enhancement mechanism and a random reselection strategy into the backtracking search algorithm, which respectively enhances the ability of the backtracking search algorithm in improving the quality of candidate solutions and maintaining population diversity. In the segmentation of breast cancer pathological images, the improved backtracking search optimization algorithm is used as the search method, and the best segmentation threshold combination obtained by Renyi entropy as the fitness objective function is used for multi-threshold segmentation, which overcomes the shortcomings of the traditional method in dealing with complex tissue structures. The shortcomings of insufficient segmentation accuracy, improved the effect of multi-threshold image segmentation, have the advantages of high segmentation accuracy, strong algorithm robustness and wide adaptability, and thus provide a powerful scientific guidance for the processing of breast cancer pathological images.

[0034] The technical solution of the present invention is not limited to the above-mentioned specific embodiments. All technical variations made according to the technical solution of the present invention fall within the protection scope of the present invention.

Claims

1. A breast cancer pathology image segmentation method, characterized in that: The following steps are involved: Acquire a breast cancer pathology image to be segmented; and initialize the segmentation threshold; Graying the breast cancer pathology image to obtain a grayscale image; and performing a non-local mean filtering operation based on the grayscale image to obtain a non-local mean image; Through non-local mean filtering of images and grayscale images, a two-dimensional histogram is constructed and Renyi entropy is selected as the fitness objective function; An improved backtracking search optimization algorithm is used to find the optimal segmentation threshold combination, and the optimal threshold combination is used to segment the breast cancer pathology image to be segmented to obtain the final segmentation result; wherein the improved backtracking search optimization algorithm is obtained by introducing a quality enhancement mechanism for improving the quality of candidate solutions and a random reselection strategy for maintaining population diversity into the backtracking search algorithm.

2. A breast cancer pathology image segmentation method as claimed in claim 1, characterized in that: The improved backtracking search optimization algorithm is used to find the optimal segmentation threshold combination, including the following steps: Initialize the parameters of the improved backtracking search optimization algorithm, including: population size, maximum number of evaluations, and dimension; Calculate the fitness values ​​of all individuals in the population, and obtain the optimal individual and the optimal fitness value; Generate an initial historical population and perform a mutation process to obtain a mutant population; Determine the crossover population and select the population with better fitness by calculating the binary integer mapping value whose size is the product of the population size and the dimension; Introduce a quality enhancement mechanism to determine whether the random number used to generate the initial historical population is less than 0.

5. If the random number is less than 0.5, the quality enhancement mechanism is executed; otherwise, the quality enhancement mechanism is skipped. Introduce a random reselection strategy to randomly select the position of a new individual and obtain a newly generated solution; Compare the newly generated solution with the current solution. If it is better than the current solution, update it, and update the optimal individual and optimal fitness value at the same time; Determine whether the current evaluation times are less than the maximum evaluation times. If so, redefine the historical population; otherwise, output the individual in the optimal position as the optimal threshold combination, and output the optimal fitness value as the maximum Renyi entropy; wherein the individual in the optimal position is a multidimensional vector, which includes multiple values, each value corresponds to an optimal threshold, and the multidimensional vector composed of the optimal thresholds corresponding to all values ​​is the optimal threshold combination.

3. A breast cancer pathology image segmentation method as claimed in claim 2, characterized in that: The acquisition of the mutant population comprises: Generate initial historical population : ; ; in, is the individual of the initial historical population; , , and Respectively represent Maximum and minimum bounds in each dimension; Used to generate random numbers between 0 and 1; At the beginning of each iteration, we redefine : ； Among them, a and b are Randomly generated internally; Indicates that in the interval Uniform distribution on In the definition After that, randomly change The order of individuals in is as follows: ; in, Indicates an update operation. represents a random shuffle function; By performing mutation operations on the initial historical population, the mutant population is obtained. The formula is as follows: ； in, is the mutation population, search direction matrix Depend on 1 control, , , is a standard normal distribution.

4. A breast cancer pathology image segmentation method as claimed in claim 2, characterized in that: The method of determining a crossover population and selecting a population with better fitness by calculating a binary integer mapping value whose size is the product of the population size and the dimension comprises: The calculated size is Binary integer value mapping : ； ； in, is the dimension after random shuffling; Determine the crossover population individuals: ； in, is a crossover population individual, is the mutant population individual, is the initial historical population individual; Select a population with better fitness. If the fitness value of the updated individual is better than the current global minimum, it will be updated as the optimal individual and the corresponding optimal fitness value will be updated. The formula is as follows: ； in, Represents the fitness function of the population, which is used to calculate the fitness value of the population.

5. A breast cancer pathology image segmentation method as claimed in claim 1 or 2, characterized in that: The quality enhancement mechanism includes: Create a solution , the formula is as follows: ; ； ； ； in, is an integer between [-1,1]; is the optimal individual, are three random individuals ( )’s average value; is a random number in the range of 0 to 1; To control the parameters, () represents the exponential function, is a random number in the range [0,5]. is the current evaluation number, is the maximum number of evaluations; When the solution is met The fitness value is greater than the current solution The fitness value of The fitness value of the solution is not better than the fitness value of the current solution, and the random number Less than the control parameter Create a new solution , the formula is as follows: ； ； ； ； in, is a random number in the range [0,2], used to emphasize the importance of the best solution; is an adaptability factor, To control the parameters, and is a constant with values ​​of 20 and 12 respectively. Generated by the Runge-Kutta method, where Calculated by the following formula: ； ； ； ； ； ； In the formula, All are random numbers between [0,1]; and Represent the optimal solution and the worst solution respectively; , represents the optimal solution the importance of Indicates the increment of position; is the scaling factor, which decreases exponentially.

6. A breast cancer pathology image segmentation method as claimed in claim 5, characterized in that: The control parameters The value of decreases with the increase of the number of evaluations. In the early stage of the improved backtracking search optimization algorithm, the control parameters When the improved backtracking search optimization algorithm performs a global search, the improved backtracking search optimization algorithm controls the parameters in the later stage. When, the improved backtracking search optimization algorithm is locally developed.

7. A breast cancer pathology image segmentation method as claimed in claim 5, characterized in that: The adaptability factor It is used to balance the relationship between exploration and development. In the early iterations of the improved backtracking search optimization algorithm, a larger fitness value is obtained to increase diversity and enhance the exploration search capability. In the later stage of the improved backtracking search optimization algorithm, its value is reduced by increasing the number of evaluations to improve the development search capability of the algorithm.

8. A breast cancer pathology image segmentation method as claimed in claim 1 or 2, characterized in that: The random reselection strategy includes: Generate new candidate solutions based on Levy flight search strategy , the formula is as follows: ； ； in, represents the current candidate solution, Indicates the flight step size; , represents normal distribution; is the Levy index; is the Gamma function; Individuals reproduce in a random manner and by probability The new candidate solution found is as follows: ； in, and is a random number that follows a uniform distribution in the range [0,1]. and are two randomly selected candidate solutions.

Images