CN110910325B - Medical image processing method and device based on artificial butterfly optimization algorithm - Google Patents

Medical image processing method and device based on artificial butterfly optimization algorithm Download PDF

Info

Publication number
CN110910325B
CN110910325B CN201911153403.9A CN201911153403A CN110910325B CN 110910325 B CN110910325 B CN 110910325B CN 201911153403 A CN201911153403 A CN 201911153403A CN 110910325 B CN110910325 B CN 110910325B
Authority
CN
China
Prior art keywords
image
butterfly
vector
value
noise
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911153403.9A
Other languages
Chinese (zh)
Other versions
CN110910325A (en
Inventor
马超
徐守祥
于成龙
谭旭
蔡圳杰
黄蓉
湛邵斌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Information Technology
Original Assignee
Shenzhen Institute of Information Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Information Technology filed Critical Shenzhen Institute of Information Technology
Priority to CN201911153403.9A priority Critical patent/CN110910325B/en
Publication of CN110910325A publication Critical patent/CN110910325A/en
Application granted granted Critical
Publication of CN110910325B publication Critical patent/CN110910325B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10132—Ultrasound image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G06T2207/20032—Median filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a butterfly based on an artificial butterflyA medical image processing method and device of a butterfly optimization algorithm are disclosed, wherein the method comprises the following steps: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i, j); constructing a kernel function based on a multi-kernel local discriminant analysis (MKLFDA) method, and extracting the noiseless image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form; using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j); and inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed. According to the scheme of the invention, the accuracy of ultrasonic image identification can be improved, and the efficiency of ultrasonic image classification and image identification can be improved.

Description

Medical image processing method and device based on artificial butterfly optimization algorithm
Technical Field
The invention relates to the technical field of medical images, in particular to a medical image processing method and device based on an artificial butterfly optimization algorithm.
Background
In recent years, medical imaging technology is rapidly developed, and ultrasonic imaging is an important branch of medical imaging, plays an important role in quantitative analysis, real-time diagnosis, surgical planning and other aspects, and can provide a basis for analysis and diagnosis of medical workers.
However, the speckle noise occurrence rate of the ultrasound image is higher than that of CT and MRI, and the accuracy of ultrasound image identification is affected. In order to accurately identify the region of interest in the ultrasound image and improve the identification accuracy, the object in the ultrasound image needs to be read for multiple times. Furthermore, in order to identify an interested region in an ultrasound image, a large number of features need to be extracted from the ultrasound image, and although the extraction of a large number of features can improve the accuracy of ultrasound image identification, irrelevant and redundant features are still extracted, which affects the efficiency of ultrasound image classification and image identification.
Disclosure of Invention
In order to solve the technical problems, the invention provides a medical image processing method and device based on an artificial butterfly optimization algorithm, which are used for solving the technical problems that in the prior art, the noise of an ultrasonic image is high, objects in the ultrasonic image are read for many times, a large number of irrelevant and redundant features are extracted when the image is identified, and the efficiency of ultrasonic image classification and image identification is influenced.
According to a first aspect of the present invention, there is provided a medical image processing method based on an artificial butterfly optimization algorithm, including:
step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i,j)；
Step S102: constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, and extracting a noise-free image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j);
step S104: and inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed.
Further, the method can be used for preparing a novel materialIn step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i, j) comprising:
firstly, arranging all pixel points in a window according to a pixel value sequence, and then replacing the pixel points with noise by using a central pixel value; then, the median of the window is calculated, and the calculation formula of the intermediate filter for calculating the median is as follows:
Figure BDA0002284177510000021
for a given image I (I, j), (r, s) ∈ (- (W-1)/2, …, (W-1)/2), (I, j) ∈ (1,2, …, H) × (1,2, …, L), H and L respectively denote the width and height of the image, W is the odd value of the window, (3,5, …), W is a set of coordinates in a rectangular sub-image window, centered at point (x, y), replacing all the central pixel values in the window with the calculated median value;
removing noise from the rest windows of the medical image I (I, j) by median filtering to obtain a noise-free image IF(i,j)。
Further, the step S102: constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, and extracting a noise-free image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form, wherein the method comprises the following steps:
step S1021: for the noiseless image IF(i, j) marking the category to which the noiseless image belongs, and obtaining the following marks:
Figure BDA0002284177510000031
wherein x isiFor the ith image sample feature, ciIs the ith category to which the image belongs,cr is a real number set for the classification category number;
step S1022: constructing the noiseless image IF(i, j) inter-class adjacency graph WbAnd intra-class adjacency graph Ww：
Figure BDA0002284177510000032
Figure BDA0002284177510000033
Where l ∈ (1,2, …, c) denotes the class to which the image belongs, and W is a similarity matrix defined as:
Figure BDA0002284177510000034
wherein N isr(x) Represents the nearest neighbor of x;
step S1023: constructing a kernel function, and initializing the coefficient of the kernel function;
constructing kernel functions
Figure BDA0002284177510000035
Namely, it is
Figure BDA0002284177510000036
Wherein the content of the first and second substances,
Figure BDA0002284177510000037
as a function of the distance measure between the data in the p-th image, xi,pFor each image xiCorresponding description feature, xj,pFor each image xjCorresponding descriptive feature, σpIs a positive constant value;
sample coefficient vector alpha, alpha of initialization kernel functionT＝1；
Step S1024: determining a sample coefficient vector α of the kernel function, comprising:
computing
Figure BDA0002284177510000041
And
Figure BDA0002284177510000042
Figure BDA0002284177510000043
Figure BDA0002284177510000044
wherein the content of the first and second substances,
Figure BDA0002284177510000045
Figure BDA0002284177510000046
W(w)is a local weighting matrix between classes, W(b)Is a local weighting matrix within the class, K(i)For the basic kernel function, T represents the matrix transpose, W is the similarity matrix, and the calculation formula is as follows:
Figure BDA0002284177510000047
wherein N isr(x) R nearest neighbors of x, and t is a set constant value;
solving for the generalized features yields a sample coefficient vector α that is appropriate for the noise-free image:
setting an objective function: maximization
Figure BDA0002284177510000048
Finding an optimal conversion matrix by maximizing the local inter-class divergence in the embedded space and minimizing the local intra-class divergence, and converting the objective function into a function with a generalized vector decomposition method:
Figure BDA0002284177510000049
wherein, tau1≤τ2≤...≤τn'Is a minimum of n characteristic values, αiIs a characteristic value tauiCorresponding feature vector constituting the ith column vector of alpha, thereby determining the sample coefficient vector alpha of the noiseless image, i.e. the vector alpha is a column vector, each value of which is represented by alphaiRepresents;
step S1025: determining a weight vector w of a basic kernel of the kernel function, including;
computing
Figure BDA00022841775100000410
And
Figure BDA00022841775100000411
Figure BDA00022841775100000412
Figure BDA0002284177510000051
solving the non-convex quadratic constraint quadratic programming problem to obtain a weight vector w of a basic kernel:
the objective function is:
Figure BDA0002284177510000052
i.e. to solve for
Figure BDA0002284177510000053
and omega is not less than 0;
Figure BDA0002284177510000054
Figure BDA0002284177510000055
step S1026: representing the noiseless image I in a two-dimensional matrix form by utilizing a determined sample coefficient vector alpha of the noiseless image and a weight vector w of a basic kernelF(i,j)；
Step S1027: and reducing the dimension of the two-dimensional matrix into a one-dimensional vector form.
Further, the step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j), wherein the extracting important features comprise:
step S1031: acquiring the noiseless image IF(i, j) in the form of one-dimensional vectors, constructing a solution vector and a fitness function, and setting the number n of individuals in the cluster and the maximum iteration number KmaxSetting the initial value of the current iteration times k as 0, setting the solution vector as an initialization expression, and randomly initializing the position solution vector of each butterfly in the cluster;
the initialization expression is as follows:
X＝rand*(ub-lb)+lb
wherein ub and lb are the upper and lower boundaries of the search space, respectively, and rand is a random number between (0, 1);
the chaotic sequence is designed to improve the capability of butterfly initialization random search, and the expression of the chaotic sequence is as follows:
ci+1＝μci(1-ci)i∈{1,2,...,N}
wherein mu is a chaotic factor, c1Is a random number between (0,1), and c1≠{0.25，0.5，0.75，1}；
Mapping the chaotic sequence generated by the chaotic sequence to the initialized formula so as to generate a corresponding new population X':
X'＝lb+ci(ub-lb)；
the fitness function is expressed as follows:
Figure BDA0002284177510000061
wherein ACC is classification accuracy, N denotes a total number of features, nsuset denotes a selected feature subset, r and s denote weight coefficients, respectively, and r + s is 1;
step S1032: calculating the fitness of all butterflies in the new population, updating the fitness of the butterflies with the fitness value higher than a preset threshold value to the fitness value, and updating the positions of the butterflies according to the following formula:
Figure BDA0002284177510000062
wherein
Figure BDA0002284177510000063
Is the position vector of the ith butterfly for the t-th iteration, t is the number of iterations, step is the flight distance, rand () is the random number generated between (0,1), xkIs randomly selected and is different from butterfly xiLb is butterfly xiLower boundary value of flight range, Ub is butterfly xiAn upper boundary value of the flight range;
Figure BDA0002284177510000064
the flight distance step is a linear decrease with increasing number of iterations;
setting perturbation operator omega xiFor the maximum fitness value butterfly xmax，
Figure BDA0002284177510000065
Where ω is an adaptively derived quantity that decreases linearly with increasing number of iterations; x is the number ofiExpressing the chaos value by the formula
xi+1＝μxi(1-xi)，μ∈[0,4]
Wherein x isiIs the ith position generated randomly;
step S1033: setting a first iteration time threshold value xi, and if the maximum fitness value is not changed between the iteration time k-xi and the current iteration time k, entering a step S1034; otherwise, go to step S1035;
step S1034: updating the position of the butterfly with the maximum fitness value;
namely, the position of the butterfly is updated according to the following formula:
Figure BDA0002284177510000071
wherein
Figure BDA0002284177510000072
Is the position vector of the ith butterfly for the t-th iteration, t is the iteration number,alinearly decreasing from 2 to 0 in the iterative process, rand () is a random number generated between (0,1), xkIs randomly selected and is different from butterfly xiThe butterfly of (1);
Figure BDA0002284177510000073
step S1035: judging whether the current iteration number K is equal to the maximum iteration number KmaxOr obtaining a global optimal solution, if yes, entering step S1036; if not, adding 1 to the current iteration number k value, and entering step S1032;
step S1036: obtaining the position vector corresponding to the butterfly with the maximum fitness value in the cluster as a solution vector, processing the solution vector, and enabling the position vector corresponding to the butterfly with the maximum fitness value to pass through a mapping function T (X)i) Convert it from continuous to discrete space for the feature selection process:
Figure BDA0002284177510000074
Figure BDA0002284177510000075
wherein r is a (0,1) interval random value, the value of the solution of feature selection is in a discrete space, namely 0 or 1, 0 represents that the feature is not selected, and 1 represents the selected feature.
According to a second aspect of the present invention, there is provided a medical image processing apparatus based on an artificial butterfly optimization algorithm, comprising:
a denoising module: the method is used for acquiring an input medical image I (I, j), dividing the image into n windows, and removing noise by adopting median filtering on each window of the n windows to obtain a noise-free image IF(i,j)；
The image feature expression module: for constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, from the noiseless image IF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
a feature selection module: for deriving said noise-free image I using an artificial butterfly optimization algorithmFExtracting important features from the one-dimensional vector form of (i, j);
an image recognition module: and the system is used for inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed.
According to a third aspect of the present invention, there is provided a medical image processing system based on an artificial butterfly optimization algorithm, comprising:
a processor for executing a plurality of instructions;
a memory to store a plurality of instructions;
wherein the instructions are used for being stored by the memory and loaded and executed by the processor to perform the medical image processing method based on the artificial butterfly optimization algorithm.
According to a fourth aspect of the present invention, there is provided a computer readable storage medium having a plurality of instructions stored therein; the instructions are used for loading and executing the medical image processing method based on the artificial butterfly optimization algorithm by the processor.
According to the scheme of the invention, the region of interest in the ultrasonic image can be rapidly identified, and the important features can be extracted again from the extracted ultrasonic image features, so that the accuracy of ultrasonic image identification is improved, and the efficiency of ultrasonic image classification and image identification is improved.
The foregoing description is only an overview of the technical solutions of the present invention, and in order to make the technical solutions of the present invention more clearly understood and to implement them in accordance with the contents of the description, the following detailed description is given with reference to the preferred embodiments of the present invention and the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention. In the drawings:
FIG. 1 is a flowchart of a medical image processing method based on an artificial butterfly optimization algorithm according to an embodiment of the present invention;
fig. 2 is a block diagram of a medical image processing apparatus based on an artificial butterfly optimization algorithm according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the specific embodiments of the present invention and the accompanying drawings. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The medical image processing method based on the artificial butterfly optimization algorithm of the present invention is described below with reference to fig. 1. Fig. 1 shows a flowchart of a medical image processing method based on an artificial butterfly optimization algorithm according to the invention. As shown in fig. 1, the method comprises the steps of:
step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i,j)；
Step S102: constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, and extracting a noise-free image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j);
step S104: and inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed.
The step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i, j) comprising:
median filtering is a non-linear technique used to remove noise in medical images. An intermediate filter is constructed for calculating a median value for eliminating speckle noise in the original image without reducing the image sharpness of the medical image.
In this embodiment, all the pixel points in one window are arranged in the order of pixel values, and then the pixel point with noise is replaced by the central pixel value. Then, the median of the window is calculated, and the calculation formula of the intermediate filter for calculating the median is as follows:
Figure BDA0002284177510000091
for a given image I (I, j), (r, s) ∈ ((W-1)/2, …, (W-1)/2), (I, j) ∈ (1,2, …, H) × (1,2, …, L), H and L respectively denote the width and height of the image, W is the odd value of the window, (3,5, …), W is a set of coordinates in a rectangular sub-image window, centered at point (x, y), all the central pixel values in the window are replaced by the calculated median value.
Removing noise from the rest windows of the medical image I (I, j) by median filtering to obtain a noise-free image IF(i,j)。
The step S102: constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, and extracting a noise-free image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form, wherein the method comprises the following steps:
step S1021: for the noiseless image IF(i, j) marking the category to which the noiseless image belongs, and obtaining the following marks:
Figure BDA0002284177510000101
wherein x isiFor the ith image sample feature, ciThe image belongs to the ith category, c is the number of the classified categories, and R is a real number set;
step S1022: constructing the noiseless image IF(i, j) inter-class adjacency graph WbAnd intra-class adjacency graph Ww：
Figure BDA0002284177510000102
Figure BDA0002284177510000103
Where l ∈ (1,2, …, c) denotes the class to which the image belongs, and W is a similarity matrix defined as:
Figure BDA0002284177510000104
wherein N isr(x) Represents the maximum of xNeighbor;
step S1023: constructing a kernel function, and initializing the coefficient of the kernel function;
constructing kernel functions
Figure BDA0002284177510000111
Namely, it is
Figure BDA0002284177510000112
Wherein the content of the first and second substances,
Figure BDA0002284177510000113
as a function of the distance measure between the data in the p-th image, xi,pFor each image xiCorresponding description feature, xj,pFor each image xjCorresponding descriptive feature, σpIs a positive constant value; sample coefficient vector alpha, alpha of initialization kernel functionT＝1。
Step S1024: determining a sample coefficient vector α of the kernel function, comprising:
computing
Figure BDA0002284177510000114
And
Figure BDA0002284177510000115
Figure BDA0002284177510000116
Figure BDA0002284177510000117
wherein the content of the first and second substances,
Figure BDA0002284177510000118
Figure BDA0002284177510000119
W(w)is a local weighting matrix between classes, W(b)Is a local weighting matrix within the class, K(i)For the basic kernel function, T represents the matrix transpose, W is the similarity matrix, and the calculation formula is as follows:
Figure BDA00022841775100001110
wherein N isr(x) R nearest neighbors of x, and t is a set constant value;
solving for the generalized features yields a sample coefficient vector α that is appropriate for the noise-free image:
setting an objective function: maximization
Figure BDA00022841775100001111
Finding an optimal conversion matrix by maximizing the local inter-class divergence in the embedded space and minimizing the local intra-class divergence, and converting the objective function into a function with a generalized vector decomposition method:
Figure BDA0002284177510000121
wherein, tau1≤τ2≤...≤τn'Is a minimum of n characteristic values, αiIs a characteristic value tauiCorresponding feature vector constituting the ith column vector of alpha, thereby determining the sample coefficient vector alpha of the noiseless image, i.e. the vector alpha is a column vector, each value of which is represented by alphaiAnd (4) showing.
Step S1025: determining a weight vector w of a basic kernel of the kernel function, including;
computing
Figure BDA0002284177510000122
And
Figure BDA0002284177510000123
Figure BDA0002284177510000124
Figure BDA0002284177510000125
solving the non-convex quadratic constraint quadratic programming problem to obtain a weight vector w of a basic kernel:
the objective function is:
Figure BDA0002284177510000126
i.e. to solve for
Figure BDA0002284177510000127
and omega is not less than 0;
Figure BDA0002284177510000128
Figure BDA0002284177510000129
step S1026: representing the noiseless image I in a two-dimensional matrix form by utilizing a determined sample coefficient vector alpha of the noiseless image and a weight vector w of a basic kernelF(i,j)；
Step S1027: and reducing the dimension of the two-dimensional matrix into a one-dimensional vector form.
Step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFAnd (i, j) extracting important features from the one-dimensional vector form.
The feature selection is an NP-hard problem, an optimal feature subset set can be found only by an exhaustive search method, the artificial butterfly optimization algorithm has the characteristic of being capable of adaptively iterating and continuously finding an optimal solution, but the artificial butterfly algorithm has some defects in convergence speed, so that the chaos theory is introduced, namely the artificial butterfly algorithm introduced with the chaos theory is used for extracting important features from a noise-free image. Important features influencing the image can be selected as much as possible, the optimizing capability is improved, and the convergence speed is improved.
The step S103 includes:
step S1031: acquiring the noiseless image IF(i, j) in the form of one-dimensional vectors, constructing a solution vector and a fitness function, and setting the number n of individuals in the cluster and the maximum iteration number KmaxSetting the initial value of the current iteration times k as 0, setting the solution vector as an initialization expression, and randomly initializing the position solution vector of each butterfly in the cluster;
the initialization expression is as follows:
X＝rand*(ub-lb)+lb
wherein ub and lb are the upper and lower boundaries of the search space, respectively, and rand is a random number between (0, 1);
the chaotic sequence is designed to improve the capability of butterfly initialization random search, and the expression of the chaotic sequence is as follows:
ci+1＝μci(1-ci)i∈{1,2,...,N}
wherein mu is a chaotic factor, c1Is a random number between (0,1), and c1≠{0.25，0.5，0.75，1}；
Mapping the chaotic sequence generated by the chaotic sequence to the initialized formula so as to generate a corresponding new population X':
X'＝lb+ci(ub-lb)；
the fitness function is expressed as follows:
Figure BDA0002284177510000131
wherein ACC is classification accuracy, N denotes a total number of features, nsuset denotes a selected feature subset, r and s denote weight coefficients, respectively, and r + s is 1;
step S1032: calculating the fitness of all butterflies in the new population, updating the fitness of the butterflies with the fitness value higher than a preset threshold value to the fitness value, and updating the positions of the butterflies according to the following formula:
Figure BDA0002284177510000132
wherein
Figure BDA0002284177510000133
Is the position vector of the ith butterfly for the t-th iteration, t is the number of iterations, step is the flight distance, rand () is the random number generated between (0,1), xkIs randomly selected and is different from butterfly xiLb is butterfly xiLower boundary value of flight range, Ub is butterfly xiAn upper boundary value of the flight range;
Figure BDA0002284177510000141
the flight distance step is a linear decrease with increasing number of iterations;
setting perturbation operator omega xiFor the maximum fitness value butterfly xmax，
Figure BDA0002284177510000142
Where ω is an adaptively derived quantity that decreases linearly with increasing number of iterations; x is the number ofiExpressing the chaos value by the formula
xi+1＝μxi(1-xi)，μ∈[0,4]
Wherein x isiIs the ith position generated randomly.
In the embodiment, the global search of butterflies is realized. In the early stage of iteration, a larger step value can provide better global search capability and diversity search in a search space; in the later period of iteration, a smaller step value can avoid a larger jump, which is beneficial to the convergence of the algorithm.
Step S1033: setting a first iteration time threshold value xi, and if the maximum fitness value is not changed between the iteration time k-xi and the current iteration time k, entering a step S1034; otherwise, go to step S1035;
step S1034: updating the position of the butterfly with the maximum fitness value;
namely, the position of the butterfly is updated according to the following formula:
Figure BDA0002284177510000143
wherein
Figure BDA0002284177510000144
Is the position vector of the ith butterfly for the t iteration, t is the number of iterations, a decreases linearly from 2 to 0 during the iteration, rand () is a random number generated between (0,1), xkIs randomly selected and is different from butterfly xiThe butterfly of (1);
Figure BDA0002284177510000145
the purpose of the step is to prevent the artificial butterfly algorithm from falling into local optimum, and further increase a variation mechanism capable of jumping out of the local optimum so as to find out global optimum.
Step S1035: judging whether the current iteration number K is equal to the maximum iteration number KmaxOr obtaining a global optimal solution, if yes, entering step S1036; if not, adding 1 to the current iteration number k value, and entering step S1032;
step S1036: obtaining the position vector corresponding to the butterfly with the maximum fitness value in the cluster as a solution vector, processing the solution vector, and enabling the position vector corresponding to the butterfly with the maximum fitness value to pass through a mapping function T (X)i) Convert it from continuous to discrete space for the feature selection process:
Figure BDA0002284177510000151
Figure BDA0002284177510000152
wherein r is a (0,1) interval random value, the value of the solution of feature selection is in a discrete space, namely 0 or 1, 0 represents that the feature is not selected, and 1 represents the selected feature.
The step S104: inputting the extracted important features into a trained relevance vector machine RVM to obtain classification and identification results of the medical image to be processed, wherein the classification and identification results comprise:
and the parameters of the trained correlation vector machine are obtained by training sample data of the medical image. And acquiring a one-dimensional vector form of the sample data, selecting important features from the sample data by using an artificial butterfly optimization algorithm, further inputting the important features into a correlation vector machine, training the correlation vector machine, and training the correlation vector machine to obtain the trained correlation vector machine. The construction mode of the related vector machine is a common construction mode of the related vector machine in the field.
And (5) inputting the features selected in the step (S103) into a trained correlation vector machine to obtain the classification and identification results of the medical images.
Specifically, the obtaining of the one-dimensional vector form of the sample data, selecting features from the sample data by using an artificial butterfly optimization algorithm, inputting the features into a correlation vector machine, and training the correlation vector machine includes: the sample image set comprises a sample image Ii(1<i is less than or equal to n), and n is the number of sample images in the sample image set. For each sample image I in the sample image setiThe step S101 is executed as described above, and for each sample image IiNoise reduction is carried out; step S102 is executed as described above: constructing a kernel function based on a multi-kernel local discriminant analysis (MKLFDA) method, and extracting the noiseless image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form; and re-executing step S103 as described above: using an artificial butterfly optimization algorithm to obtain a data structure from saidNoiseless image IFExtracting important features from the one-dimensional vector form of (i, j); further, each sample image I to be acquirediThe important features are input into a correlation vector machine to train parameters of the correlation vector machine, and when the error is smaller than a set threshold or the iteration times reach a preset threshold, the training of the correlation vector machine is finished, so that the trained correlation vector machine is obtained.
Please refer to fig. 2, which is a block diagram of a medical image processing apparatus based on an artificial butterfly optimization algorithm according to the present invention. As shown, the apparatus comprises:
a denoising module: the method is used for acquiring an input medical image I (I, j), dividing the image into n windows, and removing noise by adopting median filtering on each window of the n windows to obtain a noise-free image IF(i,j)；
The image feature expression module: for constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, from the noiseless image IF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
a feature selection module: for deriving said noise-free image I using an artificial butterfly optimization algorithmFExtracting important features from the one-dimensional vector form of (i, j);
an image recognition module: and the system is used for inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed.
The embodiment of the invention further provides a medical image processing system based on an artificial butterfly optimization algorithm, which comprises the following components:
a processor for executing a plurality of instructions;
a memory to store a plurality of instructions;
wherein the instructions are used for being stored by the memory and loaded and executed by the processor to perform the medical image processing method based on the artificial butterfly optimization algorithm.
The embodiment of the invention further provides a computer readable storage medium, wherein a plurality of instructions are stored in the storage medium; the instructions are used for loading and executing the medical image processing method based on the artificial butterfly optimization algorithm by the processor.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
In the embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions in actual implementation, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a physical machine Server, or a network cloud Server, etc., and needs to install a Windows or Windows Server operating system) to perform some steps of the method according to various embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and any simple modification, equivalent change and modification made to the above embodiment according to the technical spirit of the present invention are still within the scope of the technical solution of the present invention.

Claims (7)

1. A medical image processing method based on an artificial butterfly optimization algorithm is characterized by comprising the following steps:
step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and removing noise in each window of the n windows by adopting median filtering to obtain a noise-free image IF(i,j)；
Step S102: constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, and extracting a noise-free image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j);
step S104: and inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical image to be processed.
2. The medical image processing method based on the artificial butterfly optimization algorithm according to claim 1, wherein the step S101: acquiring an input medical image I (I, j) to be processed, dividing the image into n windows, and executing the processing on the n windowsRemoving noise in each window by median filtering to obtain a noiseless image IF(i, j) comprising:
firstly, arranging all pixel points in a window according to a pixel value sequence, and then replacing the pixel points with noise by using a central pixel value; then, the median of the window is calculated, and the calculation formula of the intermediate filter for calculating the median is as follows:
Figure FDA0002634997140000011
for a given image I (I, j), (r, s) ∈ (- (W-1)/2, …, (W-1)/2), (I, j) ∈ (1,2, …, H) × (1,2, …, L), H and L respectively denote the width and height of the image, W is the odd value of the window, (3,5, …), W is a set of coordinates in a rectangular sub-image window, centered at point (x, y), replacing all the central pixel values in the window with the calculated median value;
removing noise from the rest windows of the medical image I (I, j) by median filtering to obtain a noise-free image IF(i,j)。
3. The medical image processing method based on artificial butterfly optimization algorithm according to claim 1, wherein the step S102: constructing a kernel function based on a multi-kernel local discriminant analysis (MKLFDA) method, and extracting the noiseless image I from the kernel functionF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form, wherein the method comprises the following steps:
step S1021: for the noiseless image IF(i, j) marking the category to which the noiseless image belongs, and obtaining the following marks:
Figure FDA0002634997140000021
wherein x isiFor the ith image sample feature, ciThe image belongs to the ith category, c is the number of the classified categories, and R is a real number set;
step S1022: constructing the noiseless image IF(i, j) inter-class adjacency graph W(b)And intra-class adjacency graph W(w)：
Figure FDA0002634997140000022
Figure FDA0002634997140000023
Where l ∈ (1,2, …, c) denotes the class to which the image belongs, and W is a similarity matrix defined as:
Figure FDA0002634997140000024
wherein N isr(x) Represents the nearest neighbor of x;
step S1023: constructing a kernel function, and initializing the coefficient of the kernel function;
constructing kernel functions
Figure FDA0002634997140000031
Figure FDA0002634997140000032
Namely, it is
Wherein the content of the first and second substances,
Figure FDA0002634997140000033
as a function of the distance measure between the data in the p-th image, xi,pFor each image xiCorresponding description feature, xj,pFor each image xjCorresponding descriptive feature, σpIs a positive constant value;
sample coefficient vector alpha, alpha of initialization kernel functionT＝1；
Step S1024: determining a sample coefficient vector α of the kernel function, comprising:
computing
Figure FDA0002634997140000034
And
Figure FDA0002634997140000035
Figure FDA0002634997140000036
Figure FDA0002634997140000037
wherein the content of the first and second substances,
Figure FDA0002634997140000038
Figure FDA0002634997140000039
W(w)is a local weighting matrix between classes, W(b)Is a local weighting matrix within the class, K(i)For the basic kernel function, T represents the matrix transpose, W is the similarity matrix, and the calculation formula is as follows:
Figure FDA0002634997140000041
wherein N isr(x) R nearest neighbors of x, and t is a set constant value;
solving for the generalized features yields a sample coefficient vector α that is appropriate for the noise-free image:
setting an objective function: maximization
Figure FDA0002634997140000042
Finding an optimal conversion matrix by maximizing the local inter-class divergence in the embedded space and minimizing the local intra-class divergence, and converting the objective function into a function with a generalized vector decomposition method:
Figure FDA0002634997140000043
wherein, tau1≤τ2≤...≤τn'Is a minimum of n characteristic values, αiIs a characteristic value tauiCorresponding feature vector constituting the ith column vector of alpha, thereby determining the sample coefficient vector alpha of the noiseless image, i.e. the vector alpha is a column vector, each value of which is represented by alphaiRepresents;
step S1025: determining a weight vector w of a basic kernel of the kernel function, including;
computing
Figure FDA0002634997140000044
And
Figure FDA0002634997140000045
Figure FDA0002634997140000046
Figure FDA0002634997140000047
solving the non-convex quadratic constraint quadratic programming problem to obtain a weight vector w of a basic kernel:
the objective function is:
Figure FDA0002634997140000048
i.e. to solve for
Figure FDA0002634997140000049
A minimum of w;
Figure FDA00026349971400000410
Figure FDA00026349971400000411
step S1026: representing the noiseless image I in a two-dimensional matrix form by utilizing a determined sample coefficient vector alpha of the noiseless image and a weight vector w of a basic kernelF(i,j)；
Step S1027: and reducing the dimension of the two-dimensional matrix into a one-dimensional vector form.
4. The medical image processing method based on artificial butterfly optimization algorithm according to claim 1, wherein the step S103: using an artificial butterfly optimization algorithm to extract the noise-free image IFExtracting important features from the one-dimensional vector form of (i, j), wherein the extracting important features comprise:
step S1031: acquiring the noiseless image IF(i, j) in the form of one-dimensional vectors, constructing a solution vector and a fitness function, and setting the number n of individuals in the cluster and the maximum iteration number KmaxSetting the initial value of the current iteration times k as 0, setting the solution vector as an initialization expression, and randomly initializing the position solution vector of each butterfly in the cluster;
the initialization expression is as follows:
X＝rand*(ub-lb)+lb
wherein ub and lb are the upper and lower boundaries of the search space, respectively, and rand is a random number between (0, 1);
the chaotic sequence is designed to improve the capability of butterfly initialization random search, and the expression of the chaotic sequence is as follows:
ci+1＝μci(1-ci)i∈{1,2,...,N}
wherein mu is a chaotic factor, c1Is a random number between (0,1), and c1≠{0.25，0.5，0.75，1}；
Mapping the chaotic sequence generated by the chaotic sequence to the initialized formula so as to generate a corresponding new population X':
X'＝lb+ci(ub-lb)；
the fitness function is expressed as follows:
Figure FDA0002634997140000051
wherein ACC is classification accuracy, N denotes a total number of features, nsuset denotes a selected feature subset, r and s denote weight coefficients, respectively, and r + s is 1;
step S1032: calculating the fitness of all butterflies in the new population, updating the fitness of the butterflies with the fitness value higher than a preset threshold value to the fitness value, and updating the positions of the butterflies according to the following formula:
Figure FDA0002634997140000061
wherein
Figure FDA0002634997140000062
Is the position vector of the ith butterfly for the t-th iteration, t is the number of iterations, step is the flight distance, rand () is the random number generated between (0,1), xkIs randomly selected and is different from butterfly xiLb is butterfly xiLower boundary value of flight range, Ub is butterfly xiAn upper boundary value of the flight range;
Figure FDA0002634997140000063
the flight distance step is a linear decrease with increasing number of iterations;
setting perturbation operator omega xiFor the maximum fitness value butterfly xmax，
Figure FDA0002634997140000064
Where ω is an adaptively derived quantity that decreases linearly with increasing number of iterations; x is the number ofiExpressing the chaos value by the formula
xi+1＝μxi(1-xi)，μ∈[0,4]
Wherein x isiIs the ith position generated randomly;
step S1033: setting a first iteration time threshold value xi, and if the maximum fitness value is not changed between the iteration time k-xi and the current iteration time k, entering a step S1034; otherwise, go to step S1035;
step S1034: updating the position of the butterfly with the maximum fitness value;
namely, the position of the butterfly is updated according to the following formula:
Figure FDA0002634997140000071
wherein
Figure FDA0002634997140000072
Is the position vector of the ith butterfly for the t iteration, t is the number of iterations, a decreases linearly from 2 to 0 during the iteration, rand () is a random number generated between (0,1), xkIs randomly selected and is different from butterfly xiThe butterfly of (1);
Figure FDA0002634997140000073
step S1035: judging whether the current iteration number K is equal to the maximum iteration number KmaxOr obtaining a global optimal solution, if yes, entering step S1036; if notAdding 1 to the value of the current iteration number k, and entering step S1032;
step S1036: obtaining the position vector corresponding to the butterfly with the maximum fitness value in the cluster as a solution vector, processing the solution vector, and enabling the position vector corresponding to the butterfly with the maximum fitness value to pass through a mapping function T (X)i) Convert it from continuous to discrete space for the feature selection process:
Figure FDA0002634997140000074
Figure FDA0002634997140000075
wherein r is a (0,1) interval random value, the value of the solution of feature selection is in a discrete space, namely 0 or 1, 0 represents that the feature is not selected, and 1 represents the selected feature.
5. A medical image processing apparatus based on an artificial butterfly optimization algorithm, the apparatus comprising:
a denoising module: the method is used for acquiring an input medical image I (I, j), dividing the image into n windows, and removing noise by adopting median filtering on each window of the n windows to obtain a noise-free image IF(i,j)；
The image feature expression module: for constructing a kernel function based on a multi-kernel local Fisher discriminant analysis (MKLFDA) method, from the noiseless image IF(i, j) extracting a feature matrix of the image, and reducing the dimension of the feature matrix into a one-dimensional vector form;
a feature selection module: for deriving said noise-free image I using an artificial butterfly optimization algorithmFExtracting important features from the one-dimensional vector form of (i, j);
an image recognition module: and the system is used for inputting the extracted important features into a trained relevance vector machine RVM to obtain the classification and identification results of the medical images to be processed.
6. A medical image processing system based on an artificial butterfly optimization algorithm is characterized by comprising:
a processor for executing a plurality of instructions;
a memory to store a plurality of instructions;
wherein the instructions are stored in the memory and loaded by the processor to execute the medical image processing method based on the artificial butterfly optimization algorithm according to any one of claims 1 to 4.
7. A computer-readable storage medium having stored therein a plurality of instructions; the plurality of instructions for loading and executing the medical image processing method based on the artificial butterfly optimization algorithm according to any one of claims 1 to 4 by a processor.
CN201911153403.9A 2019-11-22 2019-11-22 Medical image processing method and device based on artificial butterfly optimization algorithm Active CN110910325B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911153403.9A CN110910325B (en) 2019-11-22 2019-11-22 Medical image processing method and device based on artificial butterfly optimization algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911153403.9A CN110910325B (en) 2019-11-22 2019-11-22 Medical image processing method and device based on artificial butterfly optimization algorithm

Publications (2)

Publication Number Publication Date
CN110910325A CN110910325A (en) 2020-03-24
CN110910325B true CN110910325B (en) 2020-12-22

Family

ID=69818874

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911153403.9A Active CN110910325B (en) 2019-11-22 2019-11-22 Medical image processing method and device based on artificial butterfly optimization algorithm

Country Status (1)

Country Link
CN (1) CN110910325B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112862750A (en) * 2020-12-29 2021-05-28 深圳信息职业技术学院 Blood vessel image processing method and device based on multi-scale fusion and meta-heuristic optimization
CN112863672A (en) * 2021-03-09 2021-05-28 中电健康云科技有限公司 Patient identity matching method based on PSO algorithm optimization
CN113778936A (en) * 2021-08-17 2021-12-10 华东计算技术研究所（中国电子科技集团公司第三十二研究所） Performance optimization method of domestic embedded DSP operating system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104102839A (en) * 2013-11-16 2014-10-15 西安电子科技大学 Alzheimer's disease cortex auto-classification method based on multi-scale mesh surface form features
CN108549912A (en) * 2018-04-19 2018-09-18 北京工业大学 A kind of medical image pulmonary nodule detection method based on machine learning

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160021369A1 (en) * 2014-07-15 2016-01-21 Shreyas HAMPALI Video coding including a stage-interdependent multi-stage butterfly integer transform
CN104156628B (en) * 2014-08-29 2017-05-31 东南大学 A kind of naval vessel radiation signal recognition methods based on Multiple Kernel Learning discriminant analysis
CN105913075A (en) * 2016-04-05 2016-08-31 浙江工业大学 Endoscopic image focus identification method based on pulse coupling nerve network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104102839A (en) * 2013-11-16 2014-10-15 西安电子科技大学 Alzheimer's disease cortex auto-classification method based on multi-scale mesh surface form features
CN108549912A (en) * 2018-04-19 2018-09-18 北京工业大学 A kind of medical image pulmonary nodule detection method based on machine learning

Also Published As

Publication number Publication date
CN110910325A (en) 2020-03-24

Similar Documents

Publication Publication Date Title
Paulin et al. Transformation pursuit for image classification
CN110910325B (en) Medical image processing method and device based on artificial butterfly optimization algorithm
JP6483667B2 (en) System and method for performing Bayesian optimization
Sun et al. Heterogeneous tensor decomposition for clustering via manifold optimization
CN110659665B (en) Model construction method of different-dimension characteristics and image recognition method and device
JP4545641B2 (en) Similar image retrieval method, similar image retrieval system, similar image retrieval program, and recording medium
CA3066029A1 (en) Image feature acquisition
Hofmanninger et al. Mapping visual features to semantic profiles for retrieval in medical imaging
US9563822B2 (en) Learning apparatus, density measuring apparatus, learning method, computer program product, and density measuring system
CN109817339B (en) Patient grouping method and device based on big data
CN111161314A (en) Target object position area determining method and device, electronic equipment and storage medium
CN109034218B (en) Model training method, device, equipment and storage medium
CN111341408A (en) Image report template generation method, computer equipment and storage medium
CN107480627B (en) Behavior recognition method and device, storage medium and processor
CN111694954A (en) Image classification method and device and electronic equipment
JP2014228995A (en) Image feature learning device, image feature learning method and program
CN110929731B (en) Medical image processing method and device based on pathfinder intelligent search algorithm
CN115222443A (en) Client group division method, device, equipment and storage medium
CN112927235B (en) Brain tumor image segmentation method based on multi-scale superpixel and nuclear low-rank representation
CN113762005B (en) Feature selection model training and object classification methods, devices, equipment and media
CN111583194B (en) High-dimensional feature selection algorithm based on Bayesian rough set and cuckoo algorithm
CN114692809A (en) Data processing method and device based on neural cluster, storage medium and processor
Dong et al. Scene-oriented hierarchical classification of blurry and noisy images
CN113177608B (en) Neighbor model feature selection method and device for incomplete data
Scrucca et al. Projection pursuit based on Gaussian mixtures and evolutionary algorithms

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant