CN107045717A - The detection method of leucocyte based on artificial bee colony algorithm - Google Patents

Abstract

The invention discloses a white blood cell detection method based on an artificial bee colony algorithm. Firstly, an edge detection is used to process an image, and all edge pixels are extracted and stored in a vector P, and then the population is initialized. Calculate the fitness value of the nectar source; hire bees to search for new nectar sources, calculate their fitness value and use the greedy method to select a better nectar source; hire bees to share information, and follow the bees to calculate the probability of the nectar source being selected according to the roulette principle. The search formula performs a fine search in the vicinity; judges whether there is a continuous limit generation of food sources that has not been updated, and if so, employs bees to become scout bees, and scout bees search to generate new honey sources; memorizes the optimal honey source location; judges whether the maximum number of iterations is reached, if When the optimal solution is reached, the output of the optimal solution is terminated, and finally the image of white blood cells is detected. The invention uses the artificial bee colony algorithm for the detection of white blood cells, and improves the detection accuracy of white blood cells while maintaining a good detection speed.

Description

Detection method of white blood cells based on artificial bee colony algorithm

technical field

The invention belongs to the field of image processing, in particular to a method for detecting white blood cells based on an artificial bee colony algorithm.

Background technique

Human blood leukocytes, also known as immune cells, play an important role in the human immune system. In clinical medicine, the changes in the number and morphology of different types of leukocytes can be used to judge the pathological conditions of human body functions. Therefore, the detection of white blood cells is an important indicator of blood routine detection. At present, microscopes and automated hematology analyzers are the main methods for leukocyte detection, but they all have their own disadvantages, such as manual microscopy, which takes a long time, heavy workload, and tedious; automated instruments cannot detect changes in the shape of leukocytes.

With the continuous progress and development of image processing and pattern recognition, these technologies can be applied to medical imaging, and can also be applied to the detection of white blood cells. White blood cell detection can be divided into two steps: edge detection and ellipse detection. The methods of ellipse detection can be roughly divided into two categories, one is voting method, such as Hough transform method and RANSAC algorithm, etc.; the other is optimization algorithm, such as least square fitting, legacy algorithm, artificial bee colony algorithm, etc.; In current practical applications, the most classic method is to use the Hough transform to detect elliptical image objects. The Hough transform method is not sensitive to the part missing and noise of the ellipse, and has high detection accuracy and robustness. However, due to the ellipse There are five own parameters, which need to be accumulated in the five-dimensional parameter space, which makes this approach impractical due to the large amount of calculation and memory requirements, so the classic Hough transform is not suitable. The optimization algorithm can overcome these problems. The methods for solving optimization problems can be divided into classical optimization methods and intelligent optimization algorithms. Classical optimization algorithms include gradient descent, simple row method, conjugate direction method, Newton method, etc. Classical optimization algorithms are difficult to achieve ideal optimization results for some complex, discrete or multi-objective systems. Compared with the classic optimization method, the artificial bee colony algorithm has almost no requirements on the objective function and constraints, hardly uses external information in the search process, and only uses the fitness function as the basis for evolution, forming an algorithm characterized by "generation + inspection". artificial intelligence technology. Intelligent optimization algorithms include ant colony optimization algorithm, particle swarm optimization algorithm, bee colony algorithm, difference algorithm, etc. Although the particle swarm optimization algorithm has a fast convergence speed, it is easy to fall into the local optimal solution; the difference algorithm is not easy to fall into the local optimal solution, but the convergence speed is slow. The artificial bee colony algorithm has the characteristics of simple operation, small control parameters, high search precision and strong robustness.

Contents of the invention

Aiming at the problems existing in the above traditional white blood cell detection, the present invention proposes a white blood cell detection method based on the artificial bee colony algorithm. The artificial bee colony algorithm can not only jump out of the local optimal solution, but also search for the performance of the global optimal solution. Effectively improve the speed and precision of operation.

In order to achieve the above object, the technical solution proposed by the present invention is the detection method of white blood cells based on the artificial bee colony algorithm, comprising the following steps:

Step 1: Process the image using edge detection, and extract all edge pixels and store them in the vector P;

Step 2: Initialize the population: initialize the honey source x _id , set the total number of honey sources NP, the control parameter limit, the maximum number of iterations, t=1;

Step 3: Calculate the fitness value of the honey source;

Step 4: Hire bees to search for new nectar sources, calculate their fitness value and use greedy method to select better nectar sources;

Step 5: Hire bees to share information, follow the bees to calculate the probability of the honey source being selected according to the roulette principle, and search nearby according to the search formula;

Step 6: Determine whether there is a continuous limit generation of food sources that has not been updated. If so, the hired bees will become scout bees, and the scout bees will search for new honey sources; if not, go to step 8;

Step 7: memorize the optimal nectar source location;

Step: 8: t=t+1; Judging whether to reach the maximum number of iterations, if reached then terminate the output optimal solution, otherwise repeat steps 3-7;

Step 9: An image of white blood cells is detected.

Further, the edge detection in the above step 1 specifically includes extracting the elliptical edge contour of the white blood cell nucleus in the picture, and then performing ellipse detection, extracting all edge pixels and storing them in the vector P, P=(P ₁ ,P ₂ ,...,P _n ), where n is the sum of all edge pixels on the image.

The position of each nectar source in the above step 2 represents an ellipse of a feasible solution, each new candidate nectar source represents a new candidate ellipse, and the vectors of all nectar sources are initialized The basic equation of an ellipse can be expressed as:

ax ² +bxy+cy ² +dx+ey+1＝0

An ellipse is formed when the values of the five free parameters a, b, c, d, e satisfy b ² -4ac<0, so each ellipse can be represented by five free parameters. In order to construct an ellipse, in the vector P Randomly extract five edge pixel points p _i1 , p _i2 , p _i3 , p _i4 , p _i5 , any three of which are not on a straight line, and can be combined to form an ellipse, so each nectar source is coded as a Ellipse X _i , when performing ellipse detection, five linear equations about parameters are composed of any five edge pixels in the vector P, respectively, and solve them. If the obtained solution meets the conditions, the artificial bee colony algorithm is used Select and optimize the solution.

The position of initializing the honey source in the above step 2 is:

x _id ＝L _d +rand(0,1)(U _d -L _d )

In the formula: L _d and U _d represent the minimum value and maximum value of the search space respectively, d=1, 2,...D, D is the dimension of the solution, since the vector solution is determined by five parameters, so D=5, the bee colony The number of nectar source scale NP and the maximum number of iterations are set according to the actual situation.

As a preference, the above-mentioned number of bee colonies, size of honey sources and maximum number of iterations are respectively set as: total number of bee colonies=60, size of honey sources NP=30, maximum number of iterations=50.

The calculation method of the fitness value in the above step 3 is:

fit _i is the fitness value of the nectar source, fi is the solution of the objective function, that is, the matching degree of the edge points of the ellipse, and the set of candidate ellipse pixel points can be expressed as X _{i =} (s ₁ ,s ₂ ,…,s _N ), N is the number of all edge points existing on the candidate ellipse, and the objective function can be expressed as the matching degree of ellipse edge points:

X _i (s _1x ,s _1y ) is defined as

where (s _1x , s _1y ) is the coordinates of any pixel in the pixel set S in the candidate ellipse.

In the above step 4, employing bees to search for new honey sources, calculating their fitness value and using the greedy method to select a better honey source specifically includes: searching and generating new honey sources near the initialized honey source, as shown in the following formula:

In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1,D], which means that the hired bee randomly selects one dimension to search; j∈{1,2,… ,NP}, j≠i, x _jd means randomly selecting a honey source not equal to i among NP honey sources; It is a random number evenly distributed in the range of [-1,1], which determines the magnitude of the disturbance. When the fitness of the nectar source v _id is better than x _id , use the greedy selection method to replace x _id with v _id , otherwise keep x _id , that is to generate a new candidate ellipse, the greedy principle is expressed as:

Among them, f(v _i ) is the objective function value of the newly generated location, f( _xi ) is the objective function value of the current honey source, if and only if the objective function value of the newly generated location is smaller than the objective function value of the original honey source location, Update the location of the current honey source.

The probability formula for each nectar source to be selected in the above step 5 is:

According to the selection probability determined by the above formula, the following bee selects the nectar source according to the roulette principle and conducts a fine search near it. The search formula is:

In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1, D], which means that the hired bee randomly selects one dimension to search.

In the above step 6, the nectar source x _id has passed the threshold limit iterations and no better nectar source is found, then the nectar source will be abandoned, and the role of the corresponding hired bee becomes a scout bee, and the scout bee will randomly generate a nectar in the search space The new nectar source replaces x _id , that is, the candidate ellipse is updated. The above process can be expressed as:

The position of the best memorized nectar source in step 7 above is the candidate ellipse that is the closest to the target ellipse to obtain the optimal solution according to the above search formulas.

Compared with prior art, the beneficial effect of the present invention is:

1. The present invention uses the artificial bee colony algorithm for the detection of white blood cells, which improves the detection accuracy of white blood cells while maintaining a good detection speed.

2. The present invention has the advantages of fast detection speed, good detection effect, and strong algorithm robustness, and better solves the problems of insufficient memory and large calculation amount of the classic Hough transform, and the particle swarm algorithm is easy to fall into a local optimal solution.

3. The present invention solves the problems that the particle swarm optimization algorithm is easy to fall into a local optimal solution and other optimization algorithms have insufficient detection accuracy and slow detection speed.

Description of drawings

FIG. 1 is a flow chart of the method for detecting white blood cells based on the artificial bee colony algorithm of the present invention.

Fig. 2 is a reference image of cells applied to leukocyte detection in the present invention.

Fig. 3 is the result of edge detection of the white blood cell reference image according to the present invention.

Figure 4 shows the detection results of the particle swarm optimization algorithm on the white blood cell reference image.

Fig. 5 is the detection result of the white blood cell reference image according to the present invention.

detailed description

In order to make the object, technical solution and advantages of the invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and examples of implementation. It should be understood that the specific implementation cases described here are only used to explain the present invention, and are not intended to limit the present invention.

As shown in Figure 1, the method includes the following steps:

Step 1: Process the image with edge detection and extract all edge pixels and store them in vector P

First, edge detection is performed on the original picture (as shown in Figure 2) (edge detection is the primary condition for white blood cell detection), as shown in Figure 3, extract the elliptical edge contour of the white blood cell nucleus in the picture, then perform ellipse detection, and extract all The edge pixels are stored in a vector P, where P=(P ₁ , P ₂ ,...,P _n ), and n is the sum of all edge pixels on the image.

Step 2: Initialize the population: initialize the nectar source x _id , set the total number of nectar sources NP, the control parameter limit, the maximum number of iterations, t=1

The position of each nectar source represents a feasible solution ellipse, and each new candidate nectar source represents a new candidate ellipse. Initialize the vector of all honey sources The basic equation of an ellipse can be expressed as:

ax ² +bxy+cy ² +dx+ey+1＝0

An ellipse is formed when the values of the five free parameters a, b, c, d, e satisfy b ² -4ac<0. Therefore, each ellipse can be represented by five free parameters. In order to construct an ellipse (honey source), five edge pixel points p _i1 , p _i2 , p _i3 , p _i4 , p _i5 (any three points are not on a straight line) can form an ellipse, so each honey source is coded as an ellipse X _i . When performing ellipse detection, the five linear equations about parameters composed of any five edge pixels in the vector P are solved. If the obtained solution meets the conditions, the artificial bee colony algorithm is used to select the solution. optimization. The purpose of the present invention is to optimize these vectors to obtain the optimal solution vector. The vector initializes the position of the candidate ellipse (honey source):

x _id ＝L _d +rand(0,1)(U _d -L _d )

In the formula: L _d and U _d represent the minimum value and maximum value of the search space respectively, d=1, 2,...D. D is the dimension of the solution. Since the vector solution is determined by five parameters, D=5. The number of bee colonies, the size of the nectar source and the maximum number of iterations are set according to the actual situation, including setting the total number of bee colonies = 60, the size of the nectar source NP = 30, and the maximum number of iterations = 50.

Step 3: Calculate the fitness value of the honey source

The calculation method of fitness value is:

fit _i is the fitness value of the nectar source, and fi is the solution of the objective function _, that is, the matching degree of the edge points of the ellipse. In the present invention, the ratio of the number of edge points on the optimized candidate ellipse coincident with the vector P to the number of all edge points in the vector P is called the matching degree of the ellipse edge points. Optimization refers to obtaining the best candidate ellipse (target ellipse) from all feasible solutions. If the target ellipse really exists, in order to calculate the matching degree of the edge points of the ellipse, we should first calculate the coordinates of the edge pixel points of the actually obtained candidate ellipse. The set of candidate ellipse pixel points can be expressed as X _i =(s ₁ , s ₂ ,...,s _N ), where N is the number of all edge points existing on the candidate ellipse. The objective function can be expressed as the matching degree of ellipse edge points as:

X _i (s _1x ,s _1y ) is defined as

where (s _1x , s _1y ) is the coordinates of any pixel in the pixel set S in the candidate ellipse. The ellipse detection method of the present invention belongs to the problem of maximum value, and the purpose is to _maximize the value of fi. According to the fitness calculation function fit _i , in order to obtain the maximum solution, the greedy selection method selects the nectar source with a small fitness, so after searching and finding a new nectar source v _id , according to the size of the fitness value, use between v _id and x _id The greedy selection method selects the nectar source with a smaller fitness value to generate a new nectar source (candidate ellipse).

Step 4: Hire bees to search for new nectar sources, calculate their fitness value and use greedy method to select better nectar sources

Searching around the initialized honey source (candidate ellipse) generates a new honey source (new candidate ellipse) formula:

In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1,D], which means that the hired bee randomly selects one dimension to search; j∈{1,2,… ,NP}, j≠i, x _jd means randomly selecting a honey source not equal to i among NP honey sources; Is a uniformly distributed random number in the range of [-1,1], which determines the magnitude of the disturbance. When the fitness of the nectar source v _id is better than x _id , use greedy selection method to replace x _id with v _id , otherwise keep x _id , that is, generate a new candidate ellipse. The greedy principle is expressed as:

Among them, f(v _i ) is the objective function value of the newly generated location, f( _xi ) is the objective function value of the current honey source, if and only if the objective function value of the newly generated location is smaller than the objective function value of the original honey source location, Update the location of the current honey source.

Step 5: Hire bees to share information, and follow the bees to calculate the probability of the nectar source being selected by using the roulette wheel principle, and at the same time search for each nectar source in the vicinity according to the search formula. The probability formula for each nectar source is:

According to the selection probability determined by the listing, the following bee selects the nectar source according to the roulette wheel principle and conducts a fine search near it. The search formula is:

In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1, D], which means that the hired bee randomly selects one dimension to search.

Step 6: Determine whether there is a continuous limit generation of food sources that has not been updated. If so, the hired bees will become scout bees. The scout bees search for new honey sources. If not, go to step 8

When the nectar source x _id reaches the threshold limit after trial times of iterative search and no better nectar source is found, the nectar source x _id will be abandoned, and the corresponding hired bees will become scout bees. The scout bees will randomly generate a new honey source in the search space to replace x _id , that is, the candidate ellipse is updated. The above process can be expressed as:

Step 7: Memorize the location of the best honey source

Memorize the position of the optimal nectar source, and get the optimal solution, that is, the candidate ellipse closest to the target ellipse, according to the above search band formulas.

Step: 8: t=t+1; judge whether to reach the maximum number of iterations, if reached then terminate the output optimal solution, otherwise repeat steps 3-7

According to the iterative algorithm set in step 2, determine whether the number of iterations has reached: if the number of iterations reaches max=50, then stop the iteration, and obtain the position of the optimal honey source and the value of the objective function. The position of the honey source corresponds to the target ellipse. If the number of iterations is not reached, repeat steps 3-7 until the maximum number of iterations set in step 2 is reached.

Step 9: The image of the ellipse is detected

The higher the objective function is, the higher the matching degree of the ellipse is, and the closer to the target ellipse, the more correct the white blood cell detection result is. Comparing Figure 4 with Figure 5, it is obvious that the detection result in Figure 5 is more ideal.

Claims (10)

1. the detection method based on the white blood cell of artificial bee colony algorithm, is characterized in that, comprises the following steps: Step 1: Process the image using edge detection, and extract all edge pixels and store them in the vector P; Step 2: Initialize the population: initialize the honey source x _id , set the total number of honey sources NP, the control parameter limit, the maximum number of iterations, t=1; Step 3: Calculate the fitness value of the honey source; Step 4: Hire bees to search for new nectar sources, calculate their fitness value and use greedy method to select better nectar sources; Step 5: Hire bees to share information, follow the bees to calculate the probability of the honey source being selected according to the roulette principle, and search nearby according to the search formula; Step 6: Determine whether there is a continuous limit generation of food sources that has not been updated. If so, the hired bees will become scout bees, and the scout bees will search for new honey sources; if not, go to step 8; Step 7: memorize the optimal nectar source location; Step: 8: t=t+1; Judging whether to reach the maximum number of iterations, if reached then terminate the output optimal solution, otherwise repeat steps 3-7; Step 9: The image of white blood cells is detected. 2. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, the edge detection in the described step 1 specifically comprises the edge contour of the leukocyte nucleus ellipse in extracting picture, then carries out ellipse detection, All edge pixels are extracted and stored in vector P, P=(P ₁ , P ₂ ,...,P _n ), where n is the sum of all edge pixels on the image. 3. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, the position of each honey source in described step 2 all represents the ellipse of a feasible solution, every time produces a new candidate The nectar source represents a new candidate ellipse, and the vectors of all nectar sources are initialized The basic equation of an ellipse can be expressed as: ax ² +bxy+cy ² +dx+ey+1＝0 An ellipse is formed when the values of the five free parameters a, b, c, d, e satisfy b ² -4ac<0, so each ellipse can be represented by five free parameters. In order to construct an ellipse, in the vector P Randomly extract five edge pixel points p _i1 , p _i2 , p _i3 , p _i4 , p _i5 , any three of which are not on a straight line, and can be combined to form an ellipse, so each nectar source is coded as a Ellipse X _i , when performing ellipse detection, five linear equations about parameters are composed of any five edge pixels in the vector P, respectively, and solve them. If the obtained solution meets the conditions, the artificial bee colony algorithm is used Select and optimize the solution. 4. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, the position of initializing nectar source in described step 2 is: x _id ＝L _d +rand(0,1)(U _d -L _d ) In the formula: L _d and U _d represent the minimum value and maximum value of the search space respectively, d=1, 2,...D, D is the dimension of the solution, since the vector solution is determined by five parameters, so D=5, the bee colony The number of , nectar source size NP and the number of iterations max are set according to the actual situation. 5. the detection method based on the white blood cell of artificial bee colony algorithm as claimed in claim 4 is characterized in that the quantity of bee colony, nectar source scale and maximum number of iterations are respectively set to: bee colony total number=60, nectar source scale NP=30 , the maximum number of iterations = 50. 6. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, the calculation method of fitness value in the described step 3 is: fit _i is the fitness value of the nectar source, fi is the solution of the objective function, that is, the matching degree of the edge points of the ellipse, and the set of candidate ellipse pixel points can be expressed as X _{i =} (s ₁ ,s ₂ ,…,s _N ), N is the number of all edge points existing on the candidate ellipse, and the objective function can be expressed as the matching degree of ellipse edge points: X _i (s _1x ,s _1y ) is defined as where (s _1x , s _1y ) is the coordinates of any pixel in the pixel set S in the candidate ellipse. 7. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, it is characterized in that, in described step 4, hire bee to search and produce new honey source, calculate its fitness value and adopt greedy method to select better honey source Specifically include: search for new honey sources near the initialized honey source, as shown in the following formula: In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1,D], which means that the hired bee randomly selects one dimension to search; j∈{1,2,… ,NP}, j≠i, x _jd means randomly selecting a honey source not equal to i among NP honey sources; It is a random number evenly distributed in the range of [-1,1], which determines the magnitude of the disturbance. When the fitness of the nectar source v _id is better than x _id , use the greedy selection method to replace x _id with v _id , otherwise keep x _id , that is to generate a new candidate ellipse, the greedy principle is expressed as: Among them, f(v _i ) is the objective function value of the newly generated location, f( _xi ) is the objective function value of the current honey source, if and only if the objective function value of the newly generated location is smaller than the objective function value of the original honey source location, Update the location of the current honey source. 8. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, the probability formula that each nectar source is selected in the described step 5 is: According to the selection probability determined by the above formula, the following bee selects the nectar source according to the roulette principle and conducts a fine search near it. The search formula is: In the formula: x _id is the position of the nectar source i randomly generated in the search space, d is a random integer in [1, D], which means that the hired bee randomly selects one dimension to search. 9. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, is characterized in that, in described step 6, nectar source x _id does not find better nectar source through threshold value limit times iteration, then this nectar source will be is abandoned, the corresponding hired bee becomes a scout bee, and the scout bee will randomly generate a new honey source in the search space to replace x _id , that is, the candidate ellipse is updated. The above process can be expressed as: 10. the detection method of the white blood cell based on artificial bee colony algorithm as claimed in claim 1, it is characterized in that, the position of the best nectar source of memory described in the described step 7 is to obtain optimal solution according to above each search formula namely optimal solution. Candidate ellipses that are close to the target ellipse.