[go: up one dir, main page]

CN112200836A - Multi-cell tracking method and system based on ant self-adjusting foraging behavior - Google Patents

Multi-cell tracking method and system based on ant self-adjusting foraging behavior Download PDF

Info

Publication number
CN112200836A
CN112200836A CN202011040382.2A CN202011040382A CN112200836A CN 112200836 A CN112200836 A CN 112200836A CN 202011040382 A CN202011040382 A CN 202011040382A CN 112200836 A CN112200836 A CN 112200836A
Authority
CN
China
Prior art keywords
ant
pheromone
ant colony
group
gaussian
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.)
Granted
Application number
CN202011040382.2A
Other languages
Chinese (zh)
Other versions
CN112200836B (en
Inventor
鲁明丽
徐本连
吴妹英
施健
刘静
王伟
朱培逸
吴迪
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.)
Fifth People's Hospital Of Suzhou
Suzhou Institute Of Technology
Original Assignee
Fifth People's Hospital Of Suzhou
Changshu Institute of 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 Fifth People's Hospital Of Suzhou, Changshu Institute of Technology filed Critical Fifth People's Hospital Of Suzhou
Priority to CN202011040382.2A priority Critical patent/CN112200836B/en
Publication of CN112200836A publication Critical patent/CN112200836A/en
Application granted granted Critical
Publication of CN112200836B publication Critical patent/CN112200836B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Catching Or Destruction (AREA)

Abstract

The invention discloses a multi-cell tracking method and system based on ant self-adjusting foraging behavior. After a cell image sequence is input, predicting a current frame according to an iteration result of an ant colony and an pheromone field of a previous frame to form a Gaussian ant colony and a Gaussian pheromone field; constructing an pheromone field by utilizing three strategies, namely, ant colony foraging range limitation, an ant colony resampling mechanism and an ant colony foraging stopping criterion, under an exponential form-based ant working mode; and then calculating the existence probability of the ant colony, fusing similar ant colonies, and extracting the cell state by considering the ant colony with the existence probability greater than a threshold value and a corresponding Gaussian pheromone field to realize accurate tracking of multiple cells.

Description

Multi-cell tracking method and system based on ant self-adjusting foraging behavior
Technical Field
The invention belongs to the field of multi-target tracking, and particularly relates to a multi-cell tracking system for self-adjusting foraging behavior based on ants.
Background
The proliferation, differentiation and migration of cells, which are the fundamental structural and functional units of an organism, are not only essential links for the development, evolution and maintenance of any organically-living embryo, but also are inseparable linked to the development and progression of disease, for example, most cancers are the spread of cancer due to the transfer of cancer cells from the original diseased tissue to the surrounding healthy tissue. Therefore, the study of cellular behavior analysis is very valuable in many fields. The characteristic information of the shape, the number, the speed, the track, the growth cycle and the like of the cells is obtained by analyzing the movement process of the cells, qualitative and quantitative analysis can be carried out on the cells, and thus, the research on the cell dynamics behavior has unpredictable significance for diagnosing and treating diseases, improving the medicine development efficiency and the like. Traditional manual tracking is still the most common tracking method in laboratories at present, but the method is not only time-consuming and error-prone, but also has high requirements on the professional knowledge and clinical experience of researchers. Therefore, the research of the cell automatic tracking method becomes an urgent problem to be solved.
The tracking of cells in a microscopic image sequence faces many difficulties, such as low signal-to-noise ratio, cell boundary blurring, cell deformation, time-varying numbers and sudden changes in the speed and direction of cell migration motion, resulting in failure of tracking, and the like. Most of the automatic tracking methods proposed by researchers in recent years are directed to specific data, and the tracking accuracy depends strongly on the result of cell segmentation, and the accuracy of cell tracking is not high in a complicated case. At present, few research theories and methods are used for cell tracking, such as deformation, different motion characteristics, time-varying number and the like in a cell image sequence. The invention aims to solve the difficult problems of multi-cell tracking such as deformation, different motion characteristics, time-varying number and the like, and utilizes ants to self-regulate foraging behaviors to construct a pheromone field so as to finally realize accurate tracking of multi-cells under complex conditions.
Disclosure of Invention
1. The object of the invention is to provide a method for producing a high-quality glass.
The invention provides a multi-cell tracking system for self-adjusting foraging behavior based on ants, aiming at solving the problems of time-varying number, deformation, different motion characteristics and adjacent multi-cell tracking in the prior art.
2. The technical scheme adopted by the invention is disclosed.
The invention discloses a multi-cell tracking method based on ant self-adjusting foraging behavior, which comprises the following steps:
an ant colony and pheromone field double-prediction step, wherein an original image is input, and based on the ant colony and the result pheromone field of the previous frame, the ant colony and the pheromone field of the current frame are predicted by utilizing a Gaussian model;
the ants self-adjust foraging step, which utilizes three strategies under an exponential-form-based ant working mode, namely limiting an ant colony foraging range, an ant colony resampling mechanism and an ant colony foraging stopping criterion to construct an pheromone field, and realizes deformation, number change, uncertain motion state and adjacent cell tracking;
a cell state estimation step, namely calculating the existence probability of each sub-ant colony based on the result pheromone field and a heuristic function, deleting the ant colony of which the existence probability is smaller than a threshold value, fusing similar ant colonies and considering the ant colony of which the existence probability is larger than the threshold value to carry out multi-cell state estimation;
wherein, the ants self-adjust foraging steps are as follows:
1) inputting a cell image sequence, and constructing an exponential-form-based ant working mode in an pheromone field; the foraging behavior of ants is limited in a specified range, and pixels in the neighborhood are selected according to the probability for searching; if an ant/in the v-th ant group is in a position of a pixel (i ', j'), considering that the exponential form is sensitive to the variation of the variable, the probability that the ant selects a certain pixel (i, j) in its available neighborhood is:
Figure BDA0002706450010000021
wherein
Figure BDA0002706450010000022
For the pheromone, η, of the t-th iteration on the v-th ant colony pixel (i, j)i,jIs a heuristic function of the pixel (i, j); alpha and beta are pheromones
Figure BDA0002706450010000023
And likelihood function ηi,jQ (i ', j') is the set of available neighborhoods for pixel (i ', j'); Ω (i ', j') is the intersection of the neighborhood of the pixel (i ', j') and the range of foraging defined by the ant;
heuristic function etai,jIs defined as
Figure BDA0002706450010000024
Wherein Δ Ii,jAnd Δ Ai,jRespectively representing the pixel intensity difference and the mean value in the neighborhood of the pixel (i, j); λ is the adjustment coefficient, determining Δ Ii,jAnd Δ Ai,jThe influence of two variables on the heuristic function value is defined as λ ═ Imean/ImaxIn which ImeanAnd ImaxRespectively representing the average intensity and the maximum intensity of the current frame pixel; gamma and kappa are both regulating coefficients to ensure that the heuristic function value is in the interval [0, 1%]Change in between; if the heuristic function value of the image background is larger than the foreground, the gamma is 1, otherwise, the gamma is 2; if the heuristic function value is larger than 1, k is equal to 1, otherwise k is equal to 0;
2) when all ants complete the search, the amount of the pheromone on pixel (i, j) is updated
Figure BDA0002706450010000025
Wherein
Figure BDA0002706450010000026
Is as followsv the amount of pheromone on pixel (i, j) for the t-th iteration of the colony, ρ (0 < ρ < 1) representing the pheromone residual coefficient,
Figure BDA0002706450010000027
the amount of pheromone released by ant l on pixel (i, j) for the t-1 st iteration;
3) after all ants in the v-th group of ant colony complete the t-th iteration, Gaussian fitting is carried out on the v-th group of pheromone field to obtain a v-th group of Gaussian pheromone field, and pheromone variables obey the mean value of
Figure BDA0002706450010000031
Covariance of
Figure BDA0002706450010000032
Gaussian distribution of
Figure BDA0002706450010000033
Wherein tau is(ν)(t) is the ν -th group of Gaussian pheromone fields after the tth iteration, τ (t) represents pheromone variables,
Figure BDA0002706450010000034
and
Figure BDA0002706450010000035
respectively mean value and covariance of the v group Gaussian pheromone field pheromone quantity after the t iteration;
4) calculating KL distance D between the v group of Gaussian pheromone fields after the t iteration and the v group of Gaussian pheromone fields after the t-1 iterationτ(t) if Dτ(t) if the t is larger than the threshold epsilon, performing Gaussian resampling on the v-th group of ant colony after the t iteration to obtain a Gaussian ant colony
Figure BDA0002706450010000036
Initial distribution for the v th ant colony for the t +1 th iteration, where x(ν,l)(t +1) denotes the status of the v-th group ant l for the t +1 th iteration, and x(ν,l)(t +1) obeys an average of
Figure BDA0002706450010000037
Covariance of
Figure BDA0002706450010000038
The distribution of the gaussian component of (a) is,
Figure BDA0002706450010000039
reconstructing the ant foraging range of the ν -th group of ant colony in the (t +1) th iteration according to the ν -th group of Gaussian pheromone field after the t-th iteration;
5) continuing the steps 1) to 4) until Dτ(t) is less than a threshold epsilon and an iteration termination condition is reached.
Preferably, the ant colony and pheromone field double prediction step includes the following specific prediction modes:
1) setting N in the k-1 th framek-1The individual cells, the ant colony and pheromone field corresponding to the cells also have Nk-1V (1., N) thk-1) The ant group comprises M ants, and the state is expressed as
Figure BDA00027064500100000310
Obey mean value of
Figure BDA00027064500100000311
Covariance of
Figure BDA00027064500100000312
Gaussian distribution of
Figure BDA00027064500100000313
Wherein l represents an ant, xk-1Is the ant state variable of the (k-1) th frame,
Figure BDA00027064500100000314
and
Figure BDA00027064500100000315
respectively mean value and covariance of the v group ant state variables of the k-1 frame; state of ants l in v-th group ant colony of frame k-1
Figure BDA00027064500100000316
Predict the k frame as
Figure BDA00027064500100000317
Wherein F is a state transition matrix and wherein,
Figure BDA00027064500100000318
representing the predicted state of ant l in the v group of ant colony of the k frame, subject to the mean value of
Figure BDA00027064500100000319
Covariance of
Figure BDA00027064500100000320
Gaussian distribution of
Figure BDA00027064500100000321
xk|k-1Indicating that the ant state variable of the k-th frame is predicted,
Figure BDA00027064500100000322
where "Σ" denotes a summation operation,
Figure BDA00027064500100000323
t represents matrix transposition;
2) suppose that the v group pheromone field of the k-1 frame
Figure BDA00027064500100000324
In accordance with a mean value of
Figure BDA00027064500100000325
Covariance of
Figure BDA00027064500100000326
Gaussian distribution of
Figure BDA00027064500100000327
Then the v-th group pheromone field predicts the k-th frame as
Figure BDA00027064500100000328
Wherein,
Figure BDA00027064500100000329
predicting the pheromone field, tau, for the kth frame of the set vk|k-1Representing the k-th frame pheromone variable.
Preferably, the cell state estimating step is performed in the following specific manner:
1) calculating the existence probability r(ν): when the iteration termination condition is reached and all ants in the v-th group of ant colony finish moving, an pheromone field tau is formed(ν)Calculating the existence probability of the v group of ant colony,
Figure BDA0002706450010000041
Figure BDA0002706450010000042
for a v-th group of pixels of an information element field
Figure BDA0002706450010000043
The amount of the pheromone on the surface,
Figure BDA0002706450010000045
is a pixel
Figure BDA0002706450010000044
Value of the heuristic function of(ν)A pixel set in the cell outline corresponding to the v-th group of ant colony of the current frame;
2) when the existence probability of the v-th ant colony is greater than a preset threshold value, the v-th ant colony is reserved, otherwise, the v-th ant colony is deleted;
3) and after fusing the space similar ant colony, performing cell state estimation.
Preferably, the presence probability of the ν -th ant colony is set to 0.5, which is greater than a preset threshold value.
The invention provides a multi-cell tracking system, which comprises a memory and a processor, wherein the memory stores a computer program and is characterized in that; the processor realizes the method steps when executing the computer program.
3. The invention has the following effects:
1) the multi-cell tracking system based on the ant self-adjusting foraging behavior provided by the invention utilizes three strategies, namely, ant colony foraging range limitation, ant colony resampling mechanism and ant colony foraging stopping criterion, under an exponential form-based ant working mode to construct an pheromone field, can solve the multi-cell tracking problems of deformation, different motion characteristics, time-varying number, close neighbor and the like, and obtains good tracking effect, and has a wider application range;
2) the method designed by the invention has high tracking precision; compared with an ant colony multicellular tracking algorithm based on pheromone prediction, a particle filter cell tracking method proposed by SMAL Boshi and a multi-Bernoulli filter and a Gaussian mixture PHD filter proposed by the teaching of REZA, precision ratios precision (P), recall ratios Recall (R) and F1-measurement are improved.
Drawings
The invention is further illustrated with reference to the following figures and examples.
Fig. 1 is a multicellular tracking system that self-regulates foraging behavior based on ants without considering data association.
Fig. 2 shows the 3D tracking result of sequence 1.
Fig. 3 shows the 3D tracking result of sequence 2.
Fig. 4 is an evolution process for defining the foraging range of an ant colony.
FIG. 5 is a Gaussian ant colony resampling evolution process in an iterative process.
Detailed Description
As shown in fig. 1, after a cell image sequence is input, a current frame is predicted according to iteration results of an ant colony and an pheromone field of a previous frame to form a gaussian ant colony and a gaussian pheromone field; constructing an pheromone field by utilizing three strategies, namely, ant colony foraging range limitation, an ant colony resampling mechanism and an ant colony foraging stopping criterion, under an exponential form-based ant working mode; and then calculating the existence probability of the ant colony, fusing similar ant colonies, and extracting the cell state by considering the ant colony with the existence probability greater than a threshold value and a corresponding Gaussian pheromone field to realize accurate tracking of multiple cells.
The prediction mode in the ant colony and pheromone field double prediction step is as follows:
1) setting N in the k-1 th framek-1The individual cells, the ant colony and pheromone field corresponding to the cells also have Nk-1N, (v ═ 1.,. N)k-1) The ant group comprises M ants, and the state is expressed as
Figure BDA0002706450010000051
Obey mean value of
Figure BDA0002706450010000052
Covariance of
Figure BDA0002706450010000053
Gaussian distribution of
Figure BDA0002706450010000054
Wherein l represents an ant, xk-1Is the ant state variable of the (k-1) th frame,
Figure BDA0002706450010000055
and
Figure BDA0002706450010000056
mean and covariance of the state variables of the v-th group of ant in the k-1 th frame, respectively. State of ants l in v-th group ant colony of frame k-1
Figure BDA0002706450010000057
Predict the k frame as
Figure BDA0002706450010000058
Wherein F is a state transition matrix and wherein,
Figure BDA0002706450010000059
representing the predicted state of ant l in the v group of ant colony of the k frame, subject to the mean value of
Figure BDA00027064500100000510
Covariance of
Figure BDA00027064500100000511
Gaussian distribution of
Figure BDA00027064500100000512
xk|k-1Indicating that the ant state variable of the k-th frame is predicted,
Figure BDA00027064500100000513
where "Σ" denotes a summation operation,
Figure BDA00027064500100000514
t represents matrix transposition;
2) suppose that the v group pheromone field of the k-1 frame
Figure BDA00027064500100000515
In accordance with a mean value of
Figure BDA00027064500100000516
Covariance of
Figure BDA00027064500100000517
Gaussian distribution of
Figure BDA00027064500100000518
The v-th group ant colony pheromone field predicts the k-th frame as
Figure BDA00027064500100000519
Wherein,
Figure BDA00027064500100000520
predicting the pheromone field, tau, for the kth frame of the set vk|k-1Represents the k frame pheromone variable;
the ant self-adjusting foraging steps are as follows:
1) inputting a cell image sequence, and constructing an exponential-form-based ant working mode in the pheromone field. The foraging behavior of ants is limited in a specified range and pixels in the neighborhood are selected according to the probability for searching. If an ant/in the v-th ant group is in a position of a pixel (i ', j'), considering that the exponential form is sensitive to the variation of the variable, the probability that the ant selects a certain pixel (i, j) in its available neighborhood is:
Figure BDA00027064500100000521
wherein
Figure BDA00027064500100000522
For the pheromone, η, of the t-th iteration on the v-th ant colony pixel (i, j)i,jIs a heuristic function of the pixel (i, j). Alpha and beta are pheromones
Figure BDA00027064500100000523
And likelihood function ηi,jQ (i ', j') is the set of available neighborhoods for pixel (i ', j'); Ω (i ', j') is the intersection of the neighborhood of the pixel (i ', j') and the range of foraging defined by an ant.
Heuristic function etai,jIs defined as
Figure BDA0002706450010000061
Wherein Δ Ii,jAnd Δ Ai,jRepresenting the pixel intensity difference and mean in the neighborhood of pixel (i, j), respectively. λ is the adjustment coefficient, determining Δ Ii,jAnd Δ Ai,jThe influence of two variables on the heuristic function value is defined as λ ═ Imean/ImaxIn which ImeanAnd ImaxRespectively representing the average intensity and the maximum intensity of the current frame pixel. Gamma and kappa are both regulating coefficients to ensure that the heuristic function value is in the interval [0, 1%]To change between. If the heuristic function value of the image background is larger than the foreground, gamma is 1, otherwise gamma is 2. Also, if the heuristic function value is greater than 1, k is equal to 1, otherwise k is equal to 0.
2) When all ants complete the search, the amount of the pheromone on pixel (i, j) is updated
Figure BDA0002706450010000062
Wherein
Figure BDA0002706450010000063
The amount of pheromone on pixel (i, j) for the t-th iteration of the v-th colony, ρ (0 < ρ < 1) represents the pheromone residual coefficient,
Figure BDA0002706450010000064
the amount of pheromone released by ant l on pixel (i, j) for the t-1 st iteration;
3) after all ants in the v-th group of ant colony complete the t-th iteration, Gaussian fitting is carried out on the v-th group of pheromone field to obtain a v-th group of Gaussian pheromone field, and pheromone variables obey the mean value of
Figure BDA0002706450010000065
Covariance of
Figure BDA0002706450010000066
Gaussian distribution of
Figure BDA0002706450010000067
Wherein tau is(ν)(t) is the ν -th group of Gaussian pheromone fields after the tth iteration, τ (t) represents pheromone variables,
Figure BDA0002706450010000068
and
Figure BDA0002706450010000069
respectively mean value and covariance of the v group Gaussian pheromone field pheromone quantity after the t iteration;
4) calculating KL distance D between the v group of Gaussian pheromone fields after the t iteration and the v group of Gaussian pheromone fields after the t-1 iterationτ(t) if Dτ(t) if the t is larger than the threshold epsilon, performing Gaussian resampling on the v-th group of ant colony after the t iteration to obtain a Gaussian ant colony
Figure BDA00027064500100000610
Initial distribution of the ν th ant colony for the t +1 th iteration, where x(ν,l)(t +1) denotes the status of the v-th group ant l for the t +1 th iteration, and x(v,l)(t +1) obeys an average of
Figure BDA00027064500100000611
Covariance of
Figure BDA00027064500100000612
The distribution of the gaussian component of (a) is,
Figure BDA00027064500100000613
meanwhile, reconstructing the foraging range of the v-th group of ant colony in the t +1 th iteration according to the v-th group of Gaussian pheromone fields after the t-th iteration;
5) continuing the steps 1) to 4) until Dτ(t) is less than a threshold epsilon and an iteration termination condition is reached.
The specific manner of the cell state estimation step is as follows:
1) calculating the existence probability r(τ): when the iteration termination condition is reached and all ants in the v-th group of ant colony finish moving, an pheromone field tau is formed(ν)Calculating the existence probability of the v-th ant colony,
Figure BDA0002706450010000071
Figure BDA0002706450010000072
for a v-th group of pixels of an information element field
Figure BDA0002706450010000073
The amount of the pheromone on the surface,
Figure BDA0002706450010000074
is a pixel
Figure BDA0002706450010000075
Value of the heuristic function of(ν)The v-th ant colony for the current frame corresponds to the set of pixels within the cell outline.
2) When the existence probability of the v-th ant colony is greater than a preset threshold value (set as 0.5 in the invention), the v-th ant colony is reserved, otherwise, the v-th ant colony is deleted;
3) and after fusing the space similar ant colony, performing cell state estimation.
The result of the multi-cell tracking through the ant colony and pheromone field double prediction step, the ant self-adjusting foraging step and the cell state estimation step is shown as follows.
Figure 1 is a multicellular tracking system based on ant self-regulated foraging behavior. Fig. 2 is a 3D tracking result of the sequence 1, and fig. 3 is a 3D tracking result of the sequence 1, which shows that the situations of cell appearance, cell disappearance, cell proximity and the like exist, and all the situations are reliably tracked. Fig. 4 is an evolution process for defining the foraging range of the ant colony, and it can be seen that the foraging range of the ant colony approaches to the cell contour as the number of iterations increases. FIG. 5 is a Gaussian ant colony resampling evolution process in an iterative process.
Compared with an ant colony multicellular tracking algorithm based on pheromone prediction, a particle filter cell tracking method proposed by SMAL doctor, a multi-Bernoulli filter and a Gaussian mixture PHD filter proposed by the teaching of REZA, the method provided by the invention has the advantages that precision (P), echo rate Recall (R) and F1-measurement are improved, and the method is shown in a table I and a table II.
TABLE-comparison of tracking Performance for different algorithms (sequence 1)
Figure BDA0002706450010000076
TABLE two different algorithms tracking Performance comparison (sequence 2)
Figure BDA0002706450010000077
In conclusion, the technical scheme of the invention can solve the problem of multi-cell tracking when the number of time-varying, deformation and movement characteristics are different. For the situations of cell near-distance interaction, cell entering or leaving view, time-varying cell movement speed and direction and the like, three strategies are considered by utilizing ant self-adjusting foraging behavior, namely, an ant colony foraging range, an ant colony resampling mechanism and an ant colony foraging stopping criterion are limited to construct an pheromone field, and accurate tracking of multiple cells under complex situations is realized.

Claims (5)

1. A multi-cell tracking method based on ant self-regulation foraging behavior is characterized by comprising the following steps:
an ant colony and pheromone field double-prediction step, wherein an original image is input, and based on the ant colony and the result pheromone field of the previous frame, the ant colony and the pheromone field of the current frame are predicted by utilizing a Gaussian model;
the ants self-adjust foraging step, which utilizes three strategies under an exponential-form-based ant working mode, namely limiting an ant colony foraging range, an ant colony resampling mechanism and an ant colony foraging stopping criterion to construct an pheromone field, and realizes deformation, number change, uncertain motion state and adjacent cell tracking;
a cell state estimation step, namely calculating the existence probability of each sub-ant colony based on the result pheromone field and a heuristic function, deleting the ant colony of which the existence probability is smaller than a threshold value, fusing similar ant colonies and considering the ant colony of which the existence probability is larger than the threshold value to carry out multi-cell state estimation;
wherein, the ants self-adjust foraging steps are as follows:
1) inputting a cell image sequence, and constructing an exponential-form-based ant working mode in an pheromone field; the foraging behavior of ants is limited in a specified range, and pixels in the neighborhood are selected according to the probability for searching; if an ant/in the v-th ant group is in a position of a pixel (i ', j'), considering that the exponential form is sensitive to the variation of the variable, the probability that the ant selects a certain pixel (i, j) in its available neighborhood is:
Figure FDA0002706450000000011
wherein
Figure FDA0002706450000000012
For the pheromone, η, of the t-th iteration on the v-th ant colony pixel (i, j)i,jIs a heuristic function of the pixel (i, j); alpha and beta are pheromones
Figure FDA0002706450000000013
And likelihood function ηi,jQ (i ', j') is the set of available neighborhoods for pixel (i ', j'); Ω (i ', j') is the intersection of the neighborhood of the pixel (i ', j') and the range of foraging defined by the ant;
heuristic function etai,jIs defined as
Figure FDA0002706450000000014
Wherein Δ Ii,jAnd Δ Ai,jRespectively representing the pixel intensity difference and the mean value in the neighborhood of the pixel (i, j); λ is the adjustment coefficient, determining Δ Ii,jAnd Δ Ai,jThe influence of two variables on the heuristic function value is defined as λ ═ Imean/ImaxIn phase ImeanAnd ImaxRespectively representing the average intensity and the maximum intensity of the current frame pixel; gamma and kappa are both regulating coefficients to ensure that the heuristic function value is in the interval [0, 1%]Change in between; if the heuristic function value of the image background is larger than the foreground, the gamma is 1, otherwise, the gamma is 2; if the heuristic function value is larger than 1, k is equal to 1, otherwise k is equal to 0;
2) when all ants complete the search, the amount of the pheromone on pixel (i, j) is updated
Figure FDA0002706450000000021
Wherein
Figure FDA0002706450000000022
The amount of pheromone on pixel (i, j) for the t-th iteration of the v-th colony, ρ (0 < ρ < 1) represents the pheromone residual coefficient,
Figure FDA0002706450000000023
the amount of pheromone released by ant l on pixel (i, j) for the t-1 st iteration;
3) after all ants in the v-th group of ant colony complete the t-th iteration, Gaussian fitting is carried out on the v-th group of pheromone field to obtain a v-th group of Gaussian pheromone field, and pheromone variables obey the mean value of
Figure FDA0002706450000000024
Covariance of
Figure FDA0002706450000000025
Gaussian distribution of
Figure FDA0002706450000000026
Wherein tau is(ν)(t) is the ν -th group of Gaussian pheromone fields after the tth iteration, τ (t) represents pheromone variables,
Figure FDA0002706450000000027
and
Figure FDA0002706450000000028
respectively mean value and covariance of the v group Gaussian pheromone field pheromone quantity after the t iteration;
4) calculating KL distance D between the v group of Gaussian pheromone fields after the t iteration and the v group of Gaussian pheromone fields after the t-1 iterationτ(t) if Dτ(t) if the t is larger than the threshold epsilon, performing Gaussian resampling on the v-th group of ant colony after the t iteration to obtain a Gaussian ant colony
Figure FDA0002706450000000029
Initial distribution for the v th ant colony for the t +1 th iteration, where x(v,l)(t +1) denotes the state of the v-th group of ant l for the t +1 th iteration, and x(v,l)(t +1) obeys an average of
Figure FDA00027064500000000210
Covariance of
Figure FDA00027064500000000211
The distribution of the gaussian component of (a) is,
Figure FDA00027064500000000212
reconstructing the ant foraging range of the ν -th group of ant colony in the (t +1) th iteration according to the ν -th group of Gaussian pheromone field after the t-th iteration;
5) continuing the steps 1) to 4) until Dτ(t) is less than a threshold epsilon and an iteration termination condition is reached.
2. The ant-based multicellular tracking method for self-adjusting foraging behavior of ants according to claim 1, wherein the ant colony and pheromone field double prediction steps are specifically as follows:
1) setting N in the k-1 th framek-1The individual cells, the ant colony and pheromone field corresponding to the cells also have Nk-1N, (v ═ 1.,. N)k-1) The ant group comprises M ants, and the state is expressed as
Figure FDA00027064500000000213
Obey mean value of
Figure FDA00027064500000000214
Covariance of
Figure FDA00027064500000000215
Gaussian distribution of
Figure FDA00027064500000000216
Wherein l represents an ant, xk-1Is the ant state variable of the (k-1) th frame,
Figure FDA00027064500000000217
and
Figure FDA00027064500000000218
respectively mean value and covariance of the v group ant state variables of the k-1 frame; state of ants l in v-th group ant colony of frame k-1
Figure FDA00027064500000000219
Predict the k frame as
Figure FDA00027064500000000220
Wherein F is a state transition matrix and wherein,
Figure FDA00027064500000000221
representing the predicted state of ant l in the v group of ant colony of the k frame, subject to the mean value of
Figure FDA00027064500000000222
Covariance of
Figure FDA00027064500000000223
Gaussian distribution of
Figure FDA0002706450000000031
xk|k-1Indicating that the ant state variable of the k-th frame is predicted,
Figure FDA0002706450000000032
where "Σ" denotes a summation operation,
Figure FDA0002706450000000033
t represents matrix transposition;
2) suppose that the v group pheromone field of the k-1 frame
Figure FDA0002706450000000034
In accordance with a mean value of
Figure FDA0002706450000000035
Covariance of
Figure FDA0002706450000000036
Gaussian distribution of
Figure FDA0002706450000000037
Then the v-th group pheromone field predicts the k-th frame as
Figure FDA0002706450000000038
Wherein,
Figure FDA0002706450000000039
predicting the pheromone field, tau, for the kth frame of the v-th groupk|k-1Representing the k-th frame pheromone variable.
3. The ant-based multicellular tracking method for self-regulating foraging behavior of claim 1, wherein the cell state estimation step is implemented in a specific manner as follows:
1) calculating the existence probability r(ν): when the iteration termination condition is reached and all ants in the v-th group of ant colony finish moving, an pheromone field tau is formed(ν)Calculating the existence probability of the v group ant colony
Figure FDA00027064500000000310
Figure FDA00027064500000000311
For a v-th group of pixels of an information element field
Figure FDA00027064500000000312
The amount of the pheromone on the surface,
Figure FDA00027064500000000314
is a pixel
Figure FDA00027064500000000313
Value of the heuristic function of(ν)A pixel set in the cell outline corresponding to the v-th group of ant colony of the current frame;
2) when the existence probability of the v-th ant colony is greater than a preset threshold value, the v-th ant colony is reserved, otherwise, the v-th ant colony is deleted;
3) and after fusing the space similar ant colony, performing cell state estimation.
4. The ant-based multicellular tracking method of self-regulating foraging behavior of claim 3, wherein: and setting the existence probability of the ν th ant colony to be greater than a preset threshold value to be 0.5.
5. A multi-cell tracking system, comprising: comprising a memory and a processor, the memory storing a computer program, characterized in that; the processor, when executing the computer program, realizes the method steps of any of claims 1-4.
CN202011040382.2A 2020-09-28 2020-09-28 Multi-cell tracking method and system based on self-adjusting foraging behavior of ants Active CN112200836B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011040382.2A CN112200836B (en) 2020-09-28 2020-09-28 Multi-cell tracking method and system based on self-adjusting foraging behavior of ants

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011040382.2A CN112200836B (en) 2020-09-28 2020-09-28 Multi-cell tracking method and system based on self-adjusting foraging behavior of ants

Publications (2)

Publication Number Publication Date
CN112200836A true CN112200836A (en) 2021-01-08
CN112200836B CN112200836B (en) 2021-10-19

Family

ID=74006763

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011040382.2A Active CN112200836B (en) 2020-09-28 2020-09-28 Multi-cell tracking method and system based on self-adjusting foraging behavior of ants

Country Status (1)

Country Link
CN (1) CN112200836B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116777850A (en) * 2023-06-09 2023-09-19 长春理工大学 A cell tracking technology method based on deep learning and ant colony algorithm
CN120013199A (en) * 2025-04-14 2025-05-16 华南理工大学 A scheduling optimization method and system based on shipbuilding

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100069832A1 (en) * 2008-08-08 2010-03-18 Neal Solomon Intelligent medical device system dynamics for biological network regulation
CN102999922A (en) * 2012-11-19 2013-03-27 常熟理工学院 Multi-cell automatic tracking method and system based on plurality of task ant systems
US20180003841A1 (en) * 2015-02-05 2018-01-04 Schlumberger Technology Corporation Seismic Attributes Derived from The Relative Geological Age Property of A Volume-Based Model
CN107871156A (en) * 2017-11-02 2018-04-03 常熟理工学院 Ant colony multi-cell tracking system based on pheromone prediction
CN109118518A (en) * 2018-07-24 2019-01-01 常熟理工学院 Double-deck more bernoulli stochastic finite ant colony many cells trackings
CN109159127A (en) * 2018-11-20 2019-01-08 广东工业大学 A kind of double welding robot intelligence paths planning methods based on ant group algorithm
US20200158898A1 (en) * 2017-05-23 2020-05-21 Schlumberger Technology Corporation Analogue Facilitated Seismic Data Interpretation System

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100069832A1 (en) * 2008-08-08 2010-03-18 Neal Solomon Intelligent medical device system dynamics for biological network regulation
CN102999922A (en) * 2012-11-19 2013-03-27 常熟理工学院 Multi-cell automatic tracking method and system based on plurality of task ant systems
US20180003841A1 (en) * 2015-02-05 2018-01-04 Schlumberger Technology Corporation Seismic Attributes Derived from The Relative Geological Age Property of A Volume-Based Model
US20200158898A1 (en) * 2017-05-23 2020-05-21 Schlumberger Technology Corporation Analogue Facilitated Seismic Data Interpretation System
CN107871156A (en) * 2017-11-02 2018-04-03 常熟理工学院 Ant colony multi-cell tracking system based on pheromone prediction
CN109118518A (en) * 2018-07-24 2019-01-01 常熟理工学院 Double-deck more bernoulli stochastic finite ant colony many cells trackings
CN109159127A (en) * 2018-11-20 2019-01-08 广东工业大学 A kind of double welding robot intelligence paths planning methods based on ant group algorithm

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
BENLIAN XU ET AL: "An Ant Colony Inspired Multi-Bernoulli", 《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS》 *
刘雄 等: "基于蚁群算法的移动机器人路径规划", 《计算机仿真》 *
林涛 等: "自适应蚁群算法在多目标跟踪中的应用", 《计算机仿真》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116777850A (en) * 2023-06-09 2023-09-19 长春理工大学 A cell tracking technology method based on deep learning and ant colony algorithm
CN116777850B (en) * 2023-06-09 2025-12-19 长春理工大学 Cell tracking technical method based on deep learning and ant colony algorithm
CN120013199A (en) * 2025-04-14 2025-05-16 华南理工大学 A scheduling optimization method and system based on shipbuilding

Also Published As

Publication number Publication date
CN112200836B (en) 2021-10-19

Similar Documents

Publication Publication Date Title
CN112700434B (en) Medical image classification method and classification device thereof
CN110264471A (en) A kind of image partition method, device, storage medium and terminal device
CN107256550A (en) A kind of retinal image segmentation method based on efficient CNN CRF networks
CN113192633A (en) Stomach cancer fine-grained classification method based on attention mechanism
CN118247508A (en) Medical image segmentation method based on deep reinforcement learning
CN111524140A (en) Medical image semantic segmentation method based on CNN and random forest method
CN112200836A (en) Multi-cell tracking method and system based on ant self-adjusting foraging behavior
CN112990359B (en) Image data processing method, device, computer and storage medium
Qian et al. Unet#: a Unet-like redesigning skip connections for medical image segmentation
CN114897922B (en) A tissue pathology image segmentation method based on deep reinforcement learning
CN112418227B (en) Monitoring video truck segmentation method based on double self-attention mechanism
CN112884737A (en) Automatic mitosis detection method in breast cancer pathological image based on multistage iteration
CN113762370A (en) Depth network set generation method combined with Gaussian random field
CN116844631A (en) A gene spatial expression prediction method based on tumor pathological images
CN119106740A (en) Interpretable data-driven semiconductor silicon single crystal growth diameter prediction method
CN119152193A (en) YOLO target detection method and system based on differential architecture search
CN108182684B (en) A kind of image segmentation method and device based on weighted kernel function fuzzy clustering
Qiu et al. Semi-supervised Strong-Teacher Consistency Learning for few-shot cardiac MRI image segmentation
CN119963062A (en) Inorganic mineral fiber production system and method based on intelligent control
CN110717930B (en) A mutation moving target tracking method based on extended Wang-landau Monte Carlo and KCF
Zhao et al. Faster mean-shift: Gpu-accelerated embedding-clustering for cell segmentation and tracking
CN112132842A (en) A brain image segmentation method based on SEEDS algorithm and GRU network
CN117115182A (en) Pathological image segmentation method based on multi-agent deep reinforcement learning
CN109146881B (en) A Cell Tracking System for Ant Colony Aggregation Based on Potential Estimation
CN117670921A (en) Multi-frame scene flow estimation method based on splatter transformation

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
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20210108

Assignee: Suzhou Yueya Biotechnology Co.,Ltd.

Assignor: CHANGSHU INSTITUTE OF TECHNOLOGY|THE FIFTH PEOPLE'S HOSPITAL OF SUZHOU

Contract record no.: X2023980036933

Denomination of invention: A Multicell Tracking Method and System Based on Ant Self Adjusting foraging Behavior

Granted publication date: 20211019

License type: Common License

Record date: 20230626

EE01 Entry into force of recordation of patent licensing contract
CP03 Change of name, title or address

Address after: 215500 Suzhou, Changshou City Province, South Ring Road No. 99, No., No. three

Patentee after: Suzhou Institute of Technology

Country or region after: China

Patentee after: THE FIFTH PEOPLE'S HOSPITAL OF SUZHOU

Address before: 215500 Suzhou, Changshou City Province, South Ring Road No. 99, No., No. three

Patentee before: CHANGSHU INSTITUTE OF TECHNOLOGY

Country or region before: China

Patentee before: THE FIFTH PEOPLE'S HOSPITAL OF SUZHOU

CP03 Change of name, title or address