Search

CN-121982708-A - Chromosome automatic identification method and system based on fuzzy logic and convolutional neural network

CN121982708ACN 121982708 ACN121982708 ACN 121982708ACN-121982708-A

Abstract

The invention discloses a chromosome automatic identification method and a chromosome automatic identification system based on fuzzy logic and a convolutional neural network, and relates to the technical field of chromosome identification, wherein the method comprises the following steps of obtaining a nuclear structure area and extracting morphological parameters related to each cell; the method comprises the steps of determining a attribution evaluation factor for microkernel attribution judgment, determining a correlation degree predicted value of a candidate microkernel relative to a main kernel, determining a category attribute of a candidate nuclear structure, and determining a final attribution result of the candidate microkernel relative to the main kernel according to the category attribute and the correlation degree predicted value. The invention builds a multi-dimensional evaluation system, can more accurately discriminate micronuclei and interferents, can effectively process complex situations that a plurality of main nuclei compete for the same micronuclei by introducing competitive attribution judging logic and combining cytological characteristic parameters of the main nuclei, simulates the judging thinking of expert, and leads attribution results to be more in accordance with biological reality.

Inventors

  • WU MENGYUN
  • ZHANG HUADONG
  • LI WEI

Assignees

  • 重庆市疾病预防控制中心(重庆市预防医学科学院)

Dates

Publication Date
20260505
Application Date
20260119

Claims (10)

  1. 1. The chromosome automatic identification method based on the fuzzy logic and the convolutional neural network is characterized by comprising the following steps of: acquiring all nuclear structure areas obtained by pretreatment in the target cell image, and extracting morphological parameters related to each cell; Determining a attribution evaluation factor for micronucleus attribution judgment according to the size and morphological parameters of the nuclear structure; acquiring local environment characteristics of an area where a nuclear structure is located, and determining a correlation degree predicted value of a candidate micronucleus relative to a main nucleus according to the attribution evaluation factor and the local environment characteristics; And determining the category attribute of the candidate nuclear structure, and determining the final attribution result of the candidate micronucleus relative to the main nucleus according to the category attribute and the association degree predicted value.
  2. 2. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network of claim 1, wherein determining a home evaluation factor for microkernel home determination based on the size and morphological parameters of the nuclear structure comprises: determining a nuclear area, a nuclear perimeter, and a nuclear shape factor included in the morphological parameters; Calculating the attribution assessment factor according to the nuclear structure size, the nuclear area, the nuclear perimeter and the nuclear shape factor.
  3. 3. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network as set forth in claim 2, wherein the obtaining the local environmental features of the region where the nuclear structure is located, and determining the association degree predicted value of the candidate microkernel with respect to the main kernel according to the attribution evaluation factor and the local environmental features comprises: Determining the space distance, the relative position included angle and the texture similarity between the candidate microkernels and the main kernels included in the local environment characteristics; and inputting the attribution evaluation factor, the spatial distance, the relative position included angle and the texture similarity into a pre-trained fuzzy logic system, and calculating to obtain a correlation pre-estimated value of the candidate microkernel relative to a main kernel.
  4. 4. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network of claim 3, wherein determining the class attribute of the candidate microkernels comprises: Determining the category attribute of the candidate nuclear structure as a main core, a microcore or an interference particle; Acquiring a preconfigured attribution judging rule, wherein the attribution judging rule comprises a relevance threshold value, a space constraint condition and a category priority; And determining the final attribution result of the candidate micronucleus relative to the nearest main nucleus through decision logic according to the category attribute, the association degree predicted value and the attribution judging rule.
  5. 5. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network of claim 4, wherein determining, by decision logic, a final attribution result of the candidate microkernel with respect to a nearest main kernel based on the category attribute, the association degree predicted value, and the attribution determination rule comprises: And under the condition that the category attribute of the candidate nuclear structure is a microkernel, the association degree pre-estimated value is larger than a first association degree threshold value, and the spatial distance between the candidate microkernel and the main kernel is smaller than a first distance threshold value, determining that the candidate microkernel belongs to the nearest main kernel according to the attribution judging rule.
  6. 6. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network of claim 5, wherein determining, by decision logic, a final attribution result of the candidate microkernel with respect to a nearest principal kernel based on the category attribute, the association degree predicted value, and the attribution determination rule further comprises: acquiring cytological characteristic parameters of each candidate main core under the condition that the category attribute of the candidate nuclear structure is microcores and a plurality of candidate main cores exist; Determining a final attribution main core of the candidate micronucleus through competitive attribution judging logic according to the association degree predicted value, the space distance and the cytological characteristic parameter; Wherein the cytological feature parameters include a principal nuclear area and staining intensity, and the competitive home decision logic preferentially selects a principal nuclear with a more significant cytological feature parameter as a home object.
  7. 7. The method for automatically identifying a chromosome based on fuzzy logic and convolutional neural network of claim 6, wherein determining, by decision logic, a final attribution result of the candidate microkernel with respect to a nearest principal kernel based on the class attribute, the association degree predicted value, and the attribution determination rule further comprises: Starting a manual rechecking marking mechanism under the condition that the category attribute of the candidate nuclear structure is microkernel, the association degree predicted value is in a preset confidence interval and the attribution cannot be definitely determined through the competitive attribution judging logic; The preset confidence interval is a range between a first association threshold and a second association threshold, and the second association threshold is higher than the first association threshold; the manual review marking mechanism comprises the steps of marking the candidate microkernels and associated main kernel candidate objects in the target cell image, and recording related morphological parameters, environment characteristic parameters and association degree pre-estimated values for an expert to review.
  8. 8. The chromosome automatic identification system based on the fuzzy logic and the convolutional neural network is applied to the chromosome automatic identification method based on the fuzzy logic and the convolutional neural network, which is disclosed in any one of claims 1 to 7, and is characterized in that the system comprises a parameter acquisition module, an evaluation factor determination module, a relevance estimation module and a attribution result determination module: the parameter acquisition module is used for acquiring all nuclear structure areas obtained by pretreatment in the target cell image and extracting morphological parameters related to each cell; the evaluation factor determining module is used for determining a attribution evaluation factor for micronucleus attribution judgment according to the size and morphological parameters of the nuclear structure; the association degree estimating module is used for acquiring local environment characteristics of the area where the nuclear structure is located and determining association degree estimated values of candidate microkernels relative to the main kernels according to the attribution evaluating factors and the local environment characteristics; the attribution result determining module is used for determining the category attribute of the candidate nuclear structure, and determining the final attribution result of the candidate micronucleus relative to the main nucleus according to the category attribute and the association degree predicted value.
  9. 9. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor executes the computer program to realize the steps of the chromosome automatic identification method based on the fuzzy logic and convolutional neural network as set forth in any one of claims 1 to 7.
  10. 10. A computer readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of the method for automatically identifying chromosomes based on fuzzy logic and convolutional neural network as set forth in any one of claims 1 to 7.

Description

Chromosome automatic identification method and system based on fuzzy logic and convolutional neural network Technical Field The invention relates to the technical field of chromosome identification, in particular to a chromosome automatic identification method and system based on fuzzy logic and convolutional neural network. Background Along with the deep research of biomedicine and genetics, the chromosome analysis has increasingly prominent roles in the fields of disease diagnosis, drug screening, radiation monitoring and the like. Among them, micronuclei are important biomarkers for chromosomal lesions, and their identification and statistics are critical for genotoxicity assessment. The traditional micronucleus identification mainly depends on observation by an artificial microscope, and the method has the problems of low efficiency, strong subjectivity, easy fatigue, poor consistency and the like. At present, although some automatic identification methods based on computer vision appear, in an attempt to locate a cell nucleus and calculate morphological parameters through an image processing technology, there are obvious limitations that firstly, the methods often depend on a single morphological feature, and it is difficult to accurately distinguish a main nucleus, a micronucleus and interfering particles (such as staining impurities or overlapping cell nuclei), so that the misjudgment rate is high, and secondly, when a plurality of candidate main nuclei exist in a complex cell environment, effective competitive attribution judgment logic is lacking, and the association relationship between the micronucleus and a specific main nucleus cannot be reliably determined. Therefore, in the related art, how to comprehensively utilize multi-dimensional features (morphology, space and texture) to simulate expert judgment logic and effectively process complex scenes and edge cases, so as to realize automatic identification and attribution judgment of chromosome micronuclei with high accuracy and high reliability, and no effective solution is proposed yet. Disclosure of Invention This section is intended to outline some aspects of embodiments of the application and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description of the application and in the title of the application, which may not be used to limit the scope of the application. The present invention has been made in view of the above-mentioned problems occurring in the prior art. In order to solve the technical problems, the invention provides the following technical scheme that the chromosome automatic identification method based on the fuzzy logic and the convolutional neural network comprises the following steps: acquiring all nuclear structure areas obtained by pretreatment in the target cell image, and extracting morphological parameters related to each cell; Determining a attribution evaluation factor for micronucleus attribution judgment according to the size and morphological parameters of the nuclear structure; acquiring local environment characteristics of an area where a nuclear structure is located, and determining a correlation degree predicted value of a candidate micronucleus relative to a main nucleus according to the attribution evaluation factor and the local environment characteristics; And determining the category attribute of the candidate nuclear structure, and determining the final attribution result of the candidate micronucleus relative to the main nucleus according to the category attribute and the association degree predicted value. As an optimal scheme of the chromosome automatic identification method and system based on the fuzzy logic and the convolutional neural network, the method for determining the attribution assessment factor for microkernel attribution judgment according to the size and the morphological parameters of the nuclear structure comprises the following steps: determining a nuclear area, a nuclear perimeter, and a nuclear shape factor included in the morphological parameters; Calculating the attribution assessment factor according to the nuclear structure size, the nuclear area, the nuclear perimeter and the nuclear shape factor. The method and the system for automatically identifying the chromosome based on the fuzzy logic and the convolutional neural network have the advantages that the local environment characteristics of the area where the nuclear structure is located are obtained, and the association degree pre-estimated value of the candidate micronucleus relative to the main nucleus is determined according to the attribution evaluation factor and the local environment characteristics, wherein the method comprises the following steps: Determining the space distance, the relative position included angle and the texture similarity between the candidate microkernels and the main kernels included in the local environment cha