CN119049049B - Dynamic detection method and system of mitochondrial fusion and fission based on image recognition - Google Patents
Dynamic detection method and system of mitochondrial fusion and fission based on image recognition Download PDFInfo
- Publication number
- CN119049049B CN119049049B CN202411550863.6A CN202411550863A CN119049049B CN 119049049 B CN119049049 B CN 119049049B CN 202411550863 A CN202411550863 A CN 202411550863A CN 119049049 B CN119049049 B CN 119049049B
- Authority
- CN
- China
- Prior art keywords
- contour
- mitochondria
- image
- serial number
- cell
- 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
Links
- 230000004927 fusion Effects 0.000 title claims abstract description 63
- 238000001514 detection method Methods 0.000 title claims abstract description 36
- 230000002438 mitochondrial effect Effects 0.000 title claims abstract description 32
- 230000004992 fission Effects 0.000 title abstract description 4
- 210000003470 mitochondria Anatomy 0.000 claims abstract description 148
- 210000004027 cell Anatomy 0.000 claims abstract description 113
- 230000004660 morphological change Effects 0.000 claims abstract description 29
- 238000000034 method Methods 0.000 claims abstract description 27
- 230000002159 abnormal effect Effects 0.000 claims abstract description 26
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 19
- 238000004458 analytical method Methods 0.000 claims description 20
- 230000000877 morphologic effect Effects 0.000 claims description 13
- 230000011218 segmentation Effects 0.000 claims description 11
- 238000013135 deep learning Methods 0.000 claims description 7
- 230000010354 integration Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000002073 fluorescence micrograph Methods 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 210000003855 cell nucleus Anatomy 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000003463 organelle Anatomy 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/693—Acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/752—Contour matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/698—Matching; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
本发明公开了一种基于图像识别的线粒体融合和分裂的动态检测方法及系统，属于图像识别技术领域，该方法包括：为每张细胞图像标记采集时间，基于采集时间生成时间图像序列；赋予每张细胞图像第一序号，基于轮廓检测算法对每张细胞图像进行轮廓识别，以定位各个线粒体的第一轮廓；在细胞图像中赋予每个第一轮廓第二序号，将各张细胞图像中的第一轮廓进行对比整合；截取包含相同第二序号的第一轮廓，整合生成每个线粒体的形态变化序列；分析形态变化序列以获取线粒体各种形态及对应的出现时间；对线粒体的形态进行分析，以定位出现异常的线粒体。本发明可以对线粒体的形态进行追踪，从而准确有效地识别线粒体在整个观察周期内的动态变化。
The present invention discloses a dynamic detection method and system of mitochondrial fusion and fission based on image recognition, which belongs to the field of image recognition technology. The method comprises: marking the acquisition time for each cell image, generating a time image sequence based on the acquisition time; assigning a first serial number to each cell image, and performing contour recognition on each cell image based on a contour detection algorithm to locate the first contour of each mitochondria; assigning a second serial number to each first contour in the cell image, and comparing and integrating the first contours in each cell image; intercepting the first contour containing the same second serial number, and integrating to generate a morphological change sequence of each mitochondria; analyzing the morphological change sequence to obtain various mitochondrial forms and corresponding appearance times; analyzing the morphology of mitochondria to locate abnormal mitochondria. The present invention can track the morphology of mitochondria, thereby accurately and effectively identifying the dynamic changes of mitochondria during the entire observation period.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a dynamic detection method and a system for mitochondrial fusion and division based on image recognition.
Background
Mitochondria are key organelles in most cells, and mitochondrial dynamics refers to the process that mitochondria continuously undergo dynamic fusion and division under the control of related proteins, thereby completing the dynamic change of the morphology thereof and further maintaining the overall stability of mitochondrial network structures. In addition, excessive fusion of mitochondria may cause abnormal expansion of mitochondrial network, excessive division may cause fragmentation of mitochondria, which may be related to occurrence of diseases, etc., and various image processing schemes have been proposed in the prior art to reduce artificial recognition pressure and accelerate recognition and identification of mitochondria in super-resolution images of cells.
The method is characterized in that a super-resolution fluorescence image is shot by a super-resolution fluorescence microscope, then an image segmentation model is built by a deep learning algorithm CNN-11 and a XGBoost model, and the image segmentation model is trained to realize the precise identification and segmentation of the mitochondrial outline in the super-resolution fluorescence image. However, the method can only realize division of the outline of mitochondria, and can not track the division and fusion process of mitochondria, so that the dynamic change of mitochondria in cells can not be accurately and effectively identified.
Disclosure of Invention
In order to solve the problems, the invention provides a method and a system for dynamically detecting mitochondrial fusion and division based on image recognition, so as to realize dynamic tracking of the mitochondrial fusion and division process.
In order to achieve the above object, the present invention provides a method for dynamically detecting mitochondrial fusion and division based on image recognition, comprising:
Acquiring cell images by using a high-resolution microscope, marking an acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time;
assigning a first serial number to each cell image in the time image sequence, and carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate a first contour of each mitochondria in the cell image;
Giving a second serial number to each first outline in the cell images, and comparing and integrating the first outlines in each cell image so that the first outlines corresponding to the same mitochondria have the same second serial number in different cell images;
Intercepting the first outline containing the same second serial number from each cell image, and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first serial number;
analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of mitochondria comprise a division state, a fusion state and a stable state;
the morphology of mitochondria was analyzed to locate abnormal mitochondria.
Further, in the morphological change sequence, if no abnormal dividing line appears inside the first contour, dividing the state of the corresponding mitochondria into the stable state;
If the abnormal dividing line appears in the first contour and the abnormal dividing line divides the first contour into a plurality of subareas, defining the first contour as a second contour;
extracting the first contour with the second serial number before and after the second contour and defining the first contour as a third contour and a fourth contour respectively, acquiring contour shapes of all the subareas in the second contour, and positioning contour curves corresponding to the contour shapes in the third contour and the fourth contour;
If the subarea exists, the profile curve corresponding to the profile shape is not existed in the third profile, the profile curve corresponding to the profile shape exists in the fourth profile, the state of mitochondria corresponding to the second profile is divided into the fusion state, and the subarea is defined as a fusion area;
If the subarea exists, the profile curve corresponding to the profile shape exists in the third profile, the profile curve corresponding to the profile shape does not exist in the fourth profile, the state of mitochondria corresponding to the second profile is divided into the division state, and the subarea is defined as a division area;
And if all the subareas do not have the contour curves corresponding to the contour shapes of the subareas in the third contour and the fourth contour, positioning the state of mitochondria corresponding to the second contour as the stable state.
Further, locating the contour curve corresponding to the contour shape includes the steps of:
Defining the cell image where the third contour is located as a first image, acquiring first coordinates of the contour shape centroids of the subareas to be compared, mapping the first coordinates into the first image, generating a detection area in the first image by taking the first coordinates as the center, intercepting a contour part of the third contour located in the detection area, selecting a plurality of calibration points on the third contour and the contour part, calculating the curvature of each calibration point, generating a first curvature distribution of the contour shape, calculating a second curvature distribution of the contour part, calculating the first curvature distribution and the second curvature distribution to obtain distribution similarity, and determining that the contour curve corresponding to the contour shape exists in the third contour if the distribution similarity is larger than a first threshold.
Further, the comparison integration of the first profile includes the following steps:
Selecting the cell images adjacent to the first serial number as a second image and a third image, selecting the first contour from the second image as a target contour, acquiring second coordinates of the second serial number and the mass center corresponding to the target contour, positioning third coordinates of the mass centers of the first contour of each mitochondria in the third image, mapping the second coordinates in the third image, defining the third coordinates with the distance from the second coordinates smaller than a second threshold value in the third image as adjacent coordinates, defining the first contour with the adjacent coordinates as the mass center as an adjacent contour, comparing the shape similarity of the target contour and each adjacent contour, correcting the second serial number of the adjacent contour with the largest shape similarity to be identical with the target contour, and repeating the steps to traverse all the first contour and the cell images.
Further, calculating the shape similarity of the target contour and the neighboring contour includes the steps of:
Covering the target contour on the adjacent contour by taking the mass center as a reference point, and calculating the shape similarity P of the target contour and the adjacent contour through a comparison formula, wherein the comparison formula is as follows: Wherein M 0 is the number of overlapping pixels after the target contour is overlaid on the adjacent contour, M 1 and M 2 are the number of pixels included in the target contour and the adjacent contour, respectively, max (M 1,M2) is the larger value returned to M 1 or M 2, and L 1 and L 2 are the corresponding contour lengths of the target contour and the adjacent contour, respectively.
Further, morphological analysis of mitochondria includes the steps of:
Defining the number of the fusion regions in the first contour as a single fusion number, defining the number of the division regions as a single division number, judging that the mitochondria are excessively fused if the single fusion number exceeds a third threshold value, and judging that the mitochondria are excessively divided if the single division number exceeds a fourth threshold value.
Further, after analyzing the morphological change sequence, counting a first number of occurrences of the division state and a second number of occurrences of the fusion state of all mitochondria based on the analysis result.
Further, after the first contour is acquired, a morphological feature of the mitochondria is calculated based on the first contour, the morphological feature including an area, a length, and a shape of the mitochondria, the shape including a linear shape and a granular shape.
Further, the contour detection algorithm is a CNN-based deep learning algorithm.
The invention also provides a mitochondrial fusion and division dynamic detection system based on image recognition, which is used for realizing the mitochondrial fusion and division dynamic detection method based on image recognition, and comprises the following steps:
the acquisition module is used for acquiring cell images by using a high-resolution microscope, marking acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time;
The segmentation module is used for giving a first serial number to each cell image in the time image sequence, carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate a first contour of each mitochondria in the cell image, giving a second serial number to each first contour in the cell image, and carrying out contrast integration on the first contour in each cell image so that the first contour corresponding to the same mitochondria has the same second serial number in different cell images;
the intercepting module is used for intercepting the first outline containing the same second serial number from each cell image and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first serial number;
The analysis module is used for analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of the mitochondria comprise a division state, a fusion state and a stable state, and the forms of the mitochondria are analyzed to locate abnormal mitochondria.
Compared with the prior art, the invention has the following beneficial effects:
According to the invention, the acquired high-resolution cell images are integrated into a time image sequence according to the acquisition time, and then contour segmentation and recognition are carried out on each cell image by combining a deep learning technology, so that the first contour belonging to mitochondria in each cell image is positioned. In order to facilitate morphological analysis of subsequent mitochondria, after contour segmentation is completed, the mitochondrial contours in each cell image are numbered, and contours belonging to the same mitochondria have the same second serial number in all cell images through contour matching. Based on the method, the mitochondrial outline with the same second serial number is intercepted to form a morphological change sequence, and finally the morphological change sequence is analyzed to realize the dynamic tracking of the mitochondrial fission state and the fusion state. In addition, the invention also counts the division times and fusion times of mitochondria in the whole observation period and the specific conditions of each division and fusion, thereby automatically positioning the mitochondria with abnormal division and fusion, and further assisting researchers in carrying out statistical analysis.
Drawings
FIG. 1 is a flow chart of steps of a dynamic detection method of mitochondrial fusion and division based on image recognition of the present invention;
FIG. 2 is a schematic illustration of contour segmentation in accordance with the present invention;
FIG. 3 is a schematic diagram of the mitochondrial status detection principle of the present invention;
FIG. 4 is a schematic diagram illustrating the generation of a detection zone according to the present invention;
FIG. 5 is a block diagram of a dynamic detection system for mitochondrial fusion and division based on image recognition of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
It will be understood that the terms "first," "second," and the like, as used herein, may be used to describe various elements, but these elements are not limited by these terms unless otherwise specified. These terms are only used to distinguish one element from another element. For example, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script, without departing from the scope of this disclosure.
As shown in fig. 1, a dynamic detection method for mitochondrial fusion and division based on image recognition includes:
S1, acquiring cell images by using a high-resolution microscope, marking acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time.
Specifically, the cell images with high resolution can be acquired at intervals of 5 seconds or can be set to other values according to experience, the acquisition time is marked for each acquisition time of the cell images, the cell images can be integrated into a time image sequence according to the acquisition time, and the cell images are ordered in the time image sequence according to the sequence of the acquisition time, so that the subsequent analysis is facilitated.
And S2, giving a first serial number to each cell image in the time image sequence, and carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate the first contour of each mitochondria in the cell image.
The contour detection algorithm is a deep learning algorithm based on CNN.
Because the cell images in the time image sequence are already ordered according to the acquisition time, the first serial number can be given according to the sequence, and the first serial number is 1,2, 3 and the like. And then, carrying out contour segmentation on each cell image through a contour detection algorithm so as to separate mitochondria, cell nuclei and the like from each other, wherein the contour detection algorithm can be a HED detection algorithm based on CNN or a fusion algorithm combining deep learning, which is proposed in the background art, as shown in figure 2, and the first contour belonging to mitochondria in the cell image can be accurately identified through figure 2, which is a segmentation effect diagram after the contour detection algorithm is used.
And S3, giving a second serial number to each first outline in the cell image, and comparing and integrating the first outlines in each cell image so that the first outlines corresponding to the same mitochondria have the same second serial number in different cell images.
Specifically, first, a second serial number is marked for the first outline in each cell image, the second serial number is in the form of A1, A2, A3, and the like, and then, the cell images adjacent to the second serial number are compared with each other, so that the same mitochondria have the same second serial number in different cell images. If the second serial number in the cell image 1 is A1 and the second serial number in the cell image 2 is A2 for the same mitochondria, the second serial number of the mitochondria in the cell image 2 is modified to be A1 after comparison and integration, so that the split and fusion state of the single mitochondria can be tracked conveniently.
S4, intercepting a first contour containing the same second sequence number from each cell image, and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first sequence number.
After the integration of the second serial number is completed, contour images of mitochondria with the same second serial number are intercepted from the cell images, and form change sequences are generated according to the collection time combination of the corresponding cell images, so that the time image sequences are split into the form change sequences only aiming at each mitochondria by taking the mitochondria as a unit, and the analysis pressure of a processor can be reduced and the analysis speed can be increased when the form analysis of the mitochondria is carried out subsequently.
S5, analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of the mitochondria comprise a division state, a fusion state and a stable state.
And S6, analyzing the morphology of mitochondria to locate abnormal mitochondria.
The present example locates the morphological changes of each mitochondria by analyzing the morphological change sequence, and the specific morphological analysis procedure will be described in detail later. The time period of the division state of the mitochondria when the division occurs, the fusion state when the fusion occurs, or the stable state when the division and fusion do not occur can be obtained through morphological analysis, then the division times and the fusion times of the mitochondria in the whole observation period are further counted based on the occurrence times of the division state and the stable state, the number of the mitochondria divided into each division and the number of the mitochondria fused each time are participated, and finally, researchers are helped to locate the positions and the number of abnormal mitochondria, and the statistical pressure is reduced, so that the aim of assisting in research is fulfilled.
According to the invention, firstly, the acquired high-resolution cell images are integrated into a time image sequence according to the acquisition time, then contour segmentation and recognition are carried out on each cell image by combining a deep learning technology, and the first contour belonging to mitochondria in each cell image is positioned. In order to facilitate morphological analysis of subsequent mitochondria, after contour segmentation is completed, the mitochondrial contours in each cell image are numbered, and contours belonging to the same mitochondria have the same second serial number in all cell images through contour matching. Based on the method, the mitochondrial outline with the same second serial number is intercepted to form a morphological change sequence, and finally the morphological change sequence is analyzed to realize the dynamic tracking of the mitochondrial fission state and the fusion state. In addition, the invention also counts the division times and fusion times of mitochondria in the whole observation period and the specific conditions of each division and fusion, thereby automatically positioning the mitochondria with abnormal division and fusion, and further assisting researchers in carrying out statistical analysis.
It is particularly noted that the present invention can track the morphology of mitochondria, thereby accurately and effectively recognizing the dynamic changes of mitochondria in the whole observation period.
The analysis of the morphological change sequence according to the present embodiment includes the steps of:
in the morphological change sequence, if no abnormal dividing line appears in the first contour, the state of the corresponding mitochondria is divided into stable states.
If an abnormal dividing line appears in the first contour, and the first contour is divided into a plurality of subareas by the abnormal dividing line, the first contour is defined as a second contour.
Next, the morphological analysis of the mitochondria according to the present invention will be described with reference to fig. 3, wherein images of the first contours of the mitochondria are first located, and if no abnormal parting line occurs therein, as shown in stage a of fig. 3, mitochondria 1 and 2 are included, and the corresponding contours are first contour A1 and first contour A2, respectively. The image of the first contour A2, which is added additionally for clarity of description of the present step, has been practically eliminated at the time of previous contour truncation and is not present in the sequence of morphological changes of the mitochondria 1. If the image is in the state shown by the first contour A1, i.e., there is no abnormal dividing line, the mitochondria in the image are considered to be in a stable state. If an abnormal parting line exists, as shown in a stage b and a stage c, the abnormal parting line appears in the first contour A1 to divide the first contour into a sub-region 1 and a sub-region 2, and the abnormal parting line appears in the stage c to divide the first contour A1 into a sub-region 1 and a sub-region 3, and in order to avoid confusion of description, the first contour with the abnormal parting line is redefined as a second contour, and in this embodiment, the following description is still performed with the first contour A1 for clarity of description.
And extracting the first contours with the second serial numbers before and after the second contours, defining the first contours as a third contour and a fourth contour, acquiring contour shapes of all subareas in the second contours, and positioning contour curves corresponding to the contour shapes in the third contour and the fourth contour.
Since the fusion or division process of single mitochondria generally lasts for a longer time, in order to extract mitochondrial images under different forms, in this embodiment, the mth first contour located before and after the second contour is extracted as the third contour and the fourth contour, respectively, for example, the first serial number of the cell image corresponding to the stage a is 1, and the first serial number of the cell image corresponding to the stage b is 4. The contour shapes of sub-areas 1 and 2 within the first contour A1 in phase b are acquired, and then contour curves approximating the contour shapes of sub-areas 1 and 2 in phase b are located in phase a and phase c, respectively.
If a subarea exists, no contour curve corresponding to the contour shape exists in the third contour, a contour curve corresponding to the contour shape exists in the fourth contour, the state of mitochondria corresponding to the second contour is divided into a fusion state, and the subarea is defined as a fusion area;
if a subarea exists, a contour curve corresponding to the contour shape of the subarea exists in the third contour, a contour curve corresponding to the contour shape of the subarea does not exist in the fourth contour, the state of mitochondria corresponding to the second contour is divided into a split state, and the subarea is defined as a split area;
And if all the subareas do not have contour curves corresponding to the contour shapes of the subareas in the third contour and the fourth contour, positioning the state of mitochondria corresponding to the second contour to be a stable state.
For the sub-region 1 in the phase b, the contour shape is similar to the upper half of the first contour A1 in the phase a and the phase c, so that the sub-region 1 is not suitable for the above-mentioned determination conditions, and for the sub-region 2 in the phase b, there is no contour curve similar to the first contour A1 in the phase a and the phase c, so that the mitochondrial state corresponding to the first contour a in the phase b is set to a stable state, and the fact is that the mitochondria 2 floating in the phase a are in exactly overlapped state in the phase b, are not actually fused and are separated from the mitochondria 1 in the phase c, so that the mitochondrial state in the phase b is a stable state.
Further, it is assumed that, in the case of the sub-region 2 of the phase b, if the contour curve of the lower half of the first contour A1 of the phase a is not similar to the contour shape of the sub-region 2, the contour curve of the lower half of the first contour A1 of the phase c is similar to the contour shape of the sub-region 2, and the mitochondria 1 in the phase b are considered to be in a fused state. I.e. the mitochondria 2 floating in phase a fuse with the mitochondria 1 and stay there.
Referring to stage c and stage d in fig. 3, for the first contour A1 of stage c, the contour shape of its subregion 3 approximates the contour curve of the upper half of the first contour A1 in stage b, and in stage d, no contour curve is found that approximates it, the state of mitochondria 1 in stage c is divided into split states. I.e. sub-region 3 in phase c is separated from mitochondria 1 and left.
Locating the contour curve corresponding to the contour shape includes the steps of:
Defining a cell image in which a third contour is positioned as a first image, acquiring first coordinates of outline shape centroids of subregions to be compared, mapping the first coordinates into the first image, generating a detection area in the first image by taking the first coordinates as the center, intercepting a contour part of the third contour in the detection area, selecting a plurality of calibration points on the third contour and the contour part, calculating the curvature of each calibration point, generating first curvature distribution of the contour shape, calculating second curvature distribution of the contour part, calculating the first curvature distribution and the second curvature distribution to obtain distribution similarity, and determining that a contour curve corresponding to the contour shape exists in the third contour if the distribution similarity is larger than a first threshold value.
The comparison process of the contour shape and the fourth contour is the same as the present process, the centroid of the contour shape can be obtained on the premise of determining the contour shape of the subarea, the same coordinate system is established in each cell image, the first coordinate of the centroid in the original cell image is obtained according to the coordinate system, the first coordinate is mapped into the first image where the third contour is located, and then the detection area is generated by taking the first coordinate as the center. As shown in fig. 4, the centroid Y of the outline shape is projected from the original cell image to the third image, and a rectangular detection region V is generated with the centroid Y as the center, then, similarly, a plurality of calibration points are selected on the outline shape of the sub-region and curvatures at the calibration points are calculated, a first curvature distribution is generated in the clockwise order of the calibration points, a plurality of calibration points are selected on the inner outline portion of the third outline located in the detection region and curvatures are calculated, and a second curvature distribution is generated in the clockwise order of the calibration points.
The first curvature distribution and the second curvature distribution are specifically a set of data sequences, as one way of calculating the distribution similarity, variance, standard deviation, average value, etc. may be adopted as the distribution similarity, or a method is used in which assuming that the first curvature distribution has N curvatures and the second curvature distribution has 2N curvatures, the 1 st to nth digits of the first curvature distribution are aligned with the 1 st to nth digits of the second curvature distribution, and then the sum of the euclidean distances of the digits at the same positions in the two data sequences is calculated as the first calculation result. And aligning the 1 st to N th digits of the first curvature distribution with the 2 nd to (n+1) th digits of the second curvature distribution, recalculating the sum of Euclidean distances, and repeating the process until the N th digits of the first curvature distribution are aligned with the 2N th digits of the first curvature distribution. Finally, the Euclidean distance X with the smallest value is obtained, and 1/e X is used as the distribution similarity.
The movement speed of the mitochondria is reduced when the mitochondria are split, the positions of the mitochondria are not changed greatly, and the positions of the two mitochondria which participate in fusion are not changed greatly when the two mitochondria are fused, so that the split mitochondria or the mitochondria which participate in fusion have the centroid position which is not far away from the positions of the splitting point and the fusion point, and therefore, the searching range for searching the contour curve in the third contour can be reduced through the step, and the processing speed is further accelerated.
The embodiment of the present invention for comparing and integrating the first contour includes the following steps:
Selecting cell images with adjacent first serial numbers as a second image and a third image respectively, selecting a first contour from the second image as a target contour, acquiring a second serial number corresponding to the target contour and a second coordinate of a centroid, positioning a third coordinate of each mitochondrial first contour centroid in the third image, mapping the second coordinate in the third image, defining a third coordinate with a distance smaller than a second threshold value from the second coordinate in the third image as an adjacent coordinate, defining a first contour with the adjacent coordinate as the centroid as an adjacent contour, comparing the shape similarity of the target contour and each adjacent contour, correcting the second serial number of the adjacent contour with the largest shape similarity to be identical with the target contour, and repeating the steps to traverse all the first contours and the cell images.
Here, the cell image 1 and the cell image 2 are taken as an example for explanation, the cell image 1 is defined as a second image, the cell image 2 is defined as a third image, a first contour is selected from the second image as a target contour, a second serial number of the target contour is A3, then a second coordinate of a centroid of the target contour is obtained, and the second coordinate is mapped from the second image to the third image. And in the third image, acquiring third coordinates of the centroid of each first contour, and screening coordinates which are closer to the second coordinates after projection from the third coordinates as adjacent coordinates. Finally, a first contour taking the adjacent coordinates as a centroid is acquired and defined as an adjacent contour, the target contour is respectively compared with each adjacent contour to find an adjacent contour most similar to the shape of the target contour, and the second serial number of the adjacent contour is corrected to be A3. Under the condition that the acquisition time interval is shorter, the position and the shape of the same mitochondria in adjacent cell images cannot be changed greatly, so that the target contour is not required to be compared with each first contour in the third image through the step, and only the target contour is required to be compared with the first contour in a certain range, the comparison range is further shortened, and finally, the mitochondria in different cell images are integrated quickly.
The method for calculating the shape similarity of the target contour and the adjacent contour comprises the following steps:
Covering the target contour on the adjacent contour by taking the mass center as a reference point, and calculating the shape similarity P of the target contour and the adjacent contour by a comparison formula, wherein the comparison formula is as follows: Wherein M 0 is the number of overlapping pixels after the target contour is overlaid on the adjacent contour, M 1 and M 2 include the number of pixels of the target contour and the adjacent contour, respectively, max (M 1,M2) is the larger value returned to M 1 or M 2, and L 1 and L 2 are the corresponding contour lengths of the target contour and the adjacent contour, respectively.
When the target contour is covered on the adjacent contour, the reference points of the two contours are overlapped, then the number of overlapped pixel points in the two contour areas after the coverage and the respective contour line lengths of the two contours are counted, and as can be seen from a comparison formula, when the target contour is covered on the adjacent contour, the more the number of overlapped pixel points is, the closer the contour line lengths of the two contours are, and the calculated shape similarity is larger.
The morphological analysis of mitochondria in this example includes the following steps:
the number of fusion areas in the first contour is defined as single fusion number, the number of division areas is defined as single division number, if the single fusion number exceeds a third threshold, the mitochondria are judged to have excessive fusion, and if the single division number exceeds a fourth threshold, the mitochondria are judged to have excessive division.
For example, if there are four division regions in the first contour, the number of single divisions is 4, meaning that the division causes the mitochondria to be divided into 4 from one division, and if the third threshold is 3, the mitochondria are considered to be excessively divided, and the process of judging the excessive fusion of the mitochondria is the same. The step can further assist researchers in locating mitochondria that exhibit excessive division and excessive fusion.
In this example, after analysis of the morphological change sequence, the first number of the split states and the second number of the fusion states of all mitochondria were counted based on the analysis result. Thereby facilitating the researchers to see if the mitochondria are frequently split or frequently fused.
In this embodiment, after the first contour is obtained, morphological characteristics of mitochondria including area, length, and shape of mitochondria including linear shape and granular shape are calculated based on the first contour.
Specifically, the area of the mitochondria is calculated through the number of pixel points included in the first contour, the length of the mitochondria is determined through obtaining the distance between the two pixel points farthest from the first contour, the shape of the mitochondria is determined according to the distribution condition of the distances by obtaining the distances between the centroid of the first contour and a plurality of points on the first contour, if the distribution of the distances is relatively equal, the mitochondria are determined to belong to a linear shape, otherwise, the mitochondria belong to a granular shape.
As shown in fig. 5, the present invention further provides a system for dynamically detecting mitochondrial fusion and division based on image recognition, which is used for implementing the above-mentioned method for dynamically detecting mitochondrial fusion and division based on image recognition, and the system comprises:
And the acquisition module is used for acquiring the cell images by using a high-resolution microscope, marking the acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time.
The segmentation module is used for giving a first serial number to each cell image in the time image sequence, carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate the first contour of each mitochondria in the cell image, giving a second serial number to each first contour in the cell image, and carrying out comparison and integration on the first contour in each cell image so that the first contour corresponding to the same mitochondria has the same second serial number in different cell images.
The intercepting module is used for intercepting a first outline containing the same second serial number from each cell image and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first serial number.
The analysis module is used for analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of the mitochondria comprise a division state, a fusion state and a stable state, and the forms of the mitochondria are analyzed to locate abnormal mitochondria.
It should be understood that the technical features of the foregoing embodiments may be combined in any manner, and for brevity, all of the possible combinations of the technical features of the foregoing embodiments are not described, however, they should be considered as being within the scope of the disclosure as long as there is no contradiction between the combinations of the technical features.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (10)
1. A dynamic detection method for mitochondrial fusion and division based on image recognition, comprising:
Acquiring cell images by using a high-resolution microscope, marking an acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time;
assigning a first serial number to each cell image in the time image sequence, and carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate a first contour of each mitochondria in the cell image;
Giving a second serial number to each first outline in the cell images, and comparing and integrating the first outlines in each cell image so that the first outlines corresponding to the same mitochondria have the same second serial number in different cell images;
Intercepting the first outline containing the same second serial number from each cell image, and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first serial number;
analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of mitochondria comprise a division state, a fusion state and a stable state;
the morphology of mitochondria was analyzed to locate abnormal mitochondria.
2. The method of claim 1, wherein analyzing the sequence of morphological changes comprises the steps of:
In the morphological change sequence, if no abnormal parting line appears in the first contour, dividing the state of the corresponding mitochondria into the stable state;
If the abnormal dividing line appears in the first contour and the abnormal dividing line divides the first contour into a plurality of subareas, defining the first contour as a second contour;
extracting the first contour with the second serial number before and after the second contour and defining the first contour as a third contour and a fourth contour respectively, acquiring contour shapes of all the subareas in the second contour, and positioning contour curves corresponding to the contour shapes in the third contour and the fourth contour;
If the subarea exists, the profile curve corresponding to the profile shape is not existed in the third profile, the profile curve corresponding to the profile shape exists in the fourth profile, the state of mitochondria corresponding to the second profile is divided into the fusion state, and the subarea is defined as a fusion area;
If the subarea exists, the profile curve corresponding to the profile shape exists in the third profile, the profile curve corresponding to the profile shape does not exist in the fourth profile, the state of mitochondria corresponding to the second profile is divided into the division state, and the subarea is defined as a division area;
And if all the subareas do not have the contour curves corresponding to the contour shapes of the subareas in the third contour and the fourth contour, positioning the state of mitochondria corresponding to the second contour as the stable state.
3. The method of claim 2, wherein locating the contour curve corresponding to the contour shape comprises the steps of:
Defining the cell image where the third contour is located as a first image, acquiring first coordinates of the contour shape centroids of the subareas to be compared, mapping the first coordinates into the first image, generating a detection area in the first image by taking the first coordinates as the center, intercepting a contour part of the third contour located in the detection area, selecting a plurality of calibration points on the third contour and the contour part, calculating the curvature of each calibration point, generating a first curvature distribution of the contour shape, calculating a second curvature distribution of the contour part, calculating the first curvature distribution and the second curvature distribution to obtain distribution similarity, and determining that the contour curve corresponding to the contour shape exists in the third contour if the distribution similarity is larger than a first threshold.
4. The method of claim 1, wherein the contrast integration of the first profile comprises the steps of:
Selecting the cell images adjacent to the first serial number as a second image and a third image, selecting the first contour from the second image as a target contour, acquiring second coordinates of the second serial number and the mass center corresponding to the target contour, positioning third coordinates of the mass centers of the first contour of each mitochondria in the third image, mapping the second coordinates in the third image, defining the third coordinates with the distance from the second coordinates smaller than a second threshold value in the third image as adjacent coordinates, defining the first contour with the adjacent coordinates as the mass center as an adjacent contour, comparing the shape similarity of the target contour and each adjacent contour, correcting the second serial number of the adjacent contour with the largest shape similarity to be identical with the target contour, and repeating the steps to traverse all the first contour and the cell images.
5. The method of claim 4, wherein calculating the shape similarity of the target contour and the neighboring contour comprises:
Covering the target contour on the adjacent contour by taking the mass center as a reference point, and calculating the shape similarity P of the target contour and the adjacent contour through a comparison formula, wherein the comparison formula is as follows: Wherein M 0 is the number of overlapping pixels after the target contour is overlaid on the adjacent contour, M 1 and M 2 are the number of pixels included in the target contour and the adjacent contour, respectively, max (M 1,M2) is the larger value returned to M 1 or M 2, and L 1 and L 2 are the corresponding contour lengths of the target contour and the adjacent contour, respectively.
6. The method of claim 2, wherein morphological analysis of mitochondria comprises the steps of:
Defining the number of the fusion regions in the first contour as a single fusion number, defining the number of the division regions as a single division number, judging that the mitochondria are excessively fused if the single fusion number exceeds a third threshold value, and judging that the mitochondria are excessively divided if the single division number exceeds a fourth threshold value.
7. The method of claim 1, wherein after analyzing the sequence of morphological changes, counting a first number of occurrences of the division state and a second number of occurrences of the fusion state for all mitochondria based on the analysis result.
8. The method of claim 1, wherein after the first contour is acquired, a morphological feature of mitochondria is calculated based on the first contour, the morphological feature including an area, a length, and a shape of mitochondria, the shape including a line shape and a granular shape.
9. The method of claim 1, wherein the contour detection algorithm is a CNN-based deep learning algorithm.
10. A dynamic detection system for mitochondrial fusion and division based on image recognition for implementing the method of any one of claims 1-9, comprising:
the acquisition module is used for acquiring cell images by using a high-resolution microscope, marking acquisition time for each cell image, and generating a time image sequence of the cell images based on the acquisition time;
The segmentation module is used for giving a first serial number to each cell image in the time image sequence, carrying out contour recognition on each cell image based on a contour detection algorithm so as to locate a first contour of each mitochondria in the cell image, giving a second serial number to each first contour in the cell image, and carrying out contrast integration on the first contour in each cell image so that the first contour corresponding to the same mitochondria has the same second serial number in different cell images;
the intercepting module is used for intercepting the first outline containing the same second serial number from each cell image and integrating and generating a morphological change sequence of each mitochondria according to the corresponding first serial number;
The analysis module is used for analyzing the morphological change sequence to obtain various forms of mitochondria and corresponding occurrence time, wherein the forms of the mitochondria comprise a division state, a fusion state and a stable state, and the forms of the mitochondria are analyzed to locate abnormal mitochondria.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411550863.6A CN119049049B (en) | 2024-11-01 | 2024-11-01 | Dynamic detection method and system of mitochondrial fusion and fission based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411550863.6A CN119049049B (en) | 2024-11-01 | 2024-11-01 | Dynamic detection method and system of mitochondrial fusion and fission based on image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119049049A CN119049049A (en) | 2024-11-29 |
| CN119049049B true CN119049049B (en) | 2025-01-28 |
Family
ID=93569533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411550863.6A Active CN119049049B (en) | 2024-11-01 | 2024-11-01 | Dynamic detection method and system of mitochondrial fusion and fission based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119049049B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119444757B (en) * | 2025-01-10 | 2025-04-22 | 深圳中旭细胞再生医学研究有限公司 | Immune cell state analysis method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117288659A (en) * | 2023-11-24 | 2023-12-26 | 深圳市前海高新国际医疗管理有限公司 | Morphological analysis system of tumor stem cells and normal cells based on AI |
| CN118608546A (en) * | 2024-08-07 | 2024-09-06 | 广东医科大学附属医院 | Fluorescence image segmentation method, system and storage medium for mitochondria |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5891561B2 (en) * | 2011-06-03 | 2016-03-23 | 学校法人自治医科大学 | Mitochondrial membrane proteins and genes encoding them |
| WO2021152562A1 (en) * | 2020-01-31 | 2021-08-05 | Stellenbosch University | Method for determining mitochondrial events |
-
2024
- 2024-11-01 CN CN202411550863.6A patent/CN119049049B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117288659A (en) * | 2023-11-24 | 2023-12-26 | 深圳市前海高新国际医疗管理有限公司 | Morphological analysis system of tumor stem cells and normal cells based on AI |
| CN118608546A (en) * | 2024-08-07 | 2024-09-06 | 广东医科大学附属医院 | Fluorescence image segmentation method, system and storage medium for mitochondria |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119049049A (en) | 2024-11-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110097093B (en) | Method for accurately matching heterogeneous images | |
| CN107633226B (en) | Human body motion tracking feature processing method | |
| CN119049049B (en) | Dynamic detection method and system of mitochondrial fusion and fission based on image recognition | |
| CN108986152B (en) | A foreign object detection method and device based on differential image | |
| CN104867137B (en) | A kind of method for registering images based on improvement RANSAC algorithms | |
| CN108491498B (en) | Bayonet image target searching method based on multi-feature detection | |
| CN111369495B (en) | Panoramic image change detection method based on video | |
| Li et al. | A complex junction recognition method based on GoogLeNet model | |
| CN114359998B (en) | Identification method of face mask in wearing state | |
| CN112966542A (en) | SLAM system and method based on laser radar | |
| CN106933816A (en) | Across camera lens object retrieval system and method based on global characteristics and local feature | |
| CN112528845B (en) | A deep learning-based physical circuit diagram recognition method and its application | |
| CN107123130B (en) | A Kernel Correlation Filtering Target Tracking Method Based on Superpixel and Hybrid Hash | |
| CN114358166B (en) | Multi-target positioning method based on self-adaptive k-means clustering | |
| CN106570475B (en) | A kind of dark-red enameled pottery seal search method | |
| CN113902991A (en) | Twin network target tracking method based on cascade characteristic fusion | |
| CN111046856A (en) | Parallel pose tracking and map creating method based on dynamic and static feature extraction | |
| CN107862319B (en) | Heterogeneous high-light optical image matching error eliminating method based on neighborhood voting | |
| CN116704490B (en) | License plate recognition method, license plate recognition device and computer equipment | |
| CN109146918A (en) | A kind of adaptive related objective localization method based on piecemeal | |
| CN111709317A (en) | A pedestrian re-identification method based on multi-scale features under saliency model | |
| CN105447869B (en) | Camera self-calibration method and device based on particle swarm optimization algorithm | |
| CN115760898A (en) | World coordinate positioning method for road sprinklers in mixed Gaussian domain | |
| CN110287959A (en) | A license plate recognition method based on re-recognition strategy | |
| CN113379777A (en) | Shape description and retrieval method based on minimum circumscribed rectangle vertical internal distance proportion |
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 |