CN111429461B - Novel segmentation method for overlapped and exfoliated epithelial cells - Google Patents
Novel segmentation method for overlapped and exfoliated epithelial cells Download PDFInfo
- Publication number
- CN111429461B CN111429461B CN201910019927.2A CN201910019927A CN111429461B CN 111429461 B CN111429461 B CN 111429461B CN 201910019927 A CN201910019927 A CN 201910019927A CN 111429461 B CN111429461 B CN 111429461B
- Authority
- CN
- China
- Prior art keywords
- image
- cell
- segmentation
- level set
- area
- 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
- 210000002919 epithelial cell Anatomy 0.000 title claims abstract description 76
- 230000011218 segmentation Effects 0.000 title claims abstract description 75
- 238000000034 method Methods 0.000 title claims abstract description 73
- 210000004027 cell Anatomy 0.000 claims abstract description 76
- 210000003855 cell nucleus Anatomy 0.000 claims abstract description 43
- 238000010586 diagram Methods 0.000 claims abstract description 30
- 230000008569 process Effects 0.000 claims description 28
- 230000002146 bilateral effect Effects 0.000 claims description 6
- 239000003086 colorant Substances 0.000 claims description 6
- 210000000805 cytoplasm Anatomy 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 5
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 3
- 210000000981 epithelium Anatomy 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 7
- 238000003745 diagnosis Methods 0.000 abstract description 4
- 206010028980 Neoplasm Diseases 0.000 abstract description 3
- 201000011510 cancer Diseases 0.000 abstract description 3
- 238000012216 screening Methods 0.000 abstract description 2
- 206010008342 Cervix carcinoma Diseases 0.000 abstract 1
- 208000005016 Intestinal Neoplasms Diseases 0.000 abstract 1
- 208000003445 Mouth Neoplasms Diseases 0.000 abstract 1
- 208000006105 Uterine Cervical Neoplasms Diseases 0.000 abstract 1
- 201000010881 cervical cancer Diseases 0.000 abstract 1
- 201000010099 disease Diseases 0.000 abstract 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 abstract 1
- 201000002313 intestinal cancer Diseases 0.000 abstract 1
- 208000012987 lip and oral cavity carcinoma Diseases 0.000 abstract 1
- 230000006872 improvement Effects 0.000 description 19
- 210000004940 nucleus Anatomy 0.000 description 15
- 238000003709 image segmentation Methods 0.000 description 9
- 238000011478 gradient descent method Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 108090000623 proteins and genes Proteins 0.000 description 2
- 230000000630 rising effect Effects 0.000 description 2
- 230000003313 weakening effect Effects 0.000 description 2
- 241001584785 Anavitrinella pampinaria Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000002380 cytological effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 210000003608 fece Anatomy 0.000 description 1
- 238000007490 hematoxylin and eosin (H&E) staining Methods 0.000 description 1
- 210000002490 intestinal epithelial cell Anatomy 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 230000035755 proliferation Effects 0.000 description 1
- 102000004169 proteins and genes Human genes 0.000 description 1
- 238000010186 staining Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a novel method for dividing overlapped and exfoliated epithelial cells, which comprises two modules, namely cell nucleus division and cell plasma division, wherein the method combines an MSER algorithm and a Kmeans algorithm to divide cell nuclei; the cytoplasma segmentation fuses two segmentation methods, namely a voronoi diagram and a level set. The invention improves the edge energy item of the traditional DRLSE level set, and the improved level set is more suitable for the segmentation of overlapped and exfoliated epithelial cells. In addition, the invention carries out edge enhancement on the level set edge indication operator, thereby furthest relieving the interference of the edges of adjacent overlapped cells on contour evolution. The method can effectively divide the exfoliated epithelial cells, provides a basis for automatically diagnosing the exfoliated epithelial cell diseases (such as pre-cancer screening or post-operation diagnosis of cervical cancer, oral cancer, intestinal cancer and the like) by a computer, and has huge social benefit.
Description
Technical Field
The invention relates to the technical field of medical image processing, in particular to a novel segmentation method for overlapped and exfoliated epithelial cells.
Background
The exfoliated epithelial cells are a sample to be tested which can be conveniently collected and obtained clinically, and screening and diagnosis of abnormal proliferation or canceration cells in the exfoliated epithelial cells can be realized by checking or analyzing the exfoliated epithelial cells and utilizing novel biotechnology such as heteroploid cell morphological characteristic parameters, genetic material DNA content change and other gene chips in cell biology.
The exfoliated epithelial cells such as cervical exfoliated epithelial cells, oral exfoliated epithelial cells or intestinal epithelial cells exfoliated in feces, etc. are subjected to conventional cytological sample treatment (such as Papanicolaou staining, HE staining, etc.), and a plurality of precancerous lesion cells are manually searched from a large number of cells by a doctor with experience under a microscope, so that the labor intensity is high, the operator is extremely easy to fatigue, the operator is required to have higher pathological knowledge and clinical experience, and the diagnosis result is influenced by various aspects such as subjective factors of the operator, and human errors are unavoidable. An accurate and efficient method of early detection of cancer cells can help save lives of more early cancer patients. In order to realize morphological analysis of precancerous lesion cells and automatic diagnosis of abnormal characteristics, cell segmentation is the first step, so that the quality of segmentation of the image of the exfoliated epithelial cells has very important influence on the accuracy of a final detection result. The ideal cell image segmentation result not only reduces the complexity of the subsequent classifier design, but also helps to improve the accuracy of final detection.
Currently, there are many cell image segmentation algorithms, which are mainly classified into two major categories, namely a cell segmentation algorithm based on region information and a segmentation algorithm based on edge information, and the general steps of the cell segmentation algorithm based on region information are to classify pixels into different categories according to information of each pixel point in an image and then to set a judgment criterion. The segmentation algorithm based on cell edge information is based on the boundary in the cell image, and the gray level of the boundary part and the gray level of the non-boundary part are generally different in numerical value, namely, the discontinuous part of the common gray level is the boundary, and the DRLSE model belongs to the algorithm. However, these methods are basically difficult to deal with the problems of edge breakage, edge blurring, etc. in the image of the exfoliated epithelial cells. In addition, due to the background area of the image of the exfoliated epithelial cells, the existence of a plurality of interference edges in the exfoliated epithelial cells, the high overlapping of the exfoliated epithelial cells and other problems, the algorithm is more difficult to accurately divide the exfoliated epithelial cells. In addition, the existing active contour algorithm for image segmentation either needs a relatively complex initial contour construction process or the initial contour construction is too simple, so that the result of contour evolution is not ideal or a large number of iterations are needed, and the cell segmentation is low in efficiency. In view of the foregoing, there is an urgent need for an algorithm that can efficiently and precisely segment exfoliated epithelial cells.
Disclosure of Invention
The invention aims to provide a novel method for overlapping and dividing exfoliated epithelial cells, so as to solve the problems in the technical background. In order to achieve the above purpose, the present invention provides the following technical solutions: a novel method for dividing overlapped and exfoliated epithelial cells.
In order to achieve the above purpose, the present invention provides the following technical solutions: a novel method for dividing overlapped and exfoliated epithelial cells comprises two modules, namely cell nucleus division and cell plasma division;
the cell nucleus segmentation method fuses two segmentation algorithms of an MSER algorithm and a Kmeans algorithm, and the steps are as follows (assuming that an original to-be-segmented and exfoliated epithelial cell image is imageOriginal):
firstly, carrying out graying and bilateral filtering pretreatment on an imageOriginal, and marking a treatment result as imagePreprocess;
dividing the cell nuclei in the image pre-process by using an MSER algorithm to obtain a cell nucleus rough division result, and marking the cell nucleus rough division result as a nucleousRough;
step three, generating an ROI picture for the rough segmentation contour of each cell nucleus in the image nucleic Rough, wherein the ROI picture is obtained by intercepting from the image process according to a rectangular region;
and step four, clustering the ROI pictures by using a Kmeans algorithm to finally obtain an accurate segmentation result of the cell nucleus, and marking the accurate segmentation result as nucleic.
Further, the image imageOriginal of the exfoliated epithelial cells is obtained by image acquisition under a microscope, the magnification of an eyepiece of the microscope is 10, and the magnification of an objective lens is 20.
Furthermore, the MSER algorithm is developed based on a watershed algorithm, and the algorithm is used for dividing the exfoliated epithelial cell nuclei by calculating the maximum stable extremum region in the image; the MSER algorithm flow is: the image is binarized, and the binarization threshold value is [0,255], so that the binarized image is subjected to a process from full black to full white, similar to an aerial view with continuously rising water level. In this process, there are some connected regions with small area variation with increasing threshold, this region is called MSER, the maximum stable extremum region, and its mathematical definition is:
min_area＜Q i ＜max_area
where Qi represents the area of the ith connected region, Δ represents a small threshold change (water filling), and when v (i) is smaller than a given threshold max_variation and the area of the connected region is within a given minimum area threshold min_area and maximum area threshold max_area, then that region is considered to be the maximum stable extremum region meeting the requirements.
Further, the principle of the MSER algorithm for positioning the exfoliated epithelial nuclei by calculating the maximum stable extremum region in the image is as follows: in the image of the exfoliated epithelial cells, the exfoliated epithelial nuclei are usually elliptical areas with lower gray values, which just symbolizes the definition of the maximum stable extremum area in the MSER, so that the exfoliated epithelial nuclei can be located by solving the maximum stable extremum area in the image of the exfoliated epithelial cells. Preferably, the MSER algorithm delta of the invention takes on a value of 2, the max_variation takes on a value of 0.5, the min_area takes on a value of 100, and the max_area takes on a value of 1500.
Further, the ROI picture is obtained by cutting from an image imageProcess according to a rectangular area, wherein the rectangular area is formed by diffusing L pixels from top to bottom, left to right respectively by a minimum circumscribed rectangle of a cell nucleus rough segmentation contour corresponding to the ROI picture, and preferably L is equal to 16.
Further, the Kmeans algorithm is used for clustering the ROI pictures, wherein the clustering number is 2, and the clustering number corresponds to the cell nucleus region and the non-cell nucleus region in the ROI pictures. According to the invention, according to the fact that the gray value of the exfoliated epithelial cell nucleus is generally lower than that of surrounding cell plasma pixel points, a region with a lower average gray value in the clustering result is extracted to serve as an accurate region of the cell nucleus.
A cell plasma segmentation method integrates two segmentation methods of a Voronoi diagram and a level set, and the segmentation steps are as follows:
step one, firstly, dividing the exfoliated epithelial cell mass by using a region growing method and combining a Kmeans algorithm, and marking the segmented epithelial cell mass as ClumpIMage.
And generating a Voronoi image of the exfoliated epithelial cell image according to the barycenter coordinates of all nuclei in the nucleousfine.
And step two, overlapping the images ClumpIMage and Voronoi image to obtain the outline of the Voronoi segment of all the exfoliated epithelial cells.
And thirdly, taking the Voronoi figure outline of the cells in the image Voronoi segment as the initial outline of the improved level set, and carrying out level set outline evolution so as to obtain the accurate outline cellFine of the cell plasma.
Further, the exfoliated epithelial cell mass refers to a region consisting of all cells in an exfoliated epithelial cell image, and the region is mutually opposite and complementary to the background region, and the exfoliated epithelial cell mass is segmented, and the specific segmentation steps are as follows:
step one, clustering image processes subjected to graying and bilateral filtering pretreatment by using a Kmeans algorithm, wherein the clustering number is 2, and the clustering number corresponds to a background area and a cell cluster area in the image process respectively to obtain a cell cluster rough segmentation result, and the cell cluster rough segmentation result is recorded as ClumpSegRough;
performing morphological corrosion operation on a background area in the image ClumpSegRough, so as to ensure that pixel points in the background area of the ClumpSegRough belong to a real background area to the greatest extent, and marking a processing result as ClumpSegRough hEROsion;
and thirdly, randomly selecting a plurality of coordinate points in a background area of the image ClumpSegRoughelctrosin, taking the coordinate points as seed points of an area growth algorithm, and carrying out area growth on the image process by using the area growth algorithm, wherein a final area growth result is a real area of the cell mass.
Further, the images ClumpImage and voronoi image are superimposed to obtain the voronoi diagram outline of all the exfoliated epithelial cells, and the method is characterized in that the superimposed mode is to add RGB channel values of pixel points corresponding to the two images ClumpImage and voronoi image and take the remainder for 255; the voronoi diagram outline is characterized in that the cytoplasms of different cells are represented by different colors, namely, the connected domains with different colors in the picture and the cytoplasms have a one-to-one correspondence.
Furthermore, the level set belongs to the active contour theory, and the basic idea is to embed an evolution curve or an evolution curved surface as a zero level set into a high-dimensional level set function, and the aim of controlling the evolution curve or the evolution curved surface is achieved through the high-dimensional level set function. The general procedure for image segmentation using level sets is as follows:
step one, providing an initial contour for a level set;
step two, constructing a level set function by using the initial contour as a zero level set;
step three, constructing an energy functional based on the level set function;
and step four, minimizing an energy functional through a gradient descent method, and realizing the evolution of a level set function, thereby realizing the evolution of a zero level set, wherein the evolution process of the zero level set is the process of the evolution of the contour of the target to be segmented.
The level set minimizes an energy function through a gradient descent method, the evolution of the level set function is realized, and then the evolution of a zero level set is realized, namely the contour evolution process of the object to be segmented. The flow of the level set gradient descent algorithm is as follows:
input: the initial Contour Contours, the gradient drops by an iteration number n.
Output: and (5) a level set function after evolution is finished.
Constructing an initial level set function Φ by Contours
for itera in n:
returnΦ
Further, the improved level set is an improvement of the traditional DRLSE level set model for segmentation of overlapping exfoliated epithelial cells according to the present invention. The improved level set of the invention, relative to the DRLSE model, mainly comprises the following three improvements:
a. a more reasonable initial profile is provided for the level set by the voronoi diagram.
b. Edge enhancement is carried out on an edge indication operator in the DRLSE, so that the interference of the edges of adjacent overlapped cells on the evolution of the contour of the target to be segmented is effectively relieved.
c. The invention provides a more reasonable edge energy item, and compared with the DRLSE edge energy item, the energy item can more accurately measure the degree of the zero level set positioned at the edge position in the image, so that the improved level set can more fully and accurately utilize the edge information of the image to carry out image segmentation.
The definition of the DRLSE edge energy term is as follows:
∫ Ω gδ(Φ)|△Φ|dx
the edge energy term can essentially be regarded as an edge indicator integral along the zero level set. When the edge of the object to be segmented is clear and the background of the object to be segmented is clean, the edge energy term can well measure the degree of the zero level set positioned at the edge pixel point in the image. In practice, however, the object to be segmented in the medical image processed by us has the conditions of blurring, crossing, breaking, etc. of edges, and many interference edges exist in the background of the object to be segmented or in the interior of the object to be segmented itself, so that such a situation may occur: the pixel points where the zero level is located are mostly interference edge pixel points, but because the perimeter of the zero level set is small, even if the zero level set does not really converge to the real edge of the object to be segmented in this case, the energy term takes the minimum value, so that the zero level set is difficult to converge to the real edge of the object to be segmented.
The improvement of the edge energy item is mainly aimed at the situation that the circumference of a zero level set in the edge energy item can bring errors to image segmentation when the original edge energy item in the DRLSE is used for segmenting the medical image because the edge information in the medical image is too complex. To this end, the invention proposes the following improved edge energy term:
the improved edge energy term, i.e. the perimeter of the zero level set divided by the edge energy term of the DRLSE source, enables the energy term to measure well the extent to which the zero level set is located at the edge pixels in the image.
Further, the present invention uses the centroid coordinates of all exfoliated epithelial nuclei to generate a voronoi diagram that generally has been compared to the true contour of the cells to be segmented, which includes two benefits: on one hand, the harsh requirement of the DRLSE model on the robustness of the initial contour is relieved; on the other hand, the level set can evolve the accurate contour of the object to be segmented only by less contour evolution times, so that the efficiency of the level set is effectively improved.
Furthermore, the edge enhancement is performed on the edge indication operator, and mainly aims at weakening the edge of the cells overlapped with the target cells to be segmented, and retaining the edge belonging to the target cells, so that the interference of the edge of the adjacent overlapped cells to the contour evolution of the target cells is prevented to the greatest extent. Edge enhancement employed in the present invention i The method of (1) is based on the following principle: cast-off epithelial cellThe cell nucleus is usually located at the center of the cell, the angle formed by the connection line between the target cell nucleus and the edge pixel point of the target cell and the gradient direction at the pixel point should be an acute angle, and the angle formed by the connection line between the target cell and the edge pixel point of the adjacent overlapped cell and the gradient direction at the pixel point should be an obtuse angle. Based on such a principle, a large number of edge pixels of cells overlapping with the target cells can be removed. The DRLSE edge indicator operator and the improved edge indicator operator of the present invention are compared as follows:
DRLSE edge indication operator:
the edge indication operator comprises the following steps:
wherein the method comprises the steps ofαAnd an included angle formed by a vector formed by a pixel point in the edge indication operator image and the cell nucleus centroid coordinates of the target cells to be segmented and the gradient direction at the pixel point is represented.
Still further, the energy function of the improved level set of the present invention is defined as follows:
E(Φ)＝μR p (Φ)+λL f (Φ)+αA f (Φ)+εE(Φ)，
wherein R is p (Φ) is a regular energy term, lf (Φ) is an edge energy term, af (Φ) is an area energy term, E (Φ) is a shape prior energy term, and the definition of each energy sub-term is as follows:
regular energy term: r is R p (Φ)＝∫ Ω p(|ΔΦ|)dx
Edge energy term:
area energy term: a is that f (Φ)＝∫ Ω fH(-Φ)dx
Shape prior energy term: the invention considers that the prior shape of the cervical cells is elliptical, and names the prior term of the shape as an elliptical term, but the invention does not explicitly define the elliptical term, but directly gives the gradient descending flow of the elliptical term, and the calculation process of the gradient descending flow of the elliptical term is as follows: the minimum circumscribed ellipse elipse of the zero level set is fitted first, and then a new level set function Φ_elipse, i.e. the gradient descent of the ellipse term, is constructed using the elipse as the new zero level set.
The level set minimizes the energy function through a gradient descent method, so that the zero level set evolves to the real outline of the object to be segmented, and the gradient descent flow of the level set function is as follows:
for the segmentation of overlapping exfoliated epithelial cells, the present invention proposes for the first time a rough segmentation using a voronoi diagram, followed by a fine segmentation using an improved level set, which has two advantages: on one hand, the Voronoi diagram segmentation result provides a reasonable initial contour for the level set, so that the result of the contour evolution of the level set is more accurate; on the other hand, the result of the voronoi diagram segmentation is usually closer to the real contour of the cell, so that the level set can be optimized with a small number of iterations, thereby reducing the time of contour evolution and improving the efficiency of cell segmentation. In addition, the invention aims at the application scene of dividing the exfoliated epithelial cells, improves the traditional DRLSE level set, and the improvement mainly comprises two aspects: the first improvement is an improvement on the edge energy term, so that the energy term can better measure the degree to which a zero level set is positioned at an edge pixel point in an image; the second improvement is the improvement of the edge indication operator, and the edge enhancement is carried out on the image of the edge indication operator, so that the interference of the edges of adjacent overlapped cells on the contour evolution is prevented to the greatest extent.
Compared with the prior art, the invention has the following advantages and remarkable advantages:
a. the two-stage segmentation process of the level set is combined with the Veno diagram, a more reasonable initial contour is provided for the level set by a high-efficiency simple method, and the efficiency and the precision of the contour evolution of the level set are improved;
b. the edge energy items of the traditional DRLSE level set are improved, and the improved level set is more suitable for the segmentation of overlapped and exfoliated epithelial cells;
c. edge enhancement is carried out on the level set edge indication operator, so that the interference of the edges of adjacent overlapped cells on contour evolution is relieved to the greatest extent;
d. an efficient cell mass segmentation algorithm based on Kmeas and region growth is presented.
Drawings
FIG. 1 is a general flow chart of a novel split epithelial cell overlap segmentation method of the present invention;
FIG. 2 is a flow chart of a cell mass segmentation method based on a region growing algorithm according to the present invention;
FIG. 3 is a schematic diagram of a method for dividing a cell nucleus based on an MSER algorithm;
FIG. 4 is a graph of a result of nuclear segmentation based on the MSER algorithm according to the present invention;
FIG. 5 is a schematic diagram of a cell mass segmentation method based on a region growing algorithm according to the present invention;
FIG. 6 is a schematic representation of a rough segmentation of exfoliated epithelial cells based on a voronoi diagram in accordance with the present invention;
FIG. 7 is a graph showing the results of rough segmentation of exfoliated epithelial cells based on a voronoi diagram in accordance with the present invention;
FIG. 8 is a pretreatment view of a novel split epithelial cell overlap segmentation method according to the present invention;
FIG. 9 is a graph comparing the level set edge indicator before and after improvement in accordance with the present invention;
FIG. 10 is a schematic diagram showing the outline evolution of a novel split epithelial cell overlap segmentation method according to the present invention;
FIG. 11 is a graph showing the segmentation results of a novel split epithelial cell overlap segmentation method according to the present invention;
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-11, the present invention provides the following technical solutions: the invention discloses a novel method for overlapping and dividing an exfoliated epithelial cell, which comprises two modules of cell nucleus division and cell plasma division, wherein the method combines an MSER algorithm and a Kmeans algorithm to divide the cell nucleus, and the steps are as follows (assuming that an image of the exfoliated epithelial cell to be divided is an imageOriginal image):
firstly, carrying out graying and bilateral filtering pretreatment on an imageOriginal, and marking a treatment result as imagePreprocess;
dividing the cell nuclei in the image pre-process by using an MSER algorithm to obtain a cell nucleus rough division result, and marking the cell nucleus rough division result as a nucleousRough;
and thirdly, generating an ROI picture for the rough segmentation contour of each cell nucleus in the image nucleic Rough, wherein the ROI picture is obtained by cutting out from the image process according to a rectangular region. Meanwhile, the image original of the exfoliated epithelial cells is obtained by taking images under the magnification of an eyepiece of 10 times and the magnification of an objective lens of 20 times of a microscope. The rectangular area is formed by diffusing 16 pixel points upwards, downwards, leftwards and rightwards by the minimum circumscribed rectangle of the cell nucleus rough segmentation outline corresponding to the ROI picture;
and step four, clustering the ROI pictures by using a Kmeans algorithm to finally obtain an accurate segmentation result of the cell nucleus, and marking the accurate segmentation result as nucleic.
The MSER algorithm is developed based on a watershed algorithm, and the algorithm divides the exfoliated epithelial cell nucleus by calculating the maximum stable extremum region in the image; the MSER algorithm flow is: the image is binarized, and the binarization threshold value is [0,255], so that the binarized image is subjected to a process from full black to full white, similar to an aerial view with continuously rising water level. In this process, there are some connected regions with small area variation with increasing threshold, this region is called MSER, the maximum stable extremum region, and its mathematical definition is:
min_area＜Q i ＜max_area
where Qi represents the area of the ith connected region, Δ represents a small threshold change (waterfilling), and when v (i) is less than a given threshold max_variation and the area of the connected region is within a given minimum area threshold min_area and maximum area threshold max_area, then that region is considered to be the required MSER.
The principle of the MSER algorithm for positioning the exfoliated epithelial nuclei by calculating the maximum stable extremum region in the image is as follows: in the image of the exfoliated epithelial cells, the exfoliated epithelial nuclei are usually elliptical areas with lower gray values, which just symbolizes the definition of the maximum stable extremum area in the MSER, so that the exfoliated epithelial nuclei can be located by solving the maximum stable extremum area in the image of the exfoliated epithelial cells. Meanwhile, the MSER algorithm delta of the invention is 2, the max_variation is 0.5, the min_area is 100, and the max_area is 1500.
The Kmeans algorithm is used for clustering the ROI pictures, the clustering number is 2, the clustering number corresponds to a cell nucleus area and a non-cell nucleus area, and the area with lower average gray value in the clustering result is extracted as the accurate area of the cell nucleus because the gray value of the cell nucleus of the detached epithelium is usually lower than that of surrounding pixel points.
A cell plasma segmentation method integrates two segmentation methods of a Voronoi diagram and a level set, and the segmentation steps are as follows:
a. firstly, dividing out an exfoliated epithelial cell mass by using a region growing method and combining a Kmeans algorithm, and marking the exfoliated epithelial cell mass as ClumpIMage;
b. generating a Voronoi image of the exfoliated epithelial cell image according to the barycenter coordinates of all the nuclei in the nucleousfine;
c. the specific operation of overlapping the images ClumpIMage and Voronoi image is that the RGB channel values of the corresponding pixel points of the two images are added and the remainder is taken for 255, so that the Voronoi image outline of all the exfoliated epithelial cells is obtained and marked as Voronoi segment; the outline of the voronoi diagram is characterized in that the cytoplasms of different cells in the diagram are expressed by different colors, namely, the connected domains with different colors in the diagram and the cytoplasms have a one-to-one correspondence;
d. taking the Voronoi segment cell voronoi diagram outline as the initial outline of the improved level set, and carrying out level set outline evolution to obtain an accurate outline cellFine of the cell plasma;
the exfoliated epithelial cell mass refers to a region consisting of all cells in an exfoliated epithelial cell image, and the region is mutually opposite and complementary with a background region, and the exfoliated epithelial cell mass is segmented by the following specific segmentation steps:
a. clustering the image process subjected to the pretreatment of graying and bilateral filtering by using a Kmeans algorithm, wherein the clustering number is 2, and the clustering number corresponds to a background area and a cell mass area in an image of the exfoliated epithelial cells respectively to obtain a cell mass rough segmentation result which is marked as ClumpSegRough;
b. performing morphological corrosion operation on a background area in the image ClumpSegRough, so as to ensure that pixel points in the roughly segmented background area belong to a real background area to the greatest extent, wherein a processing result is ClumpSegRoughererosion;
c. and randomly selecting a plurality of coordinate points in a background area of the image ClumpSegRoughearosin, taking the coordinate points as seed points of an area growth algorithm, and carrying out area growth on the image process by using the area growth algorithm, wherein the final result of the area growth is the real area of the cell mass.
The level set belongs to an active contour theory, and the basic idea is that an evolution curve or an evolution curved surface is embedded into a high-dimensional level set function as a zero level set, and the aim of controlling the evolution curve or the evolution curved surface is achieved through the high-dimensional level set function. The general procedure for image segmentation using level sets is as follows:
step one, providing an initial contour for a level set;
step two, constructing a level set function by using the initial contour as a zero level set;
step three, constructing an energy functional based on the level set function;
and step four, minimizing an energy functional through a gradient descent method, and realizing the evolution of a level set function, thereby realizing the evolution of a zero level set, wherein the evolution process of the zero level set is the process of the evolution of the contour of the target to be segmented.
The level set minimizes an energy function through a gradient descent method, the evolution of the level set function is realized, and then the evolution of a zero level set is realized, namely the contour evolution process of the object to be segmented. The flow of the level set gradient descent algorithm is as follows:
input: the initial Contour Contours, the gradient drops by an iteration number n.
Output: and (5) a level set function after evolution is finished.
Constructing an initial level set function Φ by Contours
for itera in n:
returnΦ
The improved level set is an improvement of the traditional DRLSE level set model aiming at segmentation of overlapped and exfoliated epithelial cells. The improved level set of the invention, relative to the DRLSE model, mainly comprises the following three improvements:
a. a more reasonable initial profile is provided for the level set by the voronoi diagram.
b. Edge enhancement is carried out on an edge indication operator in the DRLSE, so that the interference of the edges of adjacent overlapped cells on the evolution of the contour of the target to be segmented is effectively relieved.
c. The invention provides a more reasonable edge energy item, and compared with the DRLSE edge energy item, the energy item can more accurately measure the degree of the zero level set positioned at the edge position in the image, so that the improved level set can more fully and accurately utilize the edge information of the image to carry out image segmentation.
The definition of the DRLSE edge energy term is as follows:
∫ Ω gδ(Φ)|△Φ|dx
the edge energy term can essentially be regarded as an edge indicator integral along the zero level set. When the edge of the object to be segmented is clear and the background of the object to be segmented is clean, the edge energy term can well measure the degree of the zero level set positioned at the edge pixel point in the image. In practice, however, the object to be segmented in the medical image processed by us has the conditions of blurring, crossing, breaking, etc. of edges, and many interference edges exist in the background of the object to be segmented or in the interior of the object to be segmented itself, so that such a situation may occur: the pixel points where the zero level is located are mostly interference edge pixel points, but because the perimeter of the zero level set is small, even if the zero level set does not really converge to the real edge of the object to be segmented in this case, the energy term takes the minimum value, so that the zero level set is difficult to converge to the real edge of the object to be segmented.
The improvement of the edge energy item is mainly aimed at the situation that the circumference of a zero level set in the edge energy item can bring errors to image segmentation when the original edge energy item in the DRLSE is used for segmenting the medical image because the edge information in the medical image is too complex. To this end, the invention proposes the following improved edge energy term:
the improved edge energy term, i.e. the perimeter of the zero level set divided by the edge energy term of the DRLSE source, enables the energy term to measure well the extent to which the zero level set is located at the edge pixels in the image.
The present invention uses the centroid coordinates of all exfoliated epithelial nuclei to generate a voronoi diagram, the voronoi diagram segmentation results of which are generally already relatively close to the true contours of the cells to be segmented, which includes two benefits: on one hand, the harsh requirement of the DRLSE model on the robustness of the initial contour is relieved; on the other hand, the level set can evolve the accurate contour of the object to be segmented only by less contour evolution times, so that the efficiency of the level set is effectively improved.
The edge enhancement is carried out on the edge indication operator, and mainly aims at weakening the edge of the cell overlapped with the target cell to be segmented, and retaining the edge belonging to the target cell, so that the interference of the edge of the adjacent overlapped cell to the contour evolution of the target cell is prevented to the greatest extent. Edge enhancement employed in the present invention ii The method of (1) is based on the following principle: the exfoliated epithelial cells are usually elliptical, the nuclei are usually located at the center of the cells, the angle formed by the connection line between the target nuclei and the edge pixel points of the target cells and the direction of the gradient at the pixel points should be an acute angle, and the angle formed by the connection line between the target cells and the edge pixel points of the adjacent overlapping cells and the direction of the gradient at the pixel points should be an obtuse angle. Based on such a principle, a large number of edge pixels of cells overlapping with the target cells can be removed. The DRLSE edge indicator operator and the improved edge indicator operator of the present invention are compared as follows:
DRLSE edge indication operator:
the edge indication operator comprises the following steps:
wherein the method comprises the steps ofαAnd an included angle formed by a vector formed by a pixel point in the edge indication operator image and the cell nucleus centroid coordinates of the target cells to be segmented and the gradient direction at the pixel point is represented.
The more reasonable edge energy item is characterized in that: aiming at the problem that zero level sets are difficult to converge to a target edge when the DRLSE edge energy term is used for processing complex edge conditions, the invention provides the following more reasonable edge energy term, which is defined as follows:
the improvement of the energy item is that: dividing the edge energy term of the DRLSE primitive by the perimeter of the zero level set, thereby enabling the energy term to better measure the degree to which the zero level set is located at the edge pixels in the image.
The energy function definition of the improved level set of the present invention is as follows:
E(Φ)＝μR p (Φ)+λL f (Φ)+αA f (Φ)+EE(Φ)，
wherein R is p (Φ) is a regular energy term, lf (Φ) is an edge energy term, af (Φ) is an area energy term, E (Φ) is a shape prior energy term, and the definition of each energy sub-term is as follows:
regular energy term: r is R p (Φ)=∫ Ω p(|ΔΦ|)dx
Edge energy term:
area energy term: a is that f (Φ)＝∫ Ω fH(-Φ)dx
Shape prior energy term: the invention considers that the prior shape of the cervical cells is elliptical, and names the prior term of the shape as an elliptical term, but the invention does not explicitly define the elliptical term, but directly gives the gradient descending flow of the elliptical term, and the calculation process of the gradient descending flow of the elliptical term is as follows: the minimum circumscribed ellipse elipse of the zero level set is fitted first, and then a new level set function Φ_elipse, i.e. the gradient descent of the ellipse term, is constructed using the elipse as the new zero level set.
The level set minimizes the energy function through a gradient descent method, so that the zero level set evolves to the real outline of the object to be segmented, and the gradient descent flow of the level set function is as follows:
for the segmentation of overlapping exfoliated epithelial cells, the present invention proposes for the first time a rough segmentation using a voronoi diagram, followed by a fine segmentation using an improved level set, which has two advantages: on one hand, the Voronoi diagram segmentation result provides a reasonable initial contour for the level set, so that the result of the contour evolution of the level set is more accurate; on the other hand, the result of the voronoi diagram segmentation is usually closer to the real contour of the cell, so that the level set can be optimized with a small number of iterations, thereby reducing the time of contour evolution and improving the efficiency of cell segmentation. In addition, the invention aims at the application scene of dividing the exfoliated epithelial cells, improves the traditional DRLSE level set, and the improvement mainly comprises two aspects: the first improvement is an improvement on the edge energy term, so that the energy term can better measure the degree to which a zero level set is positioned at an edge pixel point in an image; the second improvement is the improvement of the edge indication operator, and the edge enhancement is carried out on the image of the edge indication operator, so that the interference of the edges of adjacent overlapped cells on the contour evolution is prevented to the greatest extent.
Finally, it should be noted that: the foregoing description is only a preferred embodiment of the present invention, and the present invention is not limited thereto, but it is to be understood that modifications and equivalents of some of the technical features described in the foregoing embodiments may be made by those skilled in the art, although the present invention has been described in detail with reference to the foregoing embodiments. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A novel segmentation method for overlapping and falling-off epithelial cells is characterized by comprising the following steps: comprises two modules of cell nucleus segmentation and cell plasma segmentation;
the method of combining MSER algorithm and Kmeans algorithm is used for dividing cell nuclei, and the steps are as follows:
assuming that the original image of the detached epithelial cells to be segmented is imageorigin;
firstly, carrying out graying and bilateral filtering pretreatment on an imageOriginal, and marking the result as imagePreprocess;
dividing the cell nuclei in the image pre-process by using an MSER algorithm to obtain a cell nucleus rough division result, and marking the cell nucleus rough division result as a nucleousRough;
step three, generating an ROI picture for the rough segmentation contour of each cell nucleus in the image nucleic Rough, wherein the ROI picture is obtained by cutting out from the image process according to a rectangular region, and the rectangular region is formed by diffusing L pixel points from top to bottom to left to right respectively by the minimum circumscribed rectangle of the rough segmentation contour of the cell nucleus corresponding to the ROI picture;
clustering the ROI pictures by using a Kmeans algorithm to finally obtain the accurate outline of the cell nucleus, and marking the accurate outline as nucleic;
the Kmeans clustering number is 2, and corresponds to a cell nucleus area and a non-cell nucleus area, and because the gray value of the cell nucleus of the exfoliated epithelium is generally lower than that of surrounding pixel points, the area with lower average gray value in the clustering result is extracted as an accurate area of the cell nucleus;
the cytoplasma segmentation fuses two segmentation methods of a voronoi diagram and a level set, and the segmentation steps are as follows:
firstly, dividing out an exfoliated epithelial cell mass by using a region growing method and combining a Kmeans algorithm, and marking the exfoliated epithelial cell mass as ClumpIMage;
step two, generating a Voronoi image according to the barycenter coordinates of all cell nuclei in the image nucleic, wherein the Voronoi image is formed by the Voronoi image;
step three, overlapping the images ClumpIMage and Voronoi image to obtain the outline of the Voronoi images of all the exfoliated epithelial cells, and marking the outline as Voronoi segments;
and step four, taking the Voronoi figure outline of the cells in the image Voronoi segment as the initial outline of the improved level set, and carrying out level set outline evolution so as to obtain the accurate outline cellFine of the cell plasma.
2. A novel method of dividing overlapping exfoliated epithelial cells according to claim 1, wherein: the image of the exfoliated epithelial cells is imageOriginal, the image is obtained by image acquisition under a microscope, the magnification of an ocular lens of the microscope is 10, and the magnification of an objective lens is 20.
3. The novel method for dividing overlapping and exfoliated epithelial cells according to claim 1, wherein: the cell mass segmentation process described in step one of the cell plasma segmentation method is divided into the following steps:
step one, clustering image processes subjected to graying and bilateral filtering pretreatment by using a Kmeans algorithm, wherein the clustering number is 2, and the clustering number corresponds to a background area and a cell mass area in an image of a cast-off epithelial cell respectively to obtain a cell mass rough segmentation result, and the cell mass rough segmentation result is recorded as ClumpSegRough;
performing morphological corrosion treatment on a background area in the image ClumpSegRough, so that the pixel points in the ClumpSegRough background area are ensured to belong to a real background area to the greatest extent, and the treatment result is recorded as ClumpSegRough hEROsion;
and thirdly, randomly selecting a plurality of coordinate points in a background area of the image ClumpSegRoughelctrosin, taking the coordinate points as seed points of an area growth algorithm, and carrying out area growth on the image process by using the area growth algorithm, wherein a final area growth result is an accurate area of the cell mass.
4. The novel method for dividing overlapping and exfoliated epithelial cells according to claim 1, wherein: in the third step of the cell plasma segmentation method, the images ClumpIMage and Voronoi image are mutually overlapped to obtain the Voronoi image outlines of all the exfoliated epithelial cells, wherein the mutually overlapped mode is to add the RGB channel values of the pixel points corresponding to the two images of ClumpIMage and Voronoi image and take the surplus 255; the outline of the voronoi diagram is characterized in that the cytoplasms of different cells in the diagram are expressed by different colors, namely, the connected domains with different colors in the diagram and the cytoplasms have a one-to-one correspondence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910019927.2A CN111429461B (en) | 2019-01-09 | 2019-01-09 | Novel segmentation method for overlapped and exfoliated epithelial cells |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910019927.2A CN111429461B (en) | 2019-01-09 | 2019-01-09 | Novel segmentation method for overlapped and exfoliated epithelial cells |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111429461A CN111429461A (en) | 2020-07-17 |
| CN111429461B true CN111429461B (en) | 2023-09-01 |
Family
ID=71545695
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910019927.2A Active CN111429461B (en) | 2019-01-09 | 2019-01-09 | Novel segmentation method for overlapped and exfoliated epithelial cells |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111429461B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112053355B (en) * | 2020-09-16 | 2023-03-10 | 昆明理工大学 | Cell image segmentation method |
| WO2024119326A1 (en) * | 2022-12-05 | 2024-06-13 | 深圳华大生命科学研究院 | Gene image data correction method and system, electronic device and storage medium |
| CN116580041B (en) * | 2023-05-30 | 2024-06-18 | 山东第一医科大学附属眼科研究所(山东省眼科研究所、山东第一医科大学附属青岛眼科医院) | Corneal endothelial cell boundary segmentation method and device based on voronoi diagram |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101238487A (en) * | 2005-06-15 | 2008-08-06 | 组织诺斯替斯有限责任公司 | Method for the segmentation of leukocytes |
| TW201124945A (en) * | 2010-01-06 | 2011-07-16 | Chunghwa Telecom Co Ltd | Method and system for automatically detecting representative region of pseudo-object-based image. |
| CN102831608A (en) * | 2012-08-06 | 2012-12-19 | 哈尔滨工业大学 | Unsteady measurement algorithm based image segmentation method of improved rule distance level set |
| CN102831607A (en) * | 2012-08-08 | 2012-12-19 | 深圳市迈科龙生物技术有限公司 | Method for segmenting cervix uteri liquid base cell image |
| CN103345748A (en) * | 2013-06-26 | 2013-10-09 | 福建师范大学 | Positioning and partition method for human tissue cell two-photon microscopic image |
| CN104376559A (en) * | 2014-11-14 | 2015-02-25 | 中国人民解放军军事医学科学院卫生装备研究所 | Medical image segmentation method based on improved range adjustment level set algorithm |
| CN104992435A (en) * | 2015-06-24 | 2015-10-21 | 广西师范大学 | Cervix uteri single cell image segmentation algorithm |
| CN105574528A (en) * | 2015-12-15 | 2016-05-11 | 安徽工业大学 | Synechia cell image segmenting method based on polyphase mutual exclusion level set |
| CN105976367A (en) * | 2016-04-29 | 2016-09-28 | 上海联影医疗科技有限公司 | Image segmentation method, pulmonary nodule detection method and computer-aided detection system |
| CN106650731A (en) * | 2016-12-23 | 2017-05-10 | 中山大学 | Robust license plate and logo recognition method |
| CN107730498A (en) * | 2017-10-30 | 2018-02-23 | 天津理工大学 | Novel uterine neck cell core partitioning algorithm |
| CN108596932A (en) * | 2018-04-18 | 2018-09-28 | 哈尔滨理工大学 | A kind of overlapping cervical cell image partition method |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8571278B2 (en) * | 2005-06-24 | 2013-10-29 | The University Of Iowa Research Foundation | System and methods for multi-object multi-surface segmentation |
| DE102007013971B4 (en) * | 2007-03-23 | 2011-07-14 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V., 80686 | Method and device for determining a cell contour of a cell |
| US9014480B2 (en) * | 2012-07-19 | 2015-04-21 | Qualcomm Incorporated | Identifying a maximally stable extremal region (MSER) in an image by skipping comparison of pixels in the region |
| TWI496112B (en) * | 2013-09-13 | 2015-08-11 | Univ Nat Cheng Kung | Cell image segmentation method and a nuclear-to-cytoplasmic ratio evaluation method using the same |
| US9626583B2 (en) * | 2013-12-12 | 2017-04-18 | University of Pittsburg—Of the Commonwealth System of Higher Education | Automated epithelial nuclei segmentation for computational disease detection algorithms |
| WO2015175837A1 (en) * | 2014-05-14 | 2015-11-19 | Massachusetts Institute Of Technology | Systems and methods for medical image segmentation and analysis |
| US9805248B2 (en) * | 2014-08-29 | 2017-10-31 | Definiens Ag | Applying pixelwise descriptors to a target image that are generated by segmenting objects in other images |
-
2019
- 2019-01-09 CN CN201910019927.2A patent/CN111429461B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101238487A (en) * | 2005-06-15 | 2008-08-06 | 组织诺斯替斯有限责任公司 | Method for the segmentation of leukocytes |
| TW201124945A (en) * | 2010-01-06 | 2011-07-16 | Chunghwa Telecom Co Ltd | Method and system for automatically detecting representative region of pseudo-object-based image. |
| CN102831608A (en) * | 2012-08-06 | 2012-12-19 | 哈尔滨工业大学 | Unsteady measurement algorithm based image segmentation method of improved rule distance level set |
| CN102831607A (en) * | 2012-08-08 | 2012-12-19 | 深圳市迈科龙生物技术有限公司 | Method for segmenting cervix uteri liquid base cell image |
| CN103345748A (en) * | 2013-06-26 | 2013-10-09 | 福建师范大学 | Positioning and partition method for human tissue cell two-photon microscopic image |
| CN104376559A (en) * | 2014-11-14 | 2015-02-25 | 中国人民解放军军事医学科学院卫生装备研究所 | Medical image segmentation method based on improved range adjustment level set algorithm |
| CN104992435A (en) * | 2015-06-24 | 2015-10-21 | 广西师范大学 | Cervix uteri single cell image segmentation algorithm |
| CN105574528A (en) * | 2015-12-15 | 2016-05-11 | 安徽工业大学 | Synechia cell image segmenting method based on polyphase mutual exclusion level set |
| CN105976367A (en) * | 2016-04-29 | 2016-09-28 | 上海联影医疗科技有限公司 | Image segmentation method, pulmonary nodule detection method and computer-aided detection system |
| CN106650731A (en) * | 2016-12-23 | 2017-05-10 | 中山大学 | Robust license plate and logo recognition method |
| CN107730498A (en) * | 2017-10-30 | 2018-02-23 | 天津理工大学 | Novel uterine neck cell core partitioning algorithm |
| CN108596932A (en) * | 2018-04-18 | 2018-09-28 | 哈尔滨理工大学 | A kind of overlapping cervical cell image partition method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111429461A (en) | 2020-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107256558B (en) | Unsupervised automatic cervical cell image segmentation method and system | |
| Song et al. | A deep learning based framework for accurate segmentation of cervical cytoplasm and nuclei | |
| CN103984958B (en) | Cervical cancer cell dividing method and system | |
| CN112215800B (en) | Machine learning-based overlapped chromosome recognition and segmentation method | |
| CN111429461B (en) | Novel segmentation method for overlapped and exfoliated epithelial cells | |
| Poletti et al. | A review of thresholding strategies applied to human chromosome segmentation | |
| CN107492088B (en) | Automatic identification and statistics method for white blood cells in gynecological microscopic image | |
| CN111583227A (en) | Method, device, equipment and medium for automatically counting fluorescent cells | |
| CN113570628B (en) | White blood cell segmentation method based on movable contour model | |
| CN110910367A (en) | Bioreactor cell culture quality evaluation method | |
| CN114494318B (en) | Cornea contour extraction method based on cornea dynamic deformation video of Ojin algorithm | |
| CN110544262B (en) | Cervical cell image segmentation method based on machine vision | |
| WO2022229638A1 (en) | Method & apparatus for processing microscope images of cells and tissue structures | |
| CN112508860B (en) | Artificial intelligence interpretation method and system for positive check of immunohistochemical image | |
| CN113850792A (en) | Cell classification counting method and system based on computer vision | |
| Lim et al. | Cell image processing methods for automatic cell pattern recognition and morphological analysis of mesenchymal stem cells-an algorithm for cell classification and adaptive brightness correction | |
| CN111724378A (en) | Microscopic image cell counting and posture recognition method and system | |
| Lakshmi et al. | Automated extraction of cytoplasm and nuclei from cervical cytology images by fuzzy thresholding and active contours | |
| CN113469939B (en) | HER-2 immunohistochemical automatic interpretation system based on characteristic curve | |
| CN112651368B (en) | DNA ploidy analysis method for large-size cell microscopic image | |
| CN110210578B (en) | Cervical cancer histopathology microscopic image clustering system based on graph theory | |
| Kromp et al. | Machine learning framework incorporating expert knowledge in tissue image annotation | |
| Zhifeng et al. | Chromosome Overlap Segmentation Algorithm Based on Chromosome Outline and Skeleton | |
| CN110210308B (en) | Biological tissue image identification method and device | |
| CN110675402A (en) | Colorectal polyp segmentation method based on endoscope image |
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 | ||
| CB02 | Change of applicant information |
Address after: Floor 1 and 2, unit B, C and D, building B7, medical instrument Park, 818 Gaoxin Avenue, Donghu New Technology Development Zone, Wuhan City, Hubei Province, 430000 Applicant after: Wuhan Lanting intelligent Medicine Co.,Ltd. Address before: 430073 floor 1 and 2, unit B, C and D, building B7, medical instrument Park, 818 Gaoxin Avenue, Donghu New Technology Development Zone, Hongshan District, Wuhan City, Hubei Province Applicant before: WUHAN LANDING MEDICAL HI-TECH Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |