CN113313719A - Leukocyte segmentation method based on visual attention mechanism and model fitting - Google Patents
Leukocyte segmentation method based on visual attention mechanism and model fitting Download PDFInfo
- Publication number
- CN113313719A CN113313719A CN202110652216.6A CN202110652216A CN113313719A CN 113313719 A CN113313719 A CN 113313719A CN 202110652216 A CN202110652216 A CN 202110652216A CN 113313719 A CN113313719 A CN 113313719A
- Authority
- CN
- China
- Prior art keywords
- leukocyte
- segmentation
- white blood
- image
- blood 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.)
- Pending
Links
- 210000000265 leukocyte Anatomy 0.000 title claims abstract description 120
- 230000011218 segmentation Effects 0.000 title claims abstract description 71
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000000007 visual effect Effects 0.000 title claims abstract description 23
- 210000003855 cell nucleus Anatomy 0.000 claims abstract description 21
- 238000003708 edge detection Methods 0.000 claims abstract description 14
- 210000000805 cytoplasm Anatomy 0.000 claims abstract description 12
- 238000001514 detection method Methods 0.000 claims description 17
- 230000001086 cytosolic effect Effects 0.000 claims description 10
- 230000001464 adherent effect Effects 0.000 claims description 4
- 238000000926 separation method Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 abstract description 7
- 210000004940 nucleus Anatomy 0.000 abstract description 7
- 230000023404 leukocyte cell-cell adhesion Effects 0.000 abstract description 6
- 210000005259 peripheral blood Anatomy 0.000 abstract description 4
- 239000011886 peripheral blood Substances 0.000 abstract description 4
- 230000000694 effects Effects 0.000 description 6
- 210000000601 blood cell Anatomy 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 230000005859 cell recognition Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 208000014951 hematologic disease Diseases 0.000 description 2
- 208000032839 leukemia Diseases 0.000 description 2
- 208000030507 AIDS Diseases 0.000 description 1
- 208000023141 Abnormal leukocyte morphology Diseases 0.000 description 1
- 208000019838 Blood disease Diseases 0.000 description 1
- 208000035473 Communicable disease Diseases 0.000 description 1
- 206010061218 Inflammation Diseases 0.000 description 1
- 210000003651 basophil Anatomy 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 238000004159 blood analysis Methods 0.000 description 1
- 238000004820 blood count Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 210000003979 eosinophil Anatomy 0.000 description 1
- 210000003743 erythrocyte Anatomy 0.000 description 1
- 208000018706 hematopoietic system disease Diseases 0.000 description 1
- 210000002865 immune cell Anatomy 0.000 description 1
- 230000036737 immune function Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 230000004054 inflammatory process Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 210000004698 lymphocyte Anatomy 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000001616 monocyte Anatomy 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 210000000440 neutrophil Anatomy 0.000 description 1
- 210000004976 peripheral blood cell Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- 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
-
- 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
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a leukocyte segmentation method based on a visual attention mechanism and model fitting. The invention first proposes a color space volume based on visual attention mechanism to highlight the nuclear region and then segment the nuclei using adaptive thresholding. Then, a cytoplasm segmentation result is obtained by providing a cell nucleus which is removed from a background area based on boundary priori knowledge and taking the obtained central area as an initial leukocyte area, further obtaining a leukocyte contour through edge detection, and subtracting the cell nucleus from the obtained leukocyte contour. Finally, the invention provides a method based on a model fitting strategy to solve the problem of leukocyte adhesion, effectively segments leukocytes (including cell nucleus and cytoplasm) from a peripheral blood smear image, counts and classifies the leukocytes obtained by segmentation, and effectively reduces the influence of human errors.
Description
Technical Field
The invention relates to a leukocyte segmentation method based on a visual attention mechanism and model fitting.
Background
Leukocytes are the most important immune cells of the human body and are very important for maintaining the immune function of the human body. Leukocytes include five types of cells, neutrophils, eosinophils, basophils, lymphocytes, and monocytes. The total number of leukocytes in human peripheral blood, and the ratio and morphology of various leukocytes are important indicators for diagnosing human blood diseases such as leukemia. The differential white blood cell count is an important component of routine hospital blood examinations, i.e., the total number of peripheral blood white blood cells and the percentage of each type of white blood cell. The enumeration can be used to diagnose blood disorders such as leukemia, infectious diseases, inflammation, AIDS by classifying the white blood cell enumeration and analyzing abnormal morphology.
Early hospitals achieved differential counting of leukocytes mainly by manual microscopic examination, and generally professional pathologists observed the color and morphology of leukocytes in stained blood cell images under high power microscope and counted them in differential manner. Generally, the accuracy of manual microscopic examination is high, but much time is consumed, the efficiency is low, and the professional knowledge and skill and the professional experience of an observer are determined by the accuracy of the detection result, so that the manual microscopic examination is not suitable for routine blood detection of a large number of people. The automatic blood cell analyzer is time-saving and high in efficiency, and can effectively reduce the influence of human errors. Automatic blood cell analyzers generally count leukocytes by five categories by physical, and chemical methods, but do not analyze abnormal morphology of leukocytes. In clinical application, abnormal leukocyte morphology is analyzed and checked in a manual rechecking mode.
In light of the problems set forth above, there have been several scholars in recent years proposing related leukocyte segmentation algorithms. For the traditional algorithm, there are threshold segmentation, active contour-based segmentation method and saliency-based segmentation method. The threshold segmentation method includes Otsu method, region growing method, watershed method and their combination algorithm. Cseke et al propose a method for rapidly segmenting nuclei based on automatic selection of thresholds, and the algorithm first adaptively obtains an optimal segmentation threshold based on an Otsu algorithm. Dorini et al propose an algorithm for segmenting leukocytes based on a watershed method, which achieves the whole leukocyte segmentation in two stages: firstly, extracting cell nucleuses by using a watershed algorithm based on image forest transformation, and then realizing the segmentation of cytoplasm based on basic operations such as a threshold value method, morphological operation and the like. Machine learning-based segmentation methods include supervised methods and unsupervised methods. The supervised method comprises a convolutional neural network method and an SVM method, and the unsupervised method comprises K-means, a fuzzy C mean value method, a maximum expectation algorithm and the like. Zheng et al obtains a coarse segmentation result through K-means clustering, then classifies each image pixel, and trains an SVM on the coarse segmentation result to obtain a more accurate segmentation result. Osowski et al propose a blood cell recognition algorithm based on a genetic algorithm and a support vector machine, which first uses the genetic algorithm to extract features, and then uses SVM to realize cell recognition and classification.
However, the existing methods for segmenting leukocytes all have certain limitations, and mainly have the following three problems: (1) the segmentation precision obtained by the existing algorithm is not high, and needs to be further improved (2) the existing algorithm has over-segmentation and under-segmentation phenomena. (3) The problem of adhesion is not well solved by leukocyte segmentation obtained by the existing algorithm.
Disclosure of Invention
The invention aims to provide a leukocyte segmentation method based on a visual attention mechanism and model fitting aiming at the defects of the prior art, solves the problems that the complexity of a detection result is reduced by the need of artificial naked eyes through professional knowledge and skills and practical experience, the leukocyte segmentation precision is low and the adhered leukocytes are separated in the prior art, and improves the blood leukocyte segmentation precision.
In order to achieve the purpose, the technical scheme of the invention is as follows: a leukocyte segmentation method based on visual attention mechanism and model fitting comprises the following steps:
s1, nuclear segmentation based on color space volume;
s2, background removal and cytoplasm segmentation based on boundary priori knowledge;
s3, adherent leukocyte separation based on breakpoint detection and model fitting.
In an embodiment of the present invention, the step S1 is specifically implemented as follows:
s11, passing L*a*b*And L in RGB color space*、a*The linear combination of the R and B components construct a color space volume, the formula is as follows:
b'＝B-R
the image I' is a color space volume, and (I, j) represents the coordinate position of a pixel point in the image;
s12, adaptively finding a threshold value T by using an Otsu algorithm;
s13, segmenting cell nuclei based on the threshold value T, wherein the formula is as follows:
wherein, the image Nu is a cell nucleus segmentation result.
In an embodiment of the present invention, the step S2 is specifically implemented as follows:
s21, dividing the input image I into n areas by using a SLIC algorithm, wherein the first m areas are boundary areas, and the rest n-m areas are areas which are positioned in the middle and do not contact the boundary;
s22, removing the first m boundary regions based on the boundary priori knowledge, and taking the obtained central region as an initial leukocyte region;
s23, obtaining a white blood cell outline by using a Canny edge detection algorithm, and filling the white blood cell outline with an image to obtain a final white blood cell segmentation result;
s24, obtaining a cytoplasmic region, i.e., a cytoplasmic segmentation result, by subtracting the nuclear region from the entire leukocyte region.
In an embodiment of the present invention, the step S3 is specifically implemented as follows:
s31, performing edge detection on the input white blood cell image, and marking as E;
s32, acquiring the maximum connected edge from the E as the outline of the white blood cell, and marking the outline as C;
s33, carrying out breakpoint detection on the C, and if two breakpoints exist, marking the breakpoints as A and B; otherwise, no breakpoint exists;
s34, randomly sampling some data points from the C and generating candidate circles passing through the break points; then, the best circle O having the largest number of inliers is estimated from the candidate circles, and the number of inliers of the generated candidate circle is calculated as follows:
where n is the number of data points, δ is the internal noise scale, d (x)iθ) is the data point xiA residual value from the circle θ;
s35, cutting a section of circular arc L on the best circle O obtained through fitting according to the positions of the two break points A and B;
s36, filling the intercepted arc L to the original incomplete contour to obtain a fitting contour close to the real white blood cell contour, and finally obtaining a binary image of the white blood cells through image filling.
Compared with the prior art, the invention has the following beneficial effects:
the invention can effectively segment the white blood cells (including cell nucleus and cytoplasm) from the peripheral blood smear image, and count and classify the segmented white blood cells, thereby effectively reducing the influence of human errors. First, the color space volume through the visual attention mechanism to highlight the nuclear region and suppress the background region. Then, the method of removing the background area by using the boundary prior knowledge is used, and the obtained result is used as the initial leukocyte area. And finally, acquiring the white blood cell outline through an edge detection algorithm to obtain a final white blood cell segmentation result. And, when there is a leukocyte adhesion condition, a method based on a model fitting strategy is used to solve the problem of leukocyte adhesion. By the method, the segmentation accuracy can be well improved, and the adhesion condition among white blood cells can be well solved.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a diagram of the effect of the present invention on the segmentation of nuclei based on visual attention mechanism and adaptive threshold.
FIG. 3 is a graph showing the effect of the cytoplasm segmentation according to the present invention.
FIG. 4 is a flow chart of the breakpoint detection and model fitting algorithm of the present invention.
Detailed Description
The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.
The invention relates to a leukocyte segmentation method based on a visual attention mechanism and model fitting, which comprises the following steps:
s1, nuclear segmentation based on color space volume;
s2, background removal and cytoplasm segmentation based on boundary priori knowledge;
s3, adherent leukocyte separation based on breakpoint detection and model fitting.
The following is a specific implementation process of the present invention.
The invention researches a leukocyte segmentation algorithm based on a visual attention mechanism and model fitting. First a visual attention mechanism based color space volume is proposed to highlight the nuclear region and then the adaptive thresholding method is used to segment the nuclei. Then, the algorithm provides a cytoplasm segmentation result which is obtained by taking the obtained central region as an initial leukocyte region, further obtaining a leukocyte contour through edge detection and subtracting a cell nucleus from the obtained leukocyte contour based on boundary priori knowledge and removing a background region. Finally, the invention proposes a method based on a model fitting strategy to solve the problem of leukocyte adhesion. The flow of the leukocyte segmentation algorithm based on visual attention mechanism and model fitting is shown in fig. 1.
The method comprises the following specific steps:
1. color space volume based cell nuclei
In the peripheral blood cell image, the stained cell nucleus has a darker color characteristic than the surrounding area, so the cell nucleus area can be regarded as a human interested area (salient area), and in order to improve the effect of cell nucleus segmentation, the invention provides a cell nucleus segmentation algorithm based on a visual attention mechanism and an adaptive threshold, and the steps of the algorithm are as follows:
(1) first, through L*a*b*And L in RGB color space*、a*The linear combination of the R and B components construct a color space volume, the formula is as follows:
b'＝B-R
the image I' is a color space volume, and (I, j) represents the coordinate position of a pixel point in the image;
(2) then, adaptively finding a threshold value T by using an Otsu algorithm;
(3) finally, the cell nuclei are segmented based on a threshold T, as follows:
wherein, the image Nu is a cell nucleus segmentation result.
Fig. 2 is a diagram of the effect of nuclear segmentation based on visual attention mechanism and adaptive threshold. It can be seen from the figure that the nuclear region can be well enhanced and highlighted by the color space volume based on the visual attention mechanism, and the background region can be suppressed, as shown in fig. 2 (b). The threshold T for segmenting the nuclei is easily found from the color space volume by an adaptive thresholding method. Based on the threshold T, a binary map of the cell nucleus is obtained, as shown in fig. 2 (c).
2. Background removal and cytoplasmic segmentation based on boundary prior knowledge
Secondly, the present invention introduces a method of segmenting leukocytes. Firstly, the invention uses Simple Linear Iterative Clustering (SLIC) algorithm to carry out superpixel segmentation on the input white blood cell image, removes the image boundary area based on the boundary priori knowledge, and obtains the initial white blood cell area positioned in the central position as a result. Then, a Canny edge detection algorithm is used for obtaining a white blood cell outline, and the white blood cell outline is subjected to image filling to obtain a final white blood cell segmentation result. Finally, the cytoplasmic region, i.e., the cytoplasmic segmentation result, is obtained by subtracting the nuclear region from the entire leukocyte region. The detailed procedure for cytoplasmic segmentation was as follows:
(1) the input image I is divided into n regions using the SLIC algorithm, the first m regions are boundary regions, and the remaining n-m regions are regions located in the middle without contacting the boundary.
(2) The first m border regions are removed based on the border a priori knowledge, and the obtained central region is used as the initial leukocyte region.
(3) And obtaining a white blood cell contour by using a Canny edge detection algorithm, and filling an image of the white blood cell contour to obtain a final white blood cell segmentation result.
(4) The cytoplasmic region, i.e., the cytoplasmic segmentation result, was obtained by subtracting the nuclear region from the entire leukocyte region.
FIG. 3 is a graph showing the effect of cytoplasm segmentation. Firstly, most of background areas are removed based on superpixel segmentation and background priori knowledge, and the obtained result is the whole leukocyte area and the background area including a small part of background impurities, as shown in fig. 3(b), the method can effectively remove the interference of non-target areas to enable the algorithm to focus on the segmentation of the target area, and the efficiency of the algorithm is improved to a certain extent. In order to obtain a clear and complete white blood cell contour, the present invention uses the classic Canny edge detection algorithm to extract the white blood cell contour, as shown in fig. 3 (c). In the case of complete white blood cell contour without breakpoint, the final white blood cell overall segmentation result is obtained by using image filling, as shown in fig. 3 (d). Finally, the cytoplasm segmentation result is obtained by subtracting the nucleus region from the whole leukocyte region, as shown in FIG. 3 (e).
3. Adherent leukocyte separation based on breakpoint detection and model fitting
Finally, the present invention is mainly studied to investigate the solution of the problem of leukocyte adhesion. At point 2, an algorithm for segmenting white blood cells is introduced, a white blood cell contour is extracted based on Canny detection, and a final white blood cell segmentation result is obtained through image filling. However, when adhesion between the white blood cells and other cells (e.g., red blood cells) is present, the white blood cell contour obtained is incomplete and broken arcs are present. In this case, image filling will fail, i.e., a complete binary image of white blood cells is not obtained. Based on the problem, the invention provides a method for judging whether white blood cells have adhesion or not based on breakpoint detection, when the white blood cells have the adhesion, the algorithm provided by the invention can obtain the position coordinates of two breakpoints, and a circular arc is fitted to the position of the breakpoint based on a model fitting method to make up the original gap, and the fitted final contour is close to the real white blood cell contour, so that the accuracy of white blood cell segmentation is improved.
The algorithm based on breakpoint detection and model fitting comprises the following steps:
(1) firstly, carrying out edge detection on an input leukocyte image, and marking the leukocyte image as E;
(2) then, obtaining the maximum connected edge from the E as the outline of the white blood cell, and marking the maximum connected edge as C;
(3) c, carrying out breakpoint detection, and if two breakpoints exist, marking as A and B; otherwise, no breakpoint exists;
(4) some data points are sampled randomly from C and candidate circles passing the break point are generated. Then, the best circle O having the largest number of inliers is estimated from the candidate circles, and the number of inliers of the generated candidate circle is calculated as follows:
where n is the number of data points, δ is the internal noise scale, d (x)iθ) is the data point xiA residual value from the circle θ;
(5) on the best circle O obtained by fitting, a section of circular arc L is cut out according to the positions of the two break points A and B;
(6) filling the intercepted arc L to the original incomplete contour to obtain a fitting contour close to the real white blood cell contour, and finally obtaining a binary image of the white blood cells through image filling;
fig. 4 is a flowchart of a breakpoint detection and model fitting algorithm. It can be seen from the figure that when there is adhesion problem in one leukocyte image, the leukocyte outline obtained by edge detection is incomplete and jagged. Based on the prior knowledge, the invention provides a method for judging whether the white blood cells are adhered or not based on breakpoint detection, and the method can detect two breakpoints of the outline of the adhered white blood cells, such as a break point A and a break point B in figure 4. When the adhesion condition of the white blood cells is judged, the algorithm further uses a model fitting strategy to fit a circle with the size close to the size of the outline of the real white blood cells, a section of circular arc between two break points A and B is intercepted on the fitted circle based on the detected coordinate information of the two break points, and the position of the notch of the original drawing is filled with the circular arc, so that a new outline which is clear and complete and is close to the size of the outline of the real white blood cells can be obtained. Finally, the whole white blood cell binary image can be obtained through image filling.
The method of the invention is specifically applied as follows:
the invention provides a leukocyte segmentation algorithm based on a visual attention mechanism and model fitting. First a visual attention mechanism based color space volume is proposed to highlight the nuclear region and then the adaptive thresholding method is used to segment the nuclei. Then, the algorithm provides a method based on boundary priori knowledge and background region removal, the obtained central region is used as an initial white blood cell region, and white blood cell contour is further obtained through edge detection. Finally, the invention proposes a method based on a model fitting strategy to solve the problem of leukocyte adhesion. The overall process flow is shown in figure 1.
(1) A color space volume is constructed through color space conversion, then a threshold value is found through an Ostu algorithm in a self-adaptive mode, the cell nucleus is segmented based on the threshold value, and the cell nucleus area can be well highlighted through the color space volume of a visual attention mechanism.
(2) And performing superpixel segmentation by using an SLIC clustering algorithm, removing an image boundary region based on boundary priori knowledge, then obtaining a white blood cell contour by using a Canny edge detection algorithm, filling the white blood cell contour to obtain a white blood cell segmentation result, and then subtracting a cell nucleus region from the white blood cell region to obtain a cytoplasm segmentation result.
(3) The final leukocyte segmentation result is obtained through image filling, and aiming at the condition that the leukocyte outline is incomplete, the method provides a method for judging whether the leukocytes are adhered or not based on breakpoint detection.
The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.
Claims (4)
1. A leukocyte segmentation method based on visual attention mechanism and model fitting is characterized by comprising the following steps:
s1, nuclear segmentation based on color space volume;
s2, background removal and cytoplasm segmentation based on boundary priori knowledge;
s3, adherent leukocyte separation based on breakpoint detection and model fitting.
2. The method for leukocyte segmentation based on visual attention mechanism and model fitting according to claim 1, wherein said step S1 is implemented as follows:
s11, passing L*a*b*And L in RGB color space*、a*The linear combination of the R and B components construct a color space volume, the formula is as follows:
b′＝B-R
the image I' is a color space volume, and (I, j) represents the coordinate position of a pixel point in the image;
s12, adaptively finding a threshold value T by using an Otsu algorithm;
s13, segmenting cell nuclei based on the threshold value T, wherein the formula is as follows:
wherein, the image Nu is a cell nucleus segmentation result.
3. The method for leukocyte segmentation based on visual attention mechanism and model fitting according to claim 1, wherein said step S2 is implemented as follows:
s21, dividing the input image I into n areas by using a SLIC algorithm, wherein the first m areas are boundary areas, and the rest n-m areas are areas which are positioned in the middle and do not contact the boundary;
s22, removing the first m boundary regions based on the boundary priori knowledge, and taking the obtained central region as an initial leukocyte region;
s23, obtaining a white blood cell outline by using a Canny edge detection algorithm, and filling the white blood cell outline with an image to obtain a final white blood cell segmentation result;
s24, obtaining a cytoplasmic region, i.e., a cytoplasmic segmentation result, by subtracting the nuclear region from the entire leukocyte region.
4. The method for leukocyte segmentation based on visual attention mechanism and model fitting according to claim 1, wherein said step S3 is implemented as follows:
s31, performing edge detection on the input white blood cell image, and marking as E;
s32, acquiring the maximum connected edge from the E as the outline of the white blood cell, and marking the outline as C;
s33, carrying out breakpoint detection on the C, and if two breakpoints exist, marking the breakpoints as A and B; otherwise, no breakpoint exists;
s34, randomly sampling some data points from the C and generating candidate circles passing through the break points; then, the best circle O having the largest number of inliers is estimated from the candidate circles, and the number of inliers of the generated candidate circle is calculated as follows:
where n is the number of data points, δ is the internal noise scale, d (x)iθ) is the data point xiA residual value from the circle θ;
s35, cutting a section of circular arc L on the best circle O obtained through fitting according to the positions of the two break points A and B;
s36, filling the intercepted arc L to the original incomplete contour to obtain a fitting contour close to the real white blood cell contour, and finally obtaining a binary image of the white blood cells through image filling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110652216.6A CN113313719A (en) | 2021-06-11 | 2021-06-11 | Leukocyte segmentation method based on visual attention mechanism and model fitting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110652216.6A CN113313719A (en) | 2021-06-11 | 2021-06-11 | Leukocyte segmentation method based on visual attention mechanism and model fitting |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113313719A true CN113313719A (en) | 2021-08-27 |
Family
ID=77378224
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110652216.6A Pending CN113313719A (en) | 2021-06-11 | 2021-06-11 | Leukocyte segmentation method based on visual attention mechanism and model fitting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113313719A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114937018A (en) * | 2022-05-30 | 2022-08-23 | 福建师范大学 | White blood cell segmentation method based on color space volume and super pixels |
| WO2023220913A1 (en) * | 2022-05-17 | 2023-11-23 | 京东方科技集团股份有限公司 | Cell image processing method, electronic device and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100072846A (en) * | 2008-12-22 | 2010-07-01 | 한국전자통신연구원 | Apparatus and method for image segmentation |
| CN110148126A (en) * | 2019-05-21 | 2019-08-20 | 闽江学院 | Blood leucocyte dividing method based on color component combination and contour fitting |
-
2021
- 2021-06-11 CN CN202110652216.6A patent/CN113313719A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100072846A (en) * | 2008-12-22 | 2010-07-01 | 한국전자통신연구원 | Apparatus and method for image segmentation |
| CN110148126A (en) * | 2019-05-21 | 2019-08-20 | 闽江学院 | Blood leucocyte dividing method based on color component combination and contour fitting |
Non-Patent Citations (1)
| Title |
|---|
| HAONAN ZHENG: "White Blood Cell Segmentation Based on Visual Attention Mechanism and Model Fitting", 《2020 INTERNATIONAL CONFERENCE ON COMPUTER ENGINEERING AND INTELLIGENT CONTROL (ICCEIC)》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023220913A1 (en) * | 2022-05-17 | 2023-11-23 | 京东方科技集团股份有限公司 | Cell image processing method, electronic device and storage medium |
| CN114937018A (en) * | 2022-05-30 | 2022-08-23 | 福建师范大学 | White blood cell segmentation method based on color space volume and super pixels |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Automation‐assisted cervical cancer screening in manual liquid‐based cytology with hematoxylin and eosin staining | |
| CN112070772B (en) | Blood leukocyte image segmentation method based on UNet++ and ResNet | |
| Ramoser et al. | Leukocyte segmentation and classification in blood-smear images | |
| Miao et al. | Simultaneous Segmentation of Leukocyte and Erythrocyte in Microscopic Images Using a Marker‐Controlled Watershed Algorithm | |
| NL2024774B1 (en) | Blood leukocyte segmentation method based on adaptive histogram thresholding and contour detection | |
| Alreza et al. | Design a new algorithm to count white blood cells for classification leukemic blood image using machine vision system | |
| NL2024777B1 (en) | Blood leukocyte segmentation method based on color component combination and contour fitting | |
| Mohamed et al. | An enhanced threshold based technique for white blood cells nuclei automatic segmentation | |
| CN113313719A (en) | Leukocyte segmentation method based on visual attention mechanism and model fitting | |
| Zhou et al. | Leukocyte image segmentation based on adaptive histogram thresholding and contour detection | |
| Pandit et al. | Survey on automatic rbc detection and counting | |
| Gautam et al. | White blood nucleus extraction using K-Mean clustering and mathematical morphing | |
| Huang et al. | Leukocyte nucleus segmentation and recognition in color blood-smear images | |
| Ghane et al. | Classification of chronic myeloid leukemia cell subtypes based on microscopic image analysis | |
| CN113408480A (en) | Artificial intelligent auxiliary diagnosis system for blood diseases based on bone marrow cell morphology | |
| Zheng et al. | White blood cell segmentation based on visual attention mechanism and model fitting | |
| Vale et al. | Automatic segmentation and classification of blood components in microscopic images using a fuzzy approach | |
| CN111311628A (en) | Full-automatic high-performance leukocyte segmentation method | |
| Ruberto et al. | A leucocytes count system from blood smear images. | |
| Rege et al. | Automatic leukemia identification system using otsu image segmentation and mser approach for microscopic smear image database | |
| CN110766680B (en) | Leukocyte image segmentation method based on geometric constraint | |
| Deb et al. | A noble technique for detecting anemia through classification of red blood cells in blood smear | |
| Khong et al. | A novel approach for efficient implementation of nucleus detection and segmentation using correlated dual color space | |
| CN112634338A (en) | Cerebrospinal fluid cell feature extraction method based on gray level co-occurrence matrix | |
| Zhang et al. | A robust and high-performance white blood cell segmentation algorithm |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210827 |