NL2024777B1 - Blood leukocyte segmentation method based on color component combination and contour fitting - Google Patents

Abstract

The invention relates to a blood leukocyte segmentation method based on color component combination and contour fitting. The method comprises: first, enhancing the contrast of leukocyte nuclei by using color component combination, and further segmenting the nuclei by using a classical thresholding segmentation method; second, removing the background of an image by using color prior knowledge, and performing edge detection and contour fitting to obtain a leukocyte segmentation result, and finally, subtracting the nucleus segmentation result from the leukocyte segmentation result to obtain a cytoplasm segmentation result. Experimental results on leukocyte image sets under standard and fast staining preparation conditions show that the method of the invention improves the leukocyte segmentation precision.

Description

BLOOD LEUKOCYTE SEGMENTATION METHOD BASED ON COLOR COMPONENT

COMBINATION AND CONTOUR FITTING Technical Field The invention belongs to the technical field of image processing, and particularly relates to a blood leukocyte segmentation method based on color component combination and contour fitting, intended to segment leukocytes in blood cell images collected during routine blood tests. Background Blood routine tests are common items in human health testing. An important part of the routine blood tests is to classify and count leukocytes and analyze abnormal morphology. At present, domestic hospitals usually first use a blood cell analyzer based on electrical impedance method (physical method) and flow analysis method (physical-chemical method) to perform blood cell classification and counting. Classification counting and morphological analysis of leukocytes in blood smears are of great significance in the diagnosis of blood diseases such as leukemia. When the blood cell count is abnormal or the attending doctor suspects that a patient has a blood disease, the laboratory doctor will perform push-up, staining, and microscopic examination on the blood to confirm the classification and count of leukocytes and analyze the abnormal morphology. The accuracy of manual microscopy depends on the professional skills of the doctor. It has the problems of strong subjectivity, large individual differences, and time-consuming and labor-intensive problems. It is also likely to affect the precision of the test due to the doctor's visual fatigue. Computer-assisted automatic analysis of blood cell images can not only save manpower and time, but also reduce human analysis errors caused by fatigue. Leukocyte segmentation is the basis of automatic analysis of blood cell images. The precision of leukocyte segmentation will directly affect the accuracy of subsequent leukocyte counts and morphological analysis.

Leukocyte images can be obtained by shooting blood smears with a digital imaging device. The unstained leukocytes which are similar in color to the background are difficult to recognize due to low contrast. For this reason, when preparing blood smears, staining is usually performed to enhance the contrast between leukocytes and the background and improve the degree of recognition. Standard blood smear preparation methods commonly use Wright's staining method and Giemsa staining method to stain cells, achieving a good and stable staining effect; however, staining usually takes more than ten minutes and the staining speed is slow, which cannot meet the needs of wide-scale clinical applications. Huazhong University of Science and Technology's research team, leaded by professors Liu Jianguo and Wang Guoyou, proposed a method for rapid preparation of blood smears, which has a high staining speed and shortens the staining time of cells to about ten seconds; however, with an unstable staining effect, it easily causes dark impurities and a contaminated background, and it will dissolve red blood cells that have a diagnostic effect for some blood diseases.

The challenges of leukocyte image segmentation: (1) Different staining reagents and staining time will cause leukocytes in different blood cell images to have color differences and individual differences; (2) parameter settings, shooting environment, etc., of an imaging device may cause leukocytes to have blurred edges, unclear texture, low contrast, noise and the like; and (3) in standard staining preparation, white blood cells and red blood cells and surrounding staining impurities sometimes adhere.

Leukocyte segmentation is intended to extract a region where a single leukocyte is located from an image of stained human peripheral blood cells, and then segment the nucleus and cytoplasm, as shown in FIG. 1. In recent years, scholars at home and abroad have conducted a series of studies on leukocyte segmentation. Based on the techniques used in existing leukocyte segmentation methods, we classify them as supervised leukocyte segmentation [1] and unsupervised leukocyte segmentation

[2]. The unsupervised leukocyte segmentation method directly implements segmentation based on the features of leukocytes, such as color and brightness. The most commonly used leukocyte segmentation technique is thresholding segmentation. The others are morphological transformation, fuzzy theory, clustering, deformation model, watershed segmentation, region merging, visual attention model, and edge detection in sequence. Supervised leukocyte segmentation, which considers the image segmentation problem as an image classification problem, is implemented as follows: the features (such as color and texture) of the training samples are extracted first, and then a classifier is trained using the features of the training samples, and finally the trained classifier is used to classify pixels in an image of test samples to recognize the region where leukocytes are located. The most commonly used supervised leukocyte segmentation technique is support vector machine. The others are neural network, nearest neighbor classifier, extreme learning machine, and random forest classifier in sequence.

The segmentation precision of the existing blood leukocyte image segmentation method needs to be further improved. In order to improve the segmentation precision, the invention provides a leukocyte segmentation method based on color component combination and contour fitting.

References:

[1] Gu Guanghua, Cui Dong. Flexible Combination Segmentation Algorithm for leukocyte images [J]. Chinese Journal of Scientific Instrument, 2008, 29 (9): 1977-1981.

[2] Zheng Xin, Wang Yong, Wang Guoyuo. White Blood Cell Segmentation Using Expectation-Maximization and Automatic Support Vector Machine Learning [J]. Journal of Data Acquisition and Processing, 2013, 28 (5): 614-619.

Summary of the Invention The objective of the invention is to provide a blood leukocyte segmentation method based on color component combination and contour fitting, which can improve the precision of leukocyte segmentation.

To achieve the above objective, the technical solution of the invention is: a blood leukocyte segmentation method based on color component combination and contour fitting, comprising: first, enhancing the contrast of leukocyte nuclei by using color component combination, and further segmenting the nuclei by using a classical thresholding segmentation method; second, removing the background of an image by using color prior knowledge, and performing edge detection and contour fitting to obtain a leukocyte segmentation result; and finally, subtracting the nucleus segmentation result from the leukocyte segmentation result to obtain a cytoplasm segmentation result.

In an embodiment of the invention, the Step of enhancing the contrast of leukocyte nuclei by using color component combination, and further segmenting the nuclei by using a classical thresholding segmentation method is specifically implemented as follows: step S11, enhancing a nucleus region by using color component combination to, which is formally defined as follows: 1 =1,+1,—kxlI, (1) where /,, I; and /, represent red, blue, and green components in RGB color space, respectively; step S12, after nucleus enhancement, extracting the nucleus region by using the classical image thresholding segmentation method Otsu, as follows: step S121, determination of a segmentation threshold: For a given enhanced image /, having N pixels and L gray levels {0,1, ..., L-1}, determining the segmentation threshold t* by using the Otsu algorithm, which is formally described as follows: t= AFg MAK ee; {OO (LL, — #4)" (2) where ©, represents the ratio of target pixels (pixels having gray levels {t +1, t +2, … L-1})mn the segmentation result corresponding to the segmentation threshold £ to the total pixels of the image, and ©, represents the ratio of background pixels (pixels having gray levels {0,1,...,t}) in the segmentation result corresponding to the segmentation threshold 7 to the total pixels of the image; x, and uy, represent the average gray values of the target pixels and the background pixels; step S122, image thresholding segmentation: with t * as a segmentation threshold, performing the following thresholding segmentation on image I: ejelt Fh ” lo, otherwise step S13, correction of the nucleus segmentation result based on morphological operations: first selecting a largest connected white region, i.e., a target region, and then performing image filling to correct the initial nucleus segmentation result to obtain the final nucleus segmentation result.

In an embodiment of the invention, the step of removing the background of an image by using color prior knowledge, and performing edge detection and contour fitting to obtain a leukocyte segmentation result is specifically implemented as follows: step S21, image background removal based on the color prior knowledge: removing the image background through thresholding segmentation: Hi). U { (V3) EY I(i,].0)= | |[255.255,255], otherwise (4) where t represents a threshold parameter, and 7, represents background removed, i.e., the image after it becomes white; step S22: performing edge detection on the background-removed image to extract the initial contour of a leukocyte; and to fit the contour of a leukocyte at an adhesion, first finding two breakpoints at an opening of the leukocyte, and then fitting an arc segment between the two breakpoints to achieve the separation of the leukocyte from adhesion, wherein the selection process of the breakpoint is as follows:

step S221, using a Harris corner detector to detect corner points in the contour of the leukocyte; step S222, with each corner point as the coordinate origin, determining a local Cartesian coordinate system composed of coordinate points (0, d), (0, -d), (-d, 0), and (d, 0), where d =|0.1xH 1, H represents the total number of rows of the image matrix, and the symbol | 5 represents rounding down; step S223, making four rays from the image center point to the four Cartesian coordinate points determined in step S222, and calculating the number of contour points where each ray intersects with the contour of the leukocyte; if the number of contour points through which one or more of the four rays pass is 0, determining the corresponding corner point as a breakpoint, or otherwise, determining the corresponding corner point as a non-breakpoint; similarly, if the initial contour of the leukocyte has more than two breakpoints, performing the subsequent steps; or otherwise, terminating the process and determining the initial contour of the leukocyte as the final contour of the leukocyte. step S224: calculating distances between the nucleus centroid and the two breakpoints respectively and denoting the distances as a and b; calculating a distance between the two breakpoints and denoting the distance as c; using the two breakpoints and the nucleus centroid point to define a triangle, and calculating an included angle 6 between lines between two breakpoints and the nucleus centroid point as follows: a+b cl ¢ = arccos (re) (5) step S225: finding points on the axis of the line segment between the two breakpoints having an inclined angle of 6 with the line segment between the two breakpoints, and among these points, denoting a point closest to the center point of the image as p2; and step S226, calculating a distance between the point p2 and any breakpoint and denoting the distance as r; constructing a circle with p2 as the center point and r as the radius; and separating the adherent leukocyte by using the arc between the two breakpoints as the fitted contour of the leukocyte between the two breakpoints.

Compared with the prior art, the invention has the following beneficial effects: In order to improve the segmentation precision of blood leukocyte images under standard and fasting preparation conditions, especially the segmentation precision of the adherent leukocyte, the invention provides a leukocyte segmentation algorithm based on color component combination and contour fitting. The method comprises the following steps: first, enhancing the contrast of leukocyte nuclei by using color component combination, and further segmenting the nuclei by using a classical thresholding segmentation method; second, removing the background of an image by using color prior knowledge, and further performing edge detection to obtain a maximum connected component as an initial leukocyte contour; and finally, performing contour fitting, image filling and morphological erosion operation to realize extraction of a leukocyte region, thus obtaining a cytoplasm segmentation result.

Experimental results on standard and fast stained leukocyte image data sets prove that the method provided by the invention improves the leukocyte segmentation precision under two preparation conditions.

Brief Description of the Drawings FIG. 1 shows examples of leukocyte images prepared by rapid and standard staining preparation methods.

FIG. 2 is a flowchart of the method according to the invention.

FIG. 3 shows examples of nine colors and their color components.

FIG. 4 shows examples of color components, where: (a) original image, (b) nucleus enhancement result, (c) thresholding segmentation result of sub-image (b), and (d) nucleus segmentation result.

FIG. 5 shows the background removal results of the leukocyte images, where (a) the leukocyte images under fast and standard staining conditions, (b) the green component of the sub-image (a), and (©) the background removal result of the sub-image (a).

FIG. 6 shows the edge detection results, where (a) leukocyte images under fast and standard staining conditions, (b) edge detection results, and (c) the initial contour of a leukocyte.

FIG. 7 shows breakpoint selection, where (a) the corner points indicated by the blue arrows detected from the contour points, (b) the local Cartesian coordinate system represented by the yellow cross and breakpoints indicated by the green arrows, (c) the corner points detected from the contour points, and (d) the local Cartesian coordinate system represented by the yellow cross.

FIG. 8 shows contour fitting, where (a) an image of an adherent leukocyte, (b) a triangle defined by an image center point pl and two breakpoints, (c) a selected point p2, i.e, a radius r, (d) The leukocyte contour enclosed by the fitted arc segment, (e) the results after image filling, and (f) the results of image erosion.

FIG. 9 shows the results of cytoplasm segmentation, where (a) the results of leukocyte segmentation, (b) the results of nucleus segmentation, and (c) the results of cytoplasm segmentation.

FIG. 10 shows the results of fast stained leukocyte segmentation, where the columns from lett to right are original image, the results of manual ideal segmentation, the segmentation results of Gu’s 5 method [1], the segmentation results of Zheng’s method [2], and the segmentation results of the method of the invention in sequence.

FIG. 11 shows the results of standard stained leukocyte segmentation, where the columns from left to right are original image, the results of manual ideal segmentation, the segmentation results of Gu’s method [1], the segmentation results of Zheng’s method [2], and the segmentation results of the method of the invention in sequence.

FIG. 12 shows the segmentation results of the adherent leukocyte under standard staining condition, where the columns from left to right are original image, the results of manual ideal segmentation, the segmentation results of Gu’s method [1], the segmentation results of Zheng’s method [2], and the segmentation results of the method of the invention in sequence.

Detailed Description The technical solution of the invention will be described below in detail with reference to the accompanying drawings.

As shown in FIG. 2, the invention provides a blood leukocyte segmentation method based on color component combination and contour fitting, comprising: first, enhancing the contrast of leukocyte nuclei by using color component combination, and further segmenting the nuclei by using a classical thresholding segmentation method, second, removing the background of an image by using color prior knowledge, and performing edge detection and contour fitting to obtain a leukocyte segmentation result; and finally, subtracting the nucleus segmentation result from the leukocyte segmentation result to obtain a cytoplasm segmentation result.

The implementation process of the invention is detailed as follows.

1. Nucleus segmentation

1.1 Nucleus enhancement based on color component combination The method of the invention provides a strategy for enhancing the nucleus by combining color components on the basis of observing the color composition of the nucleus of the leukocyte. Taking FIG. 3 as an example, the three values in each rectangular block in the figure represent the R, G, and

B component values of the corresponding color of the color block. It can be observed that when the R and B component values are much greater than the G component value, the corresponding color is closer to the color of the cell nucleus. Therefore, color component combination is proposed to enhance the nucleus region, which is formally described as follows: > L=1+1,—kxI, (1) where /,, 4, and I, represent red, blue, and green components in RGB color space, respectively, taking FIG. 4 (a) as an example, a color component combination image obtained after nucleus enhancement is shown in FIG. 4 (b). It can be seen from the figure that the region outside the nucleus becomes dark after the nucleus enhancement, and the contrast of the nucleus region is enhanced, which is conducive to subsequent segmentation.

1.2 Rough nucleus segmentation based on thresholding technology After the nucleus enhancement is performed, the image region outside the nucleus basically becomes a black background region, and there is a clear gray difference between the black background region and the nucleus region. Therefore, the classical image thresholding segmentation method Otsu can be used to easily extract the nucleus region. Extraction of a nucleus region is detailed as follows: 1) determination of a segmentation threshold: For a given enhanced image /; having N pixels and L gray levels {0,1, ..., L-1}, the segmentation threshold t* is determined by using the Otsu algorithm, which is formally described as follows: / , U Arg MAX eg {OO (U, —1) } (2) where o, represents the ratio of target pixels (pixels having gray levels {t+1,t +2, … L-1}) in the segmentation result corresponding to the segmentation threshold £ to the total pixels of the image, and ©; represents the ratio of background pixels (pixels having gray levels {0,1,...,t}) in the segmentation result corresponding to the segmentation threshold to the total pixels of the image; wx, and 44 represent the average gray values of the target pixels and the background pixels; 2) image thresholding segmentation: with t * as a segmentation threshold, thresholding segmentation is performed on image 1; as follows: Bi.) - a 6 0, otherwise

As shown in FIG. 3 (b), the thresholding segmentation result is shown in FIG. 3 (€).

1.3 Correction of the nucleus segmentation result based on morphological operation As shown in FIG. 3 (c), in the thresholding segmentation results, holes appear in the nucleus, and sometimes a small-area false nucleus region appears. Therefore, in the method of the invention, a > largest connected white region, i.e., a target region, is first selected and then image filling is performed to correct the initial nucleus segmentation result to obtain the final nucleus segmentation result.

2. Cytoplasm segmentation

2.1 Image background removal based on the color prior knowledge Background removal is intended to remove regions other than the leukocyte in the image, paving the way for leukocyte segmentation. It can be seen from FIG. 1 that the background of the standard-stained leukocyte image is green, so the green component value of the background region of the image is large; the background of the fast-stained leukocyte image is yellow, and it can be known from the color prior knowledge that yellow can be obtained by mixing red and green. Therefore, it can be seen that the common feature of the leukocyte image under the two staining conditions is that the value of the green component is relatively large. Based on this feature, the invention realizes image background removal through thresholding segmentation as follows: Co [gif Ii, j2)St 01) = A otherwise (4) where t represents a threshold parameter, and /, represents background removed, i.e., the image after it becomes white;

2.2 Extraction and correction of leukocyte contour Edge detection is performed on the background-removed image to extract the initial contour of the leukocyte. Taking the leukocyte images under fast staining and standard staining conditions in FIG. 5 as an example, the edge detection results are shown in FIG. 6 (b). There are false target edges in the detection results. For this reason, the method of the invention retains only the white connected components with the most pixels in the edge detection results as the initial contour of the leukocyte, and the result is shown in FIG. 6 (c). Observing the image at the top in FIG. 6 (c), it can be seen that when the leukocyte is adhered to the surrounding red blood cells, the initial contour of the leukocyte is not closed, and a contour fitting is required to close it to achieve the separation of the leukocyte from adhesion. To fit the contour of a leukocyte at an adhesion, according to the method of the invention, two breakpoints at an opening of the leukocyte are found first, and then an arc segment between the two breakpoints is fitted to achieve the separation of the leukocyte from adhesion. The selection process of the breakpoints here is as follows: 1) A Harris corner detector is used to detect corner points in the contour of the leukocyte.

2) With each corner point as the coordinate origin, a local Cartesian coordinate system composed of coordinate points (0, d), (0, -d), (-d, 0), and (d, 0) is determined, where d= | 0. IxH 1, H represents the total number of rows of the image matrix, and the symbol |i] represents rounding down; the local Cartesian coordinate system in FIGS. 7 (b) and (d) are indicated by a cross.

3) Four rays, as shown in FIGS. 7 (b) and (d), are made from the image center point to the four Cartesian coordinate points determined in Step S222, and the number of contour points where each ray intersects with the contour of the leukocyte is calculated; if the number of contour points through which one or more of the four rays pass is 0, the corresponding corner point is determined as a breakpoint, or otherwise, the corresponding corner point is determined as a non-breakpoint; similarly, the two corner points indicated by arrows in FIG. 7 (b) are breakpoints, and FIG. 7 (d) has no breakpoints; if the initial contour of the leukocyte has more than two breakpoints, the subsequent steps are performed; or otherwise, the process is terminated and the initial contour of the leukocyte is determined as the final contour of the leukocyte.

Distances between the nucleus centroid and the two breakpoints are calculated respectively and denoted as a and b; a distance between the two breakpoints is calculated and denoted as c; a triangle is defined by using the two breakpoints and the nucleus centroid point, and an included angle 8 between lines between two breakpoints and the nucleus centroid point is calculated (as shown in FIG. 8(B) as follows: a +b cl @ = arccos (re) (5) 5) Points on the axis of the line segment between the two breakpoints having an inclined angle of 6 with the line segment between the two breakpoints are found, and among these points, a point closest to the center point of the image 1s denoted as p2, as shown in FIG. 8(c).

6) A distance between the point p2 and any breakpoint is calculated and denoted as r; a circle is constructed with p2 as the center point and r as the radius; and the adherent leukocyte is separated by using the arc between the two breakpoints as the fitted contour of the leukocyte between the two breakpoints. FIG. 8 (d) shows the leukocyte contour closed by the arc segment, FIG. 8 (e) shows the result after image filling is performed on FIG. 8 (d), and Fig. 8 (f) shows the result after the image > erosion is performed on FIG. 8(e).

2.3. Cytoplasm segmentation The cytoplasm region can be obtained by subtracting the nucleus region from the previously obtained leukocyte region, and the result is shown in FIG. 9. To evaluate the performance of the leukocyte segmentation method, we perform experiments on a data set consisting of 100 fast-stained leukocyte images and a data set consisting of 50 standard-stained leukocyte images. The size of each fast-stained leukocyte image is 120x120, and the size of each standard-stained leukocyte image is 260x260. The manual ideal segmentation result of each image is given by the hospital blood examiner. The method of the invention first performs a qualitative comparison of segmentation precision on 8 representative fast-stained images and 8 standard-stained images with two existing leukocyte segmentation methods (Gu’s method [1] and Zheng’s method [2]); then, four common segmentation measures are used to quantitatively compare the average segmentation precision of the three algorithms on two data sets. These four measures are misclassification error (ME), false positive rate/false positive rate (FPR), false negative rate (FNR), and Kappa index (KI), which are defined as follows: ve 1 Be OB, |+|, OE, (6) IB, HIE fpr =P OF] (7) IB, png =n OL] (8)

IE Krall Fl (9) [E+E where, B, and #, respectively represent the background and target of the manual ideal standard segmentation result, B, and I, respectively represent the background and target in the segmentation result corresponding to the automatic segmentation algorithm, and | - | represents the number of elements in a set. The values of the four measures are all between 0 and 1. Lower ME, FPR, and FNR values represent better segmentation effects, and higher KI values represent better segmentation effects. All experiments are performed on a desktop computer with a 2.39GHz Intel Xeon W3503 ° CPU and a 6G memory.

1. Parameter selection The most important parameters of the method of the invention are k in Equation (1) and t in Equation (4). Taking all leukocyte images on the fast-stained image data sets as test objects, we explore the influence of parameters k and t on the segmentation precision of the method of the 0 invention, where the value of k is taken from a set {1, 3,5}, and the value of t is taken from a set {165, 190, 215}. For different values of parameter k, the average ME and KI values corresponding to the nucleus segmentation results obtained by the method of the invention are shown in Table 1. As can be seen from Table 1, when k = 3, the method of the invention obtains the minimum average ME and the maximum KI value, and the corresponding segmentation precision is the highest. For different values u of parameter t, the measurement results of ME and KI obtained by the method of the invention are shown in Table 2. As can be seen from Table 2, when t = 190, the method of the invention obtains the minimum average ME and the maximum KI value, and the corresponding segmentation precision is the highest. Table 1 Average KI and ME values in nucleus segmentation results obtained by the method of the invention on fast-stained image data sets under different values of parameter k k=1 k=3 k=5 KI 0.6383 0.9361 0.7586 ME 0.1527 0.0154 0.0465 Table 2 Average KI and ME values in leukocyte segmentation results obtained by the method of the invention on fast-stained image data sets under different values of parameter t t=165 t=190 t=215 KI 0.7460 0.9726 0.6443 ME 0.0872 0.0134 0.1174

2. Qualitative comparison In order to qualitatively compare the segmentation effects of the three leukocyte segmentation = methods, three sets of experiments are performed, where 8 fast-stained leukocyte images, 4 standard-stained leukocyte images without adhesion, and 4 standard-stained leukocyte images with adhesion are respectively segmented and the results are shown in FIGS. 10-12.

FIG. 10 shows the segmentation results of 8 leukocyte images under fast staining preparation condition. In FIG. 10, the five columns from left to right respectively show original image, the results of manual ideal segmentation, the segmentation results of Gu’s method {1}, the segmentation results ° of Zheng’s method {2}, and the segmentation results of the method of the invention. It can be seen from FIG. 10 that the segmentation effect of the method of the invention is generally better than that of the other two methods. In terms of nucleus segmentation, the Gu’s method obtains relatively satisfactory segmentation results on FIGS. 10 (a)-(b) and (d)-(e); the Zheng’s method obtains relatively satisfactory segmentation results on FIGS. 10 (a)-(e) and (h). In terms of cytoplasm 1 segmentation, the Gu’s method only obtains relatively satisfactory segmentation results on FIGS. 10 (b) and (d), while the Zheng’s method only obtains relatively satisfactory segmentation results on FIGS. 10 (d) and (h).

FIG. 11 shows the segmentation results of four leukocyte images without adhesion under standard staining preparation condition. In FIG. 11, the five columns from left to right respectively a show original image, the results of manual ideal segmentation, the segmentation results of Gu’s method {1}, the segmentation results of Zheng’s method {2}, and the segmentation results of the method of the invention. It can be seen from FIG. 11 that for the nucleus segmentation, the Gu’s method and the Zheng’s method obtain satisfactory segmentation results on FIGS. 11 (a) and (c)-(d), but the Gu’s method causes under-segmentation on FIG. 11 (b), and Zheng’s method causes both 2 under-segmentation and over-segmentation on FIG. 11 (b). The method of the invention obtains good segmentation results on all four images. For the cytoplasm segmentation, the Gu’s method and the method of the invention achieve better segmentation results than the Zheng’s method. The Zheng’s method causes over-segmentation in FIGS. 11 (b)-(c).

FIG. 12 shows the segmentation results of 4 leukocyte images with adhesion under standard > staining preparation condition. In FIG. 6, the five columns from left to night respectively show original image, the results of manual ideal segmentation, the segmentation results of Gu’s method {1}, the segmentation results of Zheng’s method {2}, and the segmentation results of the method of the invention. It can be seen from FIG. 12 that for nucleus segmentation, all three methods slightly cause over-segmentation, but the method of the invention and the Zheng method have better segmentation * effects than the Gu’s method. For the cytoplasm segmentation, the Gu;s method causes over-segmentation on FIGS. 12 (a) and (c), and causes under-segmentation on FIGS. 12 (b) and (d). The Zheng’s method causes over-segmentation on all four images. The method of the invention has better segmentation eftect than the other two methods. Table 3 Quantitative comparison of leukocyte nucleus segmentation results of three quantitative experiments | m8 | FR [| FR | KI ei | CT

0.056 0078 Zheng 0.021 0.007 0.121 0.914 method” The algorithm 0.015 0.001 0.115 0.936 of the invention | Experiment? | 0051 ois 04 Zheng's 0.026 0.017 0.065 0.935 method" The algorithm 0.034 0.031 0.050 0.920 of the invention To epee |

0.047 0.031 0.100 0.891 Zheng ol 0.104 0.076 0.181 0.774 method The algorithm 0.041 0.033 0.073 0.894 of the invention Table 4 Quantitative comparison of leukocyte segmentation results of three quantitative experiments | me | FR [ PNR | KI Tee |

0.166 6174

0.068 0.095 0.001 0.900 method The algorithm 0.013 0.008 0.034 0.973 of the invention ee |

0.117 0.099 0.166 0.781 Zheng > 0.070 0.095 0.013 0.894 method The algorithm 0.017 0.019 0.022 0.969 of the invention eee TT]

0.111 5052 0799

0.168 0.026 0852

| method! |

3. Quantitative comparison In order to quantitatively compare the segmentation precision of the three methods (i.e., Gu’s method [1], Zheng’s method [2], and the method of the invention), we perform experiments on a dataset consisting of 100 fast-stained leukocyte images, a dataset consisting of 30 standard-stained ° leukocyte images without adhesion and a dataset consisting of 20 standard-stained leukocyte images with adhesion, and four measures, ME, FPR, FNR and KI, are used to quantitatively evaluate the segmentation results. Tables 3 and 4 show the quantitative evaluation results of the nucleus and leukocyte segmentation results on there data sets, respectively; the best measurements in each column of data are shown in bold. For the nucleus segmentation, it can be seen from Table 3 that the method 1 of the invention has the best segmentation effect on images in the first and third sets because the corresponding KI is maximum and the corresponding ME is minimum; the segmentation effect in the second experiment is medium. For the whole leukocyte segmentation, it can be seen from Table 4 that the segmentation result obtained by the method of the invention corresponds to the minimum ME value and the maximum KI value, which indicates that the segmentation result of the present method are the best. The above are the preferred embodiments of the invention. Any changes made according to the technical solution of the invention that do not exceed the scope of the technical solution of the invention belong to the protection scope of the invention.

Claims (3)

Conclusions A blood leukocyte segmentation method based on color component combination and contour mounting, comprising: first, improving the contrast of the leukocyte nuclei using color component combination, and further segmenting the nuclei using a classical threshing segmentation method; second, removing the background from an image using color prior knowledge, and performing edge detection and contour fitting to obtain a leukocyte segmentation result; and finally, subtracting the core segmentation result from the leukocyte segmentation result to obtain a cytoplasmic segmentation result. The blood leukocyte segmentation method based on color component combination and contour mounting according to Claim 1, wherein the step of enhancing the contrast of the leukocyte nuclei using color component combination, and further segmenting the nuclei using a classical threshing segmentation method , specifically is performed as follows: step S11, enhancing a core area by using color component combination to, formally defined as follows: L = 1 + 1, -kxI, (1) Where / ,, I, and I, represent red, blue, and green components in the RGB color space, respectively; step S12, after the enhancement of the core, extracting the core region using the classical image threshing segmentation method Otsu, as follows: step S121, determination of a segmentation threshold: For a particular enhanced image 7, with N pixels and L gray levels {0, 1,…, L-1}, determination of the segmentation threshold t * using the Otsu algorithm, formally described as follows: t = Arg Max je {OO (LM, - 14,) 7} (2) Where w, the ratio of the target pixels (pixels with gray levels {t + 1, t + 2,… L-1}) in the segmentation result corresponding to the segmentation threshold £ to the total pixels of the image, and w; the ratio of the background pixels (pixels with gray levels {0,1, .... t}) in the segmentation result corresponding to the segmentation threshold 7 to the total pixels of the image; # 4 and 44 represent the mean gray values of the target and background pixels; step S122, image thresher segmentation: with t * as the segmentation threshold, the next threshing segmentation on image I; is carried out. B (i, j) Jb as LG) 3) lo, otherwise step S13, correction of the core segmentation result based on the morphological operation: first selecting a largest connected white area, i.e. a target area, and then executing it of image fill to correct the initial core segmentation result to obtain the final core segmentation result. The blood leukocyte segmentation method based on color component combination and contour mounting according to claim 1, wherein the step of removing the background of an image using color prior knowledge and performing edge detection and contour mounting to obtain a result of the leukocyte segmentation is specifically performed as following: step S21, removing the background of the image based on the prior knowledge of the color: removing the background of the image by segmentation of the thresher: H (j), ds (ij, 2) (EJ IT [255,255,255], else (4) where t represents a threshold parameter and + a background removed, ie the image after it has turned white, step S22: performing edge detection on the background removed image to determine the initial contour of a leukocyte extract; and to fit the contour of a leukocyte at a suture, first find two break points at an opening of the leukocyte, and then place an arc segment between the two break points to achieve the separation of the leukocyte and the adhesion, the break point selection process being as follows: step S221, using a Harris angle detector to locate vertices in the contour of detect the leukocyte; step S222, with each vertex as the origin of the coordinates, determining a local Cartesian coordinate system consisting of coordinate points (0, d), (0, -d), (-d, 0), and (d, 0), where d = | 0.1xH |, H represents the total number of rows of the image array, and the symbol IC] represents rounding down; step S223, making four rays from the image center to the four Cartesian coordinate points determined in step S222, and calculating the number of contour points where each ray intersects the contour of the leukocyte; if the number of contour points through which one or more of the four rays pass is 0, determining the corresponding vertex as a break point, or else, determining the corresponding vertex as a non-break point, in the same way, as the initial contour of the leukocyte has more than two breakpoints, perform the following steps; or alternatively, terminating the process of identifying the initial leukocyte contour as the final leukocyte contour; step S224: calculating distances between the core center and the two break points respectively and indicating the distances as a and b; calculating a distance between the two break points and denoting the distance as c; using the two break points and the core center to construct a triangle, and calculate an included angle between lines between two break points and the core center: 8 = arccos [er] (5) 2xaxh step S225: finding points on the axis of the line segment between the two break points with an angle of inclination of 8 with the segment between the two break points, and between these points a point closest to the center of the image is as p2; and step S226, calculating a distance between the point p2 and an arbitrary break point and indicating the distance as r; constructing a circle with p2 as center and r as radius; and separating the adherent leukocytes using the arc between the two breakpoints as the matching contour of the leukocytes between the two breakpoints.