Image-based bone marrow composition segmentation method and system

Abstract

The application relates to an image-based marrow component segmentation method and system, comprising the following steps: acquiring an original marrow biopsy digital image and preprocessing, obtaining a hematopoietic cell tissue and bone trabecula Mask mask image based on the preprocessed image; based on the original marrow biopsy digital image, sequentially performing color channel conversion, morphological open operation, isolated point elimination and hole filling to obtain a bone trabecula Mask mask image; based on the preprocessed image, obtaining a fat cell tissue Mask mask image according to the morphological characteristics of fat cells; and according to the obtained bone trabecula Mask mask image, the bone trabecula and hematopoietic cell tissue Mask mask image and the fat cell tissue Mask mask image, obtaining the area proportion of the bone trabecula, the hematopoietic cell tissue and the fat cell tissue in the original marrow biopsy digital image in each image.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, specifically to a method and system for image-based bone marrow component segmentation. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Bone marrow biopsy pathological images contain information on the distribution and proportion of bone marrow components, which can reflect the patient's physical condition to some extent. For example:

[0004] The proportion of bone trabeculae can reflect metabolic abnormalities such as osteoporosis and osteosclerosis; the proportion of hematopoietic tissue reflects abnormalities such as excessive proliferation or inhibition of hematopoietic tissue. Among them, the proportions of different types of hematopoietic cells and the proportions of hematopoietic cells at different stages have corresponding abnormalities and disease states; when the abnormal proportion of nucleated hematopoietic tissue in bone marrow biopsy images exceeds a certain range, it will reflect related abnormalities.

[0005] In existing technologies, the methods for identifying the composition and proportion of bone trabeculae, hematopoietic tissue, and fat using digital images from bone marrow biopsies still rely mainly on manual identification. This requires installing a mesh micrometer (e.g., 10×10) on the eyepiece of a microscope and randomly selecting a certain number of fields of view (e.g., 16 fields of view). The target hit by 100 points in each field of view is observed, and the various types of cells in the field of view are identified manually. The area ratio of hematopoietic tissue, bone trabeculae, and fat cells in the field of view is calculated.

[0006] The method of manually observing and assessing the area ratio of bone marrow cells, bone trabeculae, and adipose tissue under a microscope is greatly affected by subjective experience, is time-consuming and laborious, and is prone to errors. In order to save time, existing technologies usually randomly select a portion of the field of view for calculation, rather than the entire field of view, which leads to certain errors in the final results. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides an image-based bone marrow component segmentation method and system. Taking a digital image of a bone marrow biopsy as input, the method automatically segments the trabeculae, hematopoietic tissue, and adipose tissue in the image, generates three corresponding mask codes, and calculates their area and proportion respectively, thereby identifying the component proportions of this part of the bone marrow.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A first aspect of the present invention provides an image-based method for bone marrow component segmentation, comprising the following steps:

[0010] The original digital images of bone marrow biopsy were acquired and preprocessed. Based on the preprocessed images, mask images of hematopoietic tissue and bone trabeculae were obtained.

[0011] Based on the original bone marrow biopsy digital image, the mask image of the trabecular bone is obtained by sequentially performing color channel conversion, morphological opening operation, isolated point elimination and hole filling.

[0012] Based on the preprocessed image, a mask image of the adipocyte tissue is obtained by segmentation according to the morphological features of the adipocytes.

[0013] Based on the obtained trabecular bone mask image, trabecular bone and hematopoietic cell tissue mask image, and adipocyte tissue mask image, the area ratio of trabecular bone, hematopoietic cell tissue, and adipocyte tissue in the original bone marrow biopsy digital image in each image is obtained.

[0014] Preprocessing includes grayscale conversion, thresholding, and outlier removal, resulting in a binary image.

[0015] The process of obtaining masked images of hematopoietic tissue and bone trabeculae based on the preprocessed images is as follows:

[0016] In the preprocessed binary image, the adipocyte tissue and background are segmented in the 0-value pixel portion, while the hematopoietic tissue and trabeculae are segmented in the 255-value pixel portion. The 255-value pixel portion is the mask image of the hematopoietic tissue and trabeculae.

[0017] The process of obtaining a mask image of bone trabeculae based on the original bone marrow biopsy digital image, through color channel conversion, morphological opening operation, isolated point elimination, and hole filling, is as follows:

[0018] Color channel conversion includes converting the RGB color channels of the original bone marrow biopsy digital image to HSV color channels, and filtering the color space of bone trabeculae by setting HSV color space values.

[0019] Morphological opening operations include erosion and dilation, which reduce highlighted areas through erosion and increase highlighted areas through dilation.

[0020] Hole filling includes: performing boundary correction on the input image, color filling on connected regions in the image, expanding the coloring from one pixel to the surrounding pixels until all nodes in the closed region have been processed, performing NOT operation on the image mask and intersecting it with the input image, and then thresholding to obtain the complete hole-filled image.

[0021] The process of obtaining a mask image of adipocyte tissue based on the preprocessed image and its morphological features is as follows:

[0022] Non-operational processing was performed on the preprocessed binary image, and bone trabeculae and hematopoietic cell tissue were marked as background regions of 0 pixels, while adipocytes were marked as foreground regions of 255 pixels.

[0023] Segment the fat cell image and obtain the labels of all fat cells in the image.

[0024] Based on the coordinates of the pixels where the fat cell markers are located in the image, the contour attributes of each fat cell marker are calculated. The required contour attributes are then filtered according to the set values ​​to obtain the fat cell tissue mask image.

[0025] The fat cell image is segmented by performing morphological opening and erosion operations sequentially on the foreground image of the fat cells to reduce the fat cell region in the image and obtain the background marker image.

[0026] After performing distance transformation on the image, thresholding is applied to reduce the fat cell region and obtain the foreground labeled image.

[0027] The background and foreground images are differentially processed to obtain unknown region labels. The foreground and unknown regions are labeled with connected components as seed point pixels. The seed points are then irrigated with connected components based on the watershed method to obtain the segmented image of fat cells.

[0028] A second aspect of the present invention provides a system for implementing the above-described method, comprising:

[0029] The preprocessing module is configured to acquire and preprocess raw bone marrow biopsy digital images;

[0030] The first masking module is configured to: obtain a masked image of hematopoietic tissue and bone trabeculae based on the preprocessed image;

[0031] The second masking module is configured to: based on the original bone marrow biopsy digital image, sequentially perform color channel conversion, morphological opening operation, isolated point elimination and hole filling to obtain the mask image of the trabeculae;

[0032] The third masking module is configured to: obtain a mask image of adipocyte tissue by segmenting the preprocessed image according to the morphological features of adipocytes;

[0033] The output module is configured to: based on the obtained trabecular bone mask image, trabecular bone and hematopoietic cell tissue mask image, and adipocyte tissue mask image, obtain the area ratio of trabecular bone, hematopoietic cell tissue, and adipocyte tissue in the original bone marrow biopsy digital image in each image.

[0034] A third aspect of the present invention provides a computer-readable storage medium.

[0035] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the image-based bone marrow component segmentation method as described above.

[0036] A fourth aspect of the present invention provides a computer device.

[0037] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the image-based bone marrow component segmentation method described above.

[0038] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects:

[0039] 1. Using digital images of bone marrow biopsies as input, this method segments trabecular bone, hematopoietic tissue, and adipose tissue in the images, generating three corresponding mask codes and calculating their area and proportion. Compared to current manual methods, this method can automatically and quickly segment bone marrow components in digital bone marrow biopsy images, thereby identifying the proportion of bone marrow components. By calculating the entire field of view in the image, the statistics are more comprehensive and the calculation accuracy is more precise, greatly improving the work efficiency of clinical pathologists.

[0040] 2. In the original digital image of bone marrow biopsy, adipocyte tissue appears white with little color difference from the background area, while trabeculae and hematopoietic tissue appear red with a significant color difference from the background area. Therefore, in the preprocessed binary image, adipocyte tissue and background are segmented into the black part of 0-value pixels, while hematopoietic tissue and trabeculae are segmented into the white part of 255-value pixels. The resulting binary image is the mask for hematopoietic tissue and trabeculae. The trabeculae image is obtained by performing color channel transformation on the same original image and filtering the color space of trabeculae in the HSV channel, thereby segmenting hematopoietic tissue and trabeculae in the image. Attached Figure Description

[0041] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0042] Figure 1 This is a schematic diagram of the bone marrow component identification process provided in one or more embodiments of the present invention;

[0043] Figure 2 This is a schematic diagram of a process for identifying bone marrow components using digital images from a bone marrow biopsy as input, provided by one or more embodiments of the present invention.

[0044] Figures 3(a)-3(d) These are all example images of the segmentation results of trabecular bone and hematopoietic cell tissue masks, trabecular bone masks, and adipocyte tissue masks obtained from the original images during the bone marrow component identification process provided by one or more embodiments of the present invention. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0047] As described in the background section, the method of manually observing and assessing the area proportion of bone marrow cells, trabeculae, and adipose tissue under a microscope is greatly affected by subjective experience, time-consuming, labor-intensive, and prone to errors. The following embodiments present an image-based bone marrow component segmentation method and system. Taking a digital image of a bone marrow biopsy as input, it automatically segments trabeculae, hematopoietic tissue, and adipose tissue in the image, generates three corresponding mask codes, and calculates their area and proportion for each, thereby identifying the area proportion of these bone marrow components in the original image. This helps clinicians obtain the pathological state of a patient's bone marrow more efficiently.

[0048] Example 1:

[0049] like Figure 1 - As shown in Figure 3, the image-based bone marrow component segmentation method includes the following steps:

[0050] Step S1: Perform preprocessing operations such as grayscale conversion, thresholding, and isolated point removal on the digital images of bone marrow biopsy;

[0051] Step S2: Obtain the mask for hematopoietic cell tissue and bone trabeculae;

[0052] Step S3: The original bone marrow biopsy digital image is converted from RGB channels to HSV color channels, and the mask of the trabeculae is obtained through operations such as morphological opening, isolated point elimination and hole filling.

[0053] Step S4: Obtain the mask of adipocyte tissue using the watershed segmentation method and adipocyte morphological characteristics;

[0054] Step S5: After obtaining the masks of bone trabeculae, hematopoietic tissue and adipocyte tissue in the image, calculate the proportion of their area.

[0055] Step S1, as follows Figure 2 As shown, image preprocessing specifically includes:

[0056] Digital images of bone marrow biopsies are converted to grayscale by using different weighted coefficients to sum the RGB three-channel color variables, thus reducing the information in the original image and focusing only on the main features. The grayscale conversion formula is as follows:

[0057] Y = 0.2126R + 0.7152G + 0.0722B

[0058] Binarizing a grayscale image converts it from a grayscale image composed of 0-255 pixel values ​​into a black and white image with only two values, 0 and 255. This helps to highlight the outline of the target of interest and makes it easier to segment and extract the target of interest in the subsequent process.

[0059] Isolated point elimination processing is performed on binary images. Due to the presence of a lot of noise during the imaging process or in the slices themselves, the details smaller than the structural elements in the image are eliminated while the target tissue is preserved.

[0060] The isolated point elimination method is as follows: using the black part of the 0-value pixels in the binarized image as the background area and the white part of the 255-value pixels as the foreground area, find all target contours and calculate the contour area. Fill all contours with black 0-value pixels with all contours less than the threshold. In this embodiment, the area threshold is set to 8000 pixels.

[0061] Step S2, as follows Figure 2 As shown, obtaining bone trabeculae and hematopoietic cell tissue masks specifically includes:

[0062] In the original input image, adipose tissue appears white with little color difference from the background, while trabecular bone and hematopoietic tissue appear red with a significant color difference from the background. Therefore, in the preprocessed binary image, adipose tissue and the background are segmented into the black portion (0-value pixels); while hematopoietic tissue and trabecular bone are segmented into the white portion (255-value pixels). This binary image then serves as the mask for the hematopoietic tissue and trabecular bone.

[0063] Step S3, as follows Figure 2 As shown, obtaining the trabecular bone mask specifically includes:

[0064] The input raw image's RGB color channels are converted to HSV color channels. The RGB channels represent the intensity values ​​of the three color categories: R (red), G (green), and B (blue), while the HSV channels represent the intensity values ​​of the three channels: H (hue), S (saturation), and V (brightness), making them more suitable for image processing. The color space for trabeculae is filtered within the HSV channels. In this embodiment, the threshold values ​​for trabeculae in the HSV color space are defined as [0,43,46] to [10,255,255] and [146,43,46] to [180,255,255].

[0065] Morphological opening operations are performed on the HSV channel image, first erosion and then dilation. Erosion removes the highlighted areas of the image, shrinking their neighborhood and replacing adjacent regions with minima, thus reducing the highlighted area. Dilation expands the highlighted areas of the image, expanding their neighborhood and replacing adjacent regions with maxima, thus increasing the highlighted area.

[0066] By performing isolated point elimination processing on the image after the opening operation, isolated small points, burrs, and small bridges around the contour can be effectively removed.

[0067] Perform hole filling to fill the holes inside the outline and enhance the overall integrity of the outline.

[0068] The hole-filling method is as follows: First, perform boundary correction on the input image to be filled, setting the outermost pixel values ​​to 0 to represent the background and prevent foreground objects from being assigned to the background. Then, widen the image by 2 pixels in both length and width to obtain an image mask. Use the flood fill method to fill the connected regions in the image with color, starting from one pixel and expanding the color outwards until all nodes within the closed region have been processed. Perform NOT operation on the image mask and intersect it with the input image. Finally, threshold the mask to obtain a more complete hole-filled image.

[0069] The trabecular bone mask is obtained after filling the holes.

[0070] Step S4, as follows Figure 2 As shown, obtaining an adipocyte tissue mask specifically includes:

[0071] The binarized image from step S1 is processed using non-operation methods. Bone trabeculae and hematopoietic tissue are marked as black background areas of 0 pixels, and fat cells are marked as white foreground areas of 255 pixels.

[0072] The watershed method was used to segment the foreground image of fat cells to obtain the labels of all fat cells in the image.

[0073] The watershed method is as follows: Opening is performed on the foreground image of fat cells to eliminate small points and spikes; erosion is performed to reduce the fat cell region in the image to obtain the background marker image; a distance transform operation is performed on the image followed by thresholding to further reduce the fat cell region and obtain the foreground marker image; a difference operation is performed on the marked background and foreground images to obtain unknown region markers; connected component markers are applied to the foreground and unknown regions as seed pixels for the watershed method; the watershed method is used to apply connected component irrigation to the seed pixels to obtain the segmented image of fat cells.

[0074] After obtaining the fat cell markers in the image, the contour attributes of each fat cell marker are calculated based on the coordinate values ​​of the pixels in the image, including: area, perimeter, granularity, average gray value, aspect ratio, etc.

[0075] For all fat cell contours, targets that meet the above contour attributes are selected. In this embodiment, the threshold ranges for area, perimeter, roundness, average gray value, and aspect ratio are set to [500, 150000], [500, 10000], [0.5, 1], [200, 255], and [0.3, 3], respectively, thereby obtaining the fat cell tissue mask.

[0076] Step S5, as follows Figure 2 As shown, the calculation of the area ratio of bone trabeculae, hematopoietic tissue, and adipocyte tissue specifically includes:

[0077] For the three image masks obtained—the trabecular bone mask, the trabecular bone and hematopoietic cell tissue mask, and the adipocyte tissue mask—the percentage of the target pixel in each image is calculated, as shown below:

[0078]

[0079] Perecentage He =Perecentage Tra_He -Perecentage Tra

[0080] Where Tra represents trabeculae, Tra_He represents trabeculae and hematopoietic cell tissue, and Fat represents adipose cell tissue.

[0081] The obtained target pixel ratio reflects the proportion of bone trabeculae, hematopoietic cells, and adipose cells in the original image of the bone marrow biopsy digital image. Compared with the current method of dividing regions under a microscope and counting manually, it can automatically, quickly and accurately segment bone trabeculae, hematopoietic cells, and adipose cells in the bone marrow biopsy digital image, thereby identifying the proportion of these bone marrow components.

[0082] like Figures 3(a)-3(d) As shown, the process of bone marrow component identification includes the original digital image of bone marrow biopsy, masked images of trabecular bone and hematopoietic cell tissue, masked images of trabecular bone, masked images of adipocyte tissue, and the final segmentation result image. According to tests, the method in this embodiment reduces the time consumed from more than ten minutes to less than 3 seconds compared to the current manual method. Furthermore, it replaces the random selection of fields of view in the manual method with the entire field of view in the calculated image, resulting in more comprehensive statistics and more accurate calculations, which greatly improves the work efficiency of clinical pathologists.

[0083] Example 2:

[0084] A system for implementing the above method includes:

[0085] The preprocessing module is configured to acquire and preprocess raw bone marrow biopsy digital images;

[0086] The first masking module is configured to: obtain a masked image of hematopoietic tissue and bone trabeculae based on the preprocessed image;

[0087] The second masking module is configured to: based on the original bone marrow biopsy digital image, sequentially perform color channel conversion, morphological opening operation, isolated point elimination and hole filling to obtain the mask image of the trabeculae;

[0088] The third masking module is configured to: obtain a mask image of adipocyte tissue by segmenting the preprocessed image according to the morphological features of adipocytes;

[0089] The output module is configured to: based on the obtained trabecular bone mask image, trabecular bone and hematopoietic cell tissue mask image, and adipocyte tissue mask image, obtain the area ratio of trabecular bone, hematopoietic cell tissue, and adipocyte tissue in the original bone marrow biopsy digital image in each image.

[0090] The obtained target pixel ratio reflects the proportion of bone trabeculae, hematopoietic tissue, and adipose tissue in the original image of bone marrow biopsy digital image. Compared with the current manual method, it can automatically, quickly and accurately segment bone marrow components in bone marrow biopsy digital image, thereby identifying the proportion of bone marrow components. By calculating the entire field of view in the image, the statistics are more comprehensive and the calculation accuracy is more accurate, which greatly improves the working efficiency of clinical pathologists.

[0091] Example 3:

[0092] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the image-based bone marrow component segmentation method as described in Embodiment 1 above.

[0093] Example 4:

[0094] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the image-based bone marrow component segmentation method as described in Embodiment 1 above.

[0095] The steps or networks involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0096] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An image-based bone marrow composition segmentation method, characterized by, The method comprises the following steps: Obtaining an original bone marrow biopsy digital image and preprocessing, and obtaining a hematopoietic cell tissue and bone trabecula Mask image based on the preprocessed image; Based on the original bone marrow biopsy digital image, the Mask image of the bone trabecula is obtained by sequentially performing color channel conversion, morphological opening operation, isolated point elimination and hole filling; Based on the preprocessed image, the Mask image of the adipocyte tissue is segmented according to the morphological characteristics of the adipocyte, and the process comprises the following steps: Based on the preprocessed image, the Mask image of the adipocyte tissue is segmented according to the morphological characteristics of the adipocyte, and the process comprises the following steps: The preprocessed binary image is subjected to non-operation processing, the bone trabecula and the hematopoietic cell tissue are marked as 0-pixel background areas, and the adipocyte is marked as 255-pixel foreground area; Segmenting the adipocyte image to obtain the labels of all adipocytes in the image; According to the coordinate values of the pixel points where the adipocyte labels are located, the contour properties of each adipocyte label are calculated, the required contour properties are selected according to the set value, and the adipocyte tissue Mask image is obtained; The process of segmenting the adipocyte image comprises the following steps: the foreground image of the adipocyte is sequentially subjected to morphological opening operation and corrosion processing, so that the adipocyte area in the image is reduced to obtain a background label image; After distance transformation operation and thresholding processing are performed on the image, the adipocyte area is reduced to obtain a foreground label image; The labeled background image and the foreground image are subjected to difference processing to obtain an unknown area label, the foreground and the unknown area are subjected to connected domain labeling as seed point pixels, and the seed points are subjected to connected domain irrigation based on the watershed method to obtain a segmented adipocyte image; According to the obtained bone trabecula Mask image, the bone trabecula and hematopoietic cell tissue Mask image and the adipocyte tissue Mask image, the area proportions of the bone trabecula, the hematopoietic cell tissue and the adipocyte tissue in the original bone marrow biopsy digital image are obtained.

2. The image-based bone marrow composition segmentation method of claim 1, wherein, The preprocessing comprises grayscale processing, thresholding and isolated point elimination, and a binary image is obtained after preprocessing.

3. The image-based bone marrow composition segmentation method of claim 2, wherein, In the binary image obtained after preprocessing, the adipocyte tissue and the background are segmented in the 0-value pixel part, the hematopoietic cell tissue and the bone trabecula are segmented in the 255-value pixel part, and the 255-value pixel part is the Mask image of the hematopoietic cell tissue and the bone trabecula.

4. The image-based bone marrow composition segmentation method of claim 1, wherein, The process of the morphological opening operation comprises corrosion and dilation, the high-light area is reduced by corrosion, and the high-light area is increased by dilation.

5. The image-based bone marrow composition segmentation method of claim 1, wherein, The process of the hole filling comprises the following steps: boundary correction is performed on the input image, the connected areas in the image are color-filled, coloring is expanded from a pixel point to the surrounding pixel points until all the nodes in the closed area are processed, non-operation processing is performed on the image mask, the input image is intersected, and thresholding processing is performed to obtain a complete hole filling image.

6. An image-based bone marrow composition segmentation system employing the image-based bone marrow composition segmentation method according to any one of claims 1 to 5, characterized in that, It comprises: A preprocessing module configured to obtain an original bone marrow biopsy digital image and preprocess the image; A first mask module configured to obtain a hematopoietic cell tissue and bone trabecula Mask image based on the preprocessed image; The second mask module is configured to obtain a bone trabecula Mask image based on the original bone marrow biopsy digital image through color channel conversion, morphological opening operation, isolated point elimination and hole filling in sequence; The third mask module is configured to obtain a fat cell tissue Mask image based on the preprocessed image according to the morphological characteristics of the fat cells. The result output module is configured to obtain the area proportion of the bone trabecula, the hematopoietic cell tissue and the fat cell tissue in the original bone marrow biopsy digital image according to the obtained bone trabecula Mask image, the bone trabecula and hematopoietic cell tissue Mask image and the fat cell tissue Mask image. 7.A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the image-based bone marrow component segmentation method according to any one of claims 1-5. 8.A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the image-based bone marrow component segmentation method according to any one of claims 1-5 when executing the program.

Images