Image processing device and image processing system

Abstract

This invention provides an image processing method and an image processing program that generate images capable of more accurately understanding the shape of the vascular network structure. [Solution] The image processing method includes the steps of: acquiring an image of a cell structure having a colorized vascular network structure; applying a wavelet transform to the image to generate a contour image in which the contour of the vascular network structure is extracted; and repeatedly removing pixels from the boundaries of the object drawn on the contour image to generate a skeleton image in which a skeleton having a predetermined number of pixels in line width is extracted.

Description

Technical Field

[0005]

[0001] The present disclosure relates to an image processing apparatus and an image processing system.

Background Art

[0002] Patent Document 1 discloses an artificial cell structure having a vascular network structure. The cells constituting the vascular network structure of this cell structure are immunostained using an anti-CD31 antibody. As a result, the vascular network structure can be observed using a microscope.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Although the vascular network structure of the cell structure described in Patent Document 1 can be observed using a microscope, it is dense, so it is difficult to quantitatively analyze the vascular network structure. As an example of a method for quantitatively analyzing the vascular network structure, a method of counting the number of cells in the produced cell structure by performing trypsin treatment or the like on the cell structure can be considered. However, it is difficult to count only the cells constituting the vascular network structure by this method. Alternatively, in an image of the cell structure taken after immunostaining, it is also conceivable to discriminate and evaluate the vascular network structure based on the intensity of fluorescence caused by the cells constituting the vascular network structure. However, since the stained cells are not necessarily limited to the cells constituting the vascular network structure, it is insufficient for evaluating the shape such as the length of the vascular network structure. The present disclosure provides a method and a program for generating an image that can more accurately grasp the shape of the vascular network structure.

Means for Solving the Problems

[0005] An image processing method according to one aspect of the present invention includes the steps of: acquiring an image of a cell structure having a colorized vascular network structure; applying a wavelet transform to the image to generate a contour image in which the contour of the vascular network structure is extracted; and repeatedly excluding the pixels of an object from the boundary of the object recognized in the contour image to generate a skeleton image in which a skeleton having a predetermined number of pixels in line width is extracted.

[0006] This image processing method processes images of cellular structures that have a vascular network structure. First, a wavelet transform is applied to the image. This generates a contour image in which the contours of the vascular network structure are extracted. Next, the pixels of objects are repeatedly removed from the boundaries of the objects recognized in the contour image. This generates a skeletal image in which the skeleton of the vascular network structure is extracted. Because the skeletal image is generated from the contour image, a skeleton that more accurately reflects the shape of the blood vessels is obtained compared to when the skeletal image is generated directly from the image. Therefore, this image processing method can generate images that allow for a more accurate understanding of the shape of the vascular network structure.

[0007] In one embodiment, the image processing method may further include the step of detecting the shape of the vascular network structure based on a skeletal image. In this case, the image processing method can use the detected shape of the vascular network structure to evaluate the features of the vascular network structure.

[0008] In one embodiment, the shape of the vascular network structure may be defined as the length of the vascular network structure. In this case, the image processing method can evaluate the features of the vascular network structure using the detected length of the vascular network structure.

[0009] In one embodiment, the captured image may be an image of a region including the center of a cell structure. In this case, the image processing method can appropriately extract regions that are likely to contain the vascular network structure of the cell structure.

[0010] In one embodiment, the image processing method may further include a step of generating an image from which the background signal has been removed, after the acquisition step and before the contour image generation step. In this case, since the background signal of the image is removed, the image processing method can more clearly extract the skeleton showing the vascular network structure.

[0011] In one embodiment, the image processing method may be performed after the step of generating a skeletal image and further include the step of generating an extracted image in which skeletons drawn on the skeletal image whose length is greater than a threshold are extracted. In this case, since vascular structures whose skeleton length is less than or equal to the threshold are not extracted, the image processing method can remove skeletons that are unlikely to constitute a vascular network structure as noise.

[0012] In one embodiment, the image processing method may further include a step of converting the image into an image with predetermined tonal levels, which is performed after the acquisition step and before the contour image generation step. In this case, since the captured image is converted into an image with predetermined tonal levels, the image processing method can easily proceed with processing the converted image.

[0013] In one embodiment, the image processing method is performed before the acquisition step and further includes the steps of: immunostaining cells having a vascular network structure, pre-fluorescently modifying cells having a vascular network structure, or a combination of immunostaining and fluorescent modification to cause the vascular network structure to develop color; and imaging the cell structure having a vascular network structure that has developed color in the color development step, and the acquisition step may be the acquisition of the image captured in the imaging step. In this case, the image processing method can more accurately grasp the shape of the vascular network structure for each cell structure by performing the color development step and the imaging step on the cell structure having a vascular network structure.

[0014] In one embodiment, the cell structure may have vascular endothelial cells.

[0015] An image processing program relating to another aspect of this disclosure causes a computer to perform the following steps: acquire an image of a cell structure having a colorized vascular network structure; apply a wavelet transform to the image to generate a contour image in which the contour of the vascular network structure is extracted; and repeatedly remove the pixels of an object from the boundaries of the object drawn on the contour image to generate a skeleton image in which a skeleton having a predetermined number of pixels in line width is extracted.

[0016] This image processing program produces the same effect as the image processing method described above. [Effects of the Invention]

[0017] According to one aspect and embodiment of this disclosure, the shape of the vascular network structure can be more accurately determined. [Brief explanation of the drawing]

[0018] [Figure 1] Figure 1 is a block diagram showing an example of the functions of an image processing system equipped with an image processing device according to an embodiment. [Figure 2] Figure 2 is a block diagram showing an example of the hardware configuration of an image processing device. [Figure 3] Figure 3 is a flowchart showing an example of the image acquisition process performed by the image processing method according to the embodiment. [Figure 4] Figure 4 is a flowchart showing an example of an image processing method according to the embodiment. [Figure 5] Figure 5 shows an example of an image obtained by the contour image generation process and skeletal image generation process according to the embodiment. [Figure 6] Figure 6 shows an example of an image obtained through grayscale conversion processing, noise reduction processing, contour image generation processing, and skeletal image generation processing. [Figure 7] Figure 7 shows an example of an image obtained through pixel value adjustment processing, contour image generation processing, and skeletal image generation processing. [Figure 8]FIG. 8 is a diagram showing an example of an image obtained by the extraction image generation process. [Figure 9] FIGS. 9(a) to (f) are diagrams showing an example of an imaging image of a cell construct administered with cyclophosphamide, and FIGS. 9(g) to (l) are diagrams showing an example of an extraction image for the imaging image of a cell construct administered with cyclophosphamide. [Figure 10] FIG. 10(a) is a diagram showing the correspondence between the length of the skeleton and the dosing concentration of cyclophosphamide in the extraction image according to the embodiment, and FIG. 10(b) is a diagram showing the correspondence between the number of branch points of the skeleton and the dosing concentration of cyclophosphamide in the extraction image.

Mode for Carrying Out the Invention

[0019] Hereinafter, embodiments will be described with reference to the drawings. In the following description, the same or corresponding elements are denoted by the same reference numerals, and redundant descriptions are omitted.

[0020] [Overview of Cell Construct] First, an overview of the cell construct that is the imaging target of the image processing apparatus according to the embodiment will be described. The cell construct includes, for example, vascular endothelial cells, and as a specific example, it is a three-dimensional liver-like tissue (3D liver model). The 3D liver model has stromal cells including vascular endothelial cells and is a tissue capable of evaluating hepatotoxicity. The 3D liver model includes, for example, hepatocytes and further includes at least one type of cell selected from the group consisting of fibroblasts (e.g., hepatic stellate cells), immune cells, and smooth muscle cells, and vascular endothelial cells as stromal cells. The cell construct may further include an extracellular matrix component that functions as a scaffold material for the cells, and may also include components (e.g., polyelectrolytes) that support the formation of other cell constructs.

[0021] In this specification, "extracellular matrix component" refers to an aggregate of extracellular matrix molecules formed by multiple extracellular matrix molecules. The extracellular matrix refers to substances present outside the cell in living organisms. Any substance can be used as the extracellular matrix, as long as it does not adversely affect cell growth and the formation of cell aggregates. Specific examples include, but are not limited to, collagen, elastin, proteoglycan, fibronectin, hyaluronic acid, laminin, vitronectin, tenascin, enteractin, fibrillin, and cadherin. The extracellular matrix component may be used individually or in combination. For example, the extracellular matrix component may contain collagen, or may be collagen. When the extracellular matrix component is collagen, the collagen component functions as a scaffold for cell adhesion, further promoting the formation of three-dimensional cell structures. In this embodiment, the extracellular matrix component is preferably a substance present outside animal cells, i.e., an animal extracellular matrix component. Furthermore, the extracellular matrix molecules may be modified or variant forms of the extracellular matrix molecules described above, or polypeptides such as chemically synthesized peptides, as long as they do not adversely affect cell growth or the formation of cell aggregates.

[0022] The extracellular matrix component may have repeating sequences represented by Gly-XY, which are characteristic of collagen. Here, Gly represents a glycine residue, and X and Y each independently represent any amino acid residue. Multiple Gly-XY sequences may be identical or different. Having repeating sequences represented by Gly-XY reduces constraints on the arrangement of molecular chains, resulting in even better function as a scaffolding material. In an extracellular matrix component having repeating sequences represented by Gly-XY, the proportion of sequences represented by Gly-XY may be 80% or more of the total amino acid sequence, preferably 95% or more. The extracellular matrix component may also have RGD sequences. An RGD sequence refers to a sequence represented by Arg-Gly-Asp (arginine residue-glycine residue-aspartic acid residue). When an extracellular matrix component has RGD sequences, cell adhesion is further promoted, making it even more suitable as a scaffolding material. Extracellular matrix components containing sequences represented by Gly-XY and RGD sequences include collagen, fibronectin, vitronectin, laminin, cadherin, and others.

[0023] Examples of extracellular matrix component shapes include fibrous structures. Fibrous refers to a shape composed of thread-like extracellular matrix components, or a shape composed of thread-like extracellular matrix components cross-linked between molecules. At least a portion of the extracellular matrix components may be fibrous. The shape of the extracellular matrix components is the shape of a single mass of extracellular matrix components (an aggregate of extracellular matrix components) observed under a microscope, and the extracellular matrix components preferably have an average diameter and / or average length, as described later. Fibrous extracellular matrix components include thin thread-like structures (fibers) formed by the aggregation of multiple thread-like extracellular matrix molecules, thread-like structures formed by further aggregation of fine fibers, and defibrations of these thread-like structures. When extracellular matrix components with a fibrous shape are included, the RGD sequence is preserved without being destroyed in the fibrous extracellular matrix components, and they can function even more effectively as a scaffold for cell adhesion.

[0024] Polyelectrolytes, which are components that support the formation of cell structures, are polymer compounds that possess the properties of electrolytes. Examples of polyelectrolytes include, but are not limited to, heparin, chondroitin sulfate (e.g., chondroitin 4-sulfate, chondroitin 6-sulfate), heparan sulfate, dermatan sulfate, keratan sulfate, glycosaminoglycans such as hyaluronic acid; dextran sulfate, rhamnan sulfate, fucoidan, carrageenan, polystyrene sulfonic acid, polyacrylamide-2-methylpropanesulfonic acid, and polyacrylic acid, or derivatives thereof. A polyelectrolyte may consist of one of the above-mentioned substances, or it may contain a combination of two or more substances.

[0025] The polyelectrolyte is preferably a glycosaminoglycan, more preferably at least one selected from the group consisting of heparin, dextran sulfate, chondroitin sulfate, and dermatan sulfate, and even more preferably heparin. When the cell structure contains a polyelectrolyte, excessive aggregation of extracellular matrix components can be more effectively suppressed, and as a result, cell structures with superior responsiveness to hepatotoxic substances are more easily obtained. This effect is even more pronounced when the cell structure contains heparin.

[0026] The vascular network structure of cellular structures can be visualized by performing immunostaining, fluorescent modification, or a combination of immunostaining and fluorescent modification on a 3D liver model. Furthermore, by administering compounds such as cyclophosphamide to the 3D liver model in which the vascular network structure has been visualized, the degree of damage to the vascular network structure can be confirmed.

[0027] The 3D liver model (cellular structure) in which the vascular network to be imaged by the image processing device is colored is obtained more specifically by performing a preparation process to create the 3D liver model, a drug administration process to administer a drug to the 3D liver model, and a color control process to colorize the 3D liver model. If the degree of damage to the vascular network structure is not to be observed, that is, if no compound is administered to the 3D liver model, the drug administration process may not be performed.

[0028] Table 1 is a table showing an example of cultured cells to be cultured in the image processing system 2. The preparation process is carried out using methods and materials disclosed, for example, in International Publication No. 2017 / 146124 or International Publication No. 2018 / 143286. The preparation process in this embodiment is an example utilizing the method disclosed in International Publication No. 2017 / 146124. The preparation process includes, for example, a first recovery process to recover cells other than fresh human hepatocytes (PXB cells) derived from human hepatocyte chimeric mice shown in Table 1, a second recovery process to recover PXB cells, a mixture acquisition process to obtain a cell mixture, a suspension acquisition process to obtain a cell suspension, a gel formation process to form a fibrin gel, and a model acquisition process to obtain a 3D model.

[0029] [Table 1]

[0030] Table 2 shows an example of reagents and consumables used in the image processing system. Note that general-purpose research materials such as disposable pipettes are omitted from Table 2. In the first recovery process, for example, the operator prepares purchased human hepatic stellate cells (Lx2) and human hepatic sinusoidal endothelial cells (SEC) and measures the cell volume of each. Alternatively, in the first recovery process, the operator may awaken the frozen stocks of human hepatic stellate cells (Lx2) and human hepatic sinusoidal endothelial cells (SEC), culture them without subculturing according to the manufacturer's recommended protocol, and then recover them from the culture flask and petri dish using trypsin according to the usual method, after which the cell volume of each may be measured. In the second recovery process, for example, the operator prepares purchased PXB cells and measures the cell volume of the PXB cells.

[0031] [Table 2]

[0032] In the mixture acquisition process, for example, the cells recovered in the first and second recovery processes are mixed so that the total cell volume per well is 30,000 cells, and the tissue cell ratio is 65% PXB cells, 25% SEC cells, and 10% Lx2 cells, thereby obtaining a cell mixture (an example of a cell structure). This cell mixture contains multiple types of cells, including vascular endothelial cells and other stromal cells.

[0033] In the suspension acquisition process, for example, first, a heparin-collagen solution is prepared by mixing equal volumes of 1.0 mg / ml heparin solution (buffer: 100 mM Tris-HCl) and 0.3 mg / ml collagen solution (buffer: 5 mM acetate). Next, 100 μL of the prepared heparin-collagen solution is added to each cell mixture obtained in the mixture acquisition process, and the cells are suspended until they are no longer visible. After centrifugation (400 g × 2 min) is performed to obtain a viscous substance. After the supernatant is removed from the viscous substance solution, solvent is added so that the final volume of the viscous substance solution is "number of seeding wells" × 2 μL, and a cell suspension is obtained.

[0034] In the gel formation process, for example, the cell suspension obtained in the suspension acquisition process is mixed with a 10 mg / mL fibrinogen solution and a 20 U / mL thrombin solution (solvent: HCM) on a 48-well plate. For example, the solvent added to the viscous solution after the supernatant is removed in the suspension acquisition process may be a 20 U / mL thrombin solution (solvent: HCM). In this case, for example, after a droplet of 10 mg / mL fibrinogen solution is formed, the cell suspension obtained in the suspension acquisition process is added to the inside of the droplet. Alternatively, for example, 2 μL of the cell suspension obtained in the suspension acquisition process may be seeded and a droplet formed, after which the 10 mg / mL fibrinogen solution may be added. Note that the solvent added to the viscous solution in the suspension acquisition process does not have to be a 20 U / mL thrombin solution (solvent: HCM). In this case, the gel formation process may involve adding 20 U / mL thrombin solution (solvent: HCM) to a droplet containing a mixture of cell suspension and 10 mg / mL fibrinogen solution. The order in which the cell suspension, 10 mg / mL fibrinogen solution, and 20 U / mL thrombin solution (solvent: HCM) are mixed is not limited to the above and may be in any order. The droplet containing the mixture of cell suspension, 10 mg / mL fibrinogen solution, and 20 U / mL thrombin solution (solvent: HCM) is left to stand in an incubator for 40 minutes to form a fibrin gel. Note that if it is not necessary to form the model into a specific mass shape, the gel formation process may be omitted.

[0035] In the model acquisition process, for example, 0.5 mL of HCM (containing Endothelial Cell Growth Supplement) is added to each well where fibrin gel has formed during the gel formation process to obtain a 3D model. The preparation process is completed with the above steps. In the mixture acquisition process, suspension acquisition process, gel formation process, and model acquisition process within the preparation process, the values ​​such as total cell volume, tissue cell ratio, reagent concentration, reagent solution volume, centrifugation time, and standing time are not limited to the above values ​​and may be appropriately changed depending on the cell structure to be prepared.

[0036] Next, the drug administration process is carried out. In this process, the 3D liver models created in the preparation process are administered by changing the culture medium on days 1 and 4 with media containing monoclotaline at concentrations of 2000 μM, 666 μM, 222 μM, and 74 μM, respectively. Each compound is pre-dissolved in high concentrations of DMSO (Dimethyl sulfoxide) and stored, so DMSO is present in the culture medium at a concentration of 1% at the time of administration. In addition, since each administration condition contains 1% DMSO, a media change with a medium containing only 1% DMSO is also performed simultaneously as a negative control. This completes the drug administration step.

[0037] Next, a color development control process is performed. As an example, the color development control process includes fixation, permeabilization, primary antibody treatment, and secondary antibody treatment.

[0038] During the fixation process, on day 6, the 48-well plate is removed from the incubator, the culture medium is removed, and the plate is washed with PBS. After washing, 300 μL of 4% paraformaldehyde phosphate buffer (PFA) is added to each well to fix the 3D model. After fixation, the PFA is washed off.

[0039] Next, in the permeabilization stage, 100 μL of 0.2 (v / v)% TRITON / 1 (w / v)% BSA PBS solution (BSA solution) is added to the insert in each well and left to stand at room temperature for 2 hours.

[0040] Next, in the primary antibody treatment, mouse-derived anti-CD31 antibody is diluted 100-fold with BSA solution to obtain the primary antibody solution. 100 μL of this primary antibody solution is added to the insert in each well and allowed to stand at 4°C for 24 hours. After standing, the primary antibody solution is washed away.

[0041] Next, in the secondary antibody treatment, the secondary antibody is diluted 200-fold with BSA solution to obtain the secondary antibody solution. 100 μL of this secondary antibody solution is added to the insert in each well, and it is left to stand at room temperature for 1 hour in the dark. After standing, the secondary antibody solution is washed away, and 100 μL of PBS is added to each well. With these steps, the color control treatment is completed, and a 3D liver model with a colored vascular network structure can be obtained. Note that the color control treatment may be performed using, for example, a known method.

[0042] [Image Processing System] Figure 1 is a block diagram showing an example of the functions of an image processing system equipped with an image processing device according to an embodiment. As shown in Figure 1, the image processing system 2 comprises an image processing device 1, an imaging device 110, a first storage device 120, a display device 130, and a second storage device 140. Each of the imaging device 110, the first storage device 120, the display device 130, and the second storage device 140 is provided to communicate with the image processing device 1.

[0043] The imaging device 110 is a device for imaging a 3D liver model having a colorized vascular network structure. The imaging device 110 is, for example, a confocal microscope. The imaging device 110 uses, for example, a 4x lens, sets the non-confocal mode with a thick depth of field as the imaging mode, sets the filter to 647(Ex) / 512(EM), and images the 3D liver model in 10 μm increments within the range of 0 to 100 μm on the Z axis. The imaging device 110 images the region including the center of the 3D liver model. For each imaged 3D liver model, the imaging device 110 sums the pixel values ​​and color intensity of multiple images with different Z-axis positions on the confocal microscope to obtain an image of the 3D liver model. An example of a pixel value is the brightness value. The image includes the region in which the colorized vascular network structure is depicted.

[0044] The imaging device 110 has the function of outputting captured images to at least one of the image processing device 1 and the first storage device 120. The first storage device 120 is, for example, a storage medium such as a hard disk. The first storage device 120 stores the captured images described above. The first storage device 120 may also store images that have been captured in the past.

[0045] The image processing device 1 acquires images of cell structures and generates images that allow for the understanding of the shape of the vascular network structure of the cell structures. First, the hardware of the image processing device 1 will be described. Figure 2 is a block diagram showing an example of the hardware configuration of the image processing device according to the embodiment. As shown in Figure 2, the image processing device 1 is configured as a normal computer system including a CPU (Central Processing Unit) 100, main memory such as ROM (Read Only Memory) 101 and RAM (Random Access Memory) 102, an input device 103 such as a camera or keyboard, an output device 104 such as a display, and an auxiliary storage device 105 such as a hard disk.

[0046] Each function of the image processing device 1 is realized by loading predetermined computer software onto hardware such as the CPU 100, ROM 101, and RAM 102, thereby operating the input device 103 and output device 104 under the control of the CPU 100, and by reading and writing data to the main memory and auxiliary memory 105. The image processing device 1 may also include a communication module or the like.

[0047] Returning to Figure 1, the image processing device 1 comprises an acquisition unit 10, a background processing unit 20, a contour extraction unit 30, a skeleton extraction unit 40, a shape detection unit 50, and a display control unit 60.

[0048] The acquisition unit 10 acquires, for example, an image of a 3D liver model having a colorized vascular network structure captured from the imaging device 110. The acquisition unit 10 may also refer to the first storage device 120 and acquire the image from the first storage device 120. The acquisition unit 10 may acquire multiple images from the imaging device 110 with different Z-axis positions of the confocal microscope, and, on behalf of the imaging device 110, sum the pixel values ​​and color intensity of the multiple images to acquire an image.

[0049] The background processing unit 20 has a cropping function that extracts the central part of the 3D liver model captured in the captured image, a conversion function that converts it into a grayscale image which is an image with predetermined grayscale levels, and a function that removes background signals and noise from the image.

[0050] The background processing unit 20 generates an image by extracting a region from the captured image that includes the center of the 3D liver model, as part of its cropping function. The region including the center of the 3D liver model does not necessarily have to include, for example, the peripheral area of ​​the 3D liver model and the background portion of the 3D liver model. The background processing unit 20 generates a cropped image, for example, an image in which the region including the center of the 3D liver model is cropped from the captured image into a 1 mm square area. The cropped image includes the vascular network structure of the 3D liver model.

[0051] The background processing unit 20 performs a conversion function to convert the cropped image into a grayscale image of a predetermined level. The predetermined level is a set of levels that are set in advance, such as 8 levels, 16 levels, or 256 levels. As an example, the background processing unit 20 converts the cropped image into an 8-bit grayscale image. The vascular network structure that is colored in the cropped image has a high pixel value and may be converted to white in the grayscale image. Areas other than the vascular network structure have a low pixel value in the cropped image and may be converted to black in the grayscale image.

[0052] The background processing unit 20 generates a removed image, which is an image from which the background signal of the grayscale image has been removed, as a removal function. The background processing unit 20 removes the background signal by, for example, the rolling ball method or the sliding paraboloid method. An example of the radius of the ball or paraboloid is 20 pixels, but it is not limited to this. As a result, the background signal of the grayscale image is removed and the pixel values ​​of the vascular network structure are normalized.

[0053] The background processing unit 20 generates an adjusted image, which is an image with reduced noise from the removed image. The background processing unit 20 reduces noise by adjusting the contrast of the removed image and then subtracting a predetermined pixel value. For example, the background processing unit 20 stretches the removed image to 256 grayscale levels while maintaining the distribution of pixel values ​​of each pixel. As an example, the background processing unit 20 adjusts the pixel value of the pixel with the smallest pixel value in the removed image to 0 (minimum value), adjusts the pixel value of the pixel with the largest pixel value in the removed image to 255 (maximum value), and then generates an image in which the pixel values ​​of each pixel are distributed according to a distribution similar to the distribution of pixel values ​​of each pixel before adjustment. This function makes it possible to widen the difference in pixel values ​​between pixels showing a vascular network structure with high pixel values ​​and pixels showing noise with low pixel values ​​in the removed image. Subsequently, the background processing unit 20 subtracts a predetermined pixel value from each pixel value in the image with the new distribution. In the case of a 256-grayscale image, for example, 100 is subtracted from the pixel value, but the predetermined pixel value is not limited to 100. In this way, by increasing the difference in pixel values ​​between pixels showing the vascular network structure and pixels showing noise, and then subtracting a predetermined pixel value, it is possible to effectively eliminate only the noise portion.

[0054] The background processing unit 20 may adjust the distribution of pixel values ​​to be uniform rather than normalize them, and subtract a predetermined pixel value from each pixel value in the image with the new distribution. Furthermore, the background processing unit 20 may apply filters such as a maximum value filter, a minimum value filter, or a smoothing filter to the image from which the background signal has been removed to reduce noise.

[0055] The contour extraction unit 30 applies a wavelet transform to the adjusted image to generate a contour image in which the contours of the vascular network structure are extracted. The wavelet transform extracts the contours of subjects contained in the image. As a wavelet transform, the contour extraction unit 30 decomposes the adjusted image into wavelet coefficients (high-frequency components) and scaling function coefficients (low-frequency components). The contour extraction unit 30 further repeats the above decomposition on the decomposed image to extract the high-frequency components at each level and combines them into a single image.

[0056] The contour extraction unit 30 generates a contour image by binarizing the combined image, setting the intensity of pixels with pixel values ​​below a threshold to 0 and the intensity of pixels with pixel values ​​equal to or greater than the threshold to 1. This threshold is predetermined based on, for example, the properties of the sample and the observation conditions during measurement. The wavelet transform is performed using, for example, a Mexican hat filter. Since the vascular network structure is represented by pixels with high pixel values ​​in the adjusted image, it is more easily extracted as a high-frequency component compared to pixels of other subjects (elements) with low pixel values. Therefore, in the contour image generated by the contour extraction unit 30, pixels in which the vascular network structure is captured are represented as object pixels with an intensity of 1, and pixels in which the vascular network structure is not captured are represented as pixels with an intensity of 0.

[0057] The skeleton extraction unit 40 repeatedly removes object pixels from the boundaries of objects recognized in the contour image, and generates a skeleton image in which a skeleton having a predetermined number of line widths is extracted. The skeleton extraction unit 40 recognizes object regions, which are areas where multiple object pixels are clustered in the binarized contour image, as regions where a vascular network structure is drawn, and repeatedly removes object pixels from the boundaries of the object regions. Removal refers to, for example, setting the intensity of object pixels to 0.

[0058] The boundary of an object region refers, for example, to an object pixel within an object region that is adjacent to a pixel where the vascular network structure is not captured and the intensity is 0, either above, below, to the left, or to the right. The skeleton extraction unit 40 sets a predetermined number of pixels to 1 and repeatedly removes the boundary of the object region until the line width of the object region is represented by one pixel. For example, when focusing on an object pixel located at the boundary of an object region (the first object pixel), if another object pixel (the second object pixel) is adjacent to either the above or below the first object pixel, and another object pixel (the third object pixel) is adjacent to either the left or to the right of the first object pixel, and there are other object pixels (the fourth object pixel) adjacent to the second and third object pixels respectively, the first object pixel is removed. The skeleton extraction unit 40 generates a skeleton image in which the skeleton, which is a collection of object pixels where the line width of all object regions drawn in the contour image is 1 pixel, is drawn. Note that the predetermined number of pixels is not limited to 1 and can be set to an appropriate value.

[0059] The shape detection unit 50 detects the shape of the vascular network structure based on the skeleton drawn on the skeletal image. The shape of the vascular network structure is, for example, the length of the vascular network structure. The shape detection unit 50 detects the length of the skeleton drawn on the skeletal image as the length of the vascular network structure. The length of the skeleton is, for example, the number of pixels of the object pixels that make up the skeleton. If the skeleton branches into multiple parts, the number of pixels of each branched skeleton may be calculated and the sum of the number of pixels of each branched skeleton may be detected as the length (total length) of the skeleton.

[0060] The shape detection unit 50 generates an extracted image, which is an image in which skeletons drawn on the skeletal image whose length is greater than a threshold are extracted. This threshold is predetermined based on the properties of the sample, the observation conditions at the time of measurement, etc. The shape detection unit 50 obtains the extracted image by removing skeletons below the threshold from the skeletal image. In this way, small skeletons that are unlikely to be considered to constitute the vascular network structure are removed as noise from the skeletal image, and the image is screened to extract skeletons greater than the threshold. Therefore, the shape detection unit 50 can generate an extracted image in which the skeletons of the vascular network structure are appropriately extracted from the skeletal image. The shape detection unit 50 outputs the extracted image, or the length of all skeletons in the extracted image, to the display control unit 60. The shape detection unit 50 stores the extracted image, or the length of all skeletons in the extracted image, in the second storage device 140.

[0061] The shape detection unit 50 may detect the number of branching points of the vascular network structure as the shape of the vascular network structure. For example, the shape detection unit 50 detects the number of branching points as the number of object pixels that are adjacent in at least three directions (up, down, left, or right) among the object pixels that constitute the skeleton drawn on the skeleton image, minus 2. The shape detection unit 50 may generate an extracted image, which is an image in which skeletons with a number of branching points greater than a threshold are extracted from the skeleton drawn on the skeleton image. The threshold is predetermined based on the properties of the sample, the observation conditions at the time of measurement, etc. From the skeleton image, skeletons with a small number of branching points that are not likely to be considered to constitute a vascular network structure are removed as noise, and the skeletons with a number greater than the threshold are extracted, thereby generating an extracted image in which the skeleton of the vascular network structure is appropriately extracted from the skeleton image. The shape detection unit 50 may output the extracted image, or the number of branching points of all skeletons in the extracted image, to the display control unit 60. The shape detection unit 50 may store the extracted image, or the number of branching points of all skeletons in the extracted image, in the second storage device 140. The shape detection unit 50 may output the number of skeletal branching points in a specific region of the extracted image, or the number of skeletal branching points per unit area in the extracted image, to the display control unit 60, or store it in the second storage device 140.

[0062] The display control unit 60 is connected to the acquisition unit 10, background processing unit 20, contour extraction unit 30, skeleton extraction unit 40 or shape detection unit 50 within the image processing device 1, and to the external display device 130. The display control unit 60 controls the display on the display device 130. The display control unit 60 controls the display of a cropped image, grayscale image, removed image, adjusted image, contour image, skeleton image, or extracted image on the display device 130. The display control unit 60 may also cause the display device 130 to display a list of the lengths of all skeletons in the extracted image or a list of the number of branching points.

[0063] The display device 130 is connected to the display control unit 60 of the image processing device 1 and is a device that displays content controlled by the display control unit 60. The display device 130 displays a cropped image, a grayscale image, a removed image, an adjusted image, a contour image, a skeletal image, or an extracted image. The display device 130 may also display a list of the lengths of all the skeletal structures in the extracted image or a list of the number of branching points. The display device 130 is, for example, a display.

[0064] The second storage device 140 is a storage medium such as a hard disk. The second storage device 140 stores data used in the image processing device 1, such as cropped images, grayscale images, removed images, adjusted images, contour images, skeletal images, extracted images, a list of the lengths of all skeletal structures in the extracted images, or a list of the number of branching points. Note that the first storage device 120 and the second storage device 140 may be the same storage device.

[0065] [Image processing method] Next, the operation of the image processing system 2 will be described. First, the process of acquiring the captured image to be processed by the image processing device 1 will be described. Figure 3 is a flowchart showing an example of the image acquisition process targeted by the image processing method according to the embodiment. The method shown in Figure 3 is started, for example, when cell lines, reagents, and consumables related to the creation of a 3D liver model are prepared.

[0066] First, in the color development process (S11), the vascular network structure of the 3D liver model is colorized by an operator or other person. Here, the above-mentioned fabrication process, drug administration process, and color development control process are performed.

[0067] Next, in the imaging process (S13), the imaging device 110 of the image processing system 2 images a 3D liver model having a vascular network structure that has been colorized by the colorization process (S11). The imaging device 110 images a region including the center of the 3D liver model. The imaging device 110 sums the pixel values ​​and color intensity of multiple images, each with a different Z-axis position of the confocal microscope, for each imaged 3D liver model to obtain an image of the 3D liver model. The imaging device 110 outputs the image to the acquisition unit 10 and the second storage device 140.

[0068] This completes the flowchart shown in Figure 3. By executing the flowchart shown in Figure 3, the image captured by the image processing method according to the embodiment is obtained. Next, the operation of the image processing device 1 of the image processing system 2 will be explained using Figure 4. Figure 4 is a flowchart of an example of the image processing method according to the embodiment. The flowchart shown in Figure 4 is started after the completion of the flowchart shown in Figure 3, for example, in response to a start operation by the operator.

[0069] First, in the image acquisition process (S15), the acquisition unit 10 of the image processing device 1 acquires an image of a 3D liver model having a colorized vascular network structure from the imaging device 110 or the first storage device 120.

[0070] Next, in the grayscale conversion process (S17), the background processing unit 20 of the image processing device 1 generates a cropped image by extracting the region containing the center of the 3D liver model from the captured image and converts it into a grayscale image. The background processing unit 20 generates multiple cropped images from the captured image, for example, and converts each of them into a grayscale image.

[0071] Next, in the noise reduction process (S19), the background processing unit 20 generates a de-signaled image by removing the background signal from the grayscale image generated in the grayscale conversion process (S17). The background processing unit 20 removes the background signal, for example, by the sliding paravoid method.

[0072] Next, in the pixel value adjustment process (S21), the background processing unit 20 generates an adjusted image by reducing the noise in the removed image generated in the noise reduction process (S19). The background processing unit 20 widens the difference in pixel values ​​of each pixel while maintaining the distribution of pixel values ​​of each pixel in the removed image, and subtracts a predetermined pixel value to generate the adjusted image.

[0073] Next, in the contour image generation process (S23), the contour extraction unit 30 of the image processing device 1 applies a wavelet transform to the adjusted image generated by the pixel value adjustment process (S21) to generate a contour image in which the contours of the vascular network structure are extracted. The contour extraction unit 30 repeatedly decomposes using the wavelet transform, extracts the high-frequency components of each level, and combines them into a single image. The contour extraction unit 30 generates a binarized contour image by setting the intensity of pixels with pixel values ​​below a threshold to 0 and the intensity of pixels with pixel values ​​above a threshold to 1.

[0074] Next, in the skeletal image generation process (S25), the skeletal extraction unit 40 of the image processing device 1 repeatedly removes pixels from the boundaries of objects drawn on the contour image and generates a skeletal image in which skeletons having a predetermined number of pixels in line width are extracted. The skeletal extraction unit 40 repeatedly removes object pixels from the boundaries of object regions in the binarized contour image until the width of the object region becomes a predetermined number of pixels.

[0075] Next, in the length detection process (S27), the shape detection unit 50 of the image processing device 1 detects the length of the vascular network structure based on the skeleton drawn on the skeletal image. The shape detection unit 50 calculates, for example, the number of pixels for each skeleton in the skeletal image.

[0076] Next, in the extracted image generation process (S29), the shape detection unit 50 generates an extracted image, which is an image in which skeletons drawn on the skeleton image whose length is greater than a threshold are extracted.

[0077] Next, in the output processing (S31), the display control unit 60 of the image processing device 1 displays the extracted image, or a list of the lengths of all the skeletons in the extracted image, on the display device 130 of the image processing system 2. The shape detection unit 50 stores the extracted image, or the lengths of all the skeletons in the extracted image, in the second storage device 140 of the image processing system 2.

[0078] This completes the flowchart shown in Figure 4. Executing the flowchart in Figure 4 automatically outputs information regarding the length of the vascular network structure from the acquired images.

[0079] [Image processing program] This document describes an image processing program for making a computer function as an image processing device 1. The image processing program comprises a main module, an acquisition module, a contour generation module, and a skeleton generation module. The image processing program may also include a background processing module and a shape detection module. The main module is the part that comprehensively controls the device. The functions realized by executing the acquisition module, contour generation module, skeleton generation module, background processing module, and shape detection module are the same as the functions of the acquisition unit 10, background processing unit 20, contour extraction unit 30, skeleton extraction unit 40, shape detection unit 50, and display control unit 60 of the image processing device 1 described above.

[0080] [Summary of Embodiments] According to the image processing method and image processing program of this embodiment, an image of a cell structure having a vascular network structure is processed. First, a wavelet transform is applied to the image. This generates a contour image in which the contour of the vascular network structure is extracted (contour image generation process (S23)). Subsequently, pixels of the object are repeatedly removed from the boundaries of the object recognized in the contour image. This generates a skeleton image in which the skeleton of the vascular network structure is extracted (skeleton image generation process (S25)). Since the skeleton image is generated from the contour image, a skeleton that more accurately reflects the shape of the blood vessels is obtained compared to when the skeleton image is generated directly from the captured image. Therefore, this image processing method and image processing program can generate an image in which the shape of the vascular network structure can be grasped more accurately.

[0081] Furthermore, since the image processing method according to this embodiment includes a step of detecting the shape of the vascular network structure based on a skeletal image (length detection process (S27) or extracted image generation process (S29)), the characteristics of the vascular network structure can be evaluated using the detected shape of the vascular network structure.

[0082] Furthermore, since the image processing method according to this embodiment includes a step of detecting the shape of the vascular network structure (length detection process (S27) or extracted image generation process (S29)), the characteristics of the vascular network structure can be evaluated using the detected length of the vascular network structure.

[0083] Furthermore, the image processing method according to this embodiment can appropriately extract regions where the vascular network structure of the 3D liver model is likely to be captured by targeting captured images that include the central region of the 3D liver model.

[0084] Furthermore, the image processing method according to this embodiment is performed after the acquisition step (image acquisition processing (S15)) and before the contour image generation step (contour image generation processing (S23)), and includes a step to generate an image from which the background signal of the vascular network structure has been removed (noise reduction processing (S19)). As a result, the background signal of the image is removed, and the skeleton showing the vascular network structure can be extracted more clearly.

[0085] Furthermore, the image processing method according to this embodiment is executed after the step of generating a skeletal image (skeletal image generation process (S25)) and includes a step of generating an extracted image in which skeletons whose length is greater than a threshold are extracted from the skeletons drawn on the skeletal image (length detection process (S27) and extracted image generation process (S29)). Therefore, skeletons that are unlikely to constitute a vascular network structure can be removed as noise.

[0086] Furthermore, the image processing method according to this embodiment is executed after the acquisition step (image acquisition processing (S15)) and before the contour image generation step (contour image generation processing (S23)), and includes a step to generate an image that has been converted into a grayscale image of predetermined grayscale levels (grayscale conversion processing (S17)). As a result, the captured image is converted into a grayscale image, and processing of the converted image can be easily carried out.

[0087] Furthermore, the image processing method according to this embodiment is performed before the acquisition step (image acquisition processing (S15)) and further includes a step of immunostaining cells having a vascular network structure, pre-modifying cells having a vascular network structure with fluorescence, or developing color in the vascular network structure by a combination of immunostaining and fluorescence modification (color development processing (S11)), and a step of imaging the 3D liver model having the vascular network structure developed by the color development step (imaging processing (S13)). By performing color development processing (S11) and imaging processing (S13) on the 3D liver model having a vascular network structure, the shape of the vascular network structure can be accurately grasped for each 3D liver model.

[0088] Furthermore, the image processing method according to this embodiment can extract the vascular network structure in vascular endothelial cells and detect the shape of the vascular network structure by targeting a 3D liver model having vascular endothelial cells.

[0089] Although embodiments of the present disclosure have been described above, the present disclosure is not limited to the above embodiments. For example, the image processing apparatus 1 may include an imaging device 110, a first storage device 120, a display device 130, or a second storage device 140 in the image processing system 2. The image processing apparatus 1 does not have to include a background processing unit 20, a shape detection unit 50, and a display control unit 60. [Examples]

[0090] The image processing method and the effects of the image processing program will be explained with reference to Figures 5 to 10. In Figures 5 to 8, images were acquired from a control sample in which cyclophosphamide was not administered to the 3D liver model. Unless otherwise specified, it is assumed that the processes shown in the flowcharts in Figures 3 and 4 above have been performed on the acquired images.

[0091] [Contour image generation processing effect] In the flowchart shown in Figure 4, the resulting skeletons were compared between the case where the contour image generation process (S23) is performed and the case where the contour image generation process (S23) is not performed (or when other processes are performed). Figures 5(a) to (f) show examples of images obtained by the contour image generation process and the skeleton image generation process. Hereafter, the image immediately before the skeleton image generation process (S25) will be referred to as the original image.

[0092] (Comparative Example 1) Figure 5(a) is an image obtained without performing the contour image generation process (S23). Figure 5(b) is an image obtained by performing the skeletal image generation process (S25) using the image shown in Figure 5(a) as the source image (Comparative Example 1). As shown in Figure 5(b), linear shapes are displayed regardless of the vascular network structure of the source image, and the skeletal network structure was not properly extracted as a skeletal image.

[0093] (Comparative Example 2) Figure 5(c) is an image obtained by performing a binarization process on the image shown in Figure 5(a), but instead of contour image generation processing (S23), by setting a threshold for pixel values, displaying only pixels above the threshold in white, and displaying pixels below the threshold in black. Figure 5(d) is an image obtained by performing skeletal image generation processing (S25) on the image shown in Figure 5(c) as the source image (Comparative Example 2). As shown in Figure 5(d), a skeleton that matches the source image to some extent was extracted. However, for example, in areas showing small cavities in the vascular network structure of the source image, the skeleton was not accurately extracted, and a skeleton pattern significantly more complex than the corresponding area on the source image was formed. This is thought to be a problem because there are cavities in the pattern generated by the binarization process. The skeletal image generation process is a process that erodes the object recognized in the image from the edge side until it is 1 pixel wide, but in the source image shown in Figure 5(c), there are small black dots (holes) in various places within the overall pattern that was extracted in white. In the skeletal image generation process, this hole is mistakenly recognized as part of an edge, which is thought to hinder the accurate extraction of the skeleton.

[0094] (Example 1) Figure 5(e) is the image obtained by performing contour image generation processing (S23) on the image shown in Figure 5(a). Figure 5(f) is the image obtained by performing skeletal image generation processing (S25) on the image shown in Figure 5(e) as the source image (Example 1). As shown in Figure 5(e), information such as the thickness of the blood vessels is lost from the image shown in Figure 5(a), but as shown in Figure 5(f), a skeleton that matches the source image is extracted compared to Comparative Examples 1 and 2, and in particular, a skeleton that accurately represents the small cavities in the vascular network structure of the source image is extracted. Thus, it was confirmed that, in order to extract a skeleton in the skeletal image generation processing, it is not necessarily required that the vascular region of the source image be correctly captured, and that it is sufficient if an approximate contour can be extracted.

[0095] From Example 1, Comparative Example 1, and Comparative Example 2, it was confirmed that by performing the contour image generation process (S23), the skeleton of the vascular network structure in the image is extracted more appropriately in the skeleton image generation process (S25). The small black dots that were problematic in Comparative Example 2 are difficult to distinguish by definition from the mesh-like areas in the original image, and it is difficult to mechanically fill them in using an algorithm. It is also conceivable to set a threshold to prevent the formation of the small black dots in Comparative Example 2 and perform binarization, but this would increase the noise and is not practical. From Example 1 and Comparative Example 2, it was confirmed that the problems caused by the small black dots mentioned above can be solved by a new approach of extracting the skeleton after extracting the approximate contour of the original image.

[0096] [Effect of pixel value adjustment processing] The resulting skeletons were compared between cases where the pixel value adjustment process (S21) was performed and cases where the pixel value adjustment process (S21) was not performed. Figures 6(a) to (d) show examples of images obtained by the grayscale conversion process, noise reduction process, contour image generation process, and skeleton image generation process. Figures 7(a) to (d) show examples of images obtained by the pixel value adjustment process, contour image generation process, and skeleton image generation process.

[0097] (Comparative Example 3) Figure 6(a) is a grayscale image generated by the grayscale conversion process (S17). Figure 6(b) is a de-noise image obtained by performing noise reduction processing (S19) on the grayscale image shown in Figure 6(a). Figure 6(c) is an image obtained by performing contour image generation processing (S23) on the image shown in Figure 6(b) without performing pixel value adjustment processing (S21). Figure 6(d) is an image obtained by performing skeletal image generation processing (S25) on the image shown in Figure 6(c) (Comparative Example 3). As shown in Figure 6(d), when the pixel value adjustment process (S21) was not performed, the vascular network structure connected to other vascular networks was displayed even in areas where the vascular network structure was not clearly visible in the grayscale image, and the contours and skeletons of the vascular network structure were not properly extracted as contour images. This is thought to be due to the fact that in the contour image generation process (S23), actual vascular regions and noise were extracted as contours without distinction.

[0098] (Example 2) Figure 7(a) is an image obtained by adjusting the distribution of pixel values ​​for the same removed image as shown in Figure 6(b). Figure 7(b) is an image obtained by subtracting a predetermined pixel value from the pixel value of each pixel in the image shown in Figure 7(a). In other words, Figure 7(a) is an intermediate image in the pixel value adjustment process (S21), and Figure 7(b) is the adjusted image obtained after the pixel value adjustment process (S21) is completed. Figure 7(c) is a contour image obtained by performing the contour image generation process (S23) on the image shown in Figure 7(b). Figure 7(d) is an image obtained by performing the skeletal image generation process (S25) on the contour image shown in Figure 7(c) (Example 2).

[0099] As shown in Figures 7(a) and 7(b), it was confirmed that noise was removed by performing the pixel value adjustment process (S21). Furthermore, as shown in Figure 7(d), compared to Comparative Example 3 shown in Figure 6(d), the display of vascular network structures connected to other vascular networks in areas where the vascular network structure is not clearly visible in the grayscale image was suppressed, and the contours and skeletons of the vascular network structure were appropriately extracted as contour images. From the above, it was found that by performing the pixel value adjustment process (S21), the contours of the vascular network structure in the image were more appropriately extracted in the contour image generation process (S23), and the skeletons of the vascular network structure in the image were more appropriately extracted in the skeleton image generation process (S25).

[0100] [Effects of extracted image generation processing] Next, we compared the resulting skeletons when the extracted image generation process (S29) was performed and when it was not performed. Figure 8(a) is the captured image. Figure 8(b) is the skeleton image before the execution of the extracted image generation process (S29) (Comparative Example 4). Figure 8(c) is the extracted image after the execution of the extracted image generation process (S29) (Example 3).

[0101] By comparing Figures 8(b) and 8(c), it was found that when the extracted image generation process (S29) was performed, skeletons with short skeleton lengths were removed. As a result, the extracted image displayed structures similar to the vascular network structure visible in the captured image, compared to the skeleton image, indicating that the skeleton of the vascular network structure was appropriately extracted. Therefore, it was found that performing the extracted image generation process (S29) resulted in a more appropriate extraction of the skeleton of the vascular network structure in the image.

[0102] [Quantitative evaluation of the vascular network] This section describes the quantitative evaluation of the vascular network in sample groups exposed to cyclophosphamide at various concentrations to 3D liver models (cellular structures) prepared in the manufacturing process described in the embodiment. Figures 9(a) to (f) are images acquired according to the flowchart shown in Figure 3, with cyclophosphamide drug concentrations of 0 μM, 143 μM, 430 μM, 1430 μM, 4300 μM, and 14300 μM, respectively. Figures 9(g) to (l) are extracted images obtained by performing each process in the flowchart shown in Figure 4 on the images shown in Figures 9(a) to (f).

[0103] As can be seen in the images shown in Figures 9(a) to (f), the length of the vascular network structure tends to be shorter as the cyclophosphamide concentration increases. The extracted images shown in Figures 9(g) to (l) also reflect this tendency for the skeletal structure to be shorter as the cyclophosphamide concentration increases.

[0104] Figures 10(a) and 10(b) show the results of extracting the skeletal length and the number of branching points in the extracted images shown in Figures 9(g) to 9(l). Figure 10(a) shows the correspondence between the skeletal length in the extracted images related to the example and the drug concentration of cyclophosphamide, and Figure 10(b) shows the correspondence between the number of branching points in the skeletal structure in the extracted images and the drug concentration of cyclophosphamide. As shown in Figures 10(a) and 10(b), it can be seen that the shape of the vascular network structure changes significantly when cyclophosphamide exceeds a certain drug concentration, and the vascular network structure becomes less dense. Therefore, it was confirmed that the characteristics and trends of the shape of the vascular network structure can be appropriately grasped from the extracted images generated by the image processing method. [Explanation of Symbols]

[0105] 1...Image processing device, 2...Image processing system, 10...Acquisition unit, 20...Background processing unit, 30...Contour extraction unit, 40...Skeleton extraction unit, 50...Shape detection unit, 60...Display control unit, 110...Imaging device, 120...First storage device, 130...Display device, 140...Second storage device.

Claims

1. An acquisition unit that acquires images of cellular structures having a colorized vascular network structure, A contour extraction unit that applies a wavelet transform to the image and generates a contour image in which the contour of the vascular network structure is extracted, A skeleton extraction unit repeatedly removes pixels of an object from the boundary of the object recognized in the contour image and generates a skeleton image in which a skeleton having a predetermined number of pixel line widths is extracted. A shape detection unit generates an extracted image in which, among the skeletons drawn on the skeleton image generated by the skeleton extraction unit, the skeletons whose length is greater than a threshold are extracted, An image processing device equipped with the following features.

2. The image processing apparatus according to claim 1, wherein the shape detection unit detects the shape of the vascular network structure based on the skeletal image.

3. The image processing apparatus according to claim 2, wherein the shape of the vascular network structure is the length of the vascular network structure.

4. The image processing apparatus according to any one of claims 1 to 3, wherein the image is an image of a region including the center of the cell structure.

5. The image processing apparatus according to any one of claims 1 to 4, further comprising a background processing unit that generates an image from which the background signal of the image has been removed after the acquisition unit has acquired the image and before the contour extraction unit has generated the contour image.

6. The image processing apparatus according to any one of claims 1 to 5, further comprising a background processing unit that converts the image into an image of predetermined grayscale after the acquisition unit has acquired the image and before the contour extraction unit has generated the contour image.

7. The image processing apparatus according to any one of claims 1 to 6, wherein the cell structure comprises vascular endothelial cells.

8. The image processing apparatus according to any one of claims 1 to 7, An imaging device that, before the acquisition unit acquires the image, images the cell structure having the vascular network structure that has been colored by immunostaining, fluorescent modification, or a combination of immunostaining and fluorescent modification, Equipped with, The acquisition unit is an image processing system that acquires images captured by the imaging device.

Images