Automatic delineation method and system for tumor radiotherapy target area based on artificial intelligence

Abstract

The present invention relates to the technical field of tumor tissue positioning, and specifically to a method and system for automatic delineation of tumor radiotherapy target areas based on artificial intelligence. Based on RGB images and grayscale images of tumor tissue, the present invention first performs edge detection to extract grayscale connected domains. The grayscale similarity and distribution characteristics between pixels in each connected domain are analyzed, the tumor characterization degree is calculated, the high metabolic area is screened out and its position in the RGB image is located. For the first preset area, based on the difference and distribution characteristics of the pixel B channel value and the central area, a low infiltration characteristic factor is generated and the low infiltration area is delineated. Further, in the second preset area outside the low infiltration area, the diffusion characteristic factor is calculated by combining the R / B / G channel difference of the pixel, and the gradient value is fused to obtain the diffusion area, thereby automatically delineating the tumor tissue. The present invention can accurately segment the high metabolic area, low infiltration area and diffusion area in the tumor tissue, thereby improving the accuracy of automatic delineation of the tumor target area.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor tissue positioning, and in particular to an artificial intelligence-based method and system for automatically delineating tumor radiotherapy target areas. Background Art

[0002] In cancer radiotherapy, accurate delineation of the target volume is a critical step in developing personalized radiotherapy plans. It directly impacts whether the tumor is adequately irradiated and surrounding normal tissues are effectively protected, and is fundamental to ensuring radiotherapy efficacy and patient safety. However, traditional target delineation methods rely heavily on the professional experience and extensive imaging knowledge of the individual involved. Due to the wide range of individual patient variability, complex tumor morphology, and anatomical changes during treatment, this process is often time-consuming and labor-intensive. Delineation results can also vary significantly between individuals, impacting the accuracy and consistency of radiotherapy plans. With the rapid development of artificial intelligence (AI) technology, it is increasingly being applied in clinical practice. On the one hand, it can rapidly process and analyze large amounts of medical imaging data, learning the characteristic patterns of tumors and surrounding tissues, thereby enabling efficient and stable automatic target volume identification and delineation. On the other hand, automated delineation models based on deep learning can reduce errors and biases caused by human factors, improve the accuracy and reproducibility of delineation results, and provide clinicians with a more reliable reference. This facilitates more precise radiotherapy plans, enhances radiotherapy efficacy, and minimizes damage to surrounding normal tissues.

[0003] In reality, the tumor cells of invasive tumors not only grow within the primary site, but also infiltrate into the adjacent normal tissues. This growth characteristic makes it difficult to clearly distinguish between the tumor and normal tissue. Traditional image segmentation algorithms often use a relatively single threshold to distinguish the target area from the background area. However, the characteristics of invasive tumors make it impossible to achieve a perfect segmentation effect by directly using threshold segmentation, thereby reducing the accuracy of tumor radiotherapy target area delineation. Summary of the Invention

[0004] In order to solve the technical problem that tumor cells of invasive tumors not only grow in the primary site but also infiltrate into the adjacent normal tissues, resulting in the inability of traditional image segmentation algorithms to accurately segment tumor tissues, thereby reducing the accuracy of automatic delineation of tumor target areas, the purpose of the present invention is to provide an artificial intelligence-based method and system for automatic delineation of tumor radiotherapy target areas. The technical scheme adopted is as follows: an artificial intelligence-based method for automatic delineation of tumor radiotherapy target areas, the method comprising: obtaining an RGB image of tumor tissue and a grayscale image of tumor tissue in the area where the tumor tissue is located; performing edge detection on the grayscale image of the tumor tissue to obtain all grayscale connected domains; obtaining the tumor characterization degree of each grayscale connected domain based on the grayscale similarity characteristics between each pixel point and other pixel points in each grayscale connected domain and the grayscale distribution characteristics of the pixel points; screening all grayscale connected domains according to the tumor characterization degree to obtain the area where the tumor tissue is located. a high metabolism area; obtaining the regional position of the high metabolism area in the RGB image of the tumor tissue as the high metabolism area position; obtaining the low infiltration characteristic factor of all pixels in the preset first area according to the difference in B channel value between each pixel point and the pixel point at the high metabolism area position in a preset first area around the high metabolism area position, and the distribution characteristics of the B channel values ​​of the pixels in the preset first area; obtaining the low infiltration area according to the low infiltration characteristic factor of the pixels in the preset first area; obtaining the diffusion characteristic factor of all pixels in the preset second area according to the difference characteristics between the R channel value and the B channel value and the G channel value of each pixel point in a preset second area around the low infiltration area; obtaining the diffusion area according to the gradient value of the diffusion characteristic factor of the pixel point in the second area and the pixel point; and automatically outlining the area where the tumor tissue is located according to the positions of the high metabolism area, the low infiltration area and the diffusion area.

[0005] Furthermore, the method for obtaining the tumor characterization degree includes: obtaining the tumor characterization degree according to a tumor characterization degree calculation formula, and the tumor characterization degree calculation formula is as follows: Where, Indicates the degree of tumor representation of each grayscale connected domain; Represents the number of pixels in each grayscale connected domain; Indicates the first Gray value of each pixel; Indicates that in each grayscale connected domain, The number of pixels whose grayscale differences between pixels are within a preset first range; In the grayscale image of tumor tissue, The number of pixels whose grayscale differences between pixels are within a preset first range; Indicates that in each grayscale connected domain, The grayscale difference between the pixels is within the preset first range. pixels, and the The Euclidean distance between pixels.

[0006] Furthermore, the method for obtaining the high-metabolism region includes: taking the grayscale connected domain with the highest tumor representation degree as the high-metabolism region.

[0007] Furthermore, the method for obtaining the low-infiltration characteristic factor includes: obtaining the minimum bounding rectangle of the high-metabolism area; taking the minimum bounding rectangle as the center and the area where eight rectangles with the same width and height as the minimum bounding rectangle are located as the preset first area; obtaining the low-infiltration characteristic factor according to the low-infiltration characteristic factor calculation formula, the low-infiltration characteristic factor calculation formula is as follows: Where, Indicates the low infiltration characteristic factor of each pixel point in a preset first area around the location of the high metabolic area; Indicates the number of pixels within the high metabolic area; Indicates the location of the high metabolic area The B channel value of each pixel; Indicates the B channel value of each pixel point in the preset first area around the high metabolic area; Indicates the number of pixels within a preset neighborhood of each pixel in a preset first area around the location of the high metabolic area; Indicates the first region of the high metabolic area around each pixel in the preset neighborhood. The B channel value of each pixel; Represents Iverson brackets. If the condition in the brackets is met, the value is 1. If the condition in the brackets is not met, the value is 0.

[0008] Furthermore, the method for obtaining the low infiltration area includes: taking all pixel points in the preset first area whose low infiltration characteristic factor is greater than the preset first threshold as low infiltration area pixel points; and taking the area composed of the positions of all low infiltration area pixel points as the low infiltration area.

[0009] Furthermore, the method for obtaining the diffusion characteristic factor includes: obtaining the minimum bounding rectangle of the low infiltration area; taking the minimum bounding rectangle as the center and the area where eight rectangles with the same width and height as the minimum bounding rectangle are located as the preset second area; obtaining the diffusion characteristic factor according to the diffusion characteristic factor calculation formula, the diffusion characteristic factor calculation formula is as follows: Where, Represents the diffusion characteristic factor of each pixel in the preset second area; Represents the R channel value of each pixel in the preset second area; Indicates the G channel value of each pixel in the preset second area; Indicates the B channel value of each pixel in the preset second area; Represents the weight of the R channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the grayscale value of each pixel in the preset second area; Indicates the weight of the G channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the weight of the B channel value of each pixel point in the preset second area in the grayscale value conversion operation; Represents Iverson brackets. If the condition in the brackets is met, the value is 1. If the condition in the brackets is not met, the value is 0.

[0010] Furthermore, the method for obtaining the diffuse area includes: calculating the product between the diffuse characteristic factor and the gradient value of each pixel point in the preset second area as the diffuse area boundary factor of each pixel point in the preset second area; taking the pixel points in the preset second area whose diffuse area boundary factor is greater than the preset second threshold as diffuse area pixel points; and taking the locations of all diffuse area pixel points as diffuse areas.

[0011] A system for automatically delineating tumor radiotherapy target areas based on artificial intelligence, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned method for automatically delineating tumor radiotherapy target areas based on artificial intelligence are implemented.

[0012] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for automatic delineation of tumor radiotherapy target areas based on artificial intelligence.

[0013] A computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned method for automatic delineation of tumor radiotherapy target areas based on artificial intelligence are implemented.

[0014] The present invention has the following beneficial effects: the present invention obtains the RGB image of the tumor tissue in the area where the tumor tissue is located to distinguish the colors of different tumor areas, and obtains the grayscale image of the tumor tissue to improve the recognition speed of the tumor core area; since the high metabolism area appears to have a high grayscale value in the grayscale image, and the pixel points are relatively concentrated, the edge detection is first performed on the grayscale image of the tumor tissue to obtain all grayscale connected domains, and the high metabolism area can be further obtained by screening the grayscale connected domains; since the color feature differences of different tumor areas can be observed through the RGB image of the tumor tissue, that is, there is a large color difference between the low infiltration area and the high metabolism area, the B channel value difference between each pixel point and the pixel point at the high metabolism area position in the first area preset around the high metabolism area position is, And the B channel value distribution characteristics of the pixels in the preset first area are obtained to obtain the low infiltration characteristic factors of all pixels in the preset first area, so as to accurately identify the low infiltration area; since the diffuse area is obtained by further outward diffusion of the low infiltration area with relatively low infiltration degree around the high metabolic area, in the RGB three channels, compared with the high metabolic area and the low infiltration area, the R channel value is more prominent than the G channel and the B channel, so according to the difference characteristics between the R channel value and the B channel value and the G channel value of each pixel in the preset second area around the low infiltration area, the diffusion characteristic factors of all pixels in the preset second area are obtained, and then the diffusion area is obtained; according to the locations of the high metabolic area, the low infiltration area and the diffusion area, the area where the tumor tissue is located is automatically delineated. The present invention can accurately segment the high metabolic area, the low infiltration area and the diffusion area in the tumor tissue, thereby improving the accuracy of automatic delineation of the tumor target area. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flow chart of an artificial intelligence-based method for automatically delineating a tumor radiotherapy target area according to one embodiment of the present invention; Figure 2 A block diagram of an artificial intelligence-based automatic tumor radiotherapy target area delineation system provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of an artificial intelligence-based method and system for automatically delineating tumor radiotherapy target areas proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0018] Unless defined otherwise, 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 belongs.

[0019] The following describes in detail a specific solution of an artificial intelligence-based tumor radiotherapy target area automatic delineation method and system provided by the present invention with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows an artificial intelligence-based automatic delineation method for tumor radiotherapy target areas provided by an embodiment of the present invention, the method comprising: Step S1: acquiring an RGB image and a grayscale image of the tumor tissue in the area where the tumor tissue is located.

[0021] The embodiment of the present invention is mainly used in scenarios where the boundaries of invasive tumors are difficult to identify and segment. In order to accurately segment the area where the invasive tumor is located, an image of the area where the tumor tissue is located should be obtained first. Since the metabolic activity in the area where the tumor tissue is located is vigorous, the brightness presented in the grayscale image is different from that in other normal body tissue areas. In addition, in addition to the high-metabolism tumor core area, the invasive tumor also includes low-infiltration areas after tumor cell infiltration and diffuse areas that may contain diffused tumor cells. Accurate boundary identification cannot be performed through simple threshold segmentation. Therefore, in the embodiment of the present invention, the RGB image of the tumor tissue in the area where the tumor tissue is located is obtained to distinguish the colors of different tumor areas, and the grayscale image of the tumor tissue is obtained to improve the recognition speed of the tumor core area.

[0022] In the embodiment of the present invention, the patient lies on the examination bed and the device The device scans the body to collect anatomical information of various parts of the body, including detailed morphology and location information of the tissues surrounding the tumor area. The scan section begins to work by detecting signals emitted by radioactive tracers in the patient's body and generating images that reflect the metabolic activity of the tissue. Image and Image fusion acquisition Image, now The image has both the high brightness characteristics of the high metabolism area and the color characteristic differences of the low infiltration area, diffuse area and high metabolism area, so it is used as the RGB image of the tumor tissue required for subsequent operations. The image is grayscale processed to obtain the grayscale image of tumor tissue required for subsequent operations.

[0023] The embodiment of the present invention is to Image input to In the network, in the encoder part of the network, through multi-layer convolution and pooling operations, the shallow to deep features of the image are gradually extracted, including gray value distribution, edge information, etc., while skip connections are used to retain the spatial detail information of the image, providing a rich feature basis for subsequent region division. It should be noted that in other embodiments of the present invention, it is also possible to use There is no limitation on split network and other split network operations. The network is a technical means well known to those skilled in the art and will not be described in detail here.

[0024] Step S2: Perform edge detection on the grayscale image of the tumor tissue to obtain all grayscale connected domains; obtain the tumor characterization degree of each grayscale connected domain based on the grayscale similarity characteristics between each pixel point and other pixel points in each grayscale connected domain and the grayscale distribution characteristics of the pixel points; screen all grayscale connected domains based on the tumor characterization degree to obtain the high metabolism area in the area where the tumor tissue is located; obtain the regional position of the high metabolism area in the RGB image of the tumor tissue as the high metabolism area position; obtain the low infiltration characteristic factor of all pixel points in the preset first area based on the B channel value difference between each pixel point and the pixel point at the high metabolism area position in a preset first area around the high metabolism area position, and the B channel value distribution characteristics of the pixel points in the preset first area; obtain the low infiltration characteristic factor of the pixel points in the preset first area based on the low infiltration characteristic factor of the pixel points in the preset first area.

[0025] Since high metabolic areas appear as high grayscale values ​​in grayscale images and the pixel distribution is relatively concentrated, in an embodiment of the present invention, edge detection is first performed on the grayscale image of the tumor tissue to obtain all grayscale connected domains, among which high metabolic areas can be further obtained by screening the grayscale connected domains.

[0026] Preferably, in one embodiment of the present invention, the method for obtaining the tumor characterization degree includes: obtaining the tumor characterization degree according to a tumor characterization degree calculation formula, and the tumor characterization degree calculation formula is as follows: Where, Indicates the degree of tumor representation of each grayscale connected domain; Represents the number of pixels in each grayscale connected domain; Indicates the first Gray value of each pixel; Indicates that in each grayscale connected domain, The number of pixels whose grayscale differences between pixels are within a preset first range; In the grayscale image of tumor tissue, The number of pixels whose grayscale differences between pixels are within a preset first range; Indicates that in each grayscale connected domain, The grayscale difference between the pixels is within the preset first range. pixels, and the The Euclidean distance between pixels.

[0027] In one embodiment of the present invention, the preset first range is set to: The grayscale value range of each pixel point corresponds to a grayscale value floating value of 5. It should be noted that the preset first range can be set arbitrarily and is not limited here.

[0028] In the calculation formula of tumor characterization degree, the higher the overall gray value of the grayscale connected domain, that is, The higher the value, the higher the overall gray value of the gray connected domain is, which means that the gray connected domain is more likely to be a high metabolism area, and the tumor representation degree of the gray connected domain is greater. Since the pixels in the high metabolism area are more concentrated and the gray values ​​of the pixels are similar, when the gray connected domain is The number of pixels whose grayscale differences are within the preset first range occupies the same area in the grayscale image of the tumor tissue as that in the first range. The ratio of the number of pixels whose grayscale differences are within the preset first range The larger the The pixel points whose grayscale differences between the first pixel points are within the preset first range are mostly concentrated in the grayscale connected domain, and at this time The grayscale difference between each pixel in the preset first range is The distance between pixels The smaller the Pixels with similar grayscale are clustered in the same grayscale connected domain. At this time, the grayscale connected domain is more likely to be a high-metabolism area, and the tumor representation degree of the grayscale connected domain is greater.

[0029] All grayscale connected domains are screened according to the degree of tumor characterization to obtain a high metabolic region in the region where the tumor tissue is located. Preferably, in one embodiment of the present invention, the method for obtaining a high metabolic region includes: taking the grayscale connected domain with the highest degree of tumor characterization as the high metabolic region.

[0030] In actual situations, due to factors such as the difference between local hypoxia and oxygen enrichment caused by tumor vascular malformations inside the tumor and the difference in metabolic capacity of different subpopulations of cancer cells, the grayscale value inside the tumor will show a relatively high overall but uneven internal distribution physiological characteristic. Therefore, the less infiltrated part around the high metabolic area should also belong to the radiotherapy target area. The higher the infiltration level, the more obvious the brightness in the original image, that is, the color in the area tends to be white. Therefore, the high metabolic area corresponds to The values ​​of the three channels are close to , and the color of the low-infiltration area is the color after the white is attenuated. When the white is attenuated, the blue part is lost first, and the blue part corresponds to channel, through When the three channel values ​​are given different weights and converted into grayscale values The weight of the channel is the lowest, so it cannot be distinguished from the surrounding high metabolic areas directly by grayscale value, and it is necessary to use Three-channel values. It can be observed through the RGB image of tumor tissue that different tumor areas have different color characteristics, that is, there is a large color difference between the low-infiltration area and the high-metabolism area. Therefore, in the embodiment of the present invention, based on the B-channel value difference between each pixel in the preset first area around the high-metabolism area and the pixel in the high-metabolism area, and the B-channel value distribution characteristics of the pixels in the preset first area, the low-infiltration characteristic factor of all pixels in the preset first area is obtained, thereby accurately identifying the low-infiltration area.

[0031] Preferably, in one embodiment of the present invention, a method for obtaining a low-infiltration characteristic factor includes: obtaining a minimum bounding rectangle of a high-metabolism region; taking the minimum bounding rectangle as the center and an area containing eight rectangles having the same width and height as the minimum bounding rectangle as the preset first area; and obtaining the low-infiltration characteristic factor according to a low-infiltration characteristic factor calculation formula, the low-infiltration characteristic factor calculation formula being as follows: Where, Indicates the low infiltration characteristic factor of each pixel point in a preset first area around the location of the high metabolic area; Indicates the number of pixels within the high metabolic area; Indicates the location of the high metabolic area The B channel value of each pixel; The sum of the B channel values ​​of all pixels in the preset first area around the high metabolic area; Indicates the number of pixels within a preset neighborhood of each pixel in a preset first area around the location of the high metabolic area; Indicates the first region of the high metabolic area around each pixel in the preset neighborhood. The B channel value of each pixel; Represents Iverson brackets. If the condition in the brackets is met, the value is 1. If the condition in the brackets is not met, the value is 0.

[0032] In one embodiment of the present invention, the preset neighborhood is set to take each pixel point in the preset first area as the center. It should be noted that the preset neighborhood can be set by yourself and is not limited here.

[0033] In the calculation formula of low infiltration characteristic factor, since the B channel value in the low infiltration area is smaller than the B channel value in the high metabolism area, As a precondition, when the mean B channel value of all pixels in the high metabolic area is greater than the B channel value of the pixel in the preset first area, the pixel in the preset first area is analyzed. When the difference between the mean B channel value of all pixels in the high metabolic area and the B channel value of the pixel in the preset first area is larger, it means that the pixel is more likely to have low infiltration characteristics. At this time, if the difference between the B channel value of the pixel and other pixels in the preset neighborhood is greater, the pixel is analyzed. The smaller it is, the more similar the B channel values ​​of the pixel are to those of the surrounding pixels, and the more likely the pixel is to be located in a low-infiltration area.

[0034] Preferably, in one embodiment of the present invention, the method for obtaining a low infiltration area includes: taking all pixel points within a preset first area whose low infiltration characteristic factor is greater than a preset first threshold as low infiltration area pixel points; and taking an area composed of the positions of all low infiltration area pixel points as a low infiltration area.

[0035] In the embodiment of the present invention, the preset first threshold is set to 0.8, and can be set arbitrarily, and is not limited here.

[0036] Thus, a low-infiltration area is obtained.

[0037] Step S3: Based on the difference characteristics between the R channel value, B channel value, and G channel value of each pixel point in the preset second area around the low-infiltration area, the diffusion characteristic factor of all pixels in the preset second area is obtained; based on the diffusion characteristic factor of the pixel points in the second area and the gradient value of the pixel points, the diffusion area is obtained; and the area where the tumor tissue is located is automatically outlined according to the locations of the high-metabolism area, the low-infiltration area, and the diffusion area.

[0038] The outward expansion of the low-infiltration area is the diffuse area. Although the degree of metabolism in the diffuse area is not as high as that in the core area, it may still contain tumor cells, reflecting the possible subclinical lesions of the tumor. Therefore, the embodiment of the present invention analyzes the boundary of the diffuse area. Since the diffuse area is obtained by further outward diffusion of the low-infiltration area with a relatively low degree of infiltration around the high-metabolism area, in the RGB three channels, the R channel value is more prominent than the G channel and the B channel compared to the high-metabolism area and the low-infiltration area. Therefore, in the embodiment of the present invention, according to the difference characteristics between the R channel value and the B channel value and the G channel value of each pixel point in the preset second area around the low-infiltration area, the diffusion characteristic factor of all pixel points in the preset second area is obtained, and then the diffuse area is obtained.

[0039] Preferably, in one embodiment of the present invention, the method for obtaining the diffusion characteristic factor includes: obtaining the minimum bounding rectangle of the low infiltration area; taking the minimum bounding rectangle as the center and the area where 8 rectangles with the same width and height as the minimum bounding rectangle are located as the preset second area.

[0040] The diffusion characteristic factor is obtained according to the diffusion characteristic factor calculation formula. The diffusion characteristic factor calculation formula is as follows: Where, Represents the diffusion characteristic factor of each pixel in the preset second area; Represents the R channel value of each pixel in the preset second area; Indicates the G channel value of each pixel in the preset second area; Indicates the B channel value of each pixel in the preset second area; Represents the weight of the R channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the grayscale value of each pixel in the preset second area; Indicates the weight of the G channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the weight of the B channel value of each pixel point in the preset second area in the grayscale value conversion operation; Represents Iverson brackets. If the condition in the brackets is met, the value is 1. If the condition in the brackets is not met, the value is 0.

[0041] In the calculation formula of the diffusion characteristic factor, and The judgment conditions in the two brackets indicate the pixel point when they are met. The channel value is larger than the other two channel values, the greater the difference, The more obvious the channel value protrusion tendency is, the more likely the pixel is to be a pixel in the diffuse area. The grayscale value of the R channel value, the G channel value, and the B channel value of each pixel in the preset second area are converted. The smaller the difference between the grayscale value after the R channel value conversion and the grayscale value of the pixel itself, the more prominent the pixel is compared with the G channel value and the B channel value. That is, in the denominator of the formula, The smaller, The larger it is, the smaller the denominator is as a whole, which means that the pixel point is more likely to belong to the diffuse area.

[0042] Preferably, in one embodiment of the present invention, the method for obtaining the diffuse area includes: calculating the product of the diffuse characteristic factor and the gradient value of each pixel point in the preset second area as the diffuse area boundary factor of each pixel point in the preset second area.

[0043] Pixels within the predetermined second region whose diffusion region boundary factor is greater than a predetermined second threshold are defined as diffusion region pixels; and the locations of all diffusion region pixels are defined as the diffusion region. In this embodiment of the present invention, the predetermined second threshold is set to 0.9. It should be noted that the predetermined second threshold can be set arbitrarily and is not limited herein.

[0044] According to the high metabolic area, low infiltration area and diffuse area The position in the image is automatically delineated for the area where the tumor tissue is located. The specific steps are technical means well known to those skilled in the art and will not be described in detail here.

[0045] In summary, the RGB image and grayscale image of the tumor tissue in the area where the tumor tissue is located are obtained; the edge detection is performed on the grayscale image of the tumor tissue to obtain all grayscale connected domains; the tumor representation degree of each grayscale connected domain is obtained based on the grayscale similarity characteristics between each pixel point and other pixel points in each grayscale connected domain and the grayscale distribution characteristics of the pixel points; all grayscale connected domains are screened according to the tumor representation degree to obtain the high metabolism area in the area where the tumor tissue is located; the regional position of the high metabolism area in the RGB image of the tumor tissue is obtained as the high metabolism area position; according to the first area preset around the high metabolism area position, each pixel point and the high metabolism area position image are The low infiltration characteristic factors of all pixels in the preset first area are obtained based on the differences in B channel values ​​between pixels and the distribution characteristics of B channel values ​​of pixels in the preset first area; the low infiltration area is obtained based on the low infiltration characteristic factors of pixels in the preset first area; the diffusion characteristic factors of all pixels in the preset second area are obtained based on the difference characteristics between the R channel value and the B channel value and the G channel value of each pixel in the preset second area around the low infiltration area; the diffusion area is obtained based on the diffusion characteristic factors of pixels in the second area and the gradient values ​​of pixels; the area where the tumor tissue is located is automatically outlined according to the locations of the high metabolic area, low infiltration area and diffusion area.

[0046] The second purpose of one embodiment of the present invention is to provide an artificial intelligence-based automatic delineation system for tumor radiotherapy targets, the system comprising a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program runs in the processor, it can implement the method described in steps S1-S3, specifically including: an image acquisition module 101, used to acquire an RGB image and a grayscale image of the tumor tissue in the area where the tumor tissue is located.

[0047] The tumor tissue recognition module 102 is used to perform edge detection on the grayscale image of the tumor tissue to obtain all grayscale connected domains; obtain the tumor characterization degree of each grayscale connected domain based on the grayscale similarity characteristics between each pixel point and other pixel points in each grayscale connected domain and the grayscale distribution characteristics of the pixel points; screen all grayscale connected domains based on the tumor characterization degree to obtain a high-metabolism area in the area where the tumor tissue is located; obtain the regional position of the high-metabolism area in the RGB image of the tumor tissue as the high-metabolism area position; obtain the low-infiltration characteristic factor of all pixels in the preset first area based on the B channel value difference between each pixel point and the pixel point at the high-metabolism area position in a preset first area around the high-metabolism area position, and the B channel value distribution characteristics of the pixels in the preset first area; obtain the low-infiltration characteristic factor of the pixels in the preset first area based on the low-infiltration characteristic factor of the pixels in the preset first area.

[0048] The tumor tissue delineation module 103 is used to obtain the diffusion characteristic factors of all pixels in the preset second area around the low-infiltration area based on the difference characteristics between the R channel value, the B channel value, and the G channel value of each pixel point; obtain the diffusion area based on the diffusion characteristic factors of the pixels in the second area and the gradient values ​​of the pixels; and automatically delineate the area where the tumor tissue is located based on the locations of the high-metabolism area, the low-infiltration area, and the diffusion area.

[0049] The third object of an embodiment of the present invention is to provide a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the method described in steps S1-S3 is implemented when the processor executes the computer program.

[0050] A fourth object of an embodiment of the present invention is to provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in steps S1-S3 is implemented.

[0051] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0052] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. An artificial intelligence-based method for automatically delineating a tumor radiotherapy target area, characterized in that: The method includes: obtaining an RGB image and a grayscale image of the tumor tissue in an area where the tumor tissue is located; performing edge detection on the grayscale image of the tumor tissue to obtain all grayscale connected domains; obtaining a tumor characterization degree of each grayscale connected domain based on the grayscale similarity characteristics between each pixel point and other pixel points in each grayscale connected domain and the grayscale distribution characteristics of the pixel points; screening all grayscale connected domains based on the tumor characterization degree to obtain a high-metabolism area in the area where the tumor tissue is located; obtaining the regional position of the high-metabolism area in the RGB image of the tumor tissue as the high-metabolism area position; obtaining a minimum bounding rectangle of the high-metabolism area; taking the minimum bounding rectangle as the center and an area where eight rectangles having the same width and height as the minimum bounding rectangle are located as a preset first area; obtaining a low-infiltration characteristic factor according to a low-infiltration characteristic factor calculation formula, wherein the low-infiltration characteristic factor calculation formula is as follows: Where, Indicates the low infiltration characteristic factor of each pixel point in a preset first area around the location of the high metabolic area; Indicates the number of pixels within the high metabolic area; Indicates the location of the high metabolic area The B channel value of each pixel; Indicates the B channel value of each pixel point in the preset first area around the high metabolic area; Indicates the number of pixels within a preset neighborhood of each pixel in a preset first area around the location of the high metabolic area; Indicates the first region of the high metabolic area around each pixel in the preset neighborhood. The B channel value of each pixel; Represents an Iverson bracket, where if the condition in the bracket is met, the value is 1, and if the condition in the bracket is not met, the value is 0; a low infiltration area is obtained based on the low infiltration characteristic factor of the pixel points in the preset first area; a minimum bounding rectangle of the low infiltration area is obtained; an area with the minimum bounding rectangle as the center and eight rectangles with the same width and height as the minimum bounding rectangle is located around it as a preset second area; the diffusion characteristic factor is obtained according to a diffusion characteristic factor calculation formula, and the diffusion characteristic factor calculation formula is as follows: Where, Represents the diffusion characteristic factor of each pixel in the preset second area; Represents the R channel value of each pixel in the preset second area; Indicates the G channel value of each pixel in the preset second area; Indicates the B channel value of each pixel in the preset second area; Represents the weight of the R channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the grayscale value of each pixel in the preset second area; Indicates the weight of the G channel value of each pixel point in the preset second area in the grayscale value conversion operation; Indicates the weight of the B channel value of each pixel point in the preset second area in the grayscale value conversion operation; Represents an Iverson bracket, and if the condition in the bracket is met, the value is 1, and if the condition in the bracket is not met, the value is 0; the diffusion area is obtained according to the diffusion characteristic factor of the pixel point in the second area and the gradient value of the pixel point; the area where the tumor tissue is located is automatically outlined according to the locations of the high metabolic area, the low infiltration area and the diffusion area.

2. The method for automatic delineation of tumor radiotherapy target volume based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the tumor characterization degree includes: obtaining the tumor characterization degree according to a tumor characterization degree calculation formula, and the tumor characterization degree calculation formula is as follows: Where, Indicates the degree of tumor representation of each grayscale connected domain; Represents the number of pixels in each grayscale connected domain; Indicates the first Gray value of each pixel; Indicates that in each grayscale connected domain, The number of pixels whose grayscale differences between pixels are within a preset first range; In the grayscale image of tumor tissue, The number of pixels whose grayscale differences between pixels are within a preset first range; Indicates that in each grayscale connected domain, The grayscale difference between the pixels is within the preset first range. pixels, and the The Euclidean distance between pixels.

3. The method for automatic delineation of tumor radiotherapy target volume based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the high-metabolism region includes: taking the grayscale connected domain with the highest tumor representation degree as the high-metabolism region.

4. The method for automatic delineation of tumor radiotherapy target volume based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the low infiltration area includes: taking all pixel points in a preset first area whose low infiltration characteristic factor is greater than a preset first threshold as low infiltration area pixel points; and taking the area composed of the positions of all low infiltration area pixel points as the low infiltration area.

5. The method for automatic delineation of tumor radiotherapy target area based on artificial intelligence according to claim 1, characterized in that: The method for obtaining the diffuse area includes: calculating the product between the diffuse characteristic factor and the gradient value of each pixel point in the preset second area as the diffuse area boundary factor of each pixel point in the preset second area; taking the pixel points in the preset second area whose diffuse area boundary factor is greater than the preset second threshold as the diffuse area pixel points; and taking the locations of all diffuse area pixel points as the diffuse area.

6. An artificial intelligence-based automatic tumor radiotherapy target area delineation system, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for automatic delineation of tumor radiotherapy target areas based on artificial intelligence as described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for automatic delineation of tumor radiotherapy target areas based on artificial intelligence as described in any one of claims 1 to 5 are implemented.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method for automatic delineation of tumor radiotherapy target areas based on artificial intelligence as described in any one of claims 1 to 5 are implemented.

Images