Preoperative planning system for tumor interventional therapy based on artificial intelligence image recognition

Abstract

The invention relates to the technical field of edge segmentation, in particular to a tumor interventional therapy preoperative planning system based on artificial intelligence image recognition, and the method comprises the steps: preliminarily screening out an infiltration edge region based on the gray distribution asymmetry condition of a sliding window in a CT image and the texture disorder condition in an MRI image; further fusing DSA blood vessel data, and screening a concerned infiltration region by analyzing the position distribution condition of the blood vessel region of the infiltration edge region; analyzing curvature mutation conditions of the tumor associated blood vessel segments on the basis of the concerned infiltration region to generate a blood vessel anomaly index; and finally, fusing the infiltration index and the blood vessel characteristics to construct a blood vessel-infiltration coupling index, and correcting the edge strength map according to the blood vessel-infiltration coupling index, so that the boundary optimization based on risk classification is realized, the accuracy of the obtained tumor region edge structure curve is higher, and the accuracy of tumor region segmentation according to the tumor region edge structure curve is improved.

Description

Technical Field

[0001] This invention relates to the field of edge segmentation technology, and more specifically to a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition. Background Technology

[0002] Before planning interventional tumor therapy, it is usually necessary to locate the tumor lesion. Current technology typically uses the U-Net++ segmentation model based on CT images to locate the tumor region, and then uses the Canny edge algorithm to determine the segmentation edge line to segment the tumor region. However, the accurate edge of the tumor region is often affected by infiltration, resulting in blurred boundaries. This means that when using the Canny edge algorithm to segment the edge line, it can only segment strong edges and ignore the infiltrative edges under the coupling effect of cell diffusion-edema-blood vessels. In other words, the current Canny edge algorithm has low accuracy in identifying infiltrative edges in the tumor region, thus affecting the accuracy of tumor region segmentation. Summary of the Invention

[0003] To address the technical problem of low accuracy in identifying invasive edges when using the Canny edge recognition algorithm for tumor region segmentation, this application aims to provide a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition. The specific technical solution adopted is as follows: The first aspect of this application provides a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition, comprising: The data acquisition and preprocessing module is used to locate the same tumor ROI region in tumor CT images, tumor MRI images, and tumor DSA images; to traverse the tumor ROI region to determine all sliding windows; and to obtain the vascular region in the tumor DSA image based on angiography. The first determining module is used to determine the comprehensive invasion index of each sliding window based on the asymmetry of gray-level distribution in the tumor CT image and the disorder of gray-level distribution in the tumor MRI image; and to screen out the invasion edge region according to the comprehensive invasion index. The second determining module is used to determine the corresponding invasive activity feature value based on the location distribution of the vascular region in each invasive edge region of the tumor DSA image; to screen out the invasive regions of interest based on the invasive activity feature value; and to determine the corresponding vascular abnormality index based on the tortuosity of the vascular region in the invasive regions of interest. The region segmentation module is used to determine the corresponding vascular invasion coupling index based on the comprehensive invasion index, invasion activity characteristic value, and vascular abnormality index of each region of interest; to perform edge intensity map correction on the tumor ROI region based on the vascular invasion coupling index to determine the corresponding tumor region edge structure curve; and to perform tumor region segmentation based on the tumor region edge structure curve.

[0004] Furthermore, the process of obtaining the comprehensive infiltration index includes: In the tumor CT image, the corresponding cell proliferation activity index is determined based on the overall size of the gray values ​​in each sliding window and the uniformity of the gray value distribution. In the MRI image, the numerical entropy of the gray values ​​of all pixels in each sliding window is normalized to determine the corresponding tissue component mixing degree; The comprehensive infiltration index for each sliding window is determined by multiplying the negative correlation mapping value of the cell proliferation activity index with the degree of tissue component mixing.

[0005] Furthermore, the process of obtaining the cell proliferation activity index includes: In the tumor CT image, the mean gray value of all pixels in each sliding window is normalized to determine the mean cell density coefficient; the density distribution symmetry coefficient is determined by negative correlation normalization based on the skewness of the gray values ​​of all pixels in each sliding window; and the cell proliferation activity index of each sliding window is determined by the product of the mean cell density coefficient and the density distribution symmetry coefficient.

[0006] Furthermore, the process of obtaining the infiltrated edge region includes: The area corresponding to the sliding window of the comprehensive infiltration index that exceeds the preset threshold is taken as the infiltration edge area.

[0007] Furthermore, the process of obtaining the infiltration activity characteristic value includes: In the tumor DSA image, the vascular region in each infiltrative edge region is skeletonized to determine all skeleton pixels on the corresponding vascular skeleton; other pixels outside the skeleton pixels are used as reference pixels. The basic probability of vascular invasion is determined based on the spatial proximity of each reference pixel to all skeleton pixels. The corresponding invasive activity feature value is determined based on the mean of the basic vascular invasive probability of all reference pixels in each invasive edge region.

[0008] Furthermore, the process of obtaining the baseline vascular invasion probability includes: The baseline vascular invasion probability of each reference pixel is determined by negatively mapping the minimum Euclidean distance between each reference pixel and all skeleton pixels.

[0009] Furthermore, the process of obtaining the area of ​​interest includes: The wetting edge region corresponding to the wetting activity feature value that is greater than the preset activity threshold is regarded as the wetting region of interest.

[0010] Furthermore, the process of obtaining the vascular abnormality index includes: On the vascular skeleton of each region of interest, a curvature contrast value is determined based on the average curvature of all adjacent skeleton pixels of each skeleton pixel; and a local curvature deviation value of each skeleton pixel is determined based on the difference between the curvature of each skeleton pixel and the corresponding curvature contrast value. The mean value of the local curvature deviation of all skeleton pixels on the vascular skeleton of each region of interest is normalized to determine the vascular abnormality index of each region of interest.

[0011] Furthermore, the process of obtaining the vascular invasion coupling index includes: The product of the comprehensive infiltration index, infiltration activity characteristic value and vascular abnormality index of each infiltration region of interest is normalized to determine the corresponding vascular infiltration coupling index.

[0012] Furthermore, the process of obtaining the edge structure curve of the tumor region includes: In tumor CT images, the edge intensities of all pixels in the region of interest are arranged in descending order to determine the corresponding initial edge intensity sequence; the product of the positive correlation mapping value of the vascular invasion coupling index and the preset threshold number in the initial edge intensity sequence is rounded up to determine the number of edges selected for each region of interest. The edge intensity of the selected number of pixels in the edge intensity sequence is fitted with a B-spline curve to determine the edge structure curve of the tumor region.

[0013] Secondly, this application provides a computer device including a memory and a processor. The memory is used to store computer program code, and the processor is used to call and run the computer program code from the memory to execute a system as described in the first aspect of this application or any embodiment of the first aspect.

[0014] Thirdly, this application provides a computer program product, which includes computer program code that, when executed, performs a system as described in the first aspect of this application or any embodiment thereof.

[0015] Fourthly, this application provides a computer-readable storage medium that stores computer program code, which, when executed, performs a system as described in the first aspect of this application or any embodiment thereof.

[0016] This application has the following beneficial effects: This invention addresses the shortcomings of existing preoperative planning systems for interventional tumor surgery in identifying invasive edges. It initially screens invasive edge regions based on the asymmetry of grayscale distribution in CT images and the chaotic texture in MRI images using a sliding window. Further, it integrates DSA vascular data to analyze the location and distribution of blood vessels within the invasive edge regions, identifying areas of interest. Then, based on these areas, it analyzes curvature abrupt changes in tumor-associated vascular segments to generate a vascular abnormality index. Finally, it fuses the invasive index with vascular features to construct a vascular-invasive coupling index, which is used to correct the edge intensity map, achieving risk-based boundary optimization. This results in higher accuracy of the obtained tumor region edge structure curve, improving the accuracy of tumor region segmentation based on the tumor region edge structure curve. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a structural diagram of a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition, provided in one embodiment of the present invention. Figure 2 This is a schematic diagram of a computer device structure provided in one embodiment of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment, and specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific scheme for a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition provided by the present invention.

[0022] This application provides a preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition. Please refer to [link to relevant documentation]. Figure 1 The diagram shows a structural diagram of a preoperative planning system for tumor interventional therapy based on artificial intelligence image recognition, provided by an embodiment of the present invention. The system includes: a data acquisition and preprocessing module 101, a first determination module 102, a second determination module 103, and a region segmentation module 104.

[0023] The data acquisition and preprocessing module 101 is used to locate the same tumor ROI region in tumor CT images, tumor MRI images, and tumor DSA images; to traverse the tumor ROI region to determine all sliding windows; and to obtain the vascular region in the tumor DSA image based on angiography.

[0024] First, initial CT images of various tumor target organs acquired via CT scanners are obtained from a medical database. Then, a pre-trained U-Net++ segmentation model is used to define the tumor target region of interest (ROI) within all initial CT images. The initial CT images containing the tumor ROI are then used as tumor CT images. Next, tumor MRI images of the corresponding tumor target organs are acquired using a superconducting magnetic resonance imaging (MRI) device. Finally, digital subtraction angiography (DSA) images of the corresponding tumor target organs are acquired using the tumor CT images. During DSA, CT, and MRI scans, the patient's position is kept consistent to achieve spatial alignment of the images, ensuring that the same tumor ROI is located in the tumor CT, MRI, and DSA images. It should be noted that this embodiment only analyzes the tumor ROI corresponding to one tumor target organ's tumor CT, MRI, and DSA images. The analysis process is the same for other tumor target organs with corresponding ROIs and will not be elaborated further here.

[0025] Considering that this application needs to take into account the influence of the invasion edge, and the location of the invasion area is uncertain, it is necessary to use a sliding window for analysis. In a specific implementation of this invention, the sliding window is set to 5mm×5mm, which can be adjusted according to the specific implementation environment. In addition, since tumor DSA images are angiography images, the vascular region in the tumor DSA image can be directly obtained based on angiography.

[0026] The first determining module 102 is used to determine the comprehensive invasion index of each sliding window based on the asymmetry of gray-level distribution in the tumor CT image and the disorder of gray-level distribution in the tumor MRI image; and to screen out the invasion edge region according to the comprehensive invasion index.

[0027] In the tumor core area, cells proliferate densely, and the pixel grayscale value in CT images is significantly higher than that of normal tissue. Furthermore, the size and distribution of cells in the core area are relatively uniform, and the grayscale distribution tends to be symmetrical. In contrast, the infiltrated or necrotic areas exhibit higher asymmetry in grayscale distribution due to mixed or necrotic cells. The core pathological essence of tumor invasion is the disordered migration of proliferating cells into surrounding tissues; therefore, in MRI images, the grayscale texture distribution of the infiltrated area is usually quite chaotic. Therefore, this embodiment of the invention determines a comprehensive invasion index for each sliding window based on the asymmetry of grayscale distribution in the tumor CT image and the chaos of grayscale distribution in the tumor MRI image. This ensures that the larger the comprehensive invasion index, the greater the impact of invasion on the area corresponding to the sliding window.

[0028] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the comprehensive infiltration index includes: In the tumor CT image, the corresponding cell proliferation activity index is determined based on the overall size of the gray values ​​in each sliding window and the uniformity of the gray value distribution. In a specific implementation of this invention, the process of obtaining the cell proliferation activity index includes: normalizing the mean gray value of all pixels in each sliding window in the tumor CT image to determine the mean cell density coefficient; performing negative correlation normalization based on the skewness of the gray values ​​of all pixels in each sliding window to determine the density distribution symmetry coefficient; and determining the cell proliferation activity index of each sliding window based on the product of the mean cell density coefficient and the density distribution symmetry coefficient.

[0029] Compared to the tumor infiltration region, the tumor core region has a relatively higher cell density and overall higher grayscale. Therefore, the higher the average cell density coefficient, the higher the cell proliferation activity of the corresponding sliding window, and the lower the probability that it belongs to the tumor infiltration region; conversely, the lower the average cell density coefficient, the higher the probability that it belongs to the tumor infiltration region. In addition, the grayscale distribution of the tumor core region is close to equilibrium. Therefore, the greater the skewness of the grayscale values ​​of all pixels in the corresponding sliding window, the more asymmetrical the grayscale distribution, and the higher the probability that it belongs to the tumor core region, and the lower the probability that it belongs to the tumor infiltration region. Therefore, the higher the cell proliferation activity index obtained by combining the density distribution symmetry coefficient and the average cell density coefficient, the more significant the infiltration characteristics of the corresponding sliding window.

[0030] In the MRI image, the numerical entropy of the gray values ​​of all pixels in each sliding window is normalized to determine the corresponding tissue component mixing degree. Based on the principle of numerical entropy, the larger the numerical entropy of the gray values ​​of all pixels, the more chaotic the gray distribution. Therefore, when the tissue component mixing degree is greater, the infiltration characteristics of the corresponding sliding window are more significant.

[0031] Finally, based on the comprehensive correlation, the comprehensive invasion index of each sliding window is determined according to the product of the negative correlation mapping value of the cell proliferation activity index and the degree of tissue component mixing; so that the larger the comprehensive invasion index, the more likely the area corresponding to the sliding window is to belong to the invasion edge region.

[0032] In one specific implementation of this invention, the process of obtaining the comprehensive infiltration index is expressed by the following formula: ;in, For sliding windows The comprehensive infiltration index; For sliding windows Skewness of grayscale values ​​of all pixels in tumor CT images; It is a linear normalization function; For sliding windows The mean gray value of all pixels in a tumor CT image; For sliding windows The density distribution symmetry coefficient; For sliding windows The average cell density coefficient; For sliding windows The numerical entropy of the grayscale values ​​of all pixels in the MRI image; For sliding windows Cell proliferation activity index; For sliding windows The degree of mixture of tissue components.

[0033] The tumor core area exhibits dense cell proliferation but low invasive activity, with a clear edge structure. In contrast, the invasive edge region of the tumor shows high cell invasive activity, serving as a transition zone between the tumor and normal tissue, where the edge structure is prone to blurring and requires close attention. Further, invasive edge regions are screened based on the comprehensive invasive index. Since a higher comprehensive invasive index generally indicates a higher likelihood that the corresponding sliding window area belongs to the invasive edge region, preferably, in some possible implementations of this invention, the process of obtaining the invasive edge region includes: using the area corresponding to the sliding window with a comprehensive invasive index greater than a preset threshold as the invasive edge region. In a specific implementation of this invention, the preset threshold is set to 0.7, which can be adjusted according to the specific implementation environment.

[0034] The second determining module 103 is used to determine the corresponding invasive activity feature value based on the location distribution of the vascular region in each invasive edge region of the tumor DSA image; to screen out the invasive regions of interest based on the invasive activity feature value; and to determine the corresponding vascular abnormality index based on the tortuosity of the vascular region in the invasive regions of interest.

[0035] Tumor cells exhibit "angiotropy," typically migrating along perivascular spaces during cell invasion. This means that tumor cells around blood vessels in the tumor region experience more severe infiltration. Therefore, the corresponding infiltration activity characteristic value can be determined first by analyzing the location and distribution of blood vessels in each infiltration margin region in the tumor DSA image. The infiltration activity characteristic value can then be used to characterize the influence of blood vessel infiltration on the corresponding infiltration margin region.

[0036] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the infiltration activity characteristic value includes: In the tumor DSA image, the vascular region in each infiltrative edge region is skeletonized to determine all skeleton pixels on the corresponding vascular skeleton; other pixels outside the skeleton pixels are used as reference pixels; based on the spatial proximity of each reference pixel to all skeleton pixels, the corresponding basic vascular invasion probability is determined; preferably, in a specific implementation of the present invention, the process of obtaining the basic vascular invasion probability includes: The baseline vascular invasion probability of each reference pixel is determined by negatively correlated mapping of the minimum Euclidean distance between each reference pixel and all skeleton pixels. For each reference pixel, the smaller the distance between it and the blood vessel, the greater the invasion impact. In a specific implementation of this invention, the method for negatively correlated mapping of the minimum Euclidean distance between each reference pixel and all skeleton pixels is as follows: the linearly normalized value of the minimum Euclidean distance between each reference pixel and all skeleton pixels is used as a distance parameter; the difference between the real number 1 and this distance parameter is used as the output result of the negative correlation mapping, that is, the corresponding baseline vascular invasion probability.

[0037] Furthermore, by combining the basic vascular invasion probabilities of all reference pixels, the corresponding invasion activity feature value is determined based on the average of the basic vascular invasion probabilities of all reference pixels in each invasion edge region. This ensures that the larger the invasion activity feature value, the greater the impact of vascular invasion on the corresponding sliding window. Further, to facilitate subsequent analysis, the invasion region of interest is further screened based on the invasion activity feature value. Preferably, in some possible implementations of this invention, the process of obtaining the invasion region of interest includes: identifying the invasion edge region corresponding to an invasion activity feature value greater than a preset activity threshold as the invasion region of interest. In a specific implementation of this invention, the preset activity threshold is set to 0.6, which can be adjusted according to the specific implementation environment and will not be further elaborated here.

[0038] Furthermore, for vascular regions within the tumor ROI, significant variations in vascular curvature can lead to uneven stress on the vessel walls, resulting in increased permeability and facilitating cell penetration. This exacerbates the invasive influence of the corresponding vascular segment on tumor cells within the region of interest. Therefore, this embodiment of the invention further determines a corresponding vascular abnormality index based on the curvature of the vascular regions within the region of interest. A higher vascular abnormality index indicates a greater impact of tumor cell invasiveness on the corresponding region of interest in terms of vascular curvature, and a more severe degree of edge blurring in the corresponding region of interest.

[0039] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the vascular abnormality index includes: On the vascular skeleton of each region of interest, a curvature comparison value is determined based on the average curvature of all adjacent skeleton pixels. A local curvature deviation value is determined based on the difference between the curvature of each skeleton pixel and its corresponding curvature comparison value. The average local curvature deviation values ​​of all skeleton pixels on the vascular skeleton of each region of interest are normalized to determine the vascular abnormality index for each region of interest. In one specific implementation of this invention, the method for normalizing the average local curvature deviation values ​​of all skeleton pixels on the vascular skeleton of each region of interest employs linear normalization, which will not be further elaborated here.

[0040] For each skeleton pixel, the greater the curvature deviation between it and all its adjacent skeleton pixels, the more significant the tortuous change characteristics at the location of the corresponding blood vessel. This means that when the vascular abnormality index, which is determined by combining the mean of the local curvature deviation values ​​of all skeleton pixels, is larger, the vascular tortuous change characteristics are more obvious, and the corresponding area of ​​interest is more significantly affected by tumor cell infiltration and invasion.

[0041] The region segmentation module 104 is used to determine the corresponding vascular invasion coupling index based on the comprehensive invasion index, invasion activity characteristic value and the vascular abnormality index of each region of interest; to perform edge intensity map correction on the tumor ROI region based on the vascular invasion coupling index to determine the corresponding tumor region edge structure curve; and to perform tumor region segmentation based on the tumor region edge structure curve.

[0042] For each region of interest, the larger the corresponding comprehensive infiltration index, the larger the infiltration activity characteristic value, and the larger the vascular abnormality index, the greater the infiltration influence on the corresponding region of interest. The more blurred the corresponding edge is due to the infiltration influence, the more necessary it is to capture the blurred edge and select the edge line.

[0043] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the vascular invasion coupling index is expressed by the following formula: ;in, To focus on the infiltrated area The vascular infiltration coupling index; To focus on the infiltrated area The comprehensive infiltration index; To focus on the infiltrated area The characteristic value of infiltration activity; To focus on the infiltrated area Vascular abnormality index; It is a linear normalization function.

[0044] Further, in this embodiment of the invention, the edge intensity map of the tumor ROI region is corrected based on the vascular invasion coupling index to determine the corresponding tumor region edge structure curve; wherein, the process of obtaining the tumor region edge structure curve includes: In tumor CT images, the edge intensity of all pixels in the infiltrative region is arranged in descending order to determine the corresponding initial edge intensity sequence; that is, the edge intensity of all pixels in the infiltrative region is arranged in descending order to determine the corresponding initial edge intensity sequence. Existing techniques for determining edges based on edge intensity sorting typically set a threshold to filter out pixels with higher edge intensity, and then perform B-spline curve fitting on these pixels to determine the final edge line. However, if the threshold is set inappropriately, weak edges, such as those corresponding to infiltrative edges, may be ignored.

[0045] For each region of interest in invasion, the larger the corresponding vascular invasion coupling index, the more blurred the corresponding invasion edge is usually, and therefore the more edges should be selected to identify the invasion edge. Therefore, the product of the positive correlation mapping value of the vascular invasion coupling index and the preset threshold number in the initial edge intensity sequence is rounded up to determine the number of edges selected for each region of interest in invasion. In a specific implementation of this invention, the process of obtaining the preset threshold number includes: performing edge detection on the tumor ROI region to screen out all strong edge pixels; using the ratio between the number of all strong edge pixels and the total number of pixels in the tumor ROI region as a reference edge proportion coefficient; rounding up the product of the reference edge proportion coefficient and the total number of pixels in each region of interest in invasion to determine the corresponding preset threshold number; that is, determining the preset threshold number representing the strong edge threshold in the overall region of interest in invasion. Wherein, when the calculated preset threshold number is greater than 10% of the corresponding number of regions of interest in invasion, the value obtained by rounding up 20% of the number of regions of interest in invasion is used as the corresponding preset threshold number for subsequent calculations.

[0046] In one specific implementation of this invention, the process of obtaining the number of edge selections includes: ;in, To focus on the infiltrated area The number of edges selected; To focus on the infiltrated area The vascular infiltration coupling index; To focus on the infiltrated area The preset threshold number. Further, based on the principle of strong edge selection, the pixels corresponding to the edge intensities of the selected number of leading edges in the edge intensity sequence are subjected to B-spline curve fitting to determine a more accurate tumor region edge structure curve.

[0047] For areas outside the invasive region within the tumor ROI, these areas are less affected by invasion and their edges are usually more distinct. Therefore, Canny edge detection is directly performed on these areas to determine the corresponding edge structure curves of all tumor regions. Finally, tumor regions are segmented based on the edge structure curves of the tumor regions, resulting in higher accuracy of the segmented tumor regions.

[0048] In summary, the preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition addresses the insufficient ability of existing systems to identify invasive edges. It initially screens invasive edge regions based on the asymmetry of grayscale distribution in CT images and the chaotic texture in MRI images using a sliding window. Further, it integrates DSA vascular data to analyze the location and distribution of blood vessels within the invasive edge regions, identifying areas of interest. Then, based on these areas, it analyzes curvature abrupt changes in tumor-associated vascular segments to generate a vascular abnormality index. Finally, it integrates invasive indicators and vascular features to construct a vessel-invasive coupling index, which is used to correct the edge intensity map, achieving risk-based boundary optimization. This results in higher accuracy of the obtained tumor region edge structure curves and improves the accuracy of tumor region segmentation based on these curves.

[0049] This application also provides a computer device; please refer to [link / reference]. Figure 2 The illustration shows a schematic diagram of a computer device structure according to an embodiment of the present invention. The computer device includes a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202. When the processor 202 executes the computer program 203, the computer device can execute any of the aforementioned preoperative planning systems for tumor interventional therapy based on artificial intelligence image recognition.

[0050] This application also provides a computer program product that, when run on a computer device, enables the computer device to execute any of the aforementioned preoperative planning systems for interventional tumor treatment based on artificial intelligence image recognition.

[0051] This application also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer device, the computer device can execute any of the aforementioned preoperative planning systems for interventional tumor treatment based on artificial intelligence image recognition.

[0052] In the embodiments provided in this application, it should be understood that the computer device, computer program product and computer-readable storage medium provided are all used to execute the corresponding system provided above, and therefore the beneficial effects they can achieve can be referred to the beneficial effects of the system provided above, which will not be repeated here.

[0053] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0054] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition, characterized in that, The system includes: The data acquisition and preprocessing module is used to locate the same tumor ROI region in tumor CT images, tumor MRI images, and tumor DSA images; to traverse the tumor ROI region to determine all sliding windows; and to obtain the vascular region in the tumor DSA image based on angiography. The first determining module is used to determine the comprehensive invasion index of each sliding window based on the asymmetry of gray-level distribution in the tumor CT image and the disorder of gray-level distribution in the tumor MRI image; and to screen out the invasion edge region according to the comprehensive invasion index. The second determining module is used to determine the corresponding invasive activity feature value based on the location distribution of the vascular region in each invasive edge region of the tumor DSA image; to screen out the invasive regions of interest based on the invasive activity feature value; and to determine the corresponding vascular abnormality index based on the tortuosity of the vascular region in the invasive regions of interest. The region segmentation module is used to determine the corresponding vascular invasion coupling index based on the comprehensive invasion index, invasion activity characteristic value, and vascular abnormality index of each region of interest; to perform edge intensity map correction on the tumor ROI region based on the vascular invasion coupling index to determine the corresponding tumor region edge structure curve; and to perform tumor region segmentation based on the tumor region edge structure curve.

2. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the comprehensive infiltration index includes: In the tumor CT image, the corresponding cell proliferation activity index is determined based on the overall size of the gray values ​​in each sliding window and the uniformity of the gray value distribution. In the MRI image, the numerical entropy of the gray values ​​of all pixels in each sliding window is normalized to determine the corresponding tissue component mixing degree; The comprehensive infiltration index for each sliding window is determined by multiplying the negative correlation mapping value of the cell proliferation activity index with the degree of tissue component mixing.

3. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 2, characterized in that, The process of obtaining the cell proliferation activity index includes: In the tumor CT image, the mean gray value of all pixels in each sliding window is normalized to determine the mean cell density coefficient; the density distribution symmetry coefficient is determined by negative correlation normalization based on the skewness of the gray values ​​of all pixels in each sliding window; and the cell proliferation activity index of each sliding window is determined by the product of the mean cell density coefficient and the density distribution symmetry coefficient.

4. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the infiltrated edge region includes: The area corresponding to the sliding window of the comprehensive infiltration index that exceeds the preset threshold is taken as the infiltration edge area.

5. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the infiltration activity characteristic value includes: In the tumor DSA image, the vascular region in each infiltrative edge region is skeletonized to determine all skeleton pixels on the corresponding vascular skeleton; other pixels outside the skeleton pixels are used as reference pixels. The basic probability of vascular invasion is determined based on the spatial proximity of each reference pixel to all skeleton pixels. The corresponding invasive activity feature value is determined based on the mean of the basic vascular invasive probability of all reference pixels in each invasive edge region.

6. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 5, characterized in that, The process of obtaining the baseline vascular invasion probability includes: The baseline vascular invasion probability of each reference pixel is determined by negatively mapping the minimum Euclidean distance between each reference pixel and all skeleton pixels.

7. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the area of ​​interest includes: The wetting edge region corresponding to the wetting activity feature value that is greater than the preset activity threshold is regarded as the wetting region of interest.

8. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 5, characterized in that, The process of obtaining the vascular abnormality index includes: On the vascular skeleton of each region of interest, a curvature contrast value is determined based on the average curvature of all adjacent skeleton pixels of each skeleton pixel; and a local curvature deviation value of each skeleton pixel is determined based on the difference between the curvature of each skeleton pixel and the corresponding curvature contrast value. The mean value of the local curvature deviation of all skeleton pixels on the vascular skeleton of each region of interest is normalized to determine the vascular abnormality index of each region of interest.

9. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the vascular invasion coupling index includes: The product of the comprehensive infiltration index, infiltration activity characteristic value and vascular abnormality index of each infiltration region of interest is normalized to determine the corresponding vascular infiltration coupling index.

10. The preoperative planning system for interventional tumor treatment based on artificial intelligence image recognition according to claim 1, characterized in that, The process of obtaining the tumor region edge structure curve includes: In tumor CT images, the edge intensities of all pixels in the region of interest are arranged in descending order to determine the corresponding initial edge intensity sequence; the product of the positive correlation mapping value of the vascular invasion coupling index and the preset threshold number in the initial edge intensity sequence is rounded up to determine the number of edges selected for each region of interest. The edge intensity of the selected number of pixels in the edge intensity sequence is fitted with a B-spline curve to determine the edge structure curve of the tumor region.

Images