Intracranial aneurysm assessment method and system based on deep learning

Abstract

The present invention provides an intracranial aneurysm assessment method and system based on deep learning, wherein the method includes: receiving an initial angiography image sent by a medical terminal and identifying it, determining all aneurysms in the initial angiography image; assessing the rupture risk of each aneurysm based on a preset risk assessment strategy, determining all aneurysms with high risk as a primary risk set, and determining all aneurysms with low risk as a secondary risk set; removing each aneurysm in the secondary risk set from the initial angiography image to obtain a current angiography image, and performing image segmentation on the angiography image to obtain each segmented sub-image; respectively determining the number of aneurysms distributed in each segmented sub-image, and respectively determining the assessment processing method corresponding to each segmented sub-image based on the comparison result of the number of aneurysms corresponding to each segmented sub-image and the preset number. The present invention at least improves the efficiency of medical diagnosis assistance.

Description

Technical Field

[0001] The present invention relates to data processing technology, and in particular to a method and system for evaluating intracranial aneurysms based on deep learning. Background Art

[0002] Intracranial aneurysms are mostly abnormal bulges on the walls of intracranial arteries. They are the leading cause of subarachnoid hemorrhage and the third most common cause of cerebrovascular accidents, after cerebral thrombosis and hypertensive cerebral hemorrhage. They can occur at any age, but are more common in middle-aged and elderly women between 40 and 60 years old. For the treatment of intracranial aneurysms, a clear diagnosis must first be made through cerebral angiography, followed by early surgery and interventional embolization, because once an intracranial aneurysm ruptures, the mortality rate is extremely high.

[0003] The inventors found in their research that the existing method for evaluating intracranial aneurysms is often based on medical staff's viewing of angiography images, which is somewhat subjective and lacks a corresponding evaluation system to assist. Summary of the invention

[0004] Based on the above problems, the present invention is proposed to provide a method and system for evaluating intracranial aneurysms based on deep learning that overcomes the above problems or at least partially solves the above problems.

[0005] According to one aspect of the present invention, a method for evaluating intracranial aneurysms based on deep learning is provided, comprising the following steps:

[0006] receiving an initial angiography image sent by the medical terminal, and identifying the initial angiography image to determine all aneurysms distributed in the initial angiography image;

[0007] The rupture risk of each aneurysm is assessed based on a preset risk assessment strategy, and all aneurysms with high risk are identified as a primary risk set, and all aneurysms with low risk are identified as a secondary risk set;

[0008] Eliminate each aneurysm in the secondary risk set from the initial angiography image to obtain a current angiography image, and perform image segmentation on the current angiography image to obtain segmented sub-images;

[0009] The number of aneurysm bodies distributed in each divided sub-image is determined respectively, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image and a preset number, the evaluation processing method corresponding to each divided sub-image is determined respectively.

[0010] Optionally, in the method according to the present invention, the rupture risk of each aneurysm is evaluated based on a preset risk assessment strategy, and all aneurysms with high risk are determined as a primary risk set, and all aneurysms with low risk are determined as a secondary risk set, including:

[0011] Taking any image endpoint located in the initial angiography image as a reference point, and establishing an image coordinate system corresponding to the initial angiography image with the reference point as the origin;

[0012] Obtaining the aneurysm pixel points that constitute each aneurysm respectively, and determining the aneurysm pixel points corresponding to the same aneurysm as a aneurysm pixel set, to obtain the aneurysm pixel sets corresponding to each aneurysm;

[0013] In a direction parallel to the X-axis, each first transverse coordinate point with a maximum transverse coordinate value and each second transverse coordinate point with a minimum transverse coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate points and the second transverse coordinate points located in the same tumor pixel set are respectively connected by line segments to obtain a first comparison line segment;

[0014] In the direction parallel to the Y axis, each first longitudinal coordinate point with the maximum longitudinal coordinate value and each second longitudinal coordinate point with the minimum longitudinal coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate point and the second transverse coordinate point located in the same tumor pixel set are respectively connected by line segments to obtain a second comparison line segment;

[0015] Determine the one with the longer line segment value between the first comparison line segment and the second comparison line segment corresponding to the same aneurysm as a reference line segment, and obtain the reference line segment value of the reference line segment;

[0016] Comparing each baseline segment value corresponding to each aneurysm with a preset length, and determining an aneurysm with a corresponding baseline segment value greater than or equal to the preset length as having a high risk, and determining an aneurysm with a corresponding baseline segment value less than the preset length as having a low risk;

[0017] All aneurysms with high risk were identified as the primary risk group, and all aneurysms with low risk were identified as the secondary risk group.

[0018] Optionally, in the method according to the present invention, removing each aneurysm located in the secondary risk set from the initial angiography image to obtain a current angiography image comprises:

[0019] Determine each aneurysm contour corresponding to each aneurysm located in the secondary risk set, and determine each first intersection point and each second intersection point where each aneurysm contour intersects with the blood vessel contour;

[0020] Connect the first intersection point and the second intersection point corresponding to the same tumor contour to obtain the connection line of each tissue;

[0021] Determine the region consisting of the aneurysm contour and the tissue connection line corresponding to the same aneurysm as the rejection region corresponding to the aneurysm, and reject each rejection region corresponding to each aneurysm in the secondary risk set from the initial angiography image to obtain an initial angiography image including each blank region;

[0022] Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image.

[0023] Optionally, in the method according to the present invention, performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes:

[0024] Obtaining each contour pixel point constituting the tumor contour corresponding to the same removal area;

[0025] Acquire each first adjacent pixel point that is adjacent to each contour pixel point and is outside the removal area in the initial angiography image, and determine each first pixel value corresponding to each first adjacent pixel point;

[0026] Performing mean calculation based on each first pixel value to obtain a first pixel mean, and performing pixel value configuration on the blank area based on the first pixel mean;

[0027] Determine one of the first intersection point and the second intersection point corresponding to the same removal area as a reference intersection point, and obtain the intersection pixel points constituting the reference intersection point;

[0028] Acquire each second adjacent pixel point that is adjacent to each intersecting pixel point and located on the blood vessel contour in the initial angiography image, and determine each second pixel value corresponding to each second adjacent pixel point;

[0029] The second pixel mean is obtained by performing mean calculation based on each second pixel value, and the tissue connection line located in the blank area is reconfigured with pixel values ​​based on the second pixel mean to obtain the current angiography image.

[0030] Optionally, in the method according to the present invention, performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes:

[0031] Acquire a first interference layer having the same size as the initial angiography image, and overlay the first interference layer on the initial angiography image;

[0032] Determine the regional pixel points that respectively constitute each blank area, and determine the regional pixel points corresponding to the same blank area as a regional pixel set, to obtain different regional pixel sets corresponding to different blank areas;

[0033] Based on the pixel points of each region in the corresponding pixel sets of different regions, determining the interference regions corresponding to the blank regions in the first interference layer, and extracting the regions outside the interference regions in the first interference layer to obtain the second interference layer;

[0034] The second interference layer is fused with the initial angiography image to obtain a current angiography image.

[0035] Optionally, in the method according to the present invention, corresponding to the comparison result of the number of tumors in each divided sub-image and the preset number, determining the evaluation processing mode corresponding to each divided sub-image respectively includes:

[0036] Obtaining a blood vessel corresponding to each tumor in the divided sub-image, and if it is determined that one blood vessel corresponds to multiple tumors, obtaining multiple edge tumors in the blood vessel, determining blood vessel attribute information based on the edge tumors, and processing the divided sub-image based on the number of tumors, blood vessel attribute information, and morphological attributes of each tumor to determine whether it meets the overall processing conditions or meets the individual processing conditions;

[0037] Generate a first initial processing model based on the divided sub-images that meet the overall processing conditions, and simulate the first initial processing model based on a preset overall processing method to obtain a first current processing model;

[0038] The morphological attributes of each aneurysm located in the divided sub-image with the individual processing conditions are obtained, the type of each aneurysm is determined based on the corresponding morphological attributes, and the evaluation processing method corresponding to each aneurysm is determined based on the determination result.

[0039] Optionally, in the method according to the present invention, a blood vessel corresponding to each tumor in the divided sub-image is obtained, and if it is determined that one blood vessel corresponds to multiple tumors, multiple marginal tumors in the blood vessel are obtained, and blood vessel attribute information is determined based on the marginal tumors. Based on the number of tumors, the blood vessel attribute information and the morphological attributes of each tumor, the divided sub-image is processed to determine whether it meets the overall processing conditions or meets the individual processing conditions, including:

[0040] Obtaining the coordinate point of each blood vessel in the divided sub-image after coordinate processing to obtain a blood vessel coordinate set, and obtaining the coordinate point of each tumor in the divided sub-image after coordinate processing to obtain a tumor coordinate set;

[0041] Compare each tumor coordinate set with the blood vessel coordinate set, determine the tumor coordinate sets intersecting with the blood vessel coordinate set as a set group, and use the tumors corresponding to the tumor coordinate sets in the set group as the tumors of the blood vessels in the blood vessel coordinate set;

[0042] Determine edge tumor and blood vessel attribute information based on the coordinate information of the blood vessels corresponding to each tumor, and obtain the morphological attributes of each tumor based on the coordinate information of each tumor;

[0043] Based on the number of tumors, vascular attribute information, and the morphological attributes of each tumor, the divided sub-images are processed to determine whether they meet the overall processing conditions or the individual processing conditions.

[0044] Optionally, in the method according to the present invention, determining the edge tumor and blood vessel attribute information based on the coordinate information of the blood vessels corresponding to each tumor, and obtaining the morphological attributes of each tumor based on the coordinate information of each tumor, includes:

[0045] Obtaining the longitudinal coordinates in the coordinate information of the blood vessels corresponding to each tumor, and sorting all the tumors in descending order based on the longitudinal coordinates to obtain a corresponding tumor sequence;

[0046] The first tumor and the last tumor in the tumor sequence are extracted as marginal tumors;

[0047] Obtaining the center coordinate points of two edge tumors, calculating the length of a straight line of the center coordinate points based on the center coordinate points, and using the length of the straight line as blood vessel attribute information;

[0048] The maximum vertical coordinate difference of each edge tumor under each same horizontal coordinate is obtained to obtain height information, and the maximum horizontal coordinate difference of each edge tumor under each same vertical coordinate is obtained to obtain width information. The morphological attributes include height information and width information.

[0049] Optionally, in the method according to the present invention, the processing of the divided sub-images based on the number of tumors, the vascular attribute information and the morphological attributes of each tumor to determine whether the sub-images meet the overall processing conditions or meet the individual processing conditions includes:

[0050] The tumor quantity and vascular attribute information are compared and then weighted to obtain a tumor vascular coefficient value, the morphology of each tumor is calculated and then weighted to obtain a tumor morphology coefficient value, and the tumor processing coefficient value is obtained based on the tumor vascular coefficient value and the tumor morphology coefficient value;

[0051] If the tumor treatment coefficient value is greater than or equal to the preset coefficient value, it is determined that the overall treatment conditions are met; if the tumor treatment coefficient value is less than the preset coefficient value, it is determined that the individual treatment conditions are met;

[0052] The actual operation data of the surgery is obtained. If the actual operation data does not correspond to the outputted overall processing conditions or the outputted individual processing conditions, the weights are updated and trained.

[0053] Optionally, in the method according to the present invention, the steps of calculating the tumor treatment coefficient value and updating the weights through the following training processes include:

[0054] The tumor treatment coefficient value was calculated by the following formula:

[0055]

[0056] Among them, x is the tumor treatment coefficient value, s is the number of tumors, is the horizontal coordinate of the center coordinate point of the first edge tumor, is the horizontal coordinate of the center coordinate point of the second edge tumor, is the ordinate of the center coordinate point of the first edge tumor, is the ordinate of the center coordinate point of the second edge tumor, k s is the tumor vascular coefficient weight, g i is the height information of the ith tumor, d i is the width information of the i-th tumor, n is the upper limit of the number of tumors, k u is the tumor morphology coefficient weight.

[0057] Optionally, in the method according to the present invention, the actual operation data of the surgery is obtained, and if the actual operation data does not correspond to the outputted overall processing condition or the outputted individual processing condition, the weight is updated and trained, including:

[0058] If the actual operation data meets the individual processing conditions and the output meets the overall processing conditions, then the tumor vascular coefficient weight k s and tumor morphology coefficient weight k u Reduce training;

[0059] If the actual operation data meets the overall processing conditions and the output meets the individual processing conditions, then the tumor vascular coefficient weight k s and tumor morphology coefficient weight k u Increase training;

[0060] The trained tumor vascular coefficient weight k is obtained by the following formula w and tumor morphology coefficient weight k e ,

[0061]

[0062] Among them, q + is the tumor vascular coefficient weight ks Increase the number of training times, ∈ is the training constant value of the tumor vascular coefficient weight, v + is the tumor morphology coefficient weight k u Increase the number of training times, δ is the training constant value of the tumor treatment coefficient weight, q - is the tumor vascular coefficient weight k s Reduce the number of training times, v - is the tumor morphology coefficient weight k u Reduce the number of training sessions.

[0063] Optionally, in the method according to the present invention, generating a first initial processing model based on the divided sub-images having the overall processing conditions, and simulating the first initial processing model based on a preset overall processing method to obtain a first current processing model includes:

[0064] Performing three-dimensional modeling based on the divided sub-images having the overall processing conditions to obtain a first initial processing model corresponding to the divided sub-images;

[0065] Determining in the first initial processing model the aneurysm model parts and the blood vessel model parts corresponding to the blood vessels, and determining the blood vessel model parts that are connected to the aneurysm model parts as a blood vessel clipping group based on the relative position relationship between the aneurysm model parts and the blood vessel model parts;

[0066] Performing color configuration on each blood vessel model portion in the blood vessel clamping group based on a preset first processing color, and generating a first blood vessel clamping model and a second blood vessel clamping model surrounding the blood vessel model portion at a first model point and a second model point of each blood vessel model portion, to obtain a first current processing model;

[0067] There is no tumor model part in the part where the first model point of the blood vessel model part is far away from the second model point, and there is no tumor model part in the part where the second model point of the blood vessel model part is far away from the first model point.

[0068] Optionally, in the method according to the present invention, the morphological attributes of each aneurysm located in the divided sub-image with the separate processing conditions are obtained, the type of each aneurysm is determined based on the corresponding morphological attributes, and the evaluation processing method corresponding to each aneurysm is determined based on the determination result, including:

[0069] Performing three-dimensional modeling based on the divided sub-images with separate processing conditions to obtain a second initial processing model corresponding to the divided sub-images;

[0070] Determining in the second initial processing model each tumor model part including the aneurysm, wherein each tumor model part includes a tumor body model and a tumor connection model;

[0071] Determine a model height and a model width of the tumor connection model corresponding to each tumor model part, and determine a morphological attribute of each tumor model part based on the model height and the model width, wherein the morphological attribute includes a low-risk attribute and a high-risk attribute;

[0072] Based on the morphological attributes, an evaluation and processing method corresponding to each aneurysm is determined, and based on the determined evaluation and processing method, a simulation process is performed on the second initial processing model to obtain a second current processing model.

[0073] Optionally, in the method according to the present invention, determining a model height and a model width of a tumor connection model corresponding to each tumor model portion, and determining a morphological attribute of each tumor model portion based on the model height and the model width, comprises:

[0074] When the model width is greater than or equal to the preset width, the tumor model portion corresponding to the model width is divided into the primary risk assessment group, and the morphological attribute of the tumor model portion corresponding to the model width less than the preset width is determined as a high-risk attribute;

[0075] Determine the comparison results between the model heights corresponding to each tumor model part in the primary risk assessment group and the preset height, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height greater than or equal to the preset height as high-risk attributes, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height less than the preset height as low-risk attributes.

[0076] Optionally, in the method according to the present invention, determining the evaluation processing methods corresponding to the respective aneurysms based on the morphological attributes, and simulating the second initial processing model based on the determined evaluation processing methods to obtain the second current processing model, comprises:

[0077] Performing color configuration on the tumor model portion corresponding to the high-risk attribute based on the preset second processing color, and obtaining the model specification of the tumor body model in the tumor model portion corresponding to the tumor model portion;

[0078] generating an embolization model filling the tumor body model based on the model specification;

[0079] The tumor model portion corresponding to the low-risk attribute is color configured based on the preset third processing color, and a tumor clamping model surrounding the tumor connection model is generated based on the model height and model width of the tumor connection model in the corresponding tumor model portion to obtain a second current processing model.

[0080] According to another aspect of the present invention, there is provided an intracranial aneurysm assessment system based on deep learning, comprising:

[0081] an identification module configured to receive an initial angiography image sent by the medical terminal, identify the initial angiography image, and determine all aneurysms distributed in the initial angiography image;

[0082] An assessment module is configured to assess the rupture risk of each aneurysm based on a preset risk assessment strategy, and to identify all aneurysms with high risk as a primary risk set and all aneurysms with low risk as a secondary risk set;

[0083] a segmentation module configured to remove each aneurysm located in the secondary risk set from the initial angiography image to obtain a current angiography image, and perform image segmentation on the current angiography image to obtain segmented sub-images;

[0084] The determination module is configured to respectively determine the number of aneurysm bodies distributed in each divided sub-image, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image with a preset number, respectively determine the evaluation processing method corresponding to each divided sub-image.

[0085] According to the solution of the present invention, the server first receives the initial angiography image sent by the medical end, and identifies the initial angiography image, thereby determining all aneurysms distributed in the initial angiography image; then, based on the preset risk assessment strategy, the rupture risk of each aneurysm is assessed, and all aneurysms with high risk are determined as a first-level risk set, and all aneurysms with low risk are determined as a second-level risk set; then, each aneurysm in the second-level risk set is removed from the initial angiography image, thereby obtaining the current angiography image, and the angiography image is image-divided to obtain each divided sub-image; then, the number of aneurysms distributed in each divided sub-image is determined respectively, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image and the preset number, the evaluation processing method corresponding to each divided sub-image is determined respectively. The present invention can assess the rupture risk of each aneurysm and determine the corresponding evaluation processing method, with fast speed and certain accuracy, thereby providing medical staff with a certain reference and improving the corresponding medical diagnosis auxiliary efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 A flowchart of a method for evaluating intracranial aneurysms based on deep learning according to an embodiment of the present invention is shown;

[0087] Figure 2-Figure 5The evaluation and processing methods of different aneurysm types in this embodiment are shown;

[0088] Figure 6 A structural block diagram of an intracranial aneurysm assessment system based on deep learning according to another embodiment of the present invention is shown. DETAILED DESCRIPTION

[0089] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0090] Intracranial aneurysms are mostly abnormal bulges on the walls of intracranial arteries. They are the leading cause of subarachnoid hemorrhage and the third most common cause of cerebrovascular accidents, after cerebral thrombosis and hypertensive cerebral hemorrhage. They can occur at any age, but are more common in middle-aged and elderly women between 40 and 60 years old. For the treatment of intracranial aneurysms, a clear diagnosis must first be made through cerebral angiography, followed by early surgery and interventional embolization, because once an intracranial aneurysm ruptures, the mortality rate is extremely high.

[0091] The inventors found in their research that the existing method for evaluating intracranial aneurysms is often based on medical staff's viewing of angiography images, which is somewhat subjective and lacks a corresponding evaluation system to assist.

[0092] In order to solve the problems existing in the above-mentioned prior art, the inventor proposes the solution of the present invention. One embodiment of the present invention provides an intracranial aneurysm assessment method based on deep learning, which can be executed in a computing device.

[0093] Figure 1 A flowchart of a deep learning-based intracranial aneurysm assessment method according to an embodiment of the present invention is shown, and the method is suitable for execution in a computing device.

[0094] like Figure 1 As shown, the purpose of this embodiment is to implement a method for evaluating intracranial aneurysms based on deep learning, starting from step S102. In step S102, the following steps are included:

[0095] An initial angiography image sent by the medical terminal is received, and the initial angiography image is identified to determine all aneurysms distributed in the initial angiography image.

[0096] For example, in this embodiment, the medical end can be understood as a terminal device for medical staff to interact, such as a mobile phone, a computer, or a head-mounted display device, etc. Here, the server first receives the initial angiography image sent by the medical end, and identifies the received initial angiography image, thereby determining all aneurysms distributed in the initial angiography image.

[0097] It should be noted that the initial angiography image can be understood as being obtained by performing corresponding angiography through X-ray equipment, that is, medical staff can make corresponding medical diagnoses based on the corresponding angiography images; and in this embodiment, the processing of the initial angiography image can include corresponding identification, and the purpose of the identification is to determine all aneurysms distributed in the initial angiography image, thereby assisting medical staff in making corresponding medical diagnoses and providing certain diagnostic references.

[0098] Here, the process of identifying the initial angiography image can be implemented based on the corresponding machine learning model; for example, first, a machine learning model can be pre-built, and corresponding training images can be collected to form a corresponding training collection, wherein the training images include multiple angiography images including different aneurysms, and the aneurysms are correspondingly labeled in each training image; then, after completing the collection of the training collection, the angiography images in the training collection can be input into the machine learning model for repeated training, so as to obtain an image recognition model after training; finally, the initial angiography image in this embodiment can be input into the image recognition model, thereby completing the labeling of all aneurysms distributed in the initial angiography image.

[0099] In step S104, the following contents are included:

[0100] The rupture risk of each aneurysm is assessed based on the preset risk assessment strategy, and all aneurysms with high risk are identified as the first-level risk set, and all aneurysms with low risk are identified as the second-level risk set.

[0101] For example, in this embodiment, the server will evaluate the rupture risk of each aneurysm according to a preset risk assessment strategy. When the possibility of an aneurysm rupturing in a short period of time is high, the server will evaluate the aneurysm as having a high risk, and determine all aneurysms with a high risk as a first-level risk set; when the possibility of an aneurysm rupturing in a long period of time is low, the server will evaluate the aneurysm as having a low risk, and determine all aneurysms with a low risk as a second-level risk set; wherein, the basis for risk assessment is mainly based on the size specification of each aneurysm, that is, the larger the size specification, the greater the rupture risk of the corresponding aneurysm, and the smaller the size specification, the smaller the rupture risk of the corresponding aneurysm.

[0102] Furthermore, in this embodiment, the above-mentioned “assessing the rupture risk of each aneurysm based on a preset risk assessment strategy, and determining all aneurysms with high risk as a primary risk set, and determining all aneurysms with low risk as a secondary risk set” may also include the following steps:

[0103] Taking any image endpoint located in the initial angiography image as a reference point, and establishing an image coordinate system corresponding to the initial angiography image with the reference point as the origin;

[0104] Obtaining the aneurysm pixel points that constitute each aneurysm respectively, and determining the aneurysm pixel points corresponding to the same aneurysm as a aneurysm pixel set, to obtain the aneurysm pixel sets corresponding to each aneurysm;

[0105] In a direction parallel to the X-axis, each first transverse coordinate point with a maximum transverse coordinate value and each second transverse coordinate point with a minimum transverse coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate points and the second transverse coordinate points located in the same tumor pixel set are respectively connected by line segments to obtain a first comparison line segment;

[0106] In the direction parallel to the Y axis, each first longitudinal coordinate point with the maximum longitudinal coordinate value and each second longitudinal coordinate point with the minimum longitudinal coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate point and the second transverse coordinate point located in the same tumor pixel set are respectively connected by line segments to obtain a second comparison line segment;

[0107] Determine the one with the longer line segment value between the first comparison line segment and the second comparison line segment corresponding to the same aneurysm as a reference line segment, and obtain the reference line segment value of the reference line segment;

[0108] Comparing each baseline segment value corresponding to each aneurysm with a preset length, and determining an aneurysm with a corresponding baseline segment value greater than or equal to the preset length as having a high risk, and determining an aneurysm with a corresponding baseline segment value less than the preset length as having a low risk;

[0109] All aneurysms with high risk were identified as the primary risk group, and all aneurysms with low risk were identified as the secondary risk group.

[0110] For example, in this embodiment, first, after determining all aneurysms distributed in the initial angiography image by the above-mentioned method such as image recognition, the server will use any image endpoint in the initial angiography image as a reference point (here, the image endpoint can be understood as an endpoint located on the image contour, that is, when the shape of the image contour is a rectangle, there are four image endpoints), and establish a coordinate system corresponding to the initial angiography image with the reference point as the origin, that is, the image coordinate system;

[0111] Next, the server will respectively obtain the pixel points constituting each aneurysm in the initial angiography image, that is, each tumor pixel point, and determine the tumor pixel points constituting the same aneurysm as a tumor pixel set, thereby obtaining each tumor pixel set corresponding to each aneurysm.

[0112] Next, the server will determine the coordinate point with the maximum transverse coordinate value, i.e., each first transverse coordinate point, and the coordinate point with the minimum transverse coordinate value, i.e., each second transverse coordinate point, in each tumor pixel set in the direction parallel to the X-axis in the image coordinate system. For example, when the coordinate points located in the same tumor pixel set include (2, 3), (3, 4) and (5, 5), (5, 5) will be determined as the first transverse coordinate point, and (2, 3) will be determined as the second transverse coordinate point; at the same time, the first transverse coordinate point and the second transverse coordinate point in the same tumor pixel set will be connected by line segments to obtain a first comparison line segment; in the same manner as above, the coordinate point with the maximum longitudinal coordinate value, i.e., each first longitudinal coordinate point, and the coordinate point with the minimum longitudinal coordinate value, i.e., each second longitudinal coordinate point, in each tumor pixel set in the direction parallel to the Y-axis in the image coordinate system will be determined, and the first transverse coordinate point and the second transverse coordinate point in the same tumor pixel set will be connected by line segments to obtain a second comparison line segment.

[0113] Then, after obtaining the first comparison line segment and the second comparison line segment, the server will compare the lengths of the first comparison line segment and the second comparison line segment corresponding to the same aneurysm, and determine the comparison line segment with a longer line segment value as the baseline line segment, and then obtain the length of the baseline line segment, that is, the baseline line segment value.

[0114] Finally, a preset length is pre-set in the server, so the server will compare the baseline segment value corresponding to each aneurysm with the preset length. When the baseline segment value is greater than or equal to the preset length, it means that the size of the aneurysm is large, and the server will determine the aneurysm as having a high risk; when the baseline segment value is less than the preset length, it means that the size of the aneurysm is small, and the server will determine the aneurysm as having a low risk. Then the server will determine all aneurysms with high risks as a first-level risk set, and all aneurysms with low risks as a second-level risk set.

[0115] It should be noted that, in the present embodiment, different line segment values ​​of each aneurysm in two mutually perpendicular direction dimensions are obtained based on the image coordinate system, and the one with the longer line segment value is determined as the size specification corresponding to the aneurysm, thereby achieving rapid determination of the size specification of each aneurysm, while reducing the corresponding data processing amount and improving the corresponding data processing efficiency; at the same time, the size specification can be further compared with the preset length, and the auxiliary assessment of the rupture risk of each aneurysm can be completed according to the comparison result, thereby improving the corresponding accuracy, providing corresponding auxiliary reference for medical staff, and having a certain reference value.

[0116] Here, the above-mentioned preset length can be set based on human considerations. For example, medical staff can determine the setting of the preset length based on previous historical diagnosis experience. This embodiment does not specifically limit the value of the preset length.

[0117] In step S106, the following contents are included:

[0118] Each aneurysm in the secondary risk set is removed from the initial angiography image to obtain a current angiography image, and the current angiography image is divided to obtain each divided sub-image.

[0119] For example, in this embodiment, since each aneurysm in the secondary risk set has a low risk, that is, each aneurysm in the secondary risk set may not rupture accordingly, therefore, each aneurysm in the secondary risk set can be omitted accordingly, that is, there is no need for auxiliary medical staff to perform the corresponding medical diagnosis. In this case, each aneurysm in the secondary risk set can be eliminated in the initial angiography image, and only each aneurysm in the primary risk set is retained, so that the corresponding diagnostic noise can be reduced based on the image level to further improve the auxiliary efficiency of the corresponding medical diagnosis; at the same time, after completing the corresponding elimination process and obtaining the corresponding current angiography image, the current angiography image can also be divided into corresponding images to obtain each divided sub-image, and subsequent medical diagnosis assistance can be performed for each divided sub-image, thereby realizing separate processing of each divided sub-image, while improving the corresponding auxiliary efficiency based on medical diagnosis, it can also improve the accuracy of the corresponding data processing, and can also reduce the equipment requirements of the corresponding data processing equipment, and reduce the corresponding equipment cost.

[0120] Furthermore, in this embodiment, the above-mentioned “eliminating each aneurysm in the secondary risk set from the initial angiography image to obtain a current angiography image” may also include the following steps:

[0121] Determine each aneurysm contour corresponding to each aneurysm located in the secondary risk set, and determine each first intersection point and each second intersection point where each aneurysm contour intersects with the blood vessel contour;

[0122] Connect the first intersection point and the second intersection point corresponding to the same tumor contour to obtain the connection line of each tissue;

[0123] Determine the region consisting of the aneurysm contour and the tissue connection line corresponding to the same aneurysm as the rejection region corresponding to the aneurysm, and reject each rejection region corresponding to each aneurysm in the secondary risk set from the initial angiography image to obtain an initial angiography image including each blank region;

[0124] Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image.

[0125] For example, in this embodiment, after obtaining the secondary risk set, the server will determine the tumor contour of each aneurysm in the secondary risk set, and determine the intersection points where each tumor contour intersects with the blood vessel contour, that is, each first intersection point and each second intersection point, wherein the blood vessel contour can be understood as the contour of the corresponding blood vessel portion in the initial angiography image; then, the server will connect the first intersection point and the second intersection point on each tumor contour to generate a tissue connection line.

[0126] At this time, the server will determine the area composed of the aneurysm contour and tissue connection line of each aneurysm as the rejection area corresponding to each aneurysm, and eliminate the rejection area of ​​each aneurysm in the secondary risk set from the initial angiography image, so that the server will obtain an initial angiography image including various blank areas. Here, the blank area can be understood as an area that is transparent or an area that presents a single background color (such as white); and because after the corresponding rejection areas are eliminated, the blank areas included in the initial angiography image may reduce the corresponding viewing effect, in the subsequent steps, corresponding regional compensation can be performed on each blank area to obtain the corresponding current angiography image.

[0127] It should be noted that this embodiment can accurately determine the removal area of ​​each aneurysm according to the first intersection point and the second intersection point, thereby reducing the workload of the server to a certain extent and improving the corresponding work efficiency.

[0128] Furthermore, the above-mentioned “performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image” further includes the following steps:

[0129] Obtaining each contour pixel point constituting the tumor contour corresponding to the same removal area;

[0130] Acquire each first adjacent pixel point that is adjacent to each contour pixel point and is outside the removal area in the initial angiography image, and determine each first pixel value corresponding to each first adjacent pixel point;

[0131] Performing mean calculation based on each first pixel value to obtain a first pixel mean, and performing pixel value configuration on the blank area based on the first pixel mean;

[0132] Determine one of the first intersection point and the second intersection point corresponding to the same removal area as a reference intersection point, and obtain the intersection pixel points constituting the reference intersection point;

[0133] Acquire each second adjacent pixel point that is adjacent to each intersecting pixel point and located on the blood vessel contour in the initial angiography image, and determine each second pixel value corresponding to each second adjacent pixel point;

[0134] The second pixel mean is obtained by performing mean calculation based on each second pixel value, and the tissue connection line located in the blank area is reconfigured with pixel values ​​based on the second pixel mean to obtain the current angiography image.

[0135] For example, in this embodiment, after determining the rejection area of ​​each aneurysm, the server will first obtain each pixel point of the aneurysm contour in each rejection area, that is, each contour pixel point. Furthermore, the server will obtain the pixel points adjacent to each contour pixel point in the area outside the rejection area in the initial angiography image, that is, each first adjacent pixel point, and determine the pixel value of each first adjacent pixel point, that is, each first pixel value. After obtaining each first pixel value, the server will calculate the mean of all first pixel values, that is, the first pixel mean, and configure the pixel value of the blank area to the first pixel mean, thereby completing the pixel value configuration of the blank area once;

[0136] Then, the server will determine any one of the first intersection point and the second intersection point in the same exclusion area as the reference intersection point. When the blood vessel wall is thicker, the reference intersection point will be composed of multiple pixel points, so that the server will obtain all the pixel points that constitute the reference intersection point, that is, each intersection pixel point, and then obtain each second adjacent pixel point adjacent to each intersection pixel point on the blood vessel contour in the initial angiography image, and determine the second pixel value of each second adjacent pixel point respectively, and then calculate the average of all second pixel values, that is, the second pixel mean, and configure the pixel value of the tissue connection line in the blank area to the second pixel mean to complete the secondary pixel value configuration of the blank area, thereby obtaining the current angiography image.

[0137] In this embodiment, since the colors of the hemangioma and the blood vessels may be different, the server performs secondary pixel value configuration on the blank area, so that the blank area after pixel value configuration can be more integrated with the color of the adjacent area in the initial angiography image, thereby improving the corresponding visual display effect;

[0138] It should be noted that, through the above solution, after performing corresponding regional compensation, the color of each blank area can be more integrated with the color of the adjacent area, thereby increasing the corresponding consistency. However, in some cases, medical staff may want to visually confirm the location of each aneurysm in the secondary risk set in the current angiography image. Therefore, in view of this, this embodiment also discloses another alternative solution step, which specifically includes:

[0139] Acquire a first interference layer having the same size as the initial angiography image, and overlay the first interference layer on the initial angiography image;

[0140] Determine the regional pixel points that respectively constitute each blank area, and determine the regional pixel points corresponding to the same blank area as a regional pixel set, to obtain different regional pixel sets corresponding to different blank areas;

[0141] Based on the pixel points of each region in the corresponding pixel sets of different regions, determining the interference regions corresponding to the blank regions in the first interference layer, and extracting the regions outside the interference regions in the first interference layer to obtain the second interference layer;

[0142] The second interference layer is fused with the initial angiography image to obtain a current angiography image.

[0143] For example, in this embodiment, the above method may specifically include the following contents:

[0144] First, a corresponding first interference layer may be obtained, wherein the first interference layer may specifically be a layer having an interference image, such as a mosaic image or other image having an obvious image difference from the initial angiography image;

[0145] Next, the regional pixel points that respectively constitute each blank area may be determined, and the regional pixel points corresponding to the same blank area may be determined as a regional pixel set;

[0146] Finally, the interference areas corresponding to the blank areas can be determined in the first interference layer according to each regional pixel set, and the first interference layer is extracted to retain only the interference areas corresponding to the blank areas in the initial angiography image, thereby obtaining the corresponding second interference layer; finally, since the interference areas in the second interference layer correspond to the blank areas in the initial angiography image, the second interference layer can be fused with the initial angiography image to obtain the corresponding current angiography image, thereby achieving regional compensation for the blank areas while completing the regional marking of the aneurysms in the secondary risk set, thereby improving image recognition.

[0147] It should be noted that the above-mentioned scheme of "determining the interference areas corresponding to the blank areas respectively according to each regional pixel set in the first interference layer" can be implemented based on the coordinate system, that is, by establishing an image coordinate system based on the initial angiography image, the position coordinate points of each regional pixel point in each regional pixel set are obtained, so that the corresponding interference areas can be determined in the first interference layer based on the position coordinate points, thereby realizing rapid determination of each interference area and further improving the corresponding image processing efficiency.

[0148] In step S108, the following contents are included:

[0149] The number of aneurysm bodies distributed in each divided sub-image is determined respectively, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image and a preset number, the evaluation processing method corresponding to each divided sub-image is determined respectively.

[0150] For example, in this embodiment, the server will respectively determine the number of aneurysm bodies distributed in each divided sub-image, and a preset number is pre-set in the server. The server will compare the number of aneurysms in each divided sub-image with the preset number, and determine the evaluation and processing method for each divided sub-image according to the comparison result. That is, when the number of aneurysms in a certain divided sub-image is greater than the preset number, it indicates that the density of the corresponding aneurysm in the divided sub-image is relatively large; and when the number of aneurysms in a certain divided sub-image is less than or equal to the preset number, it indicates that the density of the corresponding aneurysm in the divided sub-image is relatively small.

[0151] Furthermore, the above-mentioned “determining the evaluation processing method corresponding to each divided sub-image according to the comparison result of the number of tumors in each divided sub-image with the preset number” also includes the following steps:

[0152] Obtaining a blood vessel corresponding to each tumor in the divided sub-image, and if it is determined that one blood vessel corresponds to multiple tumors, obtaining multiple edge tumors in the blood vessel, determining blood vessel attribute information based on the edge tumors, and processing the divided sub-image based on the number of tumors, blood vessel attribute information, and morphological attributes of each tumor to determine whether it meets the overall processing conditions or meets the individual processing conditions;

[0153] Generate a first initial processing model based on the divided sub-images that meet the overall processing conditions, and simulate the first initial processing model based on a preset overall processing method to obtain a first current processing model;

[0154] The morphological attributes of each aneurysm located in the divided sub-image with the individual processing conditions are obtained, the type of each aneurysm is determined based on the corresponding morphological attributes, and the evaluation processing method corresponding to each aneurysm is determined based on the determination result.

[0155] For example, in this embodiment, after determining the blood vessel corresponding to each tumor in each divided sub-image, the server will judge the corresponding tumor for each blood vessel. If it is determined that a blood vessel corresponds to multiple edge tumors, that is, there are multiple tumors at the edge position on a blood vessel, then multiple edge tumors corresponding to the blood vessel are obtained, and the vascular attribute information is further determined based on the edge tumor. Then, the divided sub-image can be processed based on the number of tumors corresponding to the blood vessel, the vascular attribute information and the morphological attributes of each tumor to determine whether it has the corresponding overall processing conditions or has the individual processing conditions.

[0156] Then, the server generates a first initial processing model based on the segmented sub-images with overall processing conditions, and simulates the generated first initial processing model based on the preset overall processing method to obtain a first current processing model. The morphological attributes of each aneurysm in the segmented sub-images with individual processing conditions are obtained, and the type of each aneurysm can be determined based on the morphological attributes, and the evaluation processing method for each aneurysm is determined based on the determination result.

[0157] This embodiment can determine the processing conditions and corresponding evaluation and processing methods for each divided sub-image according to the number of aneurysm bodies in each divided sub-image, and provide medical staff with certain medical auxiliary references.

[0158] Furthermore, in this embodiment, the above-mentioned "obtaining a blood vessel corresponding to each tumor in the divided sub-image, obtaining multiple edge tumors in the blood vessel if it is determined that one blood vessel corresponds to multiple tumors, determining blood vessel attribute information based on the edge tumors, and processing the divided sub-image to determine whether it meets the overall processing condition or meets the individual processing condition based on the number of tumors, the blood vessel attribute information and the morphological attributes of each tumor" may also include the following steps:

[0159] Obtaining the coordinate point of each blood vessel in the divided sub-image after coordinate processing to obtain a blood vessel coordinate set, and obtaining the coordinate point of each tumor in the divided sub-image after coordinate processing to obtain a tumor coordinate set;

[0160] Compare each tumor coordinate set with the blood vessel coordinate set, determine the tumor coordinate sets intersecting with the blood vessel coordinate set as a set group, and use the tumors corresponding to the tumor coordinate sets in the set group as the tumors of the blood vessels in the blood vessel coordinate set;

[0161] Determine edge tumor and blood vessel attribute information based on the coordinate information of the blood vessels corresponding to each tumor, and obtain the morphological attributes of each tumor based on the coordinate information of each tumor;

[0162] Based on the number of tumors, vascular attribute information, and the morphological attributes of each tumor, the divided sub-images are processed to determine whether they meet the overall processing conditions or the individual processing conditions.

[0163] For example, in this embodiment, the coordinate processing can be understood as establishing a corresponding image coordinate system based on the corresponding divided sub-images, and after the image coordinate system is established, the coordinate points of each blood vessel can be obtained and summarized into a blood vessel coordinate set, and the coordinate points of each tumor can be summarized into a tumor group coordinate set; by comparing the tumor coordinate set with the blood vessel coordinate set, the blood vessel coordinate set and the tumor coordinate set having an intersecting relationship are determined as a set group, and further the tumors corresponding to the tumor coordinate sets in the set group are respectively taken as the tumors corresponding to the blood vessels in the blood vessel coordinate set;

[0164] After the corresponding relationship between the blood vessels and the tumor is determined, the edge tumor and blood vessel attribute information can be determined based on the coordinate information of the blood vessels corresponding to each tumor, and the morphological attributes corresponding to each tumor can be obtained based on the coordinate information of the tumor;

[0165] Finally, the corresponding divided sub-images can be processed based on the aforementioned acquired tumor quantity, vascular attribute information, and morphological attributes of each tumor, and then determined to meet the overall processing conditions or the individual processing conditions.

[0166] Furthermore, in this embodiment, the above-mentioned “determining the edge tumor and blood vessel attribute information based on the coordinate information of the blood vessel corresponding to each tumor, and obtaining the morphological attribute of each tumor based on the coordinate information of each tumor” may also include the following steps:

[0167] Obtaining the longitudinal coordinates in the coordinate information of the blood vessels corresponding to each tumor, and sorting all the tumors in descending order based on the longitudinal coordinates to obtain a corresponding tumor sequence;

[0168] The first tumor and the last tumor in the tumor sequence are extracted as marginal tumors;

[0169] Obtaining the center coordinate points of two edge tumors, calculating the length of a straight line of the center coordinate points based on the center coordinate points, and using the length of the straight line as blood vessel attribute information;

[0170] The maximum vertical coordinate difference of each edge tumor under each same horizontal coordinate is obtained to obtain height information, and the maximum horizontal coordinate difference of each edge tumor under each same vertical coordinate is obtained to obtain width information. The morphological attributes include height information and width information.

[0171] For example, in this embodiment, since the divided sub-images are subjected to corresponding coordinate processing in the above process, the longitudinal coordinates in the coordinate information of the blood vessels corresponding to each tumor can be obtained based on the coordinate processing, and all the longitudinal coordinates are sorted in descending order to obtain the corresponding tumor sequence; the tumor located in the first position and the tumor located in the last position in the tumor sequence are taken as the corresponding edge tumors, and the center coordinate points of the two edge tumors are obtained at the same time, and the corresponding center coordinate point straight line length (that is, the distance between the center coordinate points of the two edge tumors) is calculated based on the center coordinate points, and then the corresponding straight line length is used as the corresponding blood vessel attribute information; finally, the maximum vertical coordinate difference of each edge tumor under each same horizontal coordinate can be obtained to obtain the corresponding height information, and the maximum horizontal coordinate difference of each edge tumor under each same vertical coordinate can be obtained to obtain the corresponding width information, and the obtained height information and width information are used as the corresponding morphological attributes.

[0172] Furthermore, in this embodiment, the above-mentioned “processing the divided sub-images based on the number of tumors, the vascular attribute information, and the morphological attributes of each tumor to determine whether the divided sub-images meet the overall processing conditions or meet the individual processing conditions” may also include the following steps:

[0173] The number of tumors and the vascular attribute information are compared and then weighted to obtain a tumor vascular coefficient value, the morphology of each tumor is calculated and then weighted to obtain a tumor morphology coefficient value, and the tumor processing coefficient value is obtained based on the tumor vascular coefficient value and the tumor morphology coefficient value;

[0174] If the tumor treatment coefficient value is greater than or equal to the preset coefficient value, it is determined that the overall treatment conditions are met; if the tumor treatment coefficient value is less than the preset coefficient value, it is determined that the individual treatment conditions are met;

[0175] The actual operation data of the surgery is obtained. If the actual operation data does not correspond to the outputted overall processing conditions or the outputted individual processing conditions, the weights are updated and trained.

[0176] For example, in this embodiment, first, the ratio of the number of tumors to the vascular attribute information can be determined, and then the density of the corresponding tumor can be determined based on the obtained ratio; then, the corresponding tumor vascular coefficient value is obtained by calling the tumor vascular coefficient weight for weighted processing; at the same time, the corresponding tumor morphology coefficient weight is called, and weighted processing is performed based on the morphological attributes of each tumor to obtain the corresponding tumor morphology coefficient value; then, the corresponding tumor processing coefficient value can be obtained according to the obtained tumor vascular coefficient value and tumor morphology coefficient value. Here, the numerical value can be obtained by addition calculation; after the corresponding tumor processing coefficient value is obtained, The tumor processing coefficient value can be compared with the pre-stored preset coefficient value. When the comparison result is that the tumor processing coefficient value is greater than or equal to the preset coefficient value, the corresponding divided sub-image can be determined as meeting the overall processing conditions. When the comparison result is that the tumor processing coefficient value is less than the preset coefficient value, the corresponding divided sub-image can be determined as meeting the individual processing conditions. Finally, in order to be able to iteratively update the corresponding judgment process, the actual surgical operation data can also be obtained, and if it is determined that the actual surgical operation data does not correspond to the output of meeting the overall processing conditions or meeting the individual processing conditions, the weights can be updated and trained accordingly.

[0177] It should be noted that, in this embodiment, on the one hand, since the tumor vascular coefficient value corresponds to the vascular density, and the tumor morphology coefficient value corresponds to the height and width of each tumor, the coefficient values ​​based on two different dimensions can improve the comprehensiveness of the prediction, and thus can also improve the prediction accuracy in the corresponding processing process; on the other hand, the actual surgical operation data can be understood as the actual surgical data, that is, by comparing the actual surgical operation data with the output data, the overall prediction accuracy can be judged, and when the comparison result is inconsistent, it indicates that the prediction is inaccurate. At this time, it is necessary to perform corresponding update training processing on the pre-stored weight values ​​to realize the update and iteration process, thereby continuously improving the corresponding prediction accuracy.

[0178] Furthermore, in this embodiment, the above-mentioned "calculating the tumor treatment coefficient value and updating the weight through the following steps" may also include the following steps:

[0179] The tumor treatment coefficient value was calculated by the following formula:

[0180]

[0181] Among them, x is the tumor treatment coefficient value, s is the number of tumors, is the horizontal coordinate of the center coordinate point of the first edge tumor, is the horizontal coordinate of the center coordinate point of the second edge tumor, is the ordinate of the center coordinate point of the first edge tumor, is the ordinate of the center coordinate point of the second edge tumor, k s is the tumor vascular coefficient weight, g i is the height information of the ith tumor, d i is the width information of the i-th tumor, n is the upper limit of the number of tumors, k u is the tumor morphology coefficient weight.

[0182] Furthermore, in this embodiment, the above-mentioned "obtaining actual surgical operation data, and if the actual operation data does not correspond to the outputted overall processing conditions or the outputted individual processing conditions, then updating the weights for training" may also include the following steps:

[0183] If the actual operation data meets the individual processing conditions and the output meets the overall processing conditions, then the tumor vascular coefficient weight k s and tumor morphology coefficient weight k u Reduce training;

[0184] If the actual operation data meets the overall processing conditions and the output meets the individual processing conditions, then the tumor vascular coefficient weight k s and tumor morphology coefficient weight k u Increase training;

[0185] The trained tumor vascular coefficient weight k is obtained by the following formula w and tumor morphology coefficient weight k e ,

[0186]

[0187] Among them, q + is the tumor vascular coefficient weight k s Increase the number of training times, ∈ is the training constant value of the tumor vascular coefficient weight, v + is the tumor morphology coefficient weight k u Increase the number of training times, δ is the training constant value of the tumor treatment coefficient weight, q - is the tumor vascular coefficient weight k s Reduce the number of training times, v - is the tumor morphology coefficient weight k u Reduce the number of training sessions.

[0188] For example, in this embodiment, when the actual surgical operation data obtained has individual processing conditions, but the predicted output is an overall processing condition, the pre-stored tumor vascular coefficient weights and tumor morphology coefficient weights can be trained accordingly, thereby completing the iterative update of the prediction process and improving the corresponding prediction accuracy.

[0189] Furthermore, the above-mentioned “generating a first initial processing model based on the divided sub-images having the overall processing conditions, and simulating the first initial processing model based on a preset overall processing method to obtain a first current processing model” further includes the following steps:

[0190] Performing three-dimensional modeling based on the divided sub-images having the overall processing conditions to obtain a first initial processing model corresponding to the divided sub-images;

[0191] Determining in the first initial processing model the aneurysm model parts and the blood vessel model parts corresponding to the blood vessels, and determining the blood vessel model parts that are connected to the aneurysm model parts as a blood vessel clipping group based on the relative position relationship between the aneurysm model parts and the blood vessel model parts;

[0192] Performing color configuration on each blood vessel model portion in the blood vessel clamping group based on a preset first processing color, and generating a first blood vessel clamping model and a second blood vessel clamping model surrounding the blood vessel model portion at a first model point and a second model point of each blood vessel model portion, to obtain a first current processing model;

[0193] There is no tumor model part in the part where the first model point of the blood vessel model part is far away from the second model point, and there is no tumor model part in the part where the second model point of the blood vessel model part is far away from the first model point.

[0194] For example, in this embodiment, after determining the segmented sub-images that meet the overall processing conditions, the server will perform three-dimensional modeling on these segmented sub-images to obtain the first initial processing model of each segmented sub-image. Then, in the first initial processing model, each tumor model part corresponding to the aneurysm and each blood vessel model part corresponding to the blood vessel are determined. Since in a segmented sub-image, there may be a situation where some blood vessels have aneurysms and some blood vessels do not have aneurysms, the server will identify the blood vessels with aneurysms, that is, based on the relative positional relationship between each tumor model part and each blood vessel model part, all blood vessel model parts that have a connection relationship with each tumor model part are determined as a blood vessel clamping group, and then each blood vessel model part in the blood vessel clamping group is color-configured based on the preset first processing color, for example, the preset first processing color can be blue.

[0195] Next, the server generates a first vascular clamping model and a second vascular clamping model surrounding the vascular model part at the first model point and the second model point of each vascular model part to obtain a first current processing model; wherein the selection of the first model point and the second model point is based on the following rules: there is no tumor model part in the area where the first model point in the vascular model part is far away from the second model point, and there is no tumor model part in the area where the second model point in the vascular model part is far away from the first model point. After the corresponding first model point and second model point are selected according to the above rules, it can be ensured that each tumor model part located in the same vascular model part is located in the area between the two model points, so that the first clamping model model and the second clamping model can be used to achieve model clamping of each tumor model part located in the same vascular model part.

[0196] For example, in this embodiment, for each tumor in the divided sub-images that meet the overall processing conditions, since its density is large and the risk is high, it can be used, for example Figure 2 The treatment method shown is to clamp the entire blood vessel as a whole, so that all corresponding tumors lose blood supply.

[0197] Furthermore, the above-mentioned "obtaining the morphological attributes of each aneurysm located in the divided sub-image with the separate processing conditions, determining the type of each aneurysm based on the corresponding morphological attributes, and determining the evaluation and processing methods corresponding to each aneurysm based on the determination result" also includes the following steps:

[0198] Performing three-dimensional modeling based on the divided sub-images with separate processing conditions to obtain a second initial processing model corresponding to the divided sub-images;

[0199] Determining in the second initial processing model each tumor model part including the aneurysm, wherein each tumor model part includes a tumor body model and a tumor connection model;

[0200] Determine a model height and a model width of the tumor connection model corresponding to each tumor model part, and determine a morphological attribute of each tumor model part based on the model height and the model width, wherein the morphological attribute includes a low-risk attribute and a high-risk attribute;

[0201] Based on the morphological attributes, an evaluation and processing method corresponding to each aneurysm is determined, and based on the determined evaluation and processing method, a simulation process is performed on the second initial processing model to obtain a second current processing model.

[0202] For example, in this embodiment, after determining the segmented sub-images that meet the conditions for separate processing, the server will perform three-dimensional modeling on these segmented sub-images, thereby obtaining the second initial processing models corresponding to all the segmented sub-images. Then, in the second initial processing model, each tumor model part including the aneurysm is determined, and each tumor model part includes a tumor body model and a tumor connection model connecting the tumor body model and the blood vessel model part. The server will determine the model height and model width of the tumor connection model in each tumor model part, and determine the morphological attributes of each tumor model part based on the model height and model width, and the morphological attributes include low-risk attributes and high-risk attributes. The server will determine the evaluation and processing method of each aneurysm based on the morphological attributes, and simulate the second initial processing model based on the evaluation and processing method to obtain the second current processing model.

[0203] For example, Figure 3 As shown, Figure 3 The figure includes two tumor models in different states. By comparison, it can be seen that the model height and model width of the tumor connection model corresponding to one of the tumor models are larger than those of the other tumor model. Therefore, the two can be respectively determined as having low-risk attributes and high-risk attributes.

[0204] Furthermore, the above-mentioned “determining the model height and model width of the tumor connection model corresponding to each tumor model part, and determining the morphological attributes of each tumor model part based on the model height and model width” also includes the following steps:

[0205] When the model width is greater than or equal to the preset width, the tumor model portion corresponding to the model width is divided into the primary risk assessment group, and the morphological attribute of the tumor model portion corresponding to the model width less than the preset width is determined as a high-risk attribute;

[0206] Determine the comparison results between the model heights corresponding to each tumor model part in the primary risk assessment group and the preset height, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height greater than or equal to the preset height as high-risk attributes, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height less than the preset height as low-risk attributes.

[0207] For example, in this embodiment, a preset width and a preset height are pre-set in the server. After determining the model width and model height of the tumor connection model in each tumor model part, the server will compare the model width with the preset width. When the model width is greater than or equal to the preset width, it indicates that the possibility of danger may be relatively small, and the server will classify the tumor model part corresponding to the model width into the primary risk assessment group; when the model width is less than the preset width, it indicates that the possibility of danger is relatively large, and the corresponding aneurysm is prone to rupture, and the server will determine the morphological attributes of the tumor model part corresponding to the model width as high-risk attributes.

[0208] Furthermore, after completing the comparison in the width dimension, the server will compare the model height of each tumor model part in the primary risk assessment group with the preset height, wherein when the model height is greater than or equal to the preset height, it means that the model height is longer, indicating a greater possibility of danger, that is, the corresponding aneurysm may also be prone to rupture, and thus the server will determine the morphological attributes of the tumor model part whose comparison result is a model height greater than or equal to the preset height as a high-risk attribute; and when the model height is less than the preset height, it means that the model height is shorter, and the possibility of danger is smaller, and thus the server will determine the morphological attributes of the tumor model part whose comparison result is a model height less than the preset height as a low-risk attribute.

[0209] This embodiment can determine the morphological attributes of each tumor model part based on the model height and model width of the tumor connection model of each tumor model part and compare them with the preset width and preset height. The morphological attributes of the tumor model part determined in this embodiment are relatively objective and can provide doctors with a certain reference.

[0210] Furthermore, the above-mentioned "determining the evaluation and processing methods corresponding to each aneurysm based on the morphological attributes, and simulating the second initial processing model based on the determined evaluation and processing methods to obtain the second current processing model" also includes the following steps:

[0211] Performing color configuration on the tumor model portion corresponding to the high-risk attribute based on the preset second processing color, and obtaining the model specification of the tumor body model in the tumor model portion corresponding to the tumor model portion;

[0212] generating an embolization model filling the tumor body model based on the model specification;

[0213] The tumor model portion corresponding to the low-risk attribute is color configured based on the preset third processing color, and a tumor clamping model surrounding the tumor connection model is generated based on the model height and model width of the tumor connection model in the corresponding tumor model portion to obtain a second current processing model.

[0214] For example, in this embodiment, after determining the morphological attributes of each tumor model part, the server will configure the color of the tumor model part with high-risk attributes with a preset second processing color, which can be green, and obtain the size of the tumor body model corresponding to the tumor model part, that is, the model specification, and then generate an embolization model filled into the tumor body model according to the model specification, for example Figure 4 shown.

[0215] The server also performs a color configuration of a preset third processing color on the tumor model portion with a low-risk attribute, and generates a tumor clamping model around the tumor connection model according to the model height and model width of the tumor connection model in the tumor model portion to obtain a second current processing model, for example, Figure 5 shown.

[0216] This embodiment can configure different preset processing colors for tumor model parts with different morphological attributes, thereby improving the identification to a certain extent, and can determine the corresponding evaluation and processing methods for tumor model parts with different morphological attributes, and obtain the corresponding embolization model or tumor clamping model, which can provide more intuitive reference for medical staff.

[0217] According to the solution of the present invention, the server first receives the initial angiography image sent by the medical end, and identifies the initial angiography image, thereby determining all aneurysms distributed in the initial angiography image; then, based on the preset risk assessment strategy, the rupture risk of each aneurysm is assessed, and all aneurysms with high risk are determined as a first-level risk set, and all aneurysms with low risk are determined as a second-level risk set; then, each aneurysm in the second-level risk set is removed from the initial angiography image, thereby obtaining the current angiography image, and the angiography image is image-divided to obtain each divided sub-image; then, the number of aneurysms distributed in each divided sub-image is determined respectively, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image and the preset number, the evaluation processing method corresponding to each divided sub-image is determined respectively. The present invention can assess the rupture risk of each aneurysm and determine the corresponding evaluation processing method, with fast speed and certain accuracy, thereby providing medical staff with a certain reference and improving the corresponding medical diagnosis auxiliary efficiency.

[0218] Another embodiment of the present invention provides an intracranial aneurysm assessment system based on deep learning. Figure 6 As a corresponding device block diagram, the system includes:

[0219] an identification module configured to receive an initial angiography image sent by the medical terminal, identify the initial angiography image, and determine all aneurysms distributed in the initial angiography image;

[0220] An assessment module is configured to assess the rupture risk of each aneurysm based on a preset risk assessment strategy, and to identify all aneurysms with high risk as a primary risk set and all aneurysms with low risk as a secondary risk set;

[0221] a segmentation module configured to remove each aneurysm located in the secondary risk set from the initial angiography image to obtain a current angiography image, and perform image segmentation on the current angiography image to obtain segmented sub-images;

[0222] The determination module is configured to respectively determine the number of aneurysm bodies distributed in each divided sub-image, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image with a preset number, respectively determine the evaluation processing method corresponding to each divided sub-image.

[0223] In the description provided herein, algorithms and displays are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the examples of the present invention. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to implement the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the preferred embodiment of the present invention.

[0224] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0225] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0226] Those skilled in the art will appreciate that the modules or units or components of the devices in the examples disclosed herein may be arranged in the devices described in the embodiment, or alternatively may be located in one or more devices different from the devices in the examples. The modules in the foregoing examples may be combined into one module or may be divided into multiple submodules.

[0227] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0228] Furthermore, those skilled in the art will appreciate that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present invention and to form different embodiments.

[0229] In addition, some of the embodiments are described herein as methods or combinations of method elements that can be implemented by a processor of a computer system or by other devices that perform the functions. Therefore, a processor with necessary instructions for implementing the method or method elements forms a device for implementing the method or method elements. In addition, the elements described herein of the device embodiments are examples of devices for implementing the functions performed by the elements for the purpose of implementing the invention.

[0230] As used herein, unless otherwise specified, the use of ordinal numbers "first," "second," "third," etc. to describe common objects merely indicates that different instances of similar objects are involved, and is not intended to imply that the objects so described must have a given order in time, space, order, or in any other manner.

[0231] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and teaching purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is illustrative and not restrictive with respect to the scope of the present invention, and the scope of the present invention is defined by the appended claims.

Claims

1. A method for evaluating intracranial aneurysms based on deep learning, characterized in that: The following steps are involved: receiving an initial angiography image sent by the medical terminal, and identifying the initial angiography image to determine all aneurysms distributed in the initial angiography image; The rupture risk of each aneurysm is assessed based on a preset risk assessment strategy, and all aneurysms with high risk are identified as a primary risk set, and all aneurysms with low risk are identified as a secondary risk set; Eliminate each aneurysm in the secondary risk set from the initial angiography image to obtain a current angiography image, and perform image segmentation on the current angiography image to obtain segmented sub-images; Determine the number of aneurysm bodies distributed in each divided sub-image respectively, and determine the evaluation processing method corresponding to each divided sub-image respectively based on the comparison result of the number of aneurysm bodies corresponding to each divided sub-image and the preset number; Wherein, each aneurysm in the secondary risk set is removed from the initial angiography image to obtain a current angiography image, including: Determine each aneurysm contour corresponding to each aneurysm located in the secondary risk set, and determine each first intersection point and each second intersection point where each aneurysm contour intersects with the blood vessel contour; Connect the first intersection point and the second intersection point corresponding to the same tumor contour to obtain the connection line of each tissue; Determine the region consisting of the aneurysm contour and the tissue connection line corresponding to the same aneurysm as the rejection region corresponding to the aneurysm, and reject each rejection region corresponding to each aneurysm in the secondary risk set from the initial angiography image to obtain an initial angiography image including each blank region; Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image; The process of performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes: Obtaining each contour pixel point constituting the tumor contour corresponding to the same removal area; Acquire each first adjacent pixel point that is adjacent to each contour pixel point and is outside the removal area in the initial angiography image, and determine each first pixel value corresponding to each first adjacent pixel point; Performing mean calculation based on each first pixel value to obtain a first pixel mean, and performing pixel value configuration on the blank area based on the first pixel mean; Determine one of the first intersection point and the second intersection point corresponding to the same removal area as a reference intersection point, and obtain the intersection pixel points constituting the reference intersection point; Acquire each second adjacent pixel point that is adjacent to each intersecting pixel point and located on the blood vessel contour in the initial angiography image, and determine each second pixel value corresponding to each second adjacent pixel point; Performing mean calculation based on each second pixel value to obtain a second pixel mean, and performing secondary pixel value configuration on the tissue connection line located in the blank area based on the second pixel mean to obtain a current angiography image; or, Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes: Acquire a first interference layer having the same size as the initial angiography image, and overlay the first interference layer on the initial angiography image; Determine the regional pixel points that respectively constitute each blank area, and determine the regional pixel points corresponding to the same blank area as a regional pixel set, to obtain different regional pixel sets corresponding to different blank areas; Based on the pixel points of each region in the corresponding pixel sets of different regions, determining the interference regions corresponding to the blank regions in the first interference layer, and extracting the regions outside the interference regions in the first interference layer to obtain the second interference layer; The second interference layer is fused with the initial angiography image to obtain a current angiography image.

2. The intracranial aneurysm assessment method based on deep learning according to claim 1, characterized in that: The rupture risk of each aneurysm is assessed based on the preset risk assessment strategy, and all aneurysms with high risk are identified as the first-level risk set, and all aneurysms with low risk are identified as the second-level risk set, including: Taking any image endpoint located in the initial angiography image as a reference point, and establishing an image coordinate system corresponding to the initial angiography image with the reference point as the origin; Obtaining the aneurysm pixel points that constitute each aneurysm respectively, and determining the aneurysm pixel points corresponding to the same aneurysm as a aneurysm pixel set, to obtain the aneurysm pixel sets corresponding to each aneurysm; In a direction parallel to the X-axis, each first transverse coordinate point with a maximum transverse coordinate value and each second transverse coordinate point with a minimum transverse coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate points and the second transverse coordinate points located in the same tumor pixel set are respectively connected by line segments to obtain a first comparison line segment; In the direction parallel to the Y axis, each first longitudinal coordinate point with the maximum longitudinal coordinate value and each second longitudinal coordinate point with the minimum longitudinal coordinate value in each tumor pixel set are respectively determined, and the first transverse coordinate point and the second transverse coordinate point located in the same tumor pixel set are respectively connected by line segments to obtain a second comparison line segment; Determine the one with the longer line segment value between the first comparison line segment and the second comparison line segment corresponding to the same aneurysm as a reference line segment, and obtain the reference line segment value of the reference line segment; Comparing each baseline segment value corresponding to each aneurysm with a preset length, and determining an aneurysm with a corresponding baseline segment value greater than or equal to the preset length as having a high risk, and determining an aneurysm with a corresponding baseline segment value less than the preset length as having a low risk; All aneurysms with high risk were identified as the primary risk group, and all aneurysms with low risk were identified as the secondary risk group.

3. The intracranial aneurysm assessment method based on deep learning according to claim 1, characterized in that: According to the comparison result between the number of tumors in each divided sub-image and the preset number, an evaluation processing method corresponding to each divided sub-image is determined respectively, including: Obtaining a blood vessel corresponding to each tumor in the divided sub-image, and if it is determined that one blood vessel corresponds to multiple tumors, obtaining multiple edge tumors in the blood vessel, determining blood vessel attribute information based on the edge tumors, and processing the divided sub-image based on the number of tumors, blood vessel attribute information, and morphological attributes of each tumor to determine whether it meets the overall processing conditions or meets the individual processing conditions; Generate a first initial processing model based on the divided sub-images that meet the overall processing conditions, and simulate the first initial processing model based on a preset overall processing method to obtain a first current processing model; The morphological attributes of each aneurysm located in the divided sub-image with the individual processing conditions are obtained, the type of each aneurysm is determined based on the corresponding morphological attributes, and the evaluation processing method corresponding to each aneurysm is determined based on the determination result.

4. The intracranial aneurysm assessment method based on deep learning according to claim 3, characterized in that: The blood vessels corresponding to each tumor in the divided sub-image are obtained, and if it is determined that one blood vessel corresponds to multiple tumors, multiple edge tumors in the blood vessel are obtained, and the blood vessel attribute information is determined based on the edge tumors. The divided sub-images are processed based on the number of tumors, the blood vessel attribute information, and the morphological attributes of each tumor to determine whether they meet the overall processing conditions or meet the individual processing conditions, including: Obtaining the coordinate point of each blood vessel in the divided sub-image after coordinate processing to obtain a blood vessel coordinate set, and obtaining the coordinate point of each tumor in the divided sub-image after coordinate processing to obtain a tumor coordinate set; Compare each tumor coordinate set with the blood vessel coordinate set, determine the tumor coordinate sets intersecting with the blood vessel coordinate set as a set group, and use the tumors corresponding to the tumor coordinate sets in the set group as the tumors of the blood vessels in the blood vessel coordinate set; Determine edge tumor and blood vessel attribute information based on the coordinate information of the blood vessels corresponding to each tumor, and obtain the morphological attributes of each tumor based on the coordinate information of each tumor; Based on the number of tumors, vascular attribute information, and the morphological attributes of each tumor, the divided sub-images are processed to determine whether they meet the overall processing conditions or the individual processing conditions.

5. The intracranial aneurysm assessment method based on deep learning according to claim 4, characterized in that: The step of determining the edge tumor and blood vessel attribute information based on the coordinate information of the blood vessels corresponding to each tumor, and obtaining the morphological attributes of each tumor based on the coordinate information of each tumor, includes: Obtaining the longitudinal coordinates in the coordinate information of the blood vessels corresponding to each tumor, and sorting all the tumors in descending order based on the longitudinal coordinates to obtain a corresponding tumor sequence; The first tumor and the last tumor in the tumor sequence are extracted as marginal tumors; Obtaining the center coordinate points of two edge tumors, calculating the length of a straight line of the center coordinate points based on the center coordinate points, and using the length of the straight line as blood vessel attribute information; The maximum vertical coordinate difference of each edge tumor under each same horizontal coordinate is obtained to obtain height information, and the maximum horizontal coordinate difference of each edge tumor under each same vertical coordinate is obtained to obtain width information. The morphological attributes include height information and width information.

6. The intracranial aneurysm assessment method based on deep learning according to claim 4, characterized in that: The processing of the divided sub-images based on the number of tumors, the vascular attribute information, and the morphological attributes of each tumor to determine whether the sub-images meet the overall processing conditions or meet the individual processing conditions includes: The number of tumors and the vascular attribute information are compared and then weighted to obtain a tumor vascular coefficient value, the morphology of each tumor is calculated and then weighted to obtain a tumor morphology coefficient value, and the tumor processing coefficient value is obtained based on the tumor vascular coefficient value and the tumor morphology coefficient value; If the tumor treatment coefficient value is greater than or equal to the preset coefficient value, it is determined that the overall treatment conditions are met; if the tumor treatment coefficient value is less than the preset coefficient value, it is determined that the individual treatment conditions are met; The actual operation data of the surgery is obtained. If the actual operation data does not correspond to the outputted overall processing conditions or the outputted individual processing conditions, the weights are updated and trained.

7. The intracranial aneurysm assessment method based on deep learning according to claim 3, characterized in that: Generating a first initial processing model based on the divided sub-images having the overall processing conditions, and simulating the first initial processing model based on a preset overall processing method to obtain a first current processing model, including: Performing three-dimensional modeling based on the divided sub-images having the overall processing conditions to obtain a first initial processing model corresponding to the divided sub-images; Determining in the first initial processing model the aneurysm model parts and the blood vessel model parts corresponding to the blood vessels, and determining the blood vessel model parts that are connected to the aneurysm model parts as a blood vessel clipping group based on the relative position relationship between the aneurysm model parts and the blood vessel model parts; Performing color configuration on each blood vessel model portion in the blood vessel clamping group based on a preset first processing color, and generating a first blood vessel clamping model and a second blood vessel clamping model surrounding the blood vessel model portion at a first model point and a second model point of each blood vessel model portion, to obtain a first current processing model; There is no tumor model part in the part where the first model point of the blood vessel model part is far away from the second model point, and there is no tumor model part in the part where the second model point of the blood vessel model part is far away from the first model point.

8. The intracranial aneurysm assessment method based on deep learning according to claim 3, characterized in that: Acquiring the morphological attributes of each aneurysm in the divided sub-images with separate processing conditions, determining the type of each aneurysm based on the corresponding morphological attributes, and determining the evaluation and processing methods corresponding to each aneurysm based on the determination results, including: Performing three-dimensional modeling based on the divided sub-images with separate processing conditions to obtain a second initial processing model corresponding to the divided sub-images; Determining in the second initial processing model each tumor model part including the aneurysm, wherein each tumor model part includes a tumor body model and a tumor connection model; Determine a model height and a model width of the tumor connection model corresponding to each tumor model part, and determine a morphological attribute of each tumor model part based on the model height and the model width, wherein the morphological attribute includes a low-risk attribute and a high-risk attribute; Based on the morphological attributes, an evaluation and processing method corresponding to each aneurysm is determined, and based on the determined evaluation and processing method, a simulation process is performed on the second initial processing model to obtain a second current processing model.

9. The intracranial aneurysm assessment method based on deep learning according to claim 8, characterized in that: Determining a model height and a model width of a tumor connection model corresponding to each tumor model portion, and determining a morphological attribute of each tumor model portion based on the model height and the model width, including: When the model width is greater than or equal to the preset width, the tumor model portion corresponding to the model width is divided into the primary risk assessment group, and the morphological attribute of the tumor model portion corresponding to the model width less than the preset width is determined as a high-risk attribute; Determine the comparison results between the model heights corresponding to each tumor model part in the primary risk assessment group and the preset height, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height greater than or equal to the preset height as high-risk attributes, and determine the morphological attributes of the tumor model part whose corresponding comparison result is a model height less than the preset height as low-risk attributes.

10. The intracranial aneurysm assessment method based on deep learning according to claim 9, characterized in that: Determining evaluation and processing methods corresponding to each aneurysm based on the morphological attributes, and simulating the second initial processing model based on the determined evaluation and processing methods to obtain a second current processing model, including: Performing color configuration on the tumor model portion corresponding to the high-risk attribute based on the preset second processing color, and obtaining the model specification of the tumor body model in the tumor model portion corresponding to the tumor model portion; generating an embolization model filling the tumor body model based on the model specification; The tumor model part corresponding to the low-risk attribute is color configured based on the preset third processing color, which is yellow. A tumor clamping model surrounding the tumor connection model is generated based on the model height and model width of the tumor connection model in the tumor model part, thereby obtaining the second current processing model.

11. A deep learning-based intracranial aneurysm assessment system, characterized in that: The following steps are involved: an identification module configured to receive an initial angiography image sent by the medical terminal, identify the initial angiography image, and determine all aneurysms distributed in the initial angiography image; An assessment module is configured to assess the rupture risk of each aneurysm based on a preset risk assessment strategy, and to identify all aneurysms with high risk as a primary risk set and all aneurysms with low risk as a secondary risk set; a segmentation module configured to remove each aneurysm located in the secondary risk set from the initial angiography image to obtain a current angiography image, and perform image segmentation on the current angiography image to obtain segmented sub-images; A determination module is configured to respectively determine the number of aneurysm bodies distributed in each divided sub-image, and based on the comparison result of the number of aneurysms corresponding to each divided sub-image and a preset number, respectively determine the evaluation processing method corresponding to each divided sub-image; Wherein, each aneurysm in the secondary risk set is removed from the initial angiography image to obtain a current angiography image, including: Determine each aneurysm contour corresponding to each aneurysm located in the secondary risk set, and determine each first intersection point and each second intersection point where each aneurysm contour intersects with the blood vessel contour; Connect the first intersection point and the second intersection point corresponding to the same tumor contour to obtain the connection line of each tissue; Determine the region consisting of the aneurysm contour and the tissue connection line corresponding to the same aneurysm as the rejection region corresponding to the aneurysm, and reject each rejection region corresponding to each aneurysm in the secondary risk set from the initial angiography image to obtain an initial angiography image including each blank region; Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image; The process of performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes: Obtaining each contour pixel point constituting the tumor contour corresponding to the same removal area; Acquire each first adjacent pixel point that is adjacent to each contour pixel point and is outside the removal area in the initial angiography image, and determine each first pixel value corresponding to each first adjacent pixel point; Performing mean calculation based on each first pixel value to obtain a first pixel mean, and performing pixel value configuration on the blank area based on the first pixel mean; Determine one of the first intersection point and the second intersection point corresponding to the same removal area as a reference intersection point, and obtain the intersection pixel points constituting the reference intersection point; Acquire each second adjacent pixel point that is adjacent to each intersecting pixel point and located on the blood vessel contour in the initial angiography image, and determine each second pixel value corresponding to each second adjacent pixel point; Performing mean calculation based on each second pixel value to obtain a second pixel mean, and performing secondary pixel value configuration on the tissue connection line located in the blank area based on the second pixel mean to obtain a current angiography image; or, Performing regional compensation on each blank area in the initial angiography image to obtain a current angiography image includes: Acquire a first interference layer having the same size as the initial angiography image, and overlay the first interference layer on the initial angiography image; Determine the regional pixel points that respectively constitute each blank area, and determine the regional pixel points corresponding to the same blank area as a regional pixel set, to obtain different regional pixel sets corresponding to different blank areas; Based on the pixel points of each region in the corresponding pixel sets of different regions, determining the interference regions corresponding to the blank regions in the first interference layer, and extracting the regions outside the interference regions in the first interference layer to obtain the second interference layer; The second interference layer is fused with the initial angiography image to obtain a current angiography image.

Images