A Detection Method, Device, Equipment and Storage Medium for a Bracket

Abstract

The present application provides a detection method, device, equipment and storage medium for a stent. The detection method includes: obtaining an image region where a target blood vessel is located from a medical image to be processed as a first processed image; performing threshold segmentation on the first processed image using a target CT value, and taking the image region where the first non-zero pixel point is located in the threshold segmentation result as a second processed image; obtaining an image intersection between the second processed image and a blood vessel segmentation image of the target blood vessel, and determining a sparsified representation result of the image intersection as target input data; inputting the target input data into a pre-trained stent classification model to obtain a stent detection result for the target blood vessel output by the stent classification model. In this way, the present application can quickly locate the target stent in the medical image to be processed based on characteristics such as the material density and morphology of the stent, which is beneficial to improving the detection efficiency of the stent in medical image data.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and more particularly, to a method, apparatus, device, and storage medium for detecting a stent. Background Art

[0002] In the treatment plan of vascular intervention surgery, a stent made of a metal material is usually used as the stent to be implanted. For example, commonly used bare metal stents, covered stents (obtained by covering a thin film on the basis of a bare metal stent through scientific and technological means, where the covered thin film can be a biological material or an artificial material), drug-coated stents (obtained by applying drugs to a bare metal stent through scientific methods), etc. Since restenosis may occur within the stent after the implantation surgery, that is, re-blockage may occur within the stent, patients still need to have regular check-ups after the stent implantation to timely understand the current state of the implanted stent.

[0003] Currently, the detection of stents mainly relies on the doctor's visual observation results of the patient's CT images. However, since the volume of the stent is small and not conducive to visual observation, it is very difficult to accurately locate the stent in the CT image only relying on the doctor's visual observation results, and at the same time, it will also lead to low efficiency of stent detection. Summary of the Invention

[0004] In view of this, the purpose of the present application is to provide a method, apparatus, device, and storage medium for detecting a stent, which can quickly locate the target stent in the medical image to be processed based on the material density and morphology of the stent, etc., and is beneficial to improving the detection efficiency of the stent in the medical image data.

[0005] To make the above objects, features, and advantages of the present application more obvious and understandable, the following specific preferred embodiments are given in conjunction with the accompanying drawings and described in detail as follows.

[0006] In a first aspect, an embodiment of the present application provides a method for detecting a stent, and the detection method includes:

[0007] Obtain the image area where the target blood vessel is located from the medical image to be processed as the first processed image;

[0008] Use the target CT value to perform threshold segmentation on the first processed image, and use the image area where the first non-zero pixel point in the threshold segmentation result is located as the second processed image; where the first non-zero pixel point is used to represent a pixel point whose CT value is greater than or equal to the target CT value;

[0009] Obtain the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determine the sparsified representation result of the image intersection as the target input data;

[0010] Input the target input data into a pre-trained stent classification model to obtain the stent detection result for the target blood vessel output by the stent classification model; wherein, the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

[0011] In an optional implementation manner, the obtaining the image region where the target blood vessel is located from the to-be-processed medical image as the first processed image includes:

[0012] Use the blood vessel segmentation image of the target blood vessel to perform image cropping on the to-be-processed medical image to obtain the first image cropping result of the target blood vessel;

[0013] Sample the first image cropping result according to a preset image sampling size, and use the sampling result of the first image cropping result as the first processed image.

[0014] In an optional implementation manner, the target CT value is determined by the following method:

[0015] Use the initial CT value to perform threshold segmentation on the first processed image, and dynamically adjust the initial CT value within a preset CT value distribution range according to the number of second non-zero pixel points included in the threshold segmentation result until the number of second non-zero pixel points included in the threshold segmentation result meets the preset point number screening condition; wherein, the second non-zero pixel points are used to characterize the pixel points whose CT value is greater than or equal to the initial CT value;

[0016] When the number of second non-zero pixel points included in the threshold segmentation result meets the preset point number screening condition, obtain the dynamically adjusted initial CT value as the target CT value.

[0017] In an optional implementation manner, the obtaining the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determining the sparsified representation result of the image intersection as the target input data includes:

[0018] According to the blood vessel marking result of the target blood vessel included in the blood vessel segmentation image, obtain the image intersection between the blood vessel marking result and the second processed image as the target image intersection;

[0019] Record the position information of each pixel point included in the target image intersection and the image parameter information corresponding to the pixel point, use the recording result as the sparsified representation result of the target image intersection, and determine the recording result as the target input data; wherein, the image parameter information at least includes the gray value corresponding to the pixel point.

[0020] In an alternative embodiment, inputting the target input data into a pre-trained stent classification model to obtain a stent detection result for the target blood vessel output by the stent classification model includes:

[0021] Input the target input data into a pre-trained stent classification model, and based on the morphological features of the target stent, classify and predict the image regions belonging to the target stent in the target input data through the stent classification model to obtain an initial classification prediction result;

[0022] According to the inverse operation of the sparse representation corresponding to the target input data, restore the initial classification prediction result to the image space corresponding to the second processed image, and output to obtain the stent detection result.

[0023] In an alternative embodiment, the stent detection result includes: a plurality of image connected components belonging to the target stent and a plurality of target discrete pixel points belonging to the target stent; after obtaining the stent detection result for the target blood vessel output by the stent classification model, the detection method further includes:

[0024] For each of the target discrete pixel points, perform region growing on each of the target discrete pixel points according to the stent pixel point closest to each of the target discrete pixel points in the stent detection result to obtain a first region growing result corresponding to the stent detection result; wherein, the stent pixel points are used to represent the pixel points belonging to the target stent in the stent detection result;

[0025] When the number of target discrete pixel points included in the first region growing result is greater than or equal to a preset number threshold, perform secondary region growing on each target discrete pixel point included in the first region growing result according to a preset growing strategy, and use the secondary region growing result as the final detection result for the target stent; wherein, the morphological integrity degree of the target stent shown in the final detection result is higher than that shown in the stent detection result.

[0026] In an alternative embodiment, performing secondary region growing on each target discrete pixel point included in the first region growing result according to a preset growing strategy includes:

[0027] Perform threshold segmentation on the first processed image using a first CT value, and use the image region where the third non-zero pixel points in the threshold segmentation result are located as the third processed image; wherein, the first CT value is less than the target CT value; the third non-zero pixel points are used to represent the pixel points whose CT values are greater than or equal to the first CT value;

[0028] Obtain each of the third non-zero pixel points included in the third processed image and each pixel point belonging to the target stent in the stent detection result as a first pixel point set;

[0029] For each target discrete pixel point included in the first region growth result, perform secondary region growth on each target discrete pixel point according to the pixel point in the first pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growth result.

[0030] In an optional implementation manner, the pixel points belonging to the target stent in the stent detection result are used to represent the pixel points for which the probability of belonging to the target stent in the classification prediction result of the stent classification model for the target stent is higher than a first preset threshold; the performing secondary region growth on each target discrete pixel point included in the first region growth result according to a preset growth strategy further includes:

[0031] Obtain the pixel points for which the probability of belonging to the target stent in the classification prediction result of the stent classification model for the target stent is higher than a second preset threshold as candidate pixel points; wherein, the second preset threshold is less than the first preset threshold;

[0032] Obtain each of the candidate pixel points and each pixel point belonging to the target stent in the stent detection result as a second pixel point set;

[0033] For each target discrete pixel point included in the first region growth result, perform secondary region growth on each target discrete pixel point according to the pixel point in the second pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growth result.

[0034] In a second aspect, an embodiment of the present application provides a detection device for a stent, the detection device includes:

[0035] A first processing module, configured to obtain an image region where a target blood vessel is located from a to-be-processed medical image as a first processed image;

[0036] A second processing module, configured to perform threshold segmentation on the first processed image by using a target CT value, and use the image region where non-zero pixel points are located in the threshold segmentation result as a second processed image;

[0037] A sparse representation module, configured to obtain an image intersection between the second processed image and a blood vessel segmentation image of the target blood vessel, and determine a sparsified representation result of the image intersection as target input data;

[0038] A stent detection module is configured to input the target input data into a pre-trained stent classification model, and obtain a stent detection result for the target blood vessel output by the stent classification model; wherein, the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

[0039] In a third aspect, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned stent detection method are implemented.

[0040] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the steps of the above-mentioned stent detection method are executed.

[0041] The technical solutions provided by the embodiments of the present application may include the following beneficial effects:

[0042] A stent detection method, device, equipment, and storage medium provided by an embodiment of the present application obtain an image region where a target blood vessel is located from a to-be-processed medical image as a first processed image; perform threshold segmentation on the first processed image using a target CT value, and use the image region where the first non-zero pixel point is located in the threshold segmentation result as a second processed image; obtain an image intersection between the second processed image and a blood vessel segmentation image of the target blood vessel, and determine a sparsified representation result of the image intersection as the target input data; input the target input data into a pre-trained stent classification model, and obtain a stent detection result for the target blood vessel output by the stent classification model. In this way, the present application can quickly locate the target stent in the to-be-processed medical image based on characteristics such as the material density and morphology of the stent, which is beneficial to improving the detection efficiency of stents in medical image data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0044] Figure 1 Shows a schematic flowchart of a stent detection method provided by an embodiment of the present application;

[0045] Figure 2 Shows a schematic diagram of the result of a second processed image provided by an embodiment of the present application;

[0046] Figure 3 Shows a schematic flowchart of a method for obtaining a first processed image provided by an embodiment of the present application;

[0047] Figure 4 Shows a schematic flowchart of a method for determining a target CT value provided by an embodiment of the present application;

[0048] Figure 5 Shows a schematic flowchart of a method for determining target input data provided by an embodiment of the present application;

[0049] Figure 6 Shows a schematic flowchart of a method for obtaining a stent detection result provided by an embodiment of the present application;

[0050] Figure 7a Shows a schematic flowchart of a method for performing region growing on discrete pixel points in a stent detection result provided by an embodiment of the present application;

[0051] Figure 7b Shows a schematic structural diagram of a stent detection result provided by an embodiment of the present application;

[0052] Figure 8 Shows a schematic structural diagram of a stent detection device provided by an embodiment of the present application;

[0053] Figure 9 Shows a schematic structural diagram of a computer device 900 provided by an embodiment of the present application. Detailed implementation manners

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It should be understood that the accompanying drawings in the present application are only for the purposes of illustration and description, and are not used to limit the protection scope of the present application. In addition, it should be understood that the schematic drawings are not drawn to actual scale. The flowcharts used in the present application show operations implemented according to some embodiments of the present application. It should be understood that the operations in the flowchart may not be implemented in sequence, and steps without logical context relationships may be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of the present application.

[0055] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments. The components of the embodiments of the present application usually described and illustrated in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application claimed, but only represents the selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0056] It should be noted that the term "including" will be used in the embodiments of the present application to indicate the existence of the features stated thereafter, but does not exclude the addition of other features.

[0057] Currently, the detection of stents mainly relies on the doctor's visual observation results of the patient's CT images. However, since the volume of the stent is small and not conducive to visual observation, it is very difficult to accurately locate the stent in the CT image only relying on the doctor's visual observation results, and at the same time, it will also lead to low efficiency of stent detection.

[0058] Based on this, the embodiments of the present application provide a stent detection method, device, equipment and storage medium. From the to-be-processed medical image, obtain the image area where the target blood vessel is located as the first processed image; perform threshold segmentation on the first processed image using the target CT value, and use the image area where the first non-zero pixel point is located in the threshold segmentation result as the second processed image; obtain the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determine the sparsification representation result of the image intersection as the target input data; input the target input data into the pre-trained stent classification model to obtain the stent detection result for the target blood vessel output by the stent classification model. In this way, the present application can quickly locate the target stent in the to-be-processed medical image based on the material density and morphology and other characteristics of the stent, which is beneficial to improving the detection efficiency of the stent in the medical image data.

[0059] It should be noted that the stent detection method provided in the embodiments of the present application is applicable to the stent detection device, and the detection device can be integrated in a computer device.

[0060] Specifically, the above computer device can be a terminal device, such as: mobile phone, tablet computer, notebook computer, desktop computer, etc.; the above computer device can also be a server, which can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms, but is not limited thereto.

[0061] For the convenience of understanding the embodiments of the present application, a detection method, device, equipment and storage medium for a bracket provided by the embodiments of the present application will be introduced in detail below.

[0062] Refer to Figure 1 shown in Figure 1 FIG. shows a schematic flow chart of a detection method for a bracket provided by an embodiment of the present application. The detection method includes steps S101-S104; specifically:

[0063] S101, obtain the image area where the target blood vessel is located from the to-be-processed medical image as the first processed image.

[0064] Here, due to the poor natural contrast between blood vessels and their background soft tissues, conventional CT (Computed Tomography) scans often have difficulty in showing blood vessels; based on this, the above to-be-processed medical image can be a CTA image (i.e., CT angiography) of the target site (i.e., the biological site where the target blood vessel is located). Among them, the CTA image is a medical image that can highlight blood vessels obtained by introducing a contrast agent during a CTA examination to change the image contrast between blood vessels and background tissues.

[0065] It should be noted that the target blood vessel can be the head and neck blood vessels in the head and neck area, or the coronary blood vessels in the heart area; the specific blood vessel type represented by the target blood vessel is not limited in the embodiments of the present application.

[0066] Specifically, when performing step S101, as an optional embodiment, the method of entity object detection can be used to directly segment the image area where the target blood vessel is located from the CTA image (i.e., the above to-be-processed medical image) of the target site as the first processed image with the target blood vessel as the target entity object to be detected currently.

[0067] It should be noted that when performing the above-mentioned entity object detection method in step S101, it is only necessary to be able to implement the image recognition function of determining the entity object categories included in the CTA image of the target part. On this basis, the specific entity object detection method executed in step S101 can be an image semantic segmentation processing method, or an image detection method such as a 3D box detection method (the CTA image of the target part belongs to a three-dimensional image); that is, the specific entity object detection method executed in step S101 is not unique, and the embodiments of the present application do not make any limitations in this regard.

[0068] Specifically, when performing step S101, as another optional embodiment, it is also possible to utilize the blood vessel segmentation image of the target blood vessel (wherein, it includes both the image boundary of the image region where the target blood vessel is located and the pixel point marking result of the pixels belonging to the target blood vessel in the image), and use the image boundary of the image region where the target blood vessel is located included in the blood vessel segmentation image to cut out the CTA image within the same image boundary from the CTA image of the target part as the first processed image.

[0069] Based on the above two optional embodiments, it can be seen that the method for obtaining the first processed image in step S101 is not unique, and the embodiments of the present application do not make any limitations on the specific obtaining method of the above first processed image.

[0070] S102, perform threshold segmentation on the first processed image using the target CT value, and take the image region where the first non-zero pixel point is located in the threshold segmentation result as the second processed image.

[0071] Here, the CT value is the corresponding value equivalent to the X-ray attenuation coefficient of each tissue in the medical image, that is, the magnitude of the CT value is used to reflect the absorption rate of different tissue parts in the medical image for X-rays; among them, the CT value can also be expressed in hu value (hounsfield unit).

[0072] It should be noted that based on the above definition content of the CT value, it can be seen that there is a certain correlation between the magnitude of the CT value and the material density of the detected object. Based on this, since the material of the stent is mainly composed of metal materials, the CT values shown by the stent and other biological tissues at the target blood vessel in the first processed image are significantly different (wherein, the CT value of the metal material is relatively high).

[0073] Here, the above-mentioned first non-zero pixel points are used to represent pixel points with a CT value greater than or equal to the target CT value. As an optional embodiment, during threshold segmentation, in the first processed image, pixel points belonging to the above-mentioned first non-zero pixel points can be marked with a gray value of 1 based on a 0-1 marking method, and pixel points not belonging to the above-mentioned first non-zero pixel points can be marked with a gray value of 0, so as to obtain a second processed image belonging to the highlighted area from the first processed image through the difference in gray values (i.e., brightness) in the first processed image.

[0074] It should be noted that the "0-1 marking" method shown in the above optional embodiment is only an optional threshold segmentation method. For example, in the first processed image, pixel points belonging to the above-mentioned first non-zero pixel points can also be marked with a gray value of L1, and pixel points not belonging to the above-mentioned first non-zero pixel points can be marked with a gray value of L2. As long as the value of L2 is ensured to be less than the value of L1, a second processed image belonging to the highlighted area can also be obtained from the first processed image. Based on this, the specific threshold segmentation method in step S102 is not limited in the embodiments of the present application.

[0075] Specifically, still taking the 0-1 marking method as an example, Figure 2 shows a schematic diagram of the result of a second processed image provided by an embodiment of the present application. Referring to Figure 2 as shown, based on the difference in CT values between the stent and other biological tissues at the target blood vessel shown in the first processed image, when performing step S102, a relatively large target CT value can be used to perform threshold segmentation on the first processed image. Pixel points in the first processed image with a CT value greater than or equal to the target CT value are used as the first non-zero pixel points (for example, the gray value is set to 1), and pixel points in the first processed image with a CT value less than the target CT value are used as zero pixel points (for example, the gray value is set to 0), and the highlighted area 201 corresponding to the first non-zero pixel points is obtained as the second processed image.

[0076] S103, obtain the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determine the sparsified representation result of the image intersection as the target input data.

[0077] Here, since the stent is usually implanted in the target blood vessel, impurity images (such as background images or images of other biological tissues at the target blood vessel) in the second processed image that contain some non-blood vessel areas may interfere with the detection of the stent. Based on this, when performing step S103, the blood vessel segmentation image of the target blood vessel can also be used to screen out specific image data (i.e., the above-mentioned image intersection) in the second processed image that only belongs to the target blood vessel by taking the intersection.

[0078] Specifically, as an optional embodiment, the vascular segmentation image of the target blood vessel can be obtained from the CTA image of the target site (i.e., the biological site where the target blood vessel is located, which is the medical image to be processed in step S101 above) in the manner shown in the following steps a1 - a3:

[0079] Step a1: First, roughly segment the CTA image of the target site to obtain a roughly segmented image of the area where the target blood vessel is located;

[0080] Step a2: Input the above - mentioned roughly segmented image into a multi - scale receptive field model, and output m seed points (i.e., multiple seed points) belonging to the target blood vessel in the roughly segmented image;

[0081] Step a3: Based on the obtained m seed points, use the skeleton algorithm (skeleton extraction algorithm) to obtain the center line of the target blood vessel, and then finely segment the above - mentioned roughly segmented image based on the obtained center line of the blood vessel to obtain the vascular segmentation image of the target blood vessel.

[0082] It should be noted that the existing methods for obtaining vascular segmentation images are not unique, and the embodiments of the present application do not limit the specific method for obtaining the above - mentioned vascular segmentation image.

[0083] Here, the above - mentioned image intersection is the image area in the second - processed image that belongs to the target blood vessel (as shown, it is equivalent to the image area in the highlighted area 201 that belongs to the target blood vessel). Based on the foregoing steps, it can be seen that both the second - processed image and the vascular segmentation image are three - dimensional images. Therefore, the image intersection obtained in step S103 is also a three - dimensional image. Figure 2 As shown, it is equivalent to the image area in the highlighted area 201 that belongs to the target blood vessel. Based on the foregoing steps, it can be seen that both the second - processed image and the vascular segmentation image are three - dimensional images. Therefore, the image intersection obtained in step S103 is also a three - dimensional image.

[0084] At this time, considering that it is difficult for the stent classification model to balance high - resolution processing and low video - memory consumption when processing the input three - dimensional image data (where, if low video - memory consumption is to be ensured, usually the image size of the input three - dimensional image data needs to be reduced, resulting in a decrease in the resolution of the three - dimensional image data). Therefore, after obtaining the above - mentioned image intersection, based on the idea of sparse convolution, only the part of the effective data related to stent detection (such as the position information, gray - scale value, etc. of each pixel point in the image intersection) can be extracted from the image intersection as the sparse representation result of the image intersection; and then the obtained sparse representation result is used as the target input data of the stent classification model, so that the stent classification model can balance high - resolution processing (i.e., it is not necessary to reduce the data size of the target input data) and low video - memory consumption (i.e., it is not necessary to process the complete three - dimensional image data) during the data - processing process.

[0085] S104. Input the target input data into a pre-trained stent classification model to obtain the stent detection result for the target blood vessel output by the stent classification model.

[0086] Here, the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

[0087] Specifically, most of the stents commonly used in current vascular intervention treatment plans have specific morphological characteristics (for example, when the stent is not bent, both bare metal stents and covered stents belong to cylindrical mesh structures, only the materials are different). Therefore, based on the specific morphological characteristics of the target stent, the stent classification model can perform binary classification prediction on the image region belonging to the target stent in the target input data, so as to output the stent detection result for the target stent.

[0088] It should be noted that in special cases, when the medical image to be processed contains special objects such as plaques (such as calcified plaques) that neither belong to normal biological tissues nor to stents, although the plaques also have a high CT value (that is, it is difficult to distinguish the image region where the stent is located and the image region where the plaque is located through the execution of steps S101 - S103), however, based on the obvious morphological differences between the plaques (mostly oval) and the stents, in step S104, the stent classification model can still effectively identify the morphology and position information of the target stent at the target blood vessel, so as to obtain a relatively accurate stent detection result.

[0089] The following will respectively describe in detail the specific implementation processes of the above steps in the embodiments of the present application:

[0090] Regarding the specific implementation process of the above step S101, when obtaining the first processed image by means of the vascular segmentation image of the target blood vessel, refer to Figure 3 as shown Figure 3 shows a schematic flowchart of a method for obtaining a first processed image provided by an embodiment of the present application. The method includes steps S301 - S302; specifically:

[0091] S301. Use the vascular segmentation image of the target blood vessel to perform image cropping on the medical image to be processed, and obtain the first image cropping result of the target blood vessel.

[0092] Here, taking the target blood vessel as the head and neck blood vessel as an example, the medical image to be processed can be a CTA image of the head and neck region; among them, the CTA image of the head and neck region includes an intracranial image region, a neck image region, an aortic arch image region, etc. That is, in the CTA image of the head and neck region, in addition to the image region where the head and neck blood vessels (i.e., the target blood vessels) are located, there will also be invalid image data of other biological tissues (i.e., invalid image data that does not need to be processed during stent detection).

[0093] Based on this, using the blood vessel segmentation image of the head and neck blood vessels as the cropping image boundary, cropping the CTA image of the head and neck region can obtain the image region where the head and neck blood vessels are located in the CTA image of the head and neck region (i.e., the above-mentioned first image cropping result).

[0094] S302, sample the first image cropping result according to a preset image sampling size, and use the sampling result of the first image cropping result as the first processed image.

[0095] Here, during the sampling process, the image region where the target blood vessels are located after cropping (i.e., the first image cropping result) can be sampled according to a preset image sampling size (where, since the first image cropping result also belongs to three-dimensional image data, different from using an m×n grid for image sampling in a two-dimensional image, the preset image sampling size in this embodiment of the present application can be in the form of a three-dimensional grid such as x1×y1×Z1), and the sampling result is used as the first processed image.

[0096] Specifically, the above-mentioned preset image sampling size can be adjusted according to actual image sampling requirements, and no specific limitation is made in this embodiment of the present application on the specific values of the image sampling size (i.e., the specific values of x1, y1, and Z1).

[0097] It should be noted that based on the above step S103, since the target input data as the model input data in the embodiments of the present application belongs to the sparse representation result (that is, the demand for video memory by the stent classification model has been reduced), therefore, compared with the commonly used image sampling size in the prior art (for example, a three-dimensional grid of 128×128×64 can be used for image sampling), in the above step S302 of the embodiments of the present application, a larger image sampling size (such as a three-dimensional grid of 512×512×600 can be used for image sampling) can be used to sample the above first image cropping result, so as to ensure high-resolution processing of the sampling result (that is, the first processed image) by using a larger image sampling size on the basis of reducing the demand for video memory by the stent classification model through the sparse representation result, so that when the stent classification model processes the input three-dimensional image data (that is, the target input data), it can take into account high-resolution processing and low video memory consumption.

[0098] Regarding the specific implementation process of the above step S102, considering that when performing CTA examinations in different hospitals, the scanning parameters and the dosage of the contrast agent used may vary, resulting in different CT values of the image area in the target blood vessel in different medical images to be processed (that is, CTA images of target parts with different specifications).

[0099] Based on this, in an optional implementation manner, in addition to the above method of directly selecting a relatively high target CT value (which can be selected within the range of 600-1600) in step S102 to perform threshold segmentation on the first processed image, the embodiments of the present application also provide a method for dynamically adjusting to determine the target CT value applicable to perform threshold segmentation on the first processed image (that is, the image area where the target blood vessel is located) of the current specification (determined according to the specific scanning parameters and the dosage of the contrast agent during CTA examination), referring to Figure 4 as shown Figure 4 shows a schematic flowchart of a method for determining the target CT value provided by the embodiments of the present application, and the method includes steps S401-S402; specifically:

[0100] S401, perform threshold segmentation on the first processed image using the initial CT value, and dynamically adjust the initial CT value within the preset CT value distribution range according to the number of second non-zero pixel points included in the threshold segmentation result until the number of second non-zero pixel points included in the threshold segmentation result meets the preset point selection condition.

[0101] Here, the initial CT value can be the minimum CT value within a preset CT value distribution range, or any preset value outside the preset CT value distribution range. The specific value of the above initial CT value is not limited in the embodiments of the present application.

[0102] Specifically, the above second non-zero pixel points are used to represent pixel points with a CT value greater than or equal to the initial CT value; that is, the specific implementation of threshold segmentation of the first processed image using the initial CT value in step S401 to determine the second non-zero pixel points is the same as the method of threshold segmentation of the first processed image using the target CT value in step S102 to determine the first non-zero pixel points, and the repeated parts will not be elaborated here.

[0103] It should be noted that the above preset point number screening condition can be a specific point number screening threshold (for example, taking the number of second non-zero pixel points being less than or equal to this point number screening threshold as meeting the preset point number screening condition), or a point number screening range interval (for example, taking the number of second non-zero pixel points being within the point number screening range interval [n1, n2] as meeting the preset point number screening condition); the specific setting method of the above preset point number screening condition is not limited in the embodiments of the present application.

[0104] S402. When the number of second non-zero pixel points included in the threshold segmentation result meets the preset point number screening condition, obtain the dynamically adjusted initial CT value as the target CT value.

[0105] Exemplarily, taking the above preset point number screening condition as "the number of second non-zero pixel points is less than or equal to 15000 points", the preset CT value distribution range as 600 - 1600, and the initial CT value as 600 (that is, the minimum CT value within the preset CT value distribution range) as an example, use the initial CT value 600 to perform threshold segmentation on the first processed image. If the number of second non-zero pixel points (that is, pixel points with a CT value greater than or equal to 600) obtained is 16000 (that is, does not meet the above preset point number screening condition), then the initial CT value can be dynamically adjusted from 600 to 700 in the preset CT value distribution range of 600 - 1600 at a step size of 100 CT values, and use the adjusted CT value 700 to re-perform threshold segmentation on the first processed image. If the number of second non-zero pixel points (that is, pixel points with a CT value greater than or equal to 700) obtained at this time is 14300 (that is, meets the above preset point number screening condition), then the dynamic adjustment of the initial CT value can be stopped, and the target CT value is determined to be 700.

[0106] Regarding the specific implementation process of the above step S103, in an optional implementation scheme, refer to Figure 5 as shown Figure 5The flowchart shows a method for determining target input data provided by an embodiment of the present application. The method includes steps S501 - S502; specifically:

[0107] S501, according to the blood vessel marking result of the target blood vessel included in the blood vessel segmentation image, obtain the image intersection between the blood vessel marking result and the second processed image as the target image intersection.

[0108] Here, the above blood vessel marking result is used to represent the marking result of pixel points belonging to the target blood vessel in the blood vessel segmentation image; for example, in a 0 - 1 marking manner, pixel points belonging to the target blood vessel in the blood vessel segmentation image can be marked as "1", and pixel points not belonging to the target blood vessel in the blood vessel segmentation image (such as pixel points belonging to the background image, etc.) can be marked as "0", so as to take the pixel points corresponding to the "1" marking as the blood vessel marking result of the target blood vessel, and obtain the image intersection between the blood vessel marking result and the second processed image.

[0109] S502, record the position information of each pixel point included in the target image intersection and the image parameter information corresponding to this pixel point, take the recording result as the sparse representation result of the target image intersection, and determine the recording result as the target input data.

[0110] Specifically, the image parameter information at least includes the gray value corresponding to this pixel point; that is, the above image parameter information is used to represent the image parameters effective for identifying and detecting the target stent at the target blood vessel. In this way, after obtaining the classification prediction result of the stent classification model for the image region belonging to the target stent in the target input data, based on the above position information and image parameter information recorded in step S502, through the inverse operation process of sparse representation, the image region where the finally detected target stent is located can be restored to the same image space as the initially obtained medical image to be processed, so as to be able to locate the target stent in the medical image to be processed.

[0111] It should be noted that regarding the specific recording method of the above sparse representation result, a hash table can be selected for recording and storage. The present application embodiment does not limit the specific recording method of the above sparse representation result. [[ID=1,7]]

[0112] For the stent classification model in the above step S104, in the embodiments of the present application, medical images marked with stent regions and non-stent regions can be used in advance as training sample data (where the specific types of blood vessels included in the training sample data can be determined according to actual stent detection requirements, and the embodiments of the present application do not make any limitations in this regard). The initial classification model is trained for binary classification, and the semantic loss between the classification prediction result and the true marking result is used to adjust the model parameters of the initial classification model until the initial classification model converges. The converged initial classification model is used as the trained stent classification model.

[0113] It should be noted that considering that the target input data used in the embodiments of the present application is the sparsified representation result of the image intersection based on the idea of sparse convolution, based on this, in the embodiments of the present application, as a preferred embodiment, the above stent classification model can be a classification model including a sparse convolution network (such as a sparse ResUNet classification model, etc.); in addition, the above stent classification model can also be any sparsified 3D classification model (that is, it can process sparsified three-dimensional image data), and the embodiments of the present application do not make any limitations on the specific model structure of the above stent classification model.

[0114] Based on this, in an optional implementation, refer to Figure 6 as shown Figure 6 FIG. shows a schematic flowchart of a method for obtaining a stent detection result provided by the embodiments of the present application. When performing step S104, the method includes steps S601 - S602; specifically:

[0115] S601, input the target input data into the pre-trained stent classification model, and based on the morphological characteristics of the target stent, classify and predict the image region belonging to the target stent in the target input data through the stent classification model to obtain an initial classification prediction result.

[0116] Specifically, based on the relevant analysis content in the foregoing step S104, it can be known that after threshold segmentation of the first processed image using a relatively high target CT value, the main entity objects retained in the target input data are the target stent and possible abnormal biological tissues such as calcified plaques (that is, the part with a CT value higher than other normal biological tissues at the target blood vessel and cannot be directly distinguished only by threshold segmentation). Based on this, in the stent classification model, the obvious morphological differences between plaques (mostly oblate) and stents (mostly cylindrical reticular structures) can be further utilized to identify the image region belonging to the target stent in the target input data, so as to obtain a relatively accurate recognition and detection result of the target stent (that is, the above initial classification prediction result).

[0117] It should be noted that based on the foregoing steps, the target input data belongs to the sparsified representation result (not the complete three-dimensional image data). At this time, the stent classification model can take into account both high-resolution processing and low video memory consumption. Therefore, the image size of the above initial classification prediction result output by the stent classification model is the same as the image size of the target input data in the input model (that is, the input and output of the model correspond to the same image size), and there is no need to reduce the input data / enlarge the output data as in the prior art during the data processing of the stent classification model.

[0118] S602. According to the inverse operation of the sparsified representation corresponding to the target input data, restore the initial classification prediction result to the image space corresponding to the second processed image, and output the stent detection result.

[0119] Here, based on the foregoing steps, the target input data belongs to the sparsified representation result. On this basis, the above initial classification prediction result obtained by the stent classification model also belongs to the sparsified representation result. On this basis, by executing step S602, the above initial classification prediction result can be mapped back to the image space corresponding to the second processed image, and the stent detection result is output.

[0120] It should be noted that since the sparsified representation operation performed in step S103 is for the above image intersection, the second processed image, the first processed image, and the medical image to be processed before executing this sparsified representation operation are all in the same image space. That is, the execution of the above step S602 is also equivalent to restoring the above initial classification prediction result obtained by the stent classification model to the image space corresponding to the medical image to be processed.

[0121] It should be noted that the implementation of the inverse operation process of the sparsified representation in the above step S602 can refer to the forward operation process of the sparsified representation shown in the above steps S501 - S502, and the repeated parts will not be elaborated here.

[0122] Regarding the stent detection result obtained in the above step S104, since the stent classification model performs classification prediction on each pixel point in the target input data, it is basically impossible to ensure that all pixel points in each category (that is, the target category belonging to the target stent / other categories not belonging to the target stent) are continuous (equivalent to there may be discrete pixel points that cannot form a connected domain). At this time, the stent detection result includes: multiple image connected domains belonging to the target stent (that is, connected domains composed of multiple continuous pixel points belonging to the target stent) and multiple target discrete pixel points belonging to the target stent.

[0123] Based on this, in an optional implementation scheme, refer to Figure 7a as shown.Figure 7a FIG. 1 shows a schematic flowchart of a method for region growing of discrete pixel points in the stent detection result provided by an embodiment of the present application. After step S104 is executed, the method includes steps S701-S702; specifically:

[0124] S701, for each of the target discrete pixel points, according to the stent pixel point closest to each of the target discrete pixel points in the stent detection result, perform region growing on each of the target discrete pixel points to obtain a first region growing result corresponding to the stent detection result.

[0125] Here, the stent pixel point is used to represent the pixel point belonging to the target stent in the stent detection result; that is, the stent pixel point can be a pixel point in the image connected domain belonging to the target stent, or a target discrete pixel point belonging to the target stent.

[0126] Specifically, as Figure 7b shown, Figure 7b FIG. 2 shows a schematic structural diagram of a stent detection result provided by an embodiment of the present application. In the stent detection result, there is an image connected domain 710 belonging to the target stent and multiple target discrete pixel points 720 belonging to the target stent. At this time, when executing step S701, for each target discrete pixel point 720, the stent pixel point closest to the target discrete pixel point can be determined from the pixel points belonging to the target stent (which can be other target discrete pixel points or pixel points in the nearby image connected domain), and region growing is performed on the target discrete pixel point along the direction pointing to the stent pixel point, so as to reduce the number of discrete pixel points belonging to the target stent in the stent detection result, which is beneficial to obtaining a more complete target stent in shape.

[0127] S702, when the number of target discrete pixel points included in the first region growing result is greater than or equal to a preset number threshold, perform secondary region growing on each target discrete pixel point included in the first region growing result according to a preset growing strategy, and use the secondary region growing result as the final detection result for the target stent.

[0128] Here, the above preset number threshold can be set according to actual stent detection requirements, and no specific limitation is made on the specific value of the above preset number threshold in the embodiments of the present application.

[0129] Specifically, the morphological integrity of the above-mentioned target stent shown in the final detection result is higher than that shown in the stent detection result; that is, the above-mentioned preset growth strategy is essentially a strategy for obtaining a larger number of pixel points belonging to the target stent based on the original stent detection result. In this way, based on the preset growth strategy, through the method of secondary region growing, the number of discrete pixel points belonging to the target stent in the stent detection result can be further reduced on the basis of the first region growing result, which is conducive to obtaining a target stent with a more complete morphology (i.e., the above-mentioned final detection result).

[0130] It should be noted that the above-mentioned preset growth strategy is only a strategy for obtaining a larger number of pixel points belonging to the target stent based on the original stent detection result. Based on this, the specific content of the above-mentioned preset growth strategy is not unique, and the embodiments of the present application do not make any limitations on the specific content of the above-mentioned preset growth strategy.

[0131] Two optional preset growth strategies are given below to introduce the implementation process of the above step S702 in detail. Specifically:[[]]END]

[0132] In the embodiments of the present application, when performing step S702, as an optional embodiment, the secondary region growing can be performed on each target discrete pixel point included in the first region growing result in the manner shown in the following steps b1 to b3. Specifically:

[0133] Step b1: Perform threshold segmentation on the first processed image using the first CT value, and take the image region where the third non-zero pixel point is located in the threshold segmentation result as the third processed image.

[0134] Step b2: Obtain each of the third non-zero pixel points included in the third processed image and each pixel point belonging to the target stent in the stent detection result as the first pixel point set.

[0135] Step b3: For each target discrete pixel point included in the first region growing result, perform secondary region growing on each target discrete pixel point according to the pixel point in the first pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growing result.

[0136] Regarding the above steps b1 - b3, it should be noted that the first CT value is less than the target CT value; that is, in the manner shown in steps b1 - b3, the preset growth strategy is to obtain a larger number of pixel points belonging to the target stent by re-performing threshold segmentation (i.e., the number of third non-zero pixel points is more than the number of first non-zero pixel points in the original second processed image).

[0137] Here, the third non-zero pixel points are used to represent pixel points with a CT value greater than or equal to the first CT value. Among them, in step b1, threshold segmentation is performed on the first processed image using the first CT value to determine the specific implementation of the third non-zero pixel points, which is the same as the method of using the target CT value to perform threshold segmentation on the first processed image in step S102 to determine the first non-zero pixel points. Repetitive parts will not be elaborated here.

[0138] Specifically, the implementation of secondary region growing in step b3 above is the same as the implementation of region growing in step S701 above. Repetitive parts will not be elaborated here.

[0139] In the embodiment of the present application, when performing step S702, as another optional embodiment, the secondary region growing can also be performed on each target discrete pixel point included in the first region growing result in the manner shown in steps c1 - c3 below. Specifically:

[0140] Step c1: Obtain pixel points with a probability of belonging to the target stent higher than the second preset threshold from the classification prediction result of the stent classification model for the target stent as candidate pixel points.

[0141] Here, the pixel points belonging to the target stent in the stent detection result are used to represent pixel points with a probability of belonging to the target stent higher than the first preset threshold in the classification prediction result of the stent classification model for the target stent.

[0142] Here, the second preset threshold is less than the first preset threshold. For example, if the stent detection result is obtained by the stent classification model outputting pixel points with a probability of belonging to the target stent higher than 60% (i.e., the first preset threshold), then the stent classification model can output pixel points with a probability of belonging to the target stent higher than 50% (i.e., the second preset threshold), so as to obtain more pixel points belonging to the target stent.

[0143] Step c2: Obtain each of the candidate pixel points and each pixel point belonging to the target stent in the stent detection result as the second pixel point set.

[0144] Step c3: For each target discrete pixel point included in the first region growing result, perform secondary region growing on each target discrete pixel point according to the pixel point in the second pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growing result.

[0145] Specifically, the implementation of secondary region growing in steps c2 - c3 above is the same as the implementation of region growing in step S701 above. Repetitive parts will not be elaborated here.

[0146] In addition to the above regional growth steps, in the embodiments of the present application, based on the actual stent detection requirements, other post-processing steps can also be continued for the final detection result of the target stent obtained after regional growth. The embodiments of the present application do not make any limitations on the specific post-processing steps that can be performed subsequently.

[0147] Exemplarily, when the first processed image in step S101 is obtained by the "cropping - sampling" method shown in steps S301 - S302, the final detection result of the target stent can also be resampled back to the original image size corresponding to the medical image to be processed based on the preset image sampling size in step S302 above; when the final detection result obtained after regional growth still contains discrete pixel points that cannot form a connected domain, redundant image data such as discrete pixel points / image connected domains / image classification fragments with a volume smaller than the preset volume threshold can also be filtered out from the final detection result through volume filtering, and only the effective image data that can reflect the main features such as the position and morphology of the target stent in the final detection result (i.e., the image connected domain with a volume greater than the preset volume threshold and belonging to the target stent) is retained.

[0148] Through the above stent detection method provided by the present application, the embodiments of the present application can obtain the image region where the target blood vessel is located from the medical image to be processed as the first processed image; perform threshold segmentation on the first processed image using the target CT value, and take the image region where the first non-zero pixel point is located in the threshold segmentation result as the second processed image; obtain the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determine the sparsified representation result of the image intersection as the target input data; input the target input data into the pre-trained stent classification model to obtain the stent detection result output by the stent classification model for the target blood vessel.

[0149] In this way, the present application can quickly locate the target stent in the medical image to be processed based on the characteristics such as the material density and morphology of the stent, which is beneficial to improving the detection efficiency of the stent in the medical image data.

[0150] Based on the same inventive concept, the present application also provides a detection device corresponding to the above stent detection method. Since the principle of solving problems by the detection device in the embodiments of the present application is similar to the above stent detection method in the embodiments of the present application, the implementation of the detection device can refer to the implementation of the above detection method, and the repeated parts will not be elaborated.

[0151] Refer to Figure 8 as shown in Figure 8 The structural schematic diagram of a stent detection device provided by the embodiments of the present application is shown. The detection device includes:

[0152] The first processing module 801 is configured to obtain, from the medical image to be processed, the image region where the target blood vessel is located as the first processed image;

[0153] The second processing module 802 is configured to perform threshold segmentation on the first processed image by using the target CT value, and use the image region where the non-zero pixel points are located in the threshold segmentation result as the second processed image;

[0154] The sparse representation module 803 is configured to obtain the image intersection between the second processed image and the blood vessel segmentation image of the target blood vessel, and determine the sparse representation result of the image intersection as the target input data;

[0155] The stent detection module 804 is configured to input the target input data into a pre-trained stent classification model, and obtain the stent detection result for the target blood vessel output by the stent classification model; wherein, the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

[0156] In an optional implementation manner, when obtaining, from the medical image to be processed, the image region where the target blood vessel is located as the first processed image, the first processing module 801 is configured to:

[0157] Perform image cropping on the medical image to be processed by using the blood vessel segmentation image of the target blood vessel, and obtain the first image cropping result of the target blood vessel;

[0158] Sample the first image cropping result according to a preset image sampling size, and use the sampling result of the first image cropping result as the first processed image.

[0159] In an optional implementation manner, the second processing module 802 is configured to determine the target CT value by the following method:

[0160] Perform threshold segmentation on the first processed image by using the initial CT value, and dynamically adjust the initial CT value within a preset CT value distribution range according to the number of second non-zero pixel points included in the threshold segmentation result until the number of second non-zero pixel points included in the threshold segmentation result meets a preset point number screening condition; wherein, the second non-zero pixel points are used to characterize the pixel points whose CT value is greater than or equal to the initial CT value;

[0161] When the number of second non-zero pixel points included in the threshold segmentation result meets the preset point number screening condition, obtain the dynamically adjusted initial CT value as the target CT value.

[0162] In an alternative embodiment, when obtaining an image intersection between the second processed image and the vascular segmentation image of the target blood vessel and determining the sparsified representation result of the image intersection as the target input data, the sparse representation module 803 is configured to:

[0163] Obtain an image intersection between the vascular marking result of the target blood vessel included in the vascular segmentation image and the second processed image as the target image intersection according to the vascular marking result of the target blood vessel;

[0164] Record the position information of each pixel point included in the target image intersection and the image parameter information corresponding to the pixel point, take the recording result as the sparsified representation result of the target image intersection, and determine the recording result as the target input data; wherein, the image parameter information at least includes the gray value corresponding to the pixel point.

[0165] In an alternative embodiment, when inputting the target input data into a pre-trained stent classification model to obtain the stent detection result for the target blood vessel output by the stent classification model, the stent detection module 804 is configured to:

[0166] Input the target input data into a pre-trained stent classification model, and based on the morphological features of the target stent, classify and predict the image region belonging to the target stent in the target input data through the stent classification model to obtain an initial classification prediction result;

[0167] Restore the initial classification prediction result to the image space corresponding to the second processed image according to the inverse operation of the sparsified representation corresponding to the target input data, and output the stent detection result.

[0168] In an alternative embodiment, the stent detection result includes: a plurality of image connected domains belonging to the target stent and a plurality of target discrete pixel points belonging to the target stent; after obtaining the stent detection result for the target blood vessel output by the stent classification model, the stent detection module 804 is further configured to:

[0169] For each of the target discrete pixel points, perform region growing on each of the target discrete pixel points according to the stent pixel point closest to each of the target discrete pixel points in the stent detection result to obtain a first region growing result corresponding to the stent detection result; wherein, the stent pixel points are used to represent the pixel points belonging to the target stent in the stent detection result;

[0170] When the number of target discrete pixel points included in the growth result of the first region is greater than or equal to a preset number threshold, according to a preset growth strategy, perform secondary region growth on each target discrete pixel point included in the growth result of the first region, and use the secondary region growth result as the final detection result for the target stent; wherein, the morphological integrity of the target stent shown in the final detection result is higher than that shown in the stent detection result.

[0171] In an optional implementation manner, when performing secondary region growth on each target discrete pixel point included in the growth result of the first region according to the preset growth strategy, the stent detection module 804 is configured to:

[0172] Perform threshold segmentation on the first processed image using a first CT value, and use the image region where the third non-zero pixel points in the threshold segmentation result are located as the third processed image; wherein, the first CT value is less than the target CT value; the third non-zero pixel points are used to represent pixel points whose CT values are greater than or equal to the first CT value;

[0173] Obtain each of the third non-zero pixel points included in the third processed image and each pixel point belonging to the target stent in the stent detection result as a first pixel point set;

[0174] For each target discrete pixel point included in the growth result of the first region, perform secondary region growth on each target discrete pixel point according to the pixel point in the first pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growth result.

[0175] In an optional implementation manner, the pixel points belonging to the target stent in the stent detection result are used to represent pixel points whose probability of belonging to the target stent in the classification prediction result of the stent classification model for the target stent is higher than a first preset threshold; when performing secondary region growth on each target discrete pixel point included in the growth result of the first region according to the preset growth strategy, the stent detection module 804 is further configured to:

[0176] Obtain, from the classification prediction result of the stent classification model for the target stent, pixel points whose probability of belonging to the target stent is higher than a second preset threshold as candidate pixel points; wherein, the second preset threshold is less than the first preset threshold;

[0177] Obtain each of the candidate pixel points and each pixel point belonging to the target stent in the stent detection result as a second pixel point set;

[0178] For each target discrete pixel point included in the growth result of the first region, perform secondary region growth on each target discrete pixel point according to the pixel point in the second pixel point set that is closest to each target discrete pixel point, to obtain the secondary region growth result.

[0179] Through the above-mentioned stent detection device provided by the present application, embodiments of the present application can obtain an image region where a target blood vessel is located from a to-be-processed medical image as a first processed image; perform threshold segmentation on the first processed image using a target CT value, and use the image region where the first non-zero pixel point is located in the threshold segmentation result as a second processed image; obtain an image intersection between the second processed image and a blood vessel segmentation image of the target blood vessel, and determine a sparsified representation result of the image intersection as target input data; input the target input data into a pre-trained stent classification model to obtain a stent detection result for the target blood vessel output by the stent classification model.

[0180] In this way, the present application can quickly locate the target stent in the to-be-processed medical image based on characteristics such as the material density and morphology of the stent, which is beneficial to improving the detection efficiency of the stent in medical image data.

[0181] As Figure 9 shown, embodiments of the present application provide a computer device 900 for executing the stent detection method in the present application. The device includes a memory 901, a processor 902, and a computer program stored on the memory 901 and executable on the processor 902. When the processor 902 executes the computer program, the steps of the above-mentioned stent detection method are implemented.

[0182] Specifically, the above-mentioned memory 901 and processor 902 can be general memories and processors, which are not specifically limited here. When the processor 902 runs the computer program stored in the memory 901, it can execute the above-mentioned stent detection method.

[0183] Corresponding to the stent detection method in the present application, embodiments of the present application further provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium, and when the computer program is run by a processor, the steps of the above-mentioned stent detection method are executed.

[0184] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, it can execute the above-mentioned stent detection method.

[0185] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the systems or units can be in electrical, mechanical or other forms.

[0186] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0187] In addition, each functional unit in the embodiments provided in the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0188] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0189] It should be noted that similar reference numerals and letters denote similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0190] Finally, it should be noted that the above-described embodiments are only specific implementation manners of the present application, used to illustrate the technical solutions of the present application, rather than limiting it. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the art within the technical scope disclosed by the present application can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. All should be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for detecting a stent, characterized in that: The detection method comprises: Acquire, from the medical image to be processed, an image region where the target blood vessel is located as a first processed image; Performing threshold segmentation on the first processed image using the target CT value, and using the image region where a first non-zero pixel in the threshold segmentation result is located as the second processed image; wherein the first non-zero pixel is used to represent a pixel whose CT value is greater than or equal to the target CT value; acquiring an image intersection between the second processed image and the vessel segmentation image of the target vessel, and determining a sparse representation result of the image intersection as target input data; The target input data is input into a pre-trained stent classification model to obtain a stent detection result for the target blood vessel output by the stent classification model; wherein the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

2. The detection method according to claim 1, wherein The step of acquiring, from the medical image to be processed, an image region where the target blood vessel is located as a first processed image includes: performing image cropping on the medical image to be processed using the blood vessel segmentation image of the target blood vessel to obtain a first image cropping result of the target blood vessel; The first image cropping result is sampled according to a preset image sampling size, and the sampling result of the first image cropping result is used as the first processed image.

3. The detection method according to claim 1, wherein The target CT value is determined by the following method: performing threshold segmentation on the first processed image using the initial CT value, and dynamically adjusting the initial CT value within a preset CT value distribution range based on the number of second non-zero pixels included in the threshold segmentation result until the number of second non-zero pixels included in the threshold segmentation result meets a preset point number screening condition; wherein the second non-zero pixels are used to represent pixels having a CT value greater than or equal to the initial CT value; When the number of second non-zero pixels included in the threshold segmentation result meets the preset point number screening condition, a dynamically adjusted initial CT value is obtained as the target CT value.

4. The detection method according to claim 1, wherein The acquiring of an image intersection between the second processed image and the vessel segmentation image of the target vessel, and determining a sparse representation result of the image intersection as target input data, includes: acquiring, according to the vessel labeling result of the target vessel contained in the vessel segmentation image, an image intersection between the vessel labeling result and the second processed image as a target image intersection; Record the position information of each pixel point contained in the intersection of the target images and the image parameter information corresponding to the pixel point, use the recording result as the sparse representation result of the intersection of the target images, and determine the recording result as the target input data; wherein the image parameter information at least includes the grayscale value corresponding to the pixel point.

5. The detection method according to claim 1, wherein Inputting the target input data into a pre-trained stent classification model to obtain a stent detection result for the target blood vessel output by the stent classification model includes: Inputting the target input data into a pre-trained stent classification model, and performing classification prediction on the image area belonging to the target stent in the target input data by the stent classification model based on the morphological characteristics of the target stent, to obtain an initial classification prediction result; According to the inverse operation of the sparse representation corresponding to the target input data, the initial classification prediction result is restored to the image space corresponding to the second processed image, and the bracket detection result is output.

6. The detection method according to claim 1, characterized in that The stent detection result includes: a plurality of image connected domains belonging to the target stent and a plurality of target discrete pixel points belonging to the target stent; after obtaining the stent detection result for the target blood vessel output by the stent classification model, the detection method further includes: For each of the target discrete pixel points, region growing is performed on each of the target discrete pixel points according to the bracket pixel point closest to each of the target discrete pixel points in the bracket detection result, to obtain a first region growing result corresponding to the bracket detection result; wherein the bracket pixel point is used to represent the pixel point belonging to the target bracket in the bracket detection result; When the number of target discrete pixel points contained in the first region growth result is greater than or equal to a preset number threshold, secondary region growth is performed on each target discrete pixel point contained in the first region growth result according to a preset growth strategy, and the secondary region growth result is used as the final detection result for the target bracket; wherein, the morphological integrity of the target bracket in the final detection result is higher than the morphological integrity in the bracket detection result.

7. The detection method according to claim 6, characterized in that The performing secondary region growing on each target discrete pixel point included in the first region growing result according to a preset growth strategy includes: performing threshold segmentation on the first processed image using the first CT value, and using an image region where a third non-zero pixel in the threshold segmentation result is located as a third processed image; wherein the first CT value is less than the target CT value; and the third non-zero pixel is used to represent a pixel whose CT value is greater than or equal to the first CT value; Acquire each of the third non-zero pixels contained in the third processed image and each pixel belonging to the target bracket in the bracket detection result as a first pixel set; For each target discrete pixel point included in the first region growing result, secondary region growing is performed on each target discrete pixel point based on the pixel point in the first pixel point set that is closest to each target discrete pixel point to obtain the secondary region growing result.

8. The detection method according to claim 6, characterized in that The pixel points belonging to the target bracket in the bracket detection result are used to represent the pixel points whose probability of belonging to the target bracket in the classification prediction result of the bracket classification model for the target bracket is higher than a first preset threshold; performing secondary region growing on each target discrete pixel point included in the first region growing result according to a preset growth strategy also includes: From the classification prediction result of the bracket classification model for the target bracket, obtaining pixel points whose probability of belonging to the target bracket is higher than a second preset threshold as candidate pixel points; wherein the second preset threshold is lower than the first preset threshold; Acquire each of the candidate pixel points and each pixel point belonging to the target bracket in the bracket detection result as a second pixel point set; For each target discrete pixel point included in the first region growing result, secondary region growing is performed on each target discrete pixel point based on the pixel point in the second pixel point set that is closest to each target discrete pixel point to obtain the secondary region growing result.

9. A detection device for a stent, characterized in that: The detection device comprises: A first processing module is configured to obtain, from the medical image to be processed, an image region where a target blood vessel is located as a first processed image; a second processing module, configured to perform threshold segmentation on the first processed image using the target CT value, and use the image region where the first non-zero pixel in the threshold segmentation result is located as the second processed image; wherein the first non-zero pixel is used to represent a pixel whose CT value is greater than or equal to the target CT value; a sparse representation module, configured to obtain an image intersection between the second processed image and the vessel segmentation image of the target vessel, and determine a sparse representation result of the image intersection as target input data; The stent detection module is used to input the target input data into a pre-trained stent classification model to obtain the stent detection result for the target blood vessel output by the stent classification model; wherein the stent detection result is used to characterize the morphology and position information of the target stent located at the target blood vessel.

10. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the detection method of the bracket as described in any one of claims 1 to 8 are performed.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for detecting a stent according to any one of claims 1 to 8 are executed.

Images