Coronary angiography image processing method and system based on Mama neural network

Abstract

The invention is suitable for the technical field of medical image analysis, and provides a coronary angiography image processing method and system based on a Mama neural network, and the method comprises the steps: obtaining the information of a to-be-detected image, accurately generating the information of an image segmentation result according to the information of the to-be-detected image and a preset blood vessel segmentation model, and carrying out the segmentation of the to-be-detected image. And finally, according to the image segmentation result information and a preset stenosis detection model, effectively generating detection result information. According to the method, subjectivity in diagnosis can be reduced, the accuracy of coronary angiography stenosis detection can be comprehensively improved, accurate segmentation can be achieved under the conditions of small blood vessels, background noise and the like, it is guaranteed that a complete and uninterrupted blood vessel structure is obtained, the diagnosis process is more reliable, and a more accurate reference basis is provided for clinicians.

Description

Technical Field

[0001] The present application relates to the technical field of medical image analysis, and in particular to a coronary angiography image processing method and system based on a Mamba neural network. Background Art

[0002] Coronary Angiography (CAG) is an important imaging examination method for clinical diagnosis of coronary artery disease. It can clearly show the blood flow of the coronary arteries and the degree of stenosis of the blood vessels. It is widely regarded as the "gold standard" for diagnosing coronary heart disease. Coronary angiography is to obtain vascular images after injecting contrast agents. Doctors diagnose and intervene in coronary artery lesions based on the images.

[0003] At present, traditional coronary angiography image analysis relies on experienced professional doctors to manually evaluate the images, including vascular segmentation and detection of stenosis areas. The entire manual evaluation process is easily affected by human factors, resulting in a highly subjective diagnosis. Especially in complex cases, the differences in subjective judgments will be more obvious, and there is a problem of low accuracy, which needs further improvement. Summary of the invention

[0004] Based on this, an embodiment of the present application provides a coronary angiography image processing method and system based on a Mamba neural network to solve the problem of low accuracy in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a method for processing coronary angiography images based on a Mamba neural network, the method comprising:

[0006] Obtaining image information to be detected;

[0007] Generate image segmentation result information according to the image information to be detected and a preset blood vessel segmentation model;

[0008] Detection result information is generated according to the image segmentation result information and a preset stenosis detection model.

[0009] Compared with the prior art, the beneficial effects are as follows: the coronary angiography image processing method based on the Mamba neural network provided in the embodiment of the present application, the terminal device can first obtain the image information to be detected, and then accurately generate the image segmentation result information according to the image information to be detected and the preset blood vessel segmentation model, and finally effectively generate the detection result information according to the image segmentation result information and the preset stenosis detection model, thereby reducing the subjective factors in the diagnosis process and greatly improving the accuracy of stenosis detection in coronary angiography. Whether it is small blood vessels, complex structures, or background noise interference, accurate segmentation can be achieved to obtain a continuous and unbroken blood vessel structure, which to a certain extent solves the current problem of low accuracy.

[0010] In a second aspect, an embodiment of the present application provides a coronary angiography image processing system based on a Mamba neural network, the system comprising:

[0011] The module for acquiring information of the image to be detected is used to acquire information of the image to be detected;

[0012] An image segmentation result information generating module is used to generate image segmentation result information according to the image information to be detected and a preset blood vessel segmentation model;

[0013] The detection result information generating module is used to generate the detection result information according to the image segmentation result information and the preset stenosis detection model.

[0014] In a third aspect, an embodiment of the present application provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of the first aspect described above when executing the computer program.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method of the first aspect described above are implemented.

[0016] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0018] Figure 1 is a flow chart of a coronary angiography image processing method provided in one embodiment of the present application;

[0019] Figure 2is a schematic diagram of an original coronary angiography provided by an embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of the flow chart before step S200 in the coronary angiography image processing method provided in one embodiment of the present application;

[0021] Figure 4 (a) is a schematic diagram of original blood vessel image information provided by an embodiment of the present application. Figure 4 (b) is a schematic diagram of predicted image information provided by an embodiment of the present application. Figure 4 (c) is a schematic diagram of the corrected predicted image information provided by an embodiment of the present application;

[0022] Figure 5 is a schematic diagram of the model structure of a blood vessel segmentation model provided in an embodiment of the present application;

[0023] Figure 6 is a flow chart after step S205 in the coronary angiography image processing method provided in one embodiment of the present application;

[0024] Figure 7 (a) is a schematic diagram of a target image provided by an embodiment of the present application. Figure 7 (b) is a schematic diagram of the detection results of the medical image segmentation model based on MedSegDiff provided in an embodiment of the present application. Figure 7 (c) is a schematic diagram of a detection result based on a blood vessel segmentation model provided in an embodiment of the present application. Figure 7 (d) is a schematic diagram of the detection result of the blood vessel segmentation model based on the modified objective function provided in one embodiment of the present application;

[0025] Figure 8 is a flowchart of step S200 in the coronary angiography image processing method provided in one embodiment of the present application;

[0026] Fig. 9 is a flowchart of step S220 in the coronary angiography image processing method provided in one embodiment of the present application;

[0027] Fig.10 It is a flowchart after step S202 in the coronary angiography image processing method provided in one embodiment of the present application;

[0028] Fig.11 is a schematic diagram of image segmentation result information provided by an embodiment of the present application;

[0029] Fig.12 It is a schematic diagram of the flow chart before step S300 in the coronary angiography image processing method provided in one embodiment of the present application;

[0030] Fig.13 is a schematic diagram of the model structure of a stenosis detection model provided in an embodiment of the present application;

[0031] Fig.14 (a) is a schematic diagram of an original image of a blood vessel provided in an embodiment of the present application. Fig.14 (b) is a first schematic diagram of a stenosis detection result provided in an embodiment of the present application. Fig.14 (c) is a schematic diagram of a binary image of a blood vessel provided in an embodiment of the present application. Fig.14 (d) is a second schematic diagram of a stenosis detection result provided in an embodiment of the present application. Fig.14 (e) is a schematic diagram of a binary original image of a blood vessel provided in an embodiment of the present application. Fig.14 (f) is a third schematic diagram of a stenosis detection result provided in an embodiment of the present application;

[0032] Fig.15 is a module block diagram of a coronary angiography image processing system provided by an embodiment of the present application;

[0033] Fig.16 It is a schematic diagram of a terminal device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0034] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0035] In the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0036] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0037] In order to illustrate the technical solution described in this application, a specific embodiment is provided below for illustration.

[0038] At present, with the development of computer vision and deep learning technology, medical image analysis technology based on neural networks has gradually become an effective means to improve diagnostic efficiency and accuracy. Coronary angiography stenosis detection usually includes two main steps: vessel segmentation and stenosis area detection based on segmentation results; vessel segmentation aims to accurately extract the morphology of coronary arteries from coronary angiography images, and stenosis detection identifies the lesion area by analyzing the changes in the diameter of the blood vessels.

[0039] Traditional image processing methods based on manual features have problems such as low efficiency, inability to handle complex image deformation, and limited accuracy, and face major adjustments in practical applications. In contrast, deep learning models, especially convolutional neural networks (CNNs), can automatically learn deep features in large-scale data and effectively improve the accuracy and robustness of image processing.

[0040] Currently, vascular segmentation technology mainly relies on deep learning networks such as U-Net and FCN (fully convolutional network). The angiography images generated by these deep learning networks are usually accompanied by strong noise, complex vascular curves and overlapping structures, especially in the identification of subtle blood vessels and lesion areas, it is difficult to achieve the high precision required for clinical applications.

[0041] In terms of stenosis detection, traditional methods usually use threshold algorithms based on morphological features or regression models based on deep learning for diameter assessment. However, these methods fail to comprehensively consider the global contextual information of the blood vessels and are easily disturbed by local abnormal images (such as artifacts and noise), resulting in detection errors. Other deep learning-based methods rely heavily on vascular segmentation. If the vascular segmentation is not accurate, it will directly affect the results of stenosis detection, resulting in missed diagnosis or misdiagnosis.

[0042] See also Figure 1 , Figure 1 : is a flow chart of the coronary angiography image processing method provided in the embodiment of the present application. In the present embodiment, the execution subject of the coronary angiography image processing method is a terminal device. It is understandable that the types of terminal devices include but are not limited to mobile phones, tablet computers, laptop computers, ultra-mobile personal computers (Ultra-Mobile Personal Computer, UMPC), netbooks, personal digital assistants (Personal Digital Assistant, PDA), etc., and the present embodiment of the present application does not impose any restrictions on the specific types of terminal devices.

[0043] See also Figure 1 The coronary angiography image processing method provided in the embodiment of the present application includes but is not limited to the following steps:

[0044] In S100, image information to be detected is obtained.

[0045] Specifically, the terminal device can obtain the image information to be detected. For example, see Figure 2 The image information to be detected is used to describe the image to be processed for blood vessel segmentation and stenosis area detection, that is, the original image of coronary angiography.

[0046] In S200, image segmentation result information is generated according to the image information to be detected and the preset blood vessel segmentation model.

[0047] Specifically, after the terminal device acquires the image information to be detected, the terminal device can generate image segmentation result information according to the image information to be detected and a preset blood vessel segmentation model, wherein the image segmentation result information is used to describe the image information to be detected after blood vessel segmentation processing has been performed.

[0048] In some possible implementations, in order to achieve accurate blood vessel segmentation, please refer to Figure 3 Before step S200, the method further includes but is not limited to the following steps:

[0049] In S201, blood vessel original image set information is obtained.

[0050] Specifically, the terminal device may first obtain the original blood vessel image set information, for example, see Figure 4 In (a), the blood vessel original image set information includes multiple blood vessel original image information.

[0051] In S202, based on a preset blood vessel segmentation network and according to the blood vessel original image set information, preliminary prediction image set information and image label set information are generated.

[0052] Specifically, after the terminal device obtains the vascular original image set information, the terminal device can use a preset vascular segmentation network to make a preliminary prediction of each vascular original image information in the vascular original image set information and obtain the corresponding label, thereby generating preliminary predicted image set information and image label set information, wherein the vascular segmentation network can be a pre-trained Diffus ionMedSeg.

[0053] For example, see Figure 4 In (b), the predicted image set information includes multiple predicted image information; the image label set information includes multiple image label information; the predicted image information and the original blood vessel image information correspond one to one, and the predicted image information and the image label information correspond one to one.

[0054] In S203, for each predicted image information: based on the preset medical image processing software, manual correction information is obtained.

[0055] Specifically, after the terminal device generates preliminary predicted image set information and image label set information, the terminal device can obtain manual correction information for each predicted image information based on preset medical image processing software, thereby facilitating the detection personnel to perform manual correction with the help of the medical image processing software and further optimize subsequent prediction results, wherein the medical image processing software can be ITK-SNAP.

[0056] In S204, blood vessel segmentation data set information is generated according to the predicted image information, the image label information and the manual correction information.

[0057] Specifically, after the terminal device obtains the manual correction information, the terminal device can correct the image label information in the predicted image information based on the manual correction information to generate the blood vessel segmentation data set information, thereby improving the consistency and accuracy of the label and solving the inefficiency and high cost of the coronary angiography image data labeling problem; for example, please refer to Figure 4 (c) in Figure 4 (c) in the figure is the image after manual label correction.

[0058] It should be noted that manual correction is mainly concentrated in the stage of adjusting a small amount of deviations after the initial automatic labeling, thereby effectively reducing the workload of manual labeling and improving the overall data quality.

[0059] In S205 , a blood vessel segmentation model is constructed based on the preset Mamba segmentation network and blood vessel segmentation dataset information.

[0060] Specifically, after the terminal device generates the blood vessel segmentation data set information, the terminal device can construct a blood vessel segmentation model based on the preset Mamba segmentation network and the blood vessel segmentation data set information; for example, refer to Figure 5 ,The vascular segmentation model includes a base layer and multiple feature extraction layers, Figure 5 The "Stem" in the figure represents the base layer. Figure 5 The "Samba Block" in the figure represents a feature extraction layer; the feature extraction layer includes a first feature extraction layer, a second feature extraction layer, a third feature extraction layer, and a fourth feature extraction layer, that is, Figure 5 The four "Samba Blocks" in.

[0061] Without loss of generality, the vascular segmentation model can be named the Mamba-CAGSeg model, which is based on the Mamba segmentation network and can automatically segment the vascular region in the coronary angiography image using a convolutional neural network (CNN). The Mamba segmentation network can extract vascular features in the image through multiple convolutional layers and restore the details of the image using deconvolution layers to ensure accurate extraction of the vascular region. The training process of the Mamba-CAGSeg model can use the connected region loss function, the quadtree scanning feature extraction method and FrangiFilter as joint guidance modules to ensure the integrity and accuracy of the vascular region and avoid segmentation omissions and mis-segmentation by continuously optimizing network parameters and maximizing segmentation accuracy.

[0062] It should be noted that the Mamba-CAGSeg model is based on the data characteristics of coronary angiography. Adding a quadtree scanning feature extraction module can better extract the coronary vascular structure. Adding FrangiFilter as a joint guidance module can select and enhance the vascular structure. Using vascular connectivity as the objective function can guide the training of the entire network.

[0063] In some possible implementations, to test the accuracy of the vessel segmentation model, see Figure 6 After step S205, the method further includes but is not limited to the following steps:

[0064] In S2051, first detection result image information is generated according to the blood vessel segmentation data set information and the blood vessel segmentation model.

[0065] Specifically, the terminal device may input the blood vessel segmentation data set information into the blood vessel segmentation model to generate first detection result image information, wherein the first detection result image information is used to describe the output result of the blood vessel segmentation model.

[0066] In S2052, other preset segmentation models are obtained.

[0067] Specifically, after the terminal device generates the first detection result image information, the terminal device can obtain other preset segmentation models, where the other segmentation models are used to describe other blood vessel segmentation models, and the other segmentation models can be a medical image segmentation model based on MedSegDiff or a medical image segmentation model based on DconnNet.

[0068] In S2053, second detection result image information is generated according to the blood vessel segmentation data set information and other segmentation models.

[0069] Specifically, after the terminal device acquires other segmentation models, the terminal device may input the blood vessel segmentation data set information into each other segmentation model to generate second detection result image information, wherein the second detection result image information is used to describe the output results of other segmentation models.

[0070] In S2054, segmented image comparison result information is generated according to the first detection result image information and the second detection result image information.

[0071] Specifically, after the terminal device generates the second detection result image information, the terminal device can generate segmented image comparison result information based on the first detection result image information and the second detection result image information, wherein the segmented image comparison result information is used to describe the comparison result between the first detection result image information and the second detection result image information.

[0072] For example, see Figure 2 and Figure 7 , Figure 7 (a) is the target image. Figure 7 (a) is used to illustrate the blood vessel area that is expected to be segmented; Figure 7 (b) is the detection result output by the medical image segmentation model based on MedSegDiff; Figure 7 (c) is the first detection result image information, that is, the detection result output by the blood vessel segmentation model; Figure 7 (d) in the figure is the detection result output by the blood vessel segmentation model after modifying the objective function. Figure 7 It can be seen that compared with other medical image segmentation models, the MMamba-CAGSeg model has more accurate segmentation of blood vessel areas, more complete detail retention, and significantly reduced mis-segmentation rate, especially after modifying the objective function, the segmentation effect is the best. This comparison result not only verifies the superiority of the MMamba-CAGSeg model, but also fully proves the rationality and advancement of technical innovation based on the Mamba segmentation network.

[0073] Accordingly, see Figure 8 The above step S200 includes but is not limited to the following steps:

[0074] In S210, the image information to be detected is input into the base layer of the blood vessel segmentation model to generate first feature map information.

[0075] Specifically, the terminal device can input the image information to be detected into the basic layer of the blood vessel segmentation model to generate the first feature map information.

[0076] In S220, image segmentation result information is generated according to the first feature map information and the plurality of feature extraction layers.

[0077] Specifically, after the terminal device generates the first feature map information, the terminal device can accurately generate image segmentation result information based on the first feature map information and multiple feature extraction layers.

[0078] In some possible implementations, in order to improve the accuracy of image segmentation result information, please refer to Figure 5 and Fig. 9 Step S220 includes but is not limited to the following steps:

[0079] In S221, the first feature map information is input into the first feature extraction layer to generate the second feature map information.

[0080] Specifically, the terminal device may first input the first feature map information into the first feature extraction layer to generate the second feature map information, wherein the size of the feature map information may be:

[0081]

[0082] In the formula, H is the image height information, W is the image width information, and C is the channel number information.

[0083] In S222, the second feature map information is input into the second feature extraction layer to generate third feature map information.

[0084] Specifically, after the terminal device generates the second feature map information, the terminal device may input the second feature map information into the second feature extraction layer to generate the third feature map information, wherein the size of the feature map information may be:

[0085]

[0086] In S223, the third feature map information is input into the third feature extraction layer to generate fourth feature map information.

[0087] Specifically, after the terminal device generates the third feature map information, the terminal device may input the third feature map information into the third feature extraction layer to generate the fourth feature map information, wherein the size of the fourth feature map information may be:

[0088]

[0089] In S224, the fourth feature map information is input into the fourth feature extraction layer to generate image segmentation result information.

[0090] Specifically, after the terminal device generates the fourth feature map information, the terminal device may input the fourth feature map information into the fourth feature extraction layer to generate image segmentation result information. For example, Figure 5The "Output" in represents the image segmentation result information, where the size of the image segmentation result information is:

[0091]

[0092] In some possible implementations, in order to avoid invalid data entering the neural network for detection, see Fig.10 After step S202, the method further includes but is not limited to the following steps:

[0093] In S2021, the maximum segmentation region area information is generated based on the plurality of predicted image information.

[0094] Specifically, the terminal device can generate the maximum segmentation area information based on the preliminary segmentation result of the DiffusionMedSeg neural network and the multiple predicted image information, wherein the maximum segmentation area information is used to describe the total area corresponding to the largest blood vessel segmentation area in the multiple predicted image information. It should be noted that the method for calculating the area of ​​the blood vessel segmentation area can be a segmentation method.

[0095] In S2022, for each predicted image information: obtain the actually measured segmented region area information.

[0096] Specifically, after the terminal device generates the maximum segmentation region area information, the terminal device can perform this processing for each predicted image information: obtain the measured segmentation region area information, thereby determining the binary segmentation region of each frame in the angiography sequence, wherein the measured segmentation region area information is used to describe the area of ​​the blood vessel segmentation region corresponding to the predicted image information.

[0097] In S2023, area difference information is generated based on the difference between the maximum segmented area information and the actually measured segmented area information.

[0098] Specifically, after the terminal device obtains the measured segmented region area information, the terminal device can generate region area difference information based on the difference between the maximum segmented region area information and the measured segmented region area information, thereby comparing the measured segmented region area information with the maximum statistical value.

[0099] In S2024, the regional area difference information is compared with the preset area threshold information.

[0100] Specifically, after receiving the area difference information of the terminal device, the terminal device may compare the area difference information with a preset area threshold information, wherein the area threshold information is a preset comparison threshold.

[0101] In S2025, if the region area difference information is less than the area threshold information, the predicted image information is discarded, otherwise the predicted image information is retained.

[0102] Specifically, if the regional area difference information is less than the area threshold information, the terminal device can discard the predicted image information, otherwise the terminal device can retain the predicted image information, thereby automatically selecting valid frames of the angiography sequence and avoiding invalid data from entering the neural network for detection.

[0103] In S300, detection result information is generated according to the image segmentation result information and a preset stenosis detection model.

[0104] Specifically, after the terminal device generates the image segmentation result information, the terminal device can generate the detection result information according to the image segmentation result information and the preset stenosis detection model, so as to realize the stenosis detection of the blood vessel area after the blood vessel segmentation is completed, and identify and locate the stenosis area in the blood vessel. For example, please refer to Fig.11 , the detection result information is used to describe the image segmentation result information after stenosis detection.

[0105] In some possible implementations, to improve the accuracy of the stenosis detection model, see Fig.12 Before step S300, the method further includes but is not limited to the following steps:

[0106] In S301, a preset vascular stenosis data set information is obtained.

[0107] Specifically, the terminal device may first obtain preset vascular stenosis data set information, wherein the vascular stenosis data set information includes a plurality of predicted image information associated with preset detection frames, and the preset detection frames are used to mark the stenosis area in the predicted image information.

[0108] In S302, a MambaYoLo model is constructed based on a preset YOLO-v11 algorithm.

[0109] Specifically, after the terminal device obtains the vascular stenosis data set information, the terminal device can build a MambaYoLo model based on the preset YOLO-v11 deep learning target detection algorithm, thereby achieving efficient detection of coronary artery stenosis areas by improving the network structure and feature extraction strategy.

[0110] In S303, a stenosis detection model is constructed based on the MambaYoLo model and the vascular stenosis dataset information.

[0111] For example, see Fig.13After the terminal device builds the MambaYoLo model, the terminal device can build a stenosis detection model based on the MambaYoLo model and the vascular stenosis dataset information, thereby achieving in-depth optimization for the specific needs of coronary angiography images based on the MambaYoLo model. Among them, the stenosis detection model can be named Mamba-CAGDetect model. For example, please refer to Fig.14 , Fig.14 (a) in the figure shows the original image of blood vessels. Fig.14 (b) in the figure shows the stenosis detection result image corresponding to the original blood vessel image; Fig.14 (b) in the figure represents the binary image of blood vessels. Fig.14 (e) in the figure shows the stenosis detection result image corresponding to the vascular binary image; Fig.14 (c) in the figure represents the original binary image of blood vessels. Fig.14 (f) in the figure represents the stenosis detection result image corresponding to the original binary image of the blood vessel; Fig.14 It can be seen that the stenosis detection model can accurately identify the stenosis area and intuitively mark the location and severity of the stenosis area through the detection box, providing doctors with clear and accurate diagnostic basis, which helps to achieve quick decision-making and precise treatment.

[0112] The implementation principle of the coronary angiography image processing method based on the Mamba neural network in the embodiment of the present application is as follows: the terminal device can first obtain the image information to be detected, and then accurately generate the image segmentation result information according to the image information to be detected and the preset blood vessel segmentation model, and finally effectively generate the detection result information according to the image segmentation result information and the preset stenosis detection model, thereby reducing the subjectivity of the diagnosis and significantly improving the accuracy of coronary angiography stenosis detection. Small blood vessels and background noise can be accurately excluded or accurately segmented, and a continuous and unbroken blood vessel structure can be obtained. While ensuring high accuracy, the speed of network training is greatly improved, and the labor cost of segmentation data label production is greatly reduced. Only hundreds of segmentation label data need to be manually drawn in the early stage, and the annotation work of a single image is reduced from several minutes to several seconds, with good operability and repeatability.

[0113] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0114] The embodiment of the present application also provides a coronary angiography image processing system based on the Mamba neural network. For ease of description, only the parts related to the present application are shown, such as Fig.15 As shown, the system 150 includes:

[0115] The image information acquisition module 151 is used to acquire the image information to be detected;

[0116] Image segmentation result information generating module 152: used to generate image segmentation result information according to the image information to be detected and the preset blood vessel segmentation model;

[0117] The detection result information generating module 153 is used to generate detection result information according to the image segmentation result information and the preset stenosis detection model.

[0118] Optionally, the system 150 further includes:

[0119] A blood vessel original image set information acquisition module: used to acquire blood vessel original image set information, wherein the blood vessel original image set information includes a plurality of blood vessel original image information;

[0120] A preliminary predicted image set information generation module: used to generate preliminary predicted image set information and image label set information based on a preset blood vessel segmentation network and the original blood vessel image set information, wherein the predicted image set information includes a plurality of predicted image information, the image label set information includes a plurality of image label information, the predicted image information corresponds to the original blood vessel image information one by one, and the predicted image information corresponds to the image label information one by one;

[0121] Manual correction information acquisition module: used for obtaining manual correction information for each predicted image information based on preset medical image processing software;

[0122] Blood vessel segmentation data set information generation module: used to generate blood vessel segmentation data set information according to predicted image information, image label information and manual correction information;

[0123] Vascular segmentation model construction module: used to construct a vascular segmentation model based on a preset Mamba segmentation network and vascular segmentation dataset information, wherein the vascular segmentation model includes a base layer and multiple feature extraction layers.

[0124] Accordingly, the above-mentioned image segmentation result information generating module 152 includes:

[0125] The first feature map information generating submodule is used to input the image information to be detected into the base layer of the blood vessel segmentation model to generate the first feature map information;

[0126] An image segmentation result information generating submodule: used to generate image segmentation result information according to the first feature map information and a plurality of feature extraction layers;

[0127] Among them, the feature extraction layer includes a first feature extraction layer, a second feature extraction layer, a third feature extraction layer and a fourth feature extraction layer; the above-mentioned image segmentation result information generation submodule includes:

[0128] The second feature map information generating unit is used to input the first feature map information into the first feature extraction layer to generate the second feature map information, wherein the size of the feature map information is:

[0129]

[0130] In the formula, H is the image height information, W is the image width information, and C is the channel number information;

[0131] The third feature map information generating unit is used to input the second feature map information into the second feature extraction layer to generate the third feature map information, wherein the size of the feature map information is:

[0132]

[0133] The fourth feature map information generating unit is used to input the third feature map information into the third feature extraction layer to generate the fourth feature map information, wherein the size of the fourth feature map information is:

[0134]

[0135] The image segmentation result information generating unit is used to input the fourth feature map information into the fourth feature extraction layer to generate image segmentation result information, wherein the size of the image segmentation result information is:

[0136]

[0137] Optionally, the system 150 further includes:

[0138] A maximum segmentation region area information generation module: used to generate maximum segmentation region area information according to a plurality of predicted image information;

[0139] The module for obtaining the area information of the measured segmented regions is used to obtain the area information of the measured segmented regions for each predicted image information;

[0140] A region area difference information generation module: used to generate region area difference information according to the difference between the maximum segmentation region area information and the measured segmentation region area information;

[0141] Regional area difference information comparison module: used to compare regional area difference information with preset area threshold information;

[0142] Prediction image information elimination module: used to eliminate the prediction image information if the area difference information of the region is less than the area threshold information, otherwise the prediction image information is retained.

[0143] Optionally, the system 150 further includes:

[0144] A first detection result image information generating module: used to generate first detection result image information according to the blood vessel segmentation data set information and the blood vessel segmentation model;

[0145] Other segmentation model acquisition module: used to obtain other preset segmentation models, wherein the other segmentation model is a medical image segmentation model based on MedSegDiff;

[0146] A second detection result image information generating module: used to generate second detection result image information according to the blood vessel segmentation data set information and other segmentation models;

[0147] The segmented image comparison result information generating module is used to generate the segmented image comparison result information according to the first detection result image information and the second detection result image information.

[0148] Optionally, the system 150 further includes:

[0149] A vascular stenosis data set information acquisition module: used to acquire preset vascular stenosis data set information, wherein the vascular stenosis data set information includes a plurality of predicted image information associated with preset detection frames, and the preset detection frames are used to mark the stenosis area in the predicted image information;

[0150] MambaYoLo model building module: used to build the MambaYoLo model based on the preset YOLO-v11 algorithm;

[0151] Stenosis detection model building module: used to build a stenosis detection model based on the MambaYoLo model and vascular stenosis dataset information.

[0152] It should be noted that the information interaction, execution process and other contents between the above-mentioned modules are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0153] The present application also provides a terminal device, such as Fig.16 As shown, the terminal device 160 of this embodiment includes: a processor 161, a memory 162, and a computer program 163 stored in the memory 162 and executable on the processor 161. When the processor 161 executes the computer program 163, the steps in the above-mentioned coronary angiography image processing method embodiment are implemented, for example Figure 1 Steps S100 to S300 shown; or, when the processor 161 executes the computer program 163, the functions of each module in the above device are realized, for example Fig.15 The functions of modules 151 to 153 are shown.

[0154] The terminal device 160 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server, and the terminal device 160 includes but is not limited to a processor 161 and a memory 162. Those skilled in the art will appreciate that Fig.16 It is only an example of the terminal device 160 and does not constitute a limitation of the terminal device 160. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device 160 may also include input and output devices, network access devices, buses, etc.

[0155] Among them, the processor 161 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0156] The memory 162 may be an internal storage unit of the terminal device 160, such as a hard disk or memory of the terminal device 160, or the memory 162 may be an external storage device of the terminal device 160, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the terminal device 160; further, the memory 162 may also include both an internal storage unit and an external storage device of the terminal device 160, and the memory 162 may also store a computer program 163 and other programs and data required by the terminal device 160, and the memory 162 may also be used to temporarily store data that has been output or is to be output.

[0157] One embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the steps of each of the above method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc.; the computer-readable medium may include: any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0158] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, all equivalent changes made according to the methods, principles, and structures of the present application should be included in the protection scope of the present application.

Claims

1. A coronary angiography image processing method based on Mamba neural network, characterized in that: The method comprises: Obtaining image information to be detected; Generate image segmentation result information according to the image information to be detected and a preset blood vessel segmentation model; Detection result information is generated according to the image segmentation result information and a preset stenosis detection model.

2. The method according to claim 1, characterized in that Before generating image segmentation result information according to the image information to be detected and the preset blood vessel segmentation model, the method further includes: Acquire blood vessel original image set information, wherein the blood vessel original image set information includes a plurality of blood vessel original image information; Based on a preset blood vessel segmentation network, preliminary predicted image set information and image label set information are generated according to the blood vessel original image set information, wherein the predicted image set information includes a plurality of predicted image information, the image label set information includes a plurality of image label information, the predicted image information corresponds to the blood vessel original image information in one-to-one correspondence, and the predicted image information corresponds to the image label information in one-to-one correspondence; For each of the predicted image information: obtaining manual correction information based on preset medical image processing software; generating blood vessel segmentation data set information according to the predicted image information, the image label information and the manual correction information; Based on a preset Mamba segmentation network and the blood vessel segmentation dataset information, constructing the blood vessel segmentation model, wherein the blood vessel segmentation model includes a base layer and a plurality of feature extraction layers; Correspondingly, the image segmentation result information is generated according to the image information to be detected and the preset blood vessel segmentation model, including: Inputting the image information to be detected into the base layer of the blood vessel segmentation model to generate first feature map information; Generate image segmentation result information according to the first feature map information and multiple feature extraction layers; The feature extraction layer includes a first feature extraction layer, a second feature extraction layer, a third feature extraction layer and a fourth feature extraction layer; and the image segmentation result information is generated according to the first feature map information and the plurality of feature extraction layers, including: The first feature map information is input into the first feature extraction layer to generate second feature map information, wherein the size of the feature map information is: In the formula, H is the image height information, W is the image width information, and C is the channel number information; The second feature map information is input into the second feature extraction layer to generate third feature map information, wherein the size of the feature map information is: The third feature map information is input into the third feature extraction layer to generate fourth feature map information, wherein the size of the fourth feature map information is: The fourth feature map information is input into the fourth feature extraction layer to generate image segmentation result information, wherein the size of the image segmentation result information is:

3. The method according to claim 2, characterized in that After generating preliminary predicted image set information and image label set information based on the preset blood vessel segmentation network according to the blood vessel original image set information, the method further includes: Generate maximum segmentation region area information based on the plurality of predicted image information; For each of the predicted image information: obtaining the measured segmented region area information; Generate area difference information according to the difference between the maximum segmented area information and the measured segmented area information; Comparing the area difference information of the region with preset area threshold information; If the region area difference information is smaller than the area threshold information, the predicted image information is discarded; otherwise, the predicted image information is retained.

4. The method according to claim 2, characterized in that: After constructing the blood vessel segmentation model based on the preset Mamba segmentation network and the blood vessel segmentation dataset information, the method further includes: generating first detection result image information according to the blood vessel segmentation data set information and the blood vessel segmentation model; Obtaining another preset segmentation model, wherein the other segmentation model is a medical image segmentation model based on MedSegDiff; generating second detection result image information according to the blood vessel segmentation data set information and the other segmentation models; The segmented image comparison result information is generated according to the first detection result image information and the second detection result image information.

5. The method according to claim 2, characterized in that: Before generating the detection result information according to the image segmentation result information and the preset stenosis detection model, the method further includes: Acquiring preset vascular stenosis data set information, wherein the vascular stenosis data set information includes a plurality of predicted image information associated with preset detection frames, and the preset detection frames are used to mark the stenosis area in the predicted image information; Build the MambaYoLo model based on the preset YOLO-v11 algorithm; The stenosis detection model is constructed based on the MambaYoLo model and the vascular stenosis dataset information.

6. A coronary angiography image processing system based on Mamba neural network, characterized in that: The system comprises: The module for acquiring information of the image to be detected is used to acquire information of the image to be detected; An image segmentation result information generating module is used to generate image segmentation result information according to the image information to be detected and a preset blood vessel segmentation model; Detection result information generation module: used to generate detection result information according to the image segmentation result information and a preset stenosis detection model.

7. The system according to claim 6, characterized in that The system further comprises: A blood vessel original image set information acquisition module: used to acquire blood vessel original image set information, wherein the blood vessel original image set information includes a plurality of blood vessel original image information; A preliminary predicted image set information generating module: used for generating preliminary predicted image set information and image label set information based on a preset blood vessel segmentation network and according to the blood vessel original image set information, wherein the predicted image set information includes a plurality of predicted image information, the image label set information includes a plurality of image label information, the predicted image information and the blood vessel original image information correspond one to one, and the predicted image information and the image label information correspond one to one; Manual correction information acquisition module: used for acquiring manual correction information for each predicted image information based on preset medical image processing software; A blood vessel segmentation data set information generation module: used to generate blood vessel segmentation data set information according to the predicted image information, the image label information and the manual correction information; Blood vessel segmentation model construction module: used to construct the blood vessel segmentation model based on the preset Mamba segmentation network and the blood vessel segmentation data set information, wherein the blood vessel segmentation model includes a base layer and multiple feature extraction layers.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images