An interactive MR image carotid artery analysis method based on deep learning model

Abstract

The present invention relates to the field of medical image processing, and particularly relates to an interactive MR image carotid artery analysis method based on a deep learning model. The method comprises obtaining a magnetic resonance image of an original carotid artery; interactively specifying the center position of the carotid artery vessel in an initial layer, and defining a region of interest based on the center; performing carotid artery wall segmentation on the region of interest using a segmentation model; performing tissue classification on the carotid artery wall using a vascular plaque classification model; detecting the inner and outer wall contours of the carotid artery wall and calculating the position of the vessel center; predicting the region of interest of the carotid artery wall on the next layer using an LSTM network; and repeating the steps to complete multi-layer carotid artery wall segmentation and analysis. After the user confirms the initial conditions, the present invention can automatically predict and segment most of the carotid artery wall areas on other layers through a deep learning model, thereby reducing the need for manual intervention, while ensuring the accuracy and consistency of the segmentation results, and significantly improving efficiency in clinical applications.

Description

Technical Field

[0001] The present invention relates to the field of medical image processing, and particularly to an interactive MR image carotid artery analysis method based on a deep learning model. Background Art

[0002] Magnetic resonance imaging (MRI) is an important medical imaging technology that uses strong magnetic fields and radiofrequency waves to generate detailed images of human tissue. It is widely used in the diagnosis and analysis of carotid artery disease. The carotid artery is the main artery supplying blood to the head, and segmentation of its inner and outer walls is crucial in the study of pathologies such as atherosclerosis, vascular plaque formation, and intimal thickening. Precise segmentation of the inner and outer walls of the carotid artery enables quantitative analysis of vessel wall thickness, elasticity, and stenosis rate, providing key evidence for clinical diagnosis.

[0003] Existing methods for segmenting the inner and outer walls of the carotid artery include traditional methods and deep learning-based approaches. While traditional methods such as manual segmentation, threshold-based segmentation, and edge detection have played an important role in early medical image processing, they have significant limitations. Manual segmentation is time-consuming and labor-intensive, relying on operator experience, making it difficult to scale for large-scale clinical data processing. Threshold-based methods are sensitive to image quality and threshold selection, limiting segmentation accuracy. Edge detection methods struggle with images with complex vascular morphology or low contrast. These methods are limited in their ability to process complex scenes and struggle to meet real-world clinical needs.

[0004] In recent years, the rapid development of deep learning technology has provided a new technical path for the segmentation of the inner and outer walls of the carotid artery. For example, convolutional neural networks (CNNs) have been widely used in medical image segmentation, offering advantages such as automatic feature extraction, high segmentation accuracy, and wide applicability. However, existing deep learning methods still have some shortcomings in practical applications. For example, some technologies only segment the entire carotid artery region and fail to distinguish between the inner and outer walls, resulting in the inability to quantitatively analyze carotid wall thickness and stenosis rate. Some methods use automatic detection of the entire image to identify regions of interest, which can easily misidentify blood vessels in non-target areas, thereby affecting the segmentation results. Furthermore, when the input data is fixed, existing technologies generally lack the ability to correct detection results and are unable to correct incorrectly identified regions of interest, thereby reducing the reliability of the segmentation results.

[0005] Prior art CN117078695A proposed a method for segmenting the inner and outer walls of the carotid artery based on the YOLO-V5 algorithm and the 2D-VNet network. However, it is limited to the segmentation of the inner and outer walls of the plaque area in the ultrasound image and cannot guarantee whether the detected area is the vascular area of ​​interest to the doctor. CN116485810A automatically segmented the carotid artery area in the magnetic resonance image, but failed to distinguish between the inner and outer walls of the carotid artery and could not achieve accurate quantitative analysis of the vascular wall. CN118397017A proposed a method for segmenting the carotid artery wall based on a segmentation network. However, it uses the region of interest in the entire image as input, which can easily misidentify other vascular areas and cannot correct the incorrect ROI area after detecting it. Summary of the Invention

[0006] Based on the above problems, the present invention proposes a method for segmenting the inner and outer walls of the carotid artery based on interactive deep learning. The method first obtains the original MRI image of the carotid artery, and the user confirms the position of the center point of the carotid artery at the initial level to ensure the absolute correctness of the region of interest. For the detection results of other levels, if there is a deviation, the user can re-specify the position of the center point of the carotid artery for correction. The segmentation method further distinguishes the inner and outer walls of the carotid artery through contour detection to achieve quantitative analysis of the vascular wall. In addition, an independent network is used to take the carotid artery wall information of the first several levels as input to predict the ROI area of ​​the subsequent levels, thereby eliminating the interference of non-target blood vessels and improving detection accuracy.

[0007] The present invention is achieved through the following technical solutions:

[0008] An interactive MR image carotid artery analysis method based on a deep learning model comprises the following steps:

[0009] S1. Obtain magnetic resonance images of the original carotid artery.

[0010] S2. Interactively specify the center position of the carotid artery in the initial layer and define the region of interest based on it;

[0011] S3. Segment the carotid artery wall of the region of interest using the segmentation model;

[0012] S4. Carotid artery wall tissue classification using vascular plaque classification model;

[0013] S5, detecting the inner and outer wall contours of the carotid artery and calculating the center position of the blood vessel;

[0014] S6. Predict the region of interest of the carotid artery wall at the next level using the LSTM network.

[0015] S7. Repeat steps S3 to S6 to complete multi-layer carotid artery wall segmentation and analysis.

[0016] Furthermore, S1 includes the following sub-steps:

[0017] S101, obtaining a magnetic resonance image of the original carotid artery, wherein the magnetic resonance image includes three modes: T1, T2, and TOF;

[0018] S102 : Align and preprocess the three acquired magnetic resonance images to provide multimodal image information of the carotid artery wall.

[0019] Furthermore, S2 includes the following sub-steps:

[0020] S201. Interactively specify the center position of the carotid artery on the initial layer containing the carotid artery ;

[0021] S202, according to The coordinate position of the physical distance is determined with it as the center The area within the range is regarded as the area of ​​interest;

[0022] S203 : Crop the image according to the determined region of interest to obtain a sub-image of the region of interest for segmentation analysis.

[0023] Furthermore, S3 includes the following sub-steps:

[0024] S301, using the sub-image of the region of interest as input to a carotid artery wall segmentation model to obtain a mask image containing only the carotid artery wall;

[0025] S302 : performing a bitwise AND operation on the carotid artery wall mask image output by the carotid artery wall segmentation model and the original magnetic resonance image to extract the original image of the carotid artery wall in the region of interest.

[0026] Furthermore, S4 includes the following sub-steps:

[0027] S401, inputting the normalized carotid artery wall distance information and the original image of the carotid artery wall in the region of interest into a vascular plaque classification model;

[0028] S402 : The vascular plaque classification model outputs a tissue component category label of the carotid artery wall, including 0: normal, 1: bleeding, 2: calcification, and 3: stenosis.

[0029] Furthermore, S5 includes the following sub-steps:

[0030] S501, detecting a set of inner and outer wall contour points in a carotid artery wall mask image;

[0031] S502: Calculate the center coordinates of the carotid artery wall according to the inner and outer wall contour points ,in Indicates the current level number.

[0032] Furthermore, S6 includes the following sub-steps:

[0033] S601, current layer number When , directly use the initial layer center point As the center point of the region of interest;

[0034] S602, current layer sequence number When, based on the previous The normalized carotid artery center coordinates, inner and outer wall slopes, and inner and outer wall distance features at each level are input into the LSTM network;

[0035] S603, using LSTM network to predict The center position of blood vessels at the level ;

[0036] in The number of layers is a constant. Indicates the layer number.

[0037] Furthermore, S7 predicts , redefine the region of interest with it as the center.

[0038] Beneficial effects of the present invention:

[0039] (1) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. In the initial layer, the user confirms the region of interest of the carotid artery wall to ensure the absolute correctness of the input data, effectively avoiding the misidentification problem that may occur in traditional automated methods, and providing reliable initial conditions for subsequent segmentation;

[0040] (2) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. This method does not require the cumbersome multiple scanning process of traditional qMRI technology, but only relies on conventional MRI images from a single scan. High-precision quantitative parameter maps can be generated through modality conversion, which reduces the time cost and computing resource consumption of MRI examinations and improves patient comfort and examination compliance.

[0041] (3) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. When deviations are detected in other layers, the user can recalibrate the segmentation results by specifying the center point of the carotid artery, thereby significantly improving the flexibility and reliability of the method and overcoming the deficiency of the existing technology that cannot correct the erroneous ROI area.

[0042] (4) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. By distinguishing and marking the inner and outer walls of the carotid artery, this method can accurately measure key parameters such as carotid artery wall thickness, elasticity, and stenosis rate, providing a quantitative basis for the diagnosis and research of atherosclerosis and other carotid artery-related diseases;

[0043] (5) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. This method uses an independent deep learning network to take the position information of the carotid artery wall at the first several layers as input and predict the region of interest (ROI) at the subsequent layers. This method can not only effectively eliminate the interference of other blood vessels in the non-ROI region, but also improve the overall accuracy and robustness of the segmentation model.

[0044] (6) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model. After the user confirms the initial conditions, the deep learning model can automatically predict and segment most of the carotid artery wall areas at other levels, reducing the need for manual intervention while ensuring the accuracy and consistency of the segmentation results, greatly improving the efficiency in clinical applications.

[0045] (7) The present invention proposes an interactive MR image carotid artery analysis method based on a deep learning model, which combines the advantages of user interaction and deep learning. It is not only suitable for high-quality MRI images, but also can show excellent segmentation effects in images with low contrast or complex carotid artery morphology, meeting the diverse needs in clinical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0047] Figure 1 This is a flowchart of an interactive MR image carotid artery analysis method based on a deep learning model proposed by the present invention;

[0048] Figure 2 This is a schematic diagram of a terminal device for an interactive MR image carotid artery analysis method based on a deep learning model proposed in the present invention;

[0049] Figure 3 A schematic diagram of a readable storage medium for an interactive MR image carotid artery analysis method based on a deep learning model proposed in the present invention;

[0050] In the figure, 200 - terminal device, 210 - memory, 211 - RAM, 212 - cache, 213 - ROM, 214 - program / utility, 215 - program module, 220 - processor, 230 - bus, 240 - external device, 250 - I / O interface, 260 - network adapter, 300 - program product. DETAILED DESCRIPTION

[0051] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0052] Example 1

[0053] An interactive MR image carotid artery analysis method based on a deep learning model comprises the following steps:

[0054] Step 1: Obtaining a magnetic resonance image of the original carotid artery;

[0055] Step 2: Interactively specify the center position of the carotid artery in the initial layer and define the region of interest based on it;

[0056] Step 3: Use the segmentation model to segment the carotid artery wall in the region of interest;

[0057] Step 4: Using the vascular plaque classification model to classify the carotid artery wall tissue;

[0058] Step 5: Detect the inner and outer wall contours of the carotid artery and calculate the center position of the blood vessel;

[0059] Step 6: Predict the region of interest of the carotid artery wall at the next level through the LSTM network;

[0060] Step 7: Repeat steps 3 to 6 to complete multi-layer carotid artery wall segmentation and analysis.

[0061] The step 1 includes the following sub-steps:

[0062] 101. Obtain a magnetic resonance image of the original carotid artery, wherein the magnetic resonance image includes three modes: T1, T2, and TOF;

[0063] 102. Align and preprocess the three acquired magnetic resonance images to provide multimodal image information of the carotid artery wall.

[0064] The step 2 includes the following sub-steps:

[0065] 201. Interactively specify the center position of the carotid artery on the initial layer containing the carotid artery ;

[0066] 202. According to The coordinate position of Establishing physical distance for the center The area within the range is taken as the region of interest of the carotid artery wall to be detected;

[0067] 203. Crop the image according to the determined region of interest to obtain a sub-image of the region of interest of the carotid artery wall for segmentation analysis.

[0068] The step 3 includes the following sub-steps:

[0069] 301. Using the sub-image of the region of interest as input to a carotid artery wall segmentation model to obtain a mask image containing only the carotid artery wall. The carotid artery wall segmentation model is trained using the mask image with carotid artery wall labels and the original image of the carotid artery wall.

[0070] S302, performing a bitwise AND operation on the carotid artery wall mask image output by the carotid artery wall segmentation model and the original magnetic resonance image to extract the original image of the carotid artery wall in the region of interest;

[0071] The step 4 includes the following sub-steps:

[0072] 401. Input the normalized carotid artery wall distance information and the original image of the carotid artery wall in the region of interest into a vascular plaque classification model, wherein the classification model is obtained from the original carotid artery wall image in the region of interest and the normalized carotid artery wall distance feature map.

[0073] Specifically, let the coordinates of the center of the carotid artery wall be the image coordinates (Px, Py). For a point (X, Y) on the carotid artery wall, the physical distance relative to the carotid artery wall is:

[0074] , where dx and dy are the spatial resolution of the image in cm. For the carotid artery wall sub-image I(x, y) and the carotid artery wall center position (Px, Py), the corresponding normalized distance map image D(x, y) can be obtained. The vascular plaque classification model is trained using images I and D as input.

[0075] 402. The vascular plaque classification model outputs the tissue component category labels of the carotid artery wall, including 0 normal, 1 bleeding, 2 calcification and 3 stenosis.

[0076] The step 5 includes the following sub-steps:

[0077] 501. Perform contour detection on the carotid artery wall mask image obtained in step 301 to obtain a set of contour points of the inner and outer walls of the carotid artery wall;

[0078] 502. Calculate the coordinates of the center of the carotid artery wall based on the inner and outer wall contour points ,in Indicates the current level number.

[0079] The step 6 includes the following sub-steps:

[0080] 601, current level number When , directly use the initial layer center point As the center point of the region of interest, The number of layers is a constant. Indicates the level number;

[0081] 602, current level serial number When, based on the previous The normalized carotid artery center coordinates, inner and outer wall slopes, and inner and outer wall distance features in each layer are input into the LSTM network for prediction. Specifically, the center coordinates of the blood vessels at the i-th layer are set. The slopes of the inner and outer walls are and The distance between the inner and outer walls is ;

[0082] Then the input sequence can be expressed as:

[0083] ;

[0084] in, ;

[0085] Normalize the input data so that all eigenvalues ​​are mapped to or interval to improve the training effect of the LSTM network. The normalization method is expressed as:

[0086] ;

[0087] in and Respectively represent the minimum and maximum values ​​of all level data.

[0088] 603. Using LSTM network to predict The center position of blood vessels at the level ;

[0089] Specifically, the hidden layer dimension of the LSTM network is , the input is , then the LSTM prediction is calculated as follows:

[0090] ;

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] ;

[0096] The fully connected layer (FC) is used to calculate the Predicted h position of the vessel center at the level:

[0097]

[0098] in, represents the forget gate, represents the input gate, represents the output gate, represents the candidate memory unit, Indicates the current memory unit, Indicates the current hidden state. and These are all trainable parameters of the LSTM network. is the sigmoid function, is the Hadamard product, element-wise multiplication.

[0099] Step 7: Based on the prediction of S6 , redefine the region of interest with it as the center.

[0100] Example 2

[0101] refer to Figure 2 Based on Example 1, this embodiment proposes a terminal device for an interactive MR image carotid artery analysis method based on a deep learning model. The terminal device 200 includes at least one memory 210, at least one processor 220, and a bus 230 connecting different platform systems.

[0102] The memory 210 may include a readable medium in the form of a volatile memory, such as a RAM 211 and / or a cache 212 memory, and may further include a ROM 213 .

[0103] The memory 210 also stores a computer program that can be executed by the processor 220, causing the processor 220 to execute any of the above-mentioned interactive MR image carotid artery analysis methods based on a deep learning model in the embodiments of the present application. The specific implementation method is consistent with the implementation method and technical effects achieved in the above-mentioned application embodiments, and some of the contents are not repeated here. The memory 210 may also include a program / utility 214 having a set (at least one) of program modules 215. Such program modules include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each of these examples or some combination may include the implementation of a network environment.

[0104] Accordingly, the processor 220 may execute the aforementioned computer programs, as well as the program / utility 214 .

[0105] The bus 230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0106] The terminal device 200 can also communicate with one or more external devices 240, such as keyboards, pointing devices, Bluetooth devices, etc., and can also communicate with one or more devices that can interact with the terminal device 200, and / or communicate with any device that enables the terminal device 200 to communicate with one or more other computing devices (such as routers, modems, etc.). Such communication can be carried out through the I / O interface 250. In addition, the terminal device 200 can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs) and / or public networks, such as the Internet) through the network adapter 260. The network adapter 260 can communicate with other modules of the terminal device 200 through the bus 230. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in conjunction with the terminal device 200, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0107] Example 3

[0108] refer to Figure 3 This embodiment proposes a computer-readable storage medium for an interactive MR image carotid artery analysis method based on a deep learning model. The computer-readable storage medium stores instructions. When the instructions are executed by a processor, the specific implementation method of implementing any of the above-mentioned interactive MR image carotid artery analysis methods based on a deep learning model is consistent with the implementation method and the technical effect achieved in the above-mentioned application embodiments, and some contents will not be repeated here.

[0109] Figure 3 The program product 300 provided in this embodiment for implementing the above-mentioned application is shown. It can adopt a portable compact disc read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product 300 of the present invention is not limited to this. In this embodiment, the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in conjunction with an instruction execution system, device or device. The program product 300 can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0110] A computer-readable storage medium may include a data signal transmitted in baseband or as part of a carrier wave, carrying readable program code. This transmitted data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, which can transmit, transmit, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof. The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a standalone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0111] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An interactive MR image carotid artery analysis method based on a deep learning model, characterized in that: The following steps are involved: S1. Obtain magnetic resonance images of the original carotid artery. S2. Interactively specify the center position of the carotid artery in the initial layer and define the region of interest based on it; S3. Segment the carotid artery wall of the region of interest using the segmentation model; S4. Carotid artery wall tissue classification using vascular plaque classification model; S5, detecting the inner and outer wall contours of the carotid artery and calculating the center position of the blood vessel; S6. Predict the vascular center position of the region of interest of the carotid artery wall at the next level using the LSTM network; The S6 comprises the following sub-steps: S601, when the slice number n≤N, directly use the initial slice center point P1 as the blood vessel center position; S602: When the slice number n>N, the normalized carotid artery center coordinates, inner and outer wall slopes, and inner and outer wall distance features in the first N slices are input into the LSTM network; S603, using the LSTM network to predict the center position Pn+1 of the blood vessel at the n+1th layer; Where N represents the number of layers as a constant, and n represents the layer number; S7. Based on Pn+1 predicted in S6, the region of interest is redefined with it as the center, and steps S3 to S6 are repeated to complete the multi-level segmentation and analysis of the carotid artery wall.

2. The method for interactive MR image carotid artery analysis based on a deep learning model according to claim 1, characterized in that: S1 includes the following sub-steps: S101, obtaining a magnetic resonance image of the original carotid artery, wherein the magnetic resonance image includes three modes: T1, T2, and TOF; S102 : Align and preprocess the three acquired magnetic resonance images to provide multimodal image information of the carotid artery wall.

3. The method for interactive MR image carotid artery analysis based on a deep learning model according to claim 2, characterized in that: S2 includes the following sub-steps: S201, interactively specifying a carotid artery center position P1 on an initial layer including the carotid artery; S202, based on the coordinate position of P1, determine an area within a physical distance D as the region of interest; S203 : Crop the image according to the determined region of interest to obtain a sub-image of the region of interest for segmentation analysis.

4. The method for interactive MR image carotid artery analysis based on a deep learning model according to claim 3, characterized in that: S3 includes the following sub-steps: S301, inputting the sub-image of the region of interest into the carotid artery wall segmentation model to obtain a carotid artery wall mask image; S302, extracting the original image of the carotid artery wall in the region of interest based on the carotid artery wall mask image output by the segmentation model and the original magnetic resonance image; S303 : Perform deep learning training on the segmentation model using the mask image and the original image of the carotid artery wall in the training data.

5. The method for interactive MR image carotid artery analysis based on a deep learning model according to claim 4, characterized in that: S4 includes the following sub-steps: S401, inputting the normalized carotid artery wall distance information and the original image of the carotid artery wall in the region of interest into a vascular plaque classification model; S402 : The vascular plaque classification model outputs a tissue component category label of the carotid artery wall, including 0: normal, 1: bleeding, 2: calcification, and 3: stenosis.

6. The method for interactive MR image carotid artery analysis based on a deep learning model according to claim 5, characterized in that: S5 includes the following sub-steps: S501, detecting a set of inner and outer wall contour points in a carotid artery wall mask image; S502. Calculate the coordinates Pn of the center of the carotid artery wall according to the inner and outer wall contour points, where n represents the current layer number.

Images