Intracranial aneurysm image detection method, system and device and storage medium

Abstract

The invention belongs to the technical field of medical image segmentation, and relates to an intracranial aneurysm image detection method, system and device and a storage medium. The invention designs a new network architecture, namely a multi-scale axial residual projection network, which is used for improving the accuracy of intracranial aneurysm detection and segmentation. In the aspect of model input, preprocessing steps including matched filtering, Gaussian filtering and Laplacian filtering are adopted for enhancing blood vessel structure and edge information and reducing computing resource requirements. In the aspect of a model structure, multi-scale feature extraction is realized by using a DWR module, and the segmentation capability of the model on a small target is enhanced; besides, the ACRE module is used for enhancing the foreground and inhibiting the background through the axial attention module and the context relation encoder module, and the processing capacity of the network for the edge information of the foreground is improved. According to the invention, the accuracy of intracranial aneurysm detection and segmentation can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image segmentation, and in particular relates to an intracranial aneurysm image detection method, system, device and storage medium. Background Art

[0002] Three Dimensional Time of Flight Magnetic Resonance Angiography (3D TOF-MRA) technology is safe and non-invasive because it does not use contrast agents. It has great application potential in the examination and diagnosis of intracranial aneurysms and can promptly detect unruptured intracranial aneurysms in early diagnosis. However, when 3D TOF-MRA images are used for aneurysm detection, there are problems such as insufficient display of tiny aneurysms, long examination and training time, and high memory usage. In addition, with the advancement of imaging technology, the number of layers of three-dimensional medical images continues to increase, and the workload of doctors' manual reading of films has greatly increased. High-intensity film reading work will reduce the sensitivity of radiologists' diagnosis.

[0003] The maximum intensity projection (MIP) image generated by rotating and projecting the 3D TOF-MRA image retains the density information of the original image to a great extent. Compared with the original 3D image, it has fewer layers and reduces the requirements for computing resources. However, the location of intracranial aneurysms is random, and there are problems such as vascular occlusion. Intracranial aneurysms are only easy to observe at certain projection angles, and the vascular structure of different people's brains is different, making the optimal projection angle random.

[0004] In addition, the current intracranial aneurysm detection algorithms can be divided into two categories: traditional algorithms and deep learning algorithms. Traditional algorithms usually perform detection based on one or more features of intracranial aneurysms. Deep learning algorithms can use depth information to improve the detection performance of intracranial aneurysms. By training the network model with a large amount of data, the detection effect of intracranial aneurysms is more robust than that of traditional algorithms. For the study of deep learning algorithms for intracranial aneurysms, the dimension of the input image is the primary factor to consider in intracranial aneurysm detection, because 3D images have more useful information, but require more memory and training time than 2D images.

[0005] In the study of intracranial aneurysms based on 3D images, most algorithms are implemented using the basic structure of encoders and decoders. Improvements usually focus on adjusting the network structure, modifying the size of the convolution kernel, extracting and fusing multi-scale features, adding attention mechanisms, and improving loss functions to improve the detection performance of the algorithm. In addition, some researchers have combined traditional algorithms with deep learning algorithms. Based on the original deep learning model, they have embedded the pre-processing or post-processing methods of the traditional algorithm into the overall detection process to improve the detection accuracy of the algorithm. Some other researchers have combined two or more lightweight deep learning models and proposed a multi-stage learning strategy to achieve fine segmentation and detection of intracranial aneurysms.

[0006] In summary, traditional algorithms have fewer restrictions on the number of training samples, and their intracranial aneurysm detection efficiency is better than that of deep learning methods. However, deep learning detection algorithms have stronger feature extraction capabilities, and are more effective in detecting data with sufficient sample numbers and complex intracranial aneurysm features. In addition, with the improvement of computer computing power, CAD-assisted diagnosis based on deep learning is becoming more and more widely used. Despite this, the intracranial aneurysm algorithm based on deep learning still has certain drawbacks, namely, low intracranial aneurysm detection accuracy, long training cycle of intracranial aneurysm detection algorithm, and low detection efficiency. Summary of the invention

[0007] The purpose of the present invention is to propose a method for detecting intracranial aneurysms by proposing a new network structure to improve the accuracy of intracranial aneurysm detection and segmentation, and the detection efficiency is significantly improved.

[0008] In order to achieve the above object, the present invention adopts the following technical scheme: A method for detecting intracranial aneurysm images comprises the following steps: Step 1. Perform maximum intensity projection on the 3D TOF-MRA image to obtain a MIP image; Step 2. Apply various filtering processes to the generated MIP images, resize the images after filtering to a uniform size, and then splice the images after uniform size to obtain the preprocessed image; Step 3. Input the preprocessed image into a pre-built intracranial aneurysm image detection model based on a multi-scale axial residual projection network to perform intracranial aneurysm detection and obtain intracranial aneurysm detection results; The detection model implements multi-scale feature extraction based on the DWR module, and uses the correlation-based axial contextual relationship encoder ACRE as the axial attention mechanism and residual connection to strengthen the feature map, improve the segmentation ability of small targets and complex backgrounds, achieve deeper feature fusion and information extraction, and enable the detection model to focus on the segmentation of small targets.

[0009] In addition, based on the above intracranial aneurysm image detection method, the present invention also proposes an intracranial aneurysm image detection system corresponding to the intracranial aneurysm image detection method, which adopts the following technical solutions: An intracranial aneurysm image detection system includes the following modules: An MIP image acquisition module is used to perform a maximum density projection operation on the 3D TOF-MRA image to obtain an MIP image; An image preprocessing module is used to perform various filtering processes on the generated MIP image, and adjust the size of each image after the filtering process to a uniform size, and then splice the images after the uniform size to obtain a preprocessed image; and a prediction module, which is used to input the preprocessed image into a pre-built intracranial aneurysm image detection model based on a multi-scale axial residual projection network to perform intracranial aneurysm detection and obtain intracranial aneurysm detection results; The detection model implements multi-scale feature extraction based on the DWR module, and uses the correlation-based axial contextual relationship encoder ACRE as the axial attention mechanism and residual connection to strengthen the feature map, improve the segmentation ability of small targets and complex backgrounds, achieve deeper feature fusion and information extraction, and enable the detection model to focus on the segmentation of small targets.

[0010] In addition, based on the above intracranial aneurysm image detection method, the present invention also proposes a computer device, which includes a memory and one or more processors. An executable code is stored in the memory. When the processor executes the executable code, it is used to implement the above intracranial aneurysm image detection method.

[0011] In addition, based on the above intracranial aneurysm image detection method, the present invention also proposes a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it is used to implement the above intracranial aneurysm image detection method.

[0012] The present invention has the following advantages: As described above, the present invention relates to a method for detecting intracranial aneurysm images, which designs a new network architecture, namely a multi-scale axial residual projection network MARP-Net, for improving the accuracy of intracranial aneurysm detection and segmentation. In terms of model input, by performing maximum density projection on the original 3D TOF-MRA to obtain a MIP image, the density information of the original image can be retained, and the number of image layers can be reduced, the requirements for computing resources can be reduced, and the detection efficiency can be improved. In addition, the present invention adopts preprocessing steps including matched filtering, Gaussian filtering and Laplace filtering to enhance vascular structure and edge information. By optimizing these preprocessing steps, noise and redundant information can be reduced, the demand for computing resources can be reduced, the training efficiency of the model can be improved, and the training cycle can be shortened. In terms of model structure, the present invention uses a DWR module to realize multi-scale feature extraction and enhance the model's segmentation ability for small targets. In addition, the present invention uses an ACRE module through an axial attention module and a contextual relationship encoder module to enhance the foreground and suppress the background, improve the network's processing ability for the edge information of the foreground, and proposes a more efficient proxy attention mechanism in the ACRE module. The design of the network architecture further optimizes the processing ability of the preprocessed image and reduces the demand for computing resources. The present invention can significantly improve the accuracy of intracranial aneurysm detection and segmentation, and significantly improve the detection efficiency through the improvements in the above-mentioned model input and model network structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is an overall flow chart of the intracranial aneurysm image detection method in an embodiment of the present invention; Figure 2 Schematic diagram of maximum intensity projection of 3D TOF-MRA images; Figure 3 This is a schematic diagram of the bilinear interpolation method in an embodiment of the present invention; Figure 4 This is a network architecture diagram of a multi-scale axial residual projection network MARP_Net constructed in an embodiment of the present invention; Figure 5 A network structure diagram of an improved DWR module in an embodiment of the present invention; Figure 6 2 is a network structure diagram of the improved ACRE attention module in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 like Figure 1As shown, this embodiment describes a method for detecting intracranial aneurysm images, which first performs a maximum intensity projection operation on the original 3D TOF-MRA image, and then performs a variety of different filtering processes on the MIP image generated thereby, optimizes the image quality to improve detection accuracy, suppress noise interference, and solve the inherent defects of the MIP technology, and then inputs the pre-processed image formed after resizing and splicing into a pre-built multi-scale feature and axial residual projection enhanced intracranial aneurysm segmentation network (Multi-scale Axial Residual Projection Network, MARP_Net), referred to as the multi-scale axial residual projection network, which is the intracranial aneurysm image detection model. Within the model, the DWR module is used to extract multi-scale features of the preprocessed image to effectively enhance the feature representation capability. Specifically, the present invention proposes an improved DWR module that adopts a dynamic expansion rate AdaDR based on frequency adaptation, and changes the fixed expansion rate to a dynamically adjusted expansion rate to balance the effective bandwidth and the receptive field. The multi-branch parallel calculation of the existing DWR module increases the number of parameters and the computing cost, and the direct splicing of the outputs of branches with different expansion rates may lead to insufficient fusion of multi-scale features. Subsequently, the ACRE module is used to dynamically adjust the weight distribution of different positions in the feature map using the axial attention mechanism. The traditional attention mechanism used in the ACRE module usually needs to calculate the interactions between all features, which will result in a very large amount of calculation in the high-dimensional feature space, and lacks effective regularization means, which easily leads to overfitting of the model and poor generalization ability. To address this problem, the present invention improves the ACRE module and adopts a proxy attention mechanism. By introducing proxy features, the features are first reduced or simplified, and then the attention weights are calculated. The amount of calculation in the high-dimensional feature space is significantly reduced, and the computational efficiency of the model is improved. In addition, calculating the attention weights through proxy features can have a certain regularization effect, reduce the risk of overfitting of the model, thereby highlighting the foreground target and suppressing background noise interference, while strengthening the edge information of the features, and finally deeply fusing the features from different paths (i.e., fusing the results of the axial attention processing with the foreground and background to supplement the foreground and edge information of the features, obtain the processed features, and obtain an accurate final segmentation result), generating an accurate final segmentation result.

[0015] like Figures 1 to 6 As shown, the intracranial aneurysm image detection method in this embodiment includes the following steps: Step 1. Perform maximum intensity projection on the 3D TOF-MRA image to obtain a MIP image.

[0016] like Figure 2The maximum density projection process is shown in Figure 1. The maximum density projection is generated by calculating the maximum density pixel encountered along each ray of the scanned object. When the light passes through the volume data, the pixel with the highest density is saved and projected onto a two-dimensional plane, thereby forming a MIP image, which can well display changes such as stenosis and dilation of blood vessels.

[0017] Maximum intensity projection was performed on 3D TOF-MRA images, and intracranial aneurysm detection was performed on the obtained MIP images.

[0018] The original 3D TOF-MRA image is subjected to maximum density projection to generate a MIP image. This step can retain the density information of the original image, reduce the number of image layers, and lower the requirements for computing resources.

[0019] There are quite a lot of black background areas in the original MIP image, which will take up a lot of computing resources and will not help the feature learning of intracranial aneurysms. In addition, there is interference from high-frequency noise, which affects the image clarity.

[0020] Therefore, the MIP images sent to the MARP_Net model before training were preprocessed (as shown in step 2 below). The anatomical prior knowledge that intracranial aneurysms appear attached to blood vessels was used to locate the area where intracranial aneurysms may appear through blood vessels. This can reduce the computational cost and reduce interference from targets such as the skull, thereby effectively improving the detection accuracy.

[0021] Step 2. Apply various filtering processes to the generated MIP image, resize the filtered images to a uniform size, and then splice the images with uniform size to obtain the preprocessed image.

[0022] In this embodiment, the filtering processing methods include, for example, matched filtering, Gaussian filtering and Laplace filtering. The bilinear interpolation method is used to adjust the images after each filtering process to a uniform size.

[0023] A matched filter is a filter used to detect specific signals or patterns. It performs correlation operations with the input signal to detect whether there is a part in the signal that matches the filter template. In image processing, matched filters can be used to detect specific image features or patterns. Matched filters are used to enhance the contrast of blood vessels and suppress background noise. Matched filters highlight the vascular structure and provide clear vascular images for subsequent feature extraction and image segmentation.

[0024] Matched filter kernel It is expressed as follows: ;in, Represents the position coordinates of the pixels in the original image, represents the range of the filter cross-sectional intensity, and L represents the length of the blood vessel.

[0025] For the original input image , the image after matched filtering, the formula is as follows: .

[0026] in, and Indicates the relative position offset of the filter template in the image, that is, the moving position of the filter template in the image. For example, when and When , the filter kernel is completely aligned with a pixel point in the image; when and , the filter kernel moves one pixel horizontally relative to a pixel in the image.

[0027] By calculating the position of each pixel in the image The convolution sum at , we get the image after matched filtering .

[0028] Gaussian filter is a linear smoothing filter used to remove noise from images. It achieves smoothing effect by convolving the image with Gaussian function, which is localized in space and can effectively remove high-frequency noise in images. Applying Gaussian filter removes high-frequency noise in images, smoothes the image, and provides a clearer image basis for feature extraction.

[0029] Assume that the template of the Gaussian filter is , then for the input image The image after Gaussian filtering is as follows: .

[0030] Among them, the Gaussian function The formula is .

[0031] It is the standard deviation of the Gaussian function, which determines the smoothness of the filter and ranges from 0.5 to 2.0.

[0032] when When is small, the smoothness of the Gaussian filter is low, mainly removing high-frequency noise in the image while retaining more image details; When the value is larger, the smoothness of the Gaussian filter is higher, which can remove more noise in the image, but it will also remove more image details. Since this model is mainly aimed at small targets, When it is 1, more image details are retained while removing high-frequency noise in the image.

[0033] The Laplace filter is an edge detection filter used to detect edges in an image. It achieves edge detection by calculating the second-order derivative of the image. The Laplace filter can detect the sudden change in the image and extract the edge features of the image. Using the Laplace filter to enhance the edge information in the image is helpful for subsequent image segmentation and feature extraction. For the input original image , the image after Laplace filter processing is: .

[0034] Matched filtering can detect specific features or patterns in an image, thereby extracting important information from the image; Gaussian filtering can remove high-frequency noise in an image, thereby improving the quality and clarity of the image; Laplace filtering can detect edge features in an image, thereby extracting important information from the image. The present invention can comprehensively extract multiple features and information from an image by combining these three filters for filtering processing, thereby improving the effect and accuracy of image processing.

[0035] After the original image is processed by these three filters, the feature map obtained is , each with different characteristics and advantages, can provide rich features and information for subsequent image processing and analysis.

[0036] The top of the window defines the images after matched filtering, Gaussian filtering and Laplace filtering respectively. ,The three filtered images are resized using an interpolation method to have the same size.

[0037] Bilinear interpolation is used to adjust the target image to the same size (W, H) to obtain better image quality. Bilinear interpolation can effectively reduce the aliasing and mosaic phenomenon of the image edge through two linear interpolations (x direction and y direction), and generate smoother results than the nearest neighbor interpolation. In addition, bilinear interpolation decomposes the two-dimensional interpolation into two one-dimensional linear interpolations, and the computational complexity is , significantly lower than bicubic interpolation and more efficient in computation.

[0038] For each pixel in the target image, find the source image i.e. The value of the target pixel is estimated by taking the weighted average of the four nearest pixels in the image. Figure 3 shown.

[0039] Assume that the coordinates of the target pixel in the source image are (x, y), and x and y are not necessarily integers. Use the following four formulas to find the four nearest pixel points: , , , : ; ; ; .

[0040] in, Represents the source image, the value of the target pixel It can be calculated by the following formula: .

[0041] in, , , Indicates rounding down. Indicates rounding up.

[0042] The three filtered images with unified size are spliced ​​to obtain the preprocessed image.

[0043] The filtered image is a single-channel grayscale image. Create a three-channel empty image, copy the single-channel data to each channel to get a multi-channel image, and splice the three multi-channel images of the same scale to get the preprocessed image.

[0044] The preprocessed images are labeled and used as training data sets for the MARP_Net detection model training in step 3. In the aneurysm detection task, the manually segmented label values ​​are 0 for the normal blood vessel area, 1 for the aneurysm area, and 2 for the blood vessel boundary area.

[0045] Step 3. Input the preprocessed image into the pre-built multi-scale axial residual projection network, namely the detection model MARP_Net, to perform intracranial aneurysm detection and obtain the intracranial aneurysm detection result.

[0046] like Figure 4 As shown in the figure, the detection model MARP_Net adopts a modular architecture design and consists of four core components: encoder, parallel decoder (Partial Decoder), DWR module and ACRE module.

[0047] The encoder is based on the improved Res-UNet framework, using the ResNet-101 deep residual network as the backbone network for multi-level semantic feature extraction. Through the design of a deep residual structure, the network can effectively capture rich image semantic information, especially enhance the feature representation ability of tiny anatomical structures, laying the foundation for subsequent accurate segmentation. The decoder part adopts a lightweight design concept and selectively fuses dense features at all levels through a partial decoding mechanism. This strategy significantly reduces the computational complexity while maintaining the accuracy of aneurysm segmentation, enhancing the practicality and deployment efficiency of the model.

[0048] To further improve the feature expression capability, the network introduces a dilated weighted residual module (DWR). This module achieves the fusion of multi-scale context information through an adjustable dilated residual unit, effectively solving the segmentation problem caused by the small size of the target and the blurred boundaries in intracranial aneurysm medical images. By capturing context information at different scales, the DWR module can more accurately identify and segment the aneurysm area, improving segmentation accuracy and robustness.

[0049] The correlation-based Axial Context Relation Encoder (ACRE) module achieves adaptive optimization of feature maps by dynamically modeling spatial dependencies.

[0050] In the intracranial aneurysm segmentation task, the ACRE module can capture the spatial correlation of aneurysms in different axes, further enhancing the richness and accuracy of feature representation, thereby improving the overall segmentation effect.

[0051] The preprocessed image is sent to the MARP_Net network. The MARP_Net network uses the DWR module to capture contextual information of different scales, and uses the ACRE module as an axial attention mechanism and residual connection. The DWR module captures contextual information of different scales through a multi-scale feature pyramid, providing rich feature representations for subsequent segmentation tasks. The ACRE module further enhances the relevance and consistency of features by modeling axial spatial dependencies. The combination of the two can more comprehensively capture and express the characteristics of aneurysms at different scales and in different axes. The model of the present invention can achieve the coordinated optimization of multi-scale features and axial spatial dependencies through the use of the DWR module and the ACRE module, thereby further improving the segmentation effect. This model constructs a segmentation framework that adapts to small targets and complex scenes through the multi-scale feature decomposition of the DWR module and the axial relationship modeling of the ACRE encoder. The technical core lies in the collaborative design of multi-scale hole convolution and axial residual attention, which not only enhances the perception of local details, but also ensures the stability of the deep network through residual connections.

[0052] The following is combined with Figure 4 The network structure and signal processing flow of MARP_Net constructed by the present invention are further described in detail. First, the original image is preprocessed, and feature extraction is performed on the image processed by the matched filter (MF), Gaussian filter (GF) and Laplace filter (LF), and the image is spliced ​​into a fixed-length feature representation.

[0053] The network architecture includes an encoder, a parallel decoder PD, a DWR module, and an ACRE module. There are three DWR modules and three ACRE modules. The pre-processed image is sent to the detection model MARP_Net, and its processing flow is as follows: The preprocessed image passes through the encoder and uses ordinary convolution to extract features. The image after two convolutions is As the first-level input layer, it is input into the first-level DWR module to extract and fuse multi-scale features to obtain the first-level transmission layer.

[0054] The transmission layer at this level passes through the DWR module at the next level again as the input layer of the next level.

[0055] Assume that the image output by the first-level DWR module is , the first-level transmission layer is used as the second-level input layer, then the image As the input image of the second stage , input into the second-level DWR module to obtain the image .

[0056] The second-level transmission layer is used as the third-level input layer, that is, the image output by the second-level DWR module As the input image of the third pole ,image Input to the third-level DWR module for processing to obtain the image .

[0057] A parallel decoder is used to aggregate the first, second, and third input layers, that is, the image Aggregation is performed, and the calculation formula is: , thus obtaining a global feature map containing rich information .

[0058] The ACRE module consists of an axial attention module and a contextual relation encoder module.

[0059] The axial attention module uses the attention mechanism to set different weights for the foreground and background of the feature map according to the difference in the importance of features in the image, so as to strengthen the foreground and suppress the background.

[0060] The context encoder module mainly targets the result feature map of the network at this level Expand the process and calculate the foreground of the image ,background Feature map after axial attention processing The contextual relationship between them.

[0061] Given that the pixel values ​​of intracranial aneurysms with protruding intracranial blood vessels are significantly different from those of the surrounding background, and the global feature map Only the approximate area of ​​the intracranial aneurysm can be captured. The features output by the DWR module , are processed by the ACRE module at the same level to obtain the corresponding feature maps, which are defined as feature maps .

[0062] The ACRE module can solve the problem that traditional multi-scale medical image segmentation algorithms have poor processing capabilities for lesion edges and details, which can also help the network segment small-area targets.

[0063] The global feature map The output characteristics of the third level Add together to get the third local feature map , the third local feature map The output characteristics of the second-level ACRE module Add together to get the second local feature map .

[0064] The second local feature map The output characteristics of the ACRE module with the first level Add together to get the first local feature map Finally, the Sigmoid activation function is used to obtain the intracranial aneurysm image segmentation result.

[0065] The DWR module is a multi-scale feature enhancement module designed for medical image segmentation tasks. Its core idea is to achieve differentiated capture and adaptive optimization of contextual information through multi-branch dilated convolution and dynamic weighted residual fusion.

[0066] The existing DWR module adopts a multi-branch structure with a fixed expansion rate. However, the fixed expansion rate cannot adapt to the dynamic requirements of different target scales. The multi-branch parallel calculation increases the number of parameters and the calculation cost. In addition, the direct splicing of the outputs of branches with different expansion rates may lead to insufficient fusion of multi-scale features. Therefore, the present invention improves the traditional DWR module and adopts a dynamic expansion rate AdaDR based on frequency adaptation to change the fixed expansion rate to a dynamically adjusted expansion rate.

[0067] The structure of the improved DWR module is as follows Figure 5 As shown, the processing flow is as follows: First, the location Centered on the input image, extract the local window (For example ),in Indicates the local window size, Represents the number of channels. Perform discrete Fourier transform on it to get the frequency domain representation , expressed as: .

[0068] in, represents the complex output array of the DFT, and Indicates its height and width, Representation feature map The normalized frequencies in the coordinate, height and width dimensions of and Given.

[0069] According to the Nyquist frequency , the definition cannot be affected by the current expansion rate A collection of high frequencies captured: .

[0070] right Sum the frequency components within to get the high frequency power: .

[0071] For the input feature map , extract spatial context features through a lightweight convolution layer (1×1 convolution) to generate an intermediate feature map. Use a convolution layer with a predicted dilation rate (parameters are , usually 3×3 convolution) predicts the intermediate feature map and outputs a dilation rate map with the same size as the input feature map , and the ReLU activation function is used to ensure that the expansion rate is non-negative: .

[0072] During training, the parameters are optimized by the following objectives : .

[0073] in, The top 25% of pixels with the highest high-frequency power (such as the edge of an object), The last 25% of pixels with the lowest high-frequency power (such as background or object center). In the AdaDR part, the dilation rate is dynamically adjusted according to the local frequency (frequency distribution in the local window) to balance the receptive field and the effective bandwidth, and the dilated convolution formula is obtained: .

[0074] in, is the position in the output feature map The value of is the size of the convolution kernel, is a predefined sampling offset, Each location The dynamic expansion rate.

[0075] The traditional fixed design is difficult to adapt to the difference between the high-frequency and low-frequency components of the input feature map. The dynamic adjustment of the expansion rate adopted by the present invention uses a smaller expansion rate in high-frequency areas (such as edges and textures) to retain details, and a larger expansion rate in low-frequency areas (such as background) to expand the receptive field, thereby balancing the effective bandwidth and the receptive field.

[0076] With the input image Add to get the final output result .

[0077] like Figure 6 As shown in Figure 1, the ACRE module consists of an axial attention module and a contextual relationship encoder module. The axial attention module adopts a proxy attention mechanism, which efficiently integrates key features and contextual information in the image by introducing a proxy vector, thereby enhancing the model's ability to capture image details, while reducing computational costs and improving the model's flexibility and interpretability.

[0078] Context encoder module, for this level of network The result feature is expanded to calculate the foreground of the image ,background Feature map after axial attention processing By supplementing and enhancing the details of the edge of the feature map, the network's ability to process the edge information of the foreground is improved, thereby improving the network's ability to segment the edge of the image.

[0079] Traditional attention mechanisms usually need to calculate the interactions between all features, which will result in a very large amount of calculation in high-dimensional feature space, and lack effective regularization methods, which can easily lead to model overfitting and poor generalization ability.

[0080] Therefore, the present invention adopts a proxy attention mechanism, by introducing proxy features, first reducing or simplifying the features, and then calculating the attention weights. This can significantly reduce the amount of calculation in high-dimensional feature space, improve the computational efficiency of the model, and calculate the attention weights by proxy features, which can have a certain regularization effect and reduce the risk of overfitting of the model.

[0081] Proxy Attention is a method that optimizes the efficiency of traditional attention calculations by introducing learnable proxy parameters. The core idea is to replace the high-dimensional interactions of original features with a small number of proxy vectors, thereby reducing the computational complexity while maintaining the ability to focus on key features.

[0082] Specifically, the processing flow of the axial attention module is as follows: For input features , with the proxy vector , perform similarity calculation: .

[0083] in, is the similarity matrix, N is the number of spatial locations, d is the number of channels, k is the number of agents, and d is the feature dimension. In the image segmentation task, k is set to be related to the number of target categories, and the temperature coefficient Used to stabilize gradients;

[0084] Use the similarity matrix S to perform weighted aggregation on the agent vectors to generate agent context features , the formula is as follows:

[0085] Combine the original features with the proxy context features Fusion, get fusion features , to preserve details and enhance semantic associations: .

[0086] in, Representation layer normalization processing, is the original feature, that is, the input feature.

[0087] Assume that the input image , whose shape is , the input feature Decomposed into two sub-features, respectively along the height Axis and Width The axis is decomposed, and the features are obtained through proxy attention , the formula is as follows: .

[0088] in, represents the batch size, Represents the number of channels, Represents height, Represents the width, Indicates along the height The axis is decomposed. Indicates along the width Axis decomposition.

[0089] The intermediate output result of a deeper layer in the MARP_Net network , through processing, we can get the foreground With background : ; .

[0090] Then the feature With prospects With background Perform feature fusion to supplement the foreground and edge information of the feature and obtain the processed features , the formula is as follows: ; ; .

[0091] The improved ACRE module can solve the problem that traditional multi-scale medical image segmentation algorithms have poor processing capabilities for lesion edges and details, thereby helping the network to segment small-area targets.

[0092] Intracranial aneurysms usually occupy a small part of the images containing them, and the incorrect use of loss functions will lead to class imbalance problems. The cross entropy loss function is used to test the similarity between the predicted results and the results obtained using manual segmentation masks. It is easy to make the loss reach a local minimum, which causes the model to focus on the background area during training and it is difficult to accurately predict the lesion area, while the Dice loss function is suitable for solving the class imbalance problem.

[0093] In the designed MARP-Net model, the loss function can be calculated by using a combined loss function, combining Dice loss and binary cross entropy loss to optimize the accuracy of segmentation. The Dice loss function is defined as follows: .

[0094] Where n is the number of label classes, is the model prediction value, i.e., the aneurysm detection result, is the manually segmented label value, is a small smoothing constant to prevent the denominator from becoming zero and the gradient from disappearing. In the aneurysm detection task, the manually segmented label value is 0 for the normal blood vessel area, 1 for the aneurysm area, and 2 for the blood vessel boundary area.

[0095] The convergence rate of the Dice loss function becomes lower in the later stages of training. During the learning process, due to the large data variance, instability is prone to occur, making it difficult to improve the segmentation accuracy of this method.

[0096] Therefore, a weighted combination of Dice and entropy loss functions is used in this embodiment, and the corresponding expression is: .

[0097] where α is a fractional weight used to balance the contribution of the dice and cross entropy loss functions, on a 0-1 scale.

[0098] During training, a previously prepared training dataset (i.e., multiple 3D TOF-MRA images are obtained and preprocessed according to steps 1 and 2 to obtain preprocessed images) is used, and the Dice loss and binary cross entropy loss are combined to optimize the model to optimize the accuracy of segmentation.

[0099] After the model is trained, the trained model is actually deployed.

[0100] After acquiring the 3D TOF-MRA image in real time, first use step 1 to obtain the MIP image, then use step 2 to perform multiple filtering processes on the MIP image, adjust the size to a uniform size, and splice to form a pre-processed image. The pre-processed image is input into the MARP-Net model to perform intracranial aneurysm detection and obtain the intracranial aneurysm detection result.

[0101] The present invention performs maximum density projection operation on the original 3D TOF-MRA image to generate a MIP image, and uses a joint processing process such as matched filtering, Gaussian filtering and Laplace filtering to enhance the vascular structure and edge information, providing a clearer input image for the network architecture, so that the DWR module and ACRE module can extract features and perform segmentation more effectively. Next, the DWR module enhances the segmentation ability of small targets through multi-scale feature extraction, which provides the ACRE module with richer feature information, enabling it to more accurately enhance the foreground and suppress the background through the axial attention and contextual relationship encoder modules. The ACRE module further enhances the model's segmentation ability for small targets by enhancing the edge information of the foreground. This synergy enables the model to more accurately identify boundaries when detecting and segmenting intracranial aneurysms, improving overall performance.

[0102] Example 2 This embodiment 2 describes an intracranial aneurysm image detection system, which is based on the same inventive concept as the intracranial aneurysm image detection method described in the above embodiment 1.

[0103] An intracranial aneurysm image detection system includes the following modules: An MIP image acquisition module is used to perform a maximum density projection operation on the 3D TOF-MRA image to obtain an MIP image; An image preprocessing module is used to perform various filtering processes on the generated MIP image, and adjust the size of each image after the filtering process to a uniform size, and then splice the images after the uniform size to obtain a preprocessed image; and a prediction module, which is used to input the preprocessed image into a pre-built intracranial aneurysm image detection model based on a multi-scale axial residual projection network to perform intracranial aneurysm detection and obtain intracranial aneurysm detection results; The detection model implements multi-scale feature extraction based on the DWR module, and uses the correlation-based axial contextual relationship encoder ACRE as the axial attention mechanism and residual connection to strengthen the feature map, improve the segmentation ability of small targets and complex backgrounds, achieve deeper feature fusion and information extraction, and enable the detection model to focus on the segmentation of small targets.

[0104] It should be noted that, in the intracranial aneurysm image detection system in this embodiment 2, the implementation process of the functions and effects of each functional module is detailed in the implementation process of the corresponding steps of the method in the above embodiment 1, and will not be repeated here.

[0105] Example 3 This embodiment 3 describes a computer device.

[0106] The computer device includes a memory and one or more processors. The memory stores executable code. When the processor executes the executable code, the steps of the intracranial aneurysm image detection method in the above-mentioned embodiment 1 are implemented.

[0107] In this embodiment, the computer device is any device or apparatus with data processing capability, which will not be described in detail here.

[0108] Example 4 This embodiment 4 describes a computer-readable storage medium on which a program is stored. When the program is executed by a processor, it is used to implement the steps of the intracranial aneurysm image detection method in the above-mentioned embodiment 1.

[0109] The computer-readable storage medium may be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc., equipped on the device.

[0110] Of course, the above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiments. It should be noted that all equivalent substitutions and obvious deformation forms made by any technician familiar with the field under the guidance of this specification fall within the essential scope of this specification and should be protected by the present invention.

Claims

1. A method for detecting intracranial aneurysm images, characterized in that: The steps include: Step 1. Perform maximum intensity projection on the 3D TOF-MRA image to obtain a MIP image; Step 2. Apply various filtering processes to the generated MIP images, resize the images after filtering to a uniform size, and then splice the images after uniform size to obtain the preprocessed image; Step 3. Input the preprocessed image into a pre-built intracranial aneurysm image detection model based on a multi-scale axial residual projection network to perform intracranial aneurysm detection and obtain intracranial aneurysm detection results; Among them, the detection model implements multi-scale feature extraction based on the DWR module, and uses the correlation-based axial contextual relationship encoder, namely the ACRE module, as the axial attention mechanism and residual connection to strengthen the feature map.

2. The intracranial aneurysm image detection method according to claim 1, characterized in that: In step 2, the filtering processing methods include at least matched filtering, Gaussian filtering and Laplace filtering, and the images after each filtering processing are adjusted to a uniform size using bilinear interpolation; The filtered image is a single-channel grayscale image. Create a three-channel empty image, copy the single-channel data to each channel to get a multi-channel image, and splice the three multi-channel images of the same scale to get the preprocessed image.

3. The intracranial aneurysm image detection method according to claim 1, characterized in that: In step 3, the detection model includes an encoder, a parallel decoder, a DWR module and an ACRE module, wherein there are three DWR modules and three ACRE modules, and the processing flow of the detection model is as follows: The preprocessed image passes through the encoder, uses two convolutions to extract features, and inputs the convolution-processed image into the first-level DWR module as the first-level input layer to extract and fuse multi-scale features to obtain the first-level transmission layer; The first-level transmission layer is used as the second-level input layer and input into the second-level DWR module to obtain the second-level transmission layer; the second-level transmission layer is used as the third-level input layer and input into the third-level DWR module for processing; A parallel decoder is used to aggregate the first, second and third input layers to obtain a global feature map; the features output by each level of DWR module are processed by the ACRE module at the corresponding level to obtain the corresponding feature map; The global feature map is added to the output feature of the third-level DWR module to obtain a third local feature map, and the third local feature map is added to the output feature of the second-level ACRE module to obtain a second local feature map; The second local feature map is added to the output feature of the first-level ACRE module to obtain the first local feature map. Finally, the first local feature map is processed using the Sigmoid activation function to obtain the intracranial aneurysm image segmentation result.

4. The intracranial aneurysm image detection method according to claim 1 or 3, characterized in that: The DWR module is obtained by improving the traditional fixed expansion rate DWR module, which is changed to adopt a dynamic expansion rate based on frequency adaptation, and the fixed expansion rate is changed to a dynamically adjusted expansion rate to balance the effective bandwidth and the receptive field.

5. The intracranial aneurysm image detection method according to claim 4, characterized in that: The processing flow of the improved DWR module is as follows: First, the location Centered on the input image, extract the local window , perform discrete Fourier transform on it and get the frequency domain representation , where s represents the local window size and C represents the number of channels; According to the Nyquist frequency , the definition cannot be affected by the current expansion rate Captured high frequency collection for: ； in, denote the normalized frequencies in height and width dimensions respectively; right The frequency components within are summed to obtain the high frequency power : ； For the input feature map , extract spatial context features through a 1×1 convolution, generate an intermediate feature map, and use a parameter of , a 3×3 convolutional layer with predicted dilation rate, predicts the intermediate feature map, and outputs a dilation rate map of the same size as the input feature map , and the ReLU activation function is used to ensure that the expansion rate is non-negative: ； During training, the parameters are optimized by the following objectives : ； in, The first 25% of pixels with the highest high-frequency power are the edge positions of objects. The last 25% of pixels with the lowest high-frequency power are the background or the center of the object; for the input feature map In the AdaDR part, the dilation rate is dynamically adjusted according to the frequency distribution in the local window to balance the receptive field and the effective bandwidth, and the dilated convolution formula is obtained: ； in, is the position in the output feature map The value of is the size of the convolution kernel, is a predefined sampling offset, Each location The dynamic expansion rate of Dynamically adjust the dilation rate. Use a smaller dilation rate in high-frequency areas where edges and textures are located to preserve details, and a larger dilation rate in low-frequency areas where the background is located to expand the receptive field. With the input image Add to get the final output result , the formula is: .

6. The intracranial aneurysm image detection method according to claim 3, characterized in that: The ACRE module consists of an axial attention module and a contextual relationship encoder module; the axial attention module adopts a proxy attention mechanism, which introduces a proxy vector to efficiently integrate key features and contextual information in the image, thereby enhancing the model's ability to capture image details; The context encoder module processes the result features of the network at this level and calculates the contextual relationship between the foreground and background of the image and the feature map after axial attention processing.

7. The intracranial aneurysm image detection method according to claim 6, characterized in that: In step 3, after the proxy attention mechanism is introduced, the processing flow of the axial attention module is as follows: For input features , with the proxy vector , perform similarity calculation: ； in, is the similarity matrix, N is the number of spatial locations, d is the number of channels, k is the number of agents, and d is the feature dimension. In the image segmentation task, k is set to be related to the number of target categories, and the temperature coefficient Used to stabilize gradients; Use the similarity matrix S to perform weighted aggregation on the agent vectors to generate agent context features , the formula is as follows: ； Combine the original features with the proxy context features Fusion, get fusion features , to preserve details and enhance semantic associations: ； in, Representation layer normalization processing, is the original feature, i.e., the input feature; Assume that the input image , whose shape is , the input feature Decomposed into two sub-features, respectively along the height Axis and Width The axis is decomposed, and the features are obtained through proxy attention , the formula is as follows: ； in, represents the batch size, Represents the number of channels, Represents height, Represents the width, Indicates along the height The axis is decomposed. Indicates along the width Axis decomposition.

8. An intracranial aneurysm image detection system, characterized in that: Includes the following modules: An MIP image acquisition module is used to perform a maximum density projection operation on the 3D TOF-MRA image to obtain an MIP image; An image preprocessing module is used to perform various filtering processes on the generated MIP image, and adjust the size of each image after the filtering process to a uniform size, and then splice the images after the uniform size to obtain a preprocessed image; and a prediction module, which is used to input the preprocessed image into a pre-built intracranial aneurysm image detection model based on a multi-scale axial residual projection network to perform intracranial aneurysm detection and obtain intracranial aneurysm detection results; The detection model implements multi-scale feature extraction based on the DWR module, and uses the correlation-based axial contextual relationship encoder ACRE as the axial attention mechanism and residual connection to strengthen the feature map, improve the segmentation ability of small targets and complex backgrounds, achieve deeper feature fusion and information extraction, and enable the detection model to focus on the segmentation of small targets.

9. A computer device comprising a memory and one or more processors; an executable code is stored in the memory; characterized in that: When the processor executes the executable code, it is used to implement the steps of the intracranial aneurysm image detection method described in any one of claims 1 to 7.

10. A computer-readable storage medium having a program stored thereon; characterized in that: When the program is executed by a processor, it is used to implement the steps of the intracranial aneurysm image detection method described in any one of claims 1 to 7.

Images