Flat-scanning CT (Computed Tomography) multi-organ segmentation method based on multi-overlapping grouping attention mechanism

Abstract

The invention discloses a plain-scan CT multi-organ segmentation method based on a multi-overlap grouping attention mechanism, and relates to the technical field of image processing, and the method comprises the steps: carrying out the extraction of a plain-scan CT image, and obtaining a feature map; determining the grouping number of the feature map according to the channel number of the image segmentation model; according to the grouping number, overlapping grouping is carried out on the feature maps, so that channel overlapping is carried out between features of adjacent groups, and overlapping grouping features are obtained; extracting the overlapped grouping features by using a channel attention mechanism in the image segmentation model to obtain channel features; extracting the channel features by using a spatial attention mechanism in the image segmentation model to obtain spatial features; fusing the channel feature and the spatial feature of each channel to obtain a fused feature; and performing organ segmentation according to the fusion features by using an image segmentation model. According to the method, the feature maps are overlapped and grouped, and then serial processing of channel attention processing and space attention processing is performed, so that the precision and robustness of image segmentation can be improved.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method for multi-organ segmentation of plain scan CT based on a multi-overlapping grouping attention mechanism. Background Art

[0002] In recent years, with the rapid development of deep learning technology, medical image segmentation models based on convolutional neural networks have been widely used in the automatic segmentation of multiple organs. Accurately segmenting organ structures has become a key research direction in intelligent healthcare, particularly in CT image-assisted diagnosis and surgical planning. However, existing multi-organ segmentation models still face significant bottlenecks in generalization and segmentation accuracy. This is particularly true in the abdominal region, where plain CT images have low quality, blurred tissue boundaries, and high structural similarity between different organs. Summary of the Invention

[0003] The main purpose of the embodiments of the present application is to propose a multi-organ segmentation method for plain scan CT based on a multi-overlapping grouping attention mechanism to improve the accuracy of organ segmentation in plain scan CT images.

[0004] To achieve the above objectives, an embodiment of the present application proposes a method for multi-organ segmentation in plain scan CT based on a multi-overlapping group attention mechanism, the method comprising the following steps:

[0005] Extract the feature map from the plain scan CT image;

[0006] Determining the number of groups of the feature map according to the number of channels of the image segmentation model;

[0007] Overlapping and grouping the feature maps according to the number of groups so that channels of features of adjacent groups overlap to obtain overlapping group features;

[0008] Extracting the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features;

[0009] Extracting the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features;

[0010] Fusing the channel features and the spatial features of each channel to obtain a fused feature;

[0011] The image segmentation model is used to perform organ segmentation according to the fusion features.

[0012] In some embodiments, determining the number of groups of the feature map according to the number of channels of the image segmentation model comprises the following steps:

[0013] Determine the number of channels corresponding to the layer with the least number of channels in the image segmentation model as the number of grouping channels;

[0014] Determining the number of groups according to the number of grouping channels;

[0015] The expression of the number of groups is:

[0016]

[0017] Wherein, G is the number of groups, C is the number of channels in each group, and c is the number of channels in each group.

[0018] In some embodiments, overlapping the feature maps according to the number of groups so as to overlap the channels of features of adjacent groups to obtain overlapping group features includes the following steps:

[0019] The feature map is overlapped according to the number of groups so that 1 / 2 channels of features of adjacent groups overlap to obtain the overlapping group features.

[0020] In some embodiments, extracting the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features includes the following steps:

[0021] The feature statistics of each channel in each group are averaged to obtain the statistical information of each group;

[0022] The statistical information is:

[0023]

[0024] Wherein, s is the statistical information, H represents the height of the feature map, L represents the length of the feature map, and W represents the width of the feature map; F x is the characteristic of grouping;

[0025] The statistical information is transformed according to a channel scale factor and a channel bias factor, and then subjected to nonlinear activation to obtain a channel attention parameter;

[0026] The channel attention parameters are:

[0027] CA x =δ(F c (s))=δ(S1s+b1));

[0028] Among them, CA x is the channel attention parameter, S1 is the channel scale factor, b1 is the channel bias factor, F c (s) is the grouped channel feature; δ is the nonlinear activation;

[0029] Multiplying the channel attention parameter by the feature of the corresponding group to obtain the channel feature;

[0030] The channel characteristics are:

[0031] F x '=CA x ·F x ;

[0032] Among them, F x ' is the channel feature.

[0033] In some embodiments, extracting the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features includes the following steps:

[0034] Calculating the mean and variance of the channel features according to the grouping to perform regularization;

[0035] The regularized channel features are transformed according to the spatial scale factor and the spatial bias factor, and then the spatial attention parameters are obtained through nonlinear activation;

[0036] The spatial attention parameters are:

[0037] SA x =δ(F c (GN(F x ')))=δ(S2·GN(F x ')+b2));

[0038] Among them, SA x is the spatial attention parameter, S2 is the spatial scale factor, b2 is the spatial bias factor, GN represents the regularization; F c (s) is the channel feature of the group; δ is the nonlinear activation; F x ' is the channel feature;

[0039] Multiplying the spatial attention parameter by the corresponding channel feature to obtain the spatial feature;

[0040] The spatial characteristics are:

[0041] F x ”=SA·F x ';

[0042] Among them, F x ” is the spatial feature.

[0043] In some embodiments, fusing the channel features and the spatial features of each channel to obtain a fused feature comprises the following steps:

[0044] Define the adjacent grouping features as Fx and F x+1 , the channel attention parameters are CA x , CA x+1 , the spatial attention parameters are SA x , SA x+1 , the overlapping characteristics of adjacent groups are The channel attention parameters corresponding to the overlapping features are The spatial attention parameters corresponding to overlapping features

[0045] The attention weight of the feature map is calculated by the channel attention parameter and the spatial attention parameter as W x_x+1 =CA x_x+1 ·SA x_x+1 , W x+1_x =CA x+1_x ·SA x+1_x ;

[0046] W x_x+1 and W x+1_x Performing weighted summation on the weights to obtain the fusion feature;

[0047] The fusion features are:

[0048]

[0049] Wherein, F is the fusion feature.

[0050] In some embodiments, before fusing the channel features and the spatial features of each channel to obtain a fused feature, the method further includes the following steps:

[0051] The channel features and the spatial features corresponding to each channel are randomly permuted to adjust the order between channels.

[0052] To achieve the above objectives, another aspect of the present invention provides a non-contrast CT multi-organ segmentation device based on a multi-overlapping group attention mechanism, the device comprising:

[0053] A feature extraction unit, configured to extract a feature map from a plain scan CT image;

[0054] A grouping number determining unit, configured to determine the number of groups of the feature map according to the number of channels of the image segmentation model;

[0055] an overlapping grouping unit, configured to perform overlapping grouping on the feature map according to the grouping number so as to overlap the channels of features of adjacent groups and obtain overlapping group features;

[0056] A channel attention processing unit, configured to extract the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features;

[0057] A spatial attention processing unit, configured to extract the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features;

[0058] a feature fusion unit, configured to fuse the channel features and the spatial features of each channel to obtain a fused feature;

[0059] An image segmentation unit is used to perform organ segmentation according to the fusion features using the image segmentation model.

[0060] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method when executing the computer program.

[0061] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above-mentioned method when executed by a processor.

[0062] The embodiments of the present application include at least the following beneficial effects:

[0063] The present application can extract feature maps from plain scan CT images; determine the number of groups of the feature maps according to the number of channels of the image segmentation model; overlap and group the feature maps according to the number of groups so that the features of adjacent groups overlap in channels to obtain overlapping group features; use the channel attention mechanism in the image segmentation model to extract the overlapping group features to obtain channel features; use the spatial attention mechanism in the image segmentation model to extract the channel features to obtain spatial features; fuse the channel features and spatial features of each channel to obtain fused features; use the image segmentation model to perform organ segmentation based on the fused features. The present application can improve the accuracy and robustness of the image segmentation model in segmenting different organs by overlapping and grouping the feature maps and then performing serial processing of first channel attention processing and then spatial attention processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application, 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 application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0065] Figure 1 A flowchart of a method for multi-organ segmentation in plain scan CT based on a multi-overlapping group attention mechanism provided in an embodiment of the present application;

[0066] Figure 2 Schematic diagram of the algorithm flow of the attention mechanism for multiple overlapping groups provided in an embodiment of the present application;

[0067] Figure 3 A schematic diagram of the overlapping grouping process provided in an embodiment of the present application;

[0068] Figure 4 Schematic diagram of the channel attention mechanism provided in an embodiment of the present application;

[0069] Figure 5 Schematic diagram of the spatial attention mechanism provided in an embodiment of the present application;

[0070] Figure 6 Schematic diagram of channel weighted summation and channel random permutation provided in an embodiment of the present application;

[0071] Figure 7 A schematic diagram of the structure of a non-contrast CT multi-organ segmentation device based on a multi-overlapping group attention mechanism provided in an embodiment of the present application;

[0072] Figure 8 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0074] It will be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0075] The terms "at least one", "plurality", "each", "any", etc. used in this application include "at least one", "two" or more, "plurality" or "each", "any" or "any one", "each" or "any one" as used herein.

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0077] Before describing the embodiments of the present application in detail, some of the related technologies involved in the embodiments of the present application are first described as follows:

[0078] Taking puncture surgery as an example, clinical practice often only provides access to plain CT images of patients. In these images, some abdominal organs exhibit low grayscale contrast and unclear edges, severely hindering the accurate identification of key organs by image-based automatic segmentation systems. While traditional segmentation network architectures such as U-Net and V-Net perform well in certain tasks, they often fail to fully extract and utilize critical contextual information when processing fine-grained features across complex regions and organs. This results in insufficient model accuracy and robustness, making it difficult to meet high-precision clinical requirements. To improve segmentation performance, researchers have attempted to introduce attention mechanisms to enhance the model's focus on key features. However, most current attention mechanism designs remain relatively crude, suffering from several major issues: First, attention weight calculation often employs a fixed channel grouping approach, lacking modeling of redundancy and shared information between channels; second, channel and spatial attention mechanisms often employ parallel structures, failing to effectively model the hierarchical relationship between them; and third, there is a lack of comprehensive optimization strategies for channel redundancy, feature fusion efficiency, and parameter selection adaptability within the network architecture. Furthermore, traditional convolutional modules often process all image regions in a fixed pattern during feature extraction, making it difficult to account for the significant differences in scale, morphology, and texture among different tissues and organs. Current segmentation methods also rarely consider feature perturbations and channel redundancy, resulting in insufficient model generalization and poor performance in new scenarios or cross-center data. Therefore, building a segmentation network that can adapt to the anatomical structures of different regions and possess high-precision feature extraction and robust feature representation capabilities has become a key issue in current multi-organ segmentation research.

[0079] The embodiment of the present application provides a method for multi-organ segmentation of plain scan CT based on a multi-overlapping grouping attention mechanism, which relates to the field of image processing technology. The method for multi-organ segmentation of plain scan CT based on a multi-overlapping grouping attention mechanism provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a car terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements a method for multi-organ segmentation of plain scan CT based on a multi-overlapping grouping attention mechanism, etc., but is not limited to the above forms.

[0080] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0081] Reference Figure 1 The embodiment of the present application provides a method for multi-organ segmentation of plain scan CT based on a multi-overlap grouping attention mechanism. The method may include but is not limited to S100 to S160, as follows:

[0082] S100: extracting a feature map from the plain scan CT image;

[0083] S110: Determine the number of groups of the feature map according to the number of channels of the image segmentation model;

[0084] S120: performing overlapping grouping on the feature map according to the grouping number so as to make channels of features of adjacent groups overlap, thereby obtaining overlapping group features;

[0085] S130: Utilizing the channel attention mechanism in the image segmentation model to extract the overlapping grouping features to obtain channel features;

[0086] S140: Extracting the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features;

[0087] S150: Fusing the channel features and the spatial features of each channel to obtain a fused feature;

[0088] S160: Utilize the image segmentation model to perform organ segmentation according to the fusion features.

[0089] Optionally, determining the number of groups of the feature map according to the number of channels of the image segmentation model comprises the following steps:

[0090] Determine the number of channels corresponding to the layer with the least number of channels in the image segmentation model as the number of grouping channels;

[0091] Determining the number of groups according to the number of grouping channels;

[0092] The expression of the number of groups is:

[0093]

[0094] Wherein, G is the number of groups, C is the number of channels in each group, and c is the number of channels in each group.

[0095] Optionally, performing overlapping grouping on the feature map according to the number of groups so as to make channels of features of adjacent groups overlap to obtain overlapping group features comprises the following steps:

[0096] The feature map is overlapped according to the number of groups so that 1 / 2 channels of features of adjacent groups overlap to obtain the overlapping group features.

[0097] Optionally, the extracting the overlapping grouping features to obtain channel features using a channel attention mechanism in the image segmentation model comprises the following steps:

[0098] The feature statistics of each channel in each group are averaged to obtain the statistical information of each group;

[0099] The statistical information is:

[0100]

[0101] Wherein, s is the statistical information, H represents the height of the feature map, L represents the length of the feature map, and W represents the width of the feature map; Fx is the characteristic of grouping;

[0102] The statistical information is transformed according to a channel scale factor and a channel bias factor, and then subjected to nonlinear activation to obtain a channel attention parameter;

[0103] The channel attention parameters are:

[0104] CA x =δ(F c (s))=δ(S1s+b1));

[0105] Among them, CA x is the channel attention parameter, S1 is the channel scale factor, b1 is the channel bias factor, F c (s) is the grouped channel feature; δ is the nonlinear activation;

[0106] Multiplying the channel attention parameter by the feature of the corresponding group to obtain the channel feature;

[0107] The channel characteristics are:

[0108] F x '=CA x ·F x ;

[0109] Among them, F x ' is the channel feature.

[0110] Optionally, extracting the channel features using a spatial attention mechanism in the image segmentation model to obtain spatial features comprises the following steps:

[0111] Calculating the mean and variance of the channel features according to the grouping to perform regularization;

[0112] The regularized channel features are transformed according to the spatial scale factor and the spatial bias factor, and then the spatial attention parameters are obtained through nonlinear activation;

[0113] The spatial attention parameters are:

[0114] SA x =δ(F c (GN(F x ')))=δ(S2·GN(F x ')+b2));

[0115] Among them, SA x is the spatial attention parameter, S2 is the spatial scale factor, b2 is the spatial bias factor, GN represents the regularization; F c(s) is the channel feature of the group; δ is the nonlinear activation; F x ' is the channel feature;

[0116] Multiplying the spatial attention parameter by the corresponding channel feature to obtain the spatial feature;

[0117] The spatial characteristics are:

[0118] F x ”=SA·F x ';

[0119] Among them, F x ” is the spatial feature.

[0120] Optionally, fusing the channel features and the spatial features of each channel to obtain a fused feature comprises the following steps:

[0121] Define the adjacent grouping features as F x and F x+1 , the channel attention parameters are CA x , CA x+1 , the spatial attention parameters are SA x , SA x+1 , the overlapping characteristics of adjacent groups are The channel attention parameters corresponding to the overlapping features are The spatial attention parameters corresponding to overlapping features

[0122] The attention weight of the feature map is calculated by the channel attention parameter and the spatial attention parameter as W x_x+1 =CA x_x+1 ·SA x_x+1 , W x+1_x =CA x+1_x ·SA x+1_x ;

[0123] W x_x+1 and W x+1_x Performing weighted summation on the weights to obtain the fusion feature;

[0124] The fusion features are:

[0125]

[0126] Wherein, F is the fusion feature.

[0127] Optionally, before fusing the channel features and the spatial features of the respective channels to obtain a fused feature, the method further comprises the following steps:

[0128] The channel features and the spatial features corresponding to each channel are randomly permuted to adjust the order between channels.

[0129] Next, the solution of the embodiment of the present application will be introduced and explained in detail with reference to specific application examples.

[0130] This embodiment discloses a multi-overlapping grouping attention mechanism for multi-organ segmentation of plain CT. In the scenario of multiple punctures in the chest and abdomen, the features of different regions are quite different. In particular, only plain CT can be obtained during the puncture surgery, so it is very dependent on the accuracy and generalization of the segmentation algorithm. In particular, under plain CT of the abdomen, the features of some organs are even less obvious. The method of this embodiment takes into account the problems of low segmentation accuracy and poor generalization of the existing multi-organ segmentation network. The method of this embodiment mainly consists of three parts: multi-overlapping grouping, serial attention modeling, and channel weighted fusion and random permutation. By using this embodiment, the accuracy and robustness of the segmentation network for different segmentation tasks can be improved.

[0131] Specifically, this embodiment includes the following technical solutions:

[0132] 1. Multiple overlapping groups.

[0133] How to improve the performance of the segmentation network and improve the learning ability and generalization performance of the model is an important research direction in the field of image segmentation. Unlike the traditional fixed channel grouping method, this method adopts overlapping grouping to enable adjacent feature groups to share part of the channel information, thereby improving the feature fusion effect. Each group is processed by channel attention and spatial attention in turn to strengthen important features and enhance the model's perception of spatial information. Subsequently, the channel feature weighted fusion strategy is used to solve the feature integration problem in the overlapping area, avoid information loss, and enhance the generalization ability of the model by introducing a disordered strategy. For example, the attention mechanism algorithm flow diagram of multiple overlapping groups in this embodiment is shown in the figure below. Figure 2 shown.

[0134] Instead of simply grouping given features by channel, multiple overlapping feature grouping adopts an overlapping design paradigm to increase feature fusion between different channels. For each input feature map, the flowchart of overlapping grouping is as follows: Figure 3 shown.

[0135] For each feature map F∈R in the middle of the segmentation network C×H×L×W, where C represents the number of channels, H represents the height of the feature map, L represents the length of the feature map, and W represents the width of the feature map. Assume that the selected features are divided into G groups of channel attention for processing, and the number of channels in each group is c. For adjacent groups, 1 / 2 of the channels will be overlapped. Therefore, the relationship between the number of channels and the number of groups is:

[0136]

[0137] Therefore, before actually dividing the groups, it is necessary to determine the number of groups in the entire network based on the network's minimum feature layer. Since the initial number of channels in the first layer of the encoder input is the least, the number of feature channels in this layer is selected for calculation. By ensuring that the channel quality of each group is sufficiently divisible by the minimum initial number of channels, the number of groups in the entire network can be determined. Assume that the features after grouping become F = [F1,…,F G ],F x ∈R c×H×L×W ,F x With F x+1 There will be a 1 / 2 feature overlap between them.

[0138] 2. Channel-space serial attention modeling.

[0139] When overlapping features are grouped, channel attention and spatial attention will be added to each group of features in sequence to complete feature enhancement, and then the features will be fused after enhancement. Traditional attention modeling divides each group of features into two channels and adds channel attention and spatial attention respectively. The serial attention modeling proposed in this embodiment is to obtain a better fusion of channel and spatial information by first adding a channel attention mechanism to each group, and then adding spatial attention serially. Because for invalid channels, the network naturally does not need to pay attention to their corresponding spatial information, so the two attentions are processed serially.

[0140] 2.1. Channel Attention Modeling.

[0141] The key to increasing a network's learning capabilities lies in improving its ability to extract features. The attention mechanism allows the network to better focus on the features it needs to learn. Therefore, the sequential learning of the dual mechanisms of channel attention and spatial attention can greatly enhance the network's ability to capture key channel and spatial information.

[0142] First, global average pooling is performed on the channel features to ensure that the channel attention is enhanced without increasing the computational complexity of the model too much. For example, the channel attention mechanism diagram is as follows: Figure 4 shown.

[0143] For each feature group F x , which contains c channels, namely F x1 ,…,Fxc Calculate each channel to obtain the statistical average information of the channel, and finally calculate the total statistical information s∈R c×1×1×1 , which can be expressed as:

[0144]

[0145] This statistical information can be understood as the knowledge attached to the corresponding channel, but whether this knowledge is valid and should be retained is a skill that the segmentation network needs to learn to master. Therefore, after obtaining the statistical information, two learnable parameters need to be introduced to transform the statistical information. One is the scaling factor S1∈R c×1×1×1 , one is the bias factor b1∈R c ×1×1×1 , and perform nonlinear activation.

[0146] First, by introducing parameters and nonlinear activation, the network can learn the value of the corresponding channel information, scale the channel information using the proportional factor, add the bias factor, and then pass the sigmoid function to obtain the attention that needs to be given to the channel, and obtain the final channel attention parameter CA x . Reuse the obtained attention parameters CA x Multiply it with the original feature, reconstruct the channel information value of the original feature, and obtain the new feature F x ', the specific calculation process can be expressed as:

[0147] F x '=CA x ·F x (3)

[0148] CA x =δ(F c (s))=δ(S1s+b1)) (4)

[0149] 2.2. Spatial Attention Modeling.

[0150] After adding the channel attention mechanism, we get the new feature F x ', and then add the spatial attention mechanism to the feature serially. Different from channel attention, spatial attention focuses on spatial related information. F is obtained through group regularization strategy x ' Spatial related information. For example, the spatial attention mechanism diagram is as follows Figure 5 shown.

[0151] Group regularization is proposed to solve the problem of increased error in batch regularization. It achieves regularization by calculating the mean and variance of the data within the group. The specific calculation of the spatial attention mechanism is as follows Figure 5As shown. After group regularization, the unbiased spatial information of the group is obtained. Similar to the previous channel attention mechanism, this study needs to introduce learnable parameters to enable the network to understand which information in the space should be paid attention to through training. Therefore, similarly, the spatial attention scaling factor S2∈R is introduced. c×1×1×1 , spatial attention bias factor b2∈R c×1×1×1 The acquired spatial information is scaled and biased, and then nonlinear activation is completed through the sigmoid function to obtain the final spatial attention parameter SA x , the obtained parameter SA x In the case of x 'Multiply to get the final feature... The specific calculation method is as follows:

[0152] F x ”=SA·F x ' (5)

[0153] SA x =δ(F c (GN(F x ')))=δ(S2·GN(F x ')+b2)) (6)

[0154] In a serial manner, the channel attention mechanism and the spatial attention mechanism are added to the original features of each group. The channel attention mechanism allows the model to learn which channel information is more important, and the spatial attention mechanism allows the model to know where the information is richer. Through the fusion of the two, the model can better learn related features. The attention parameter CA obtained at the same time x with SA x It will be used for the subsequent channel feature weighted fusion.

[0155] 3. Weighted fusion of channel features and random permutation of channels.

[0156] After processing each channel, the features of the different groups need to be fused. Unlike simple concatenation, due to the different group attention mechanisms used in previous studies, this method uses partial overlap between the groups. This also means that this method requires fusing overlapping channels during the feature output stage. If the overlapping parts are simply summed, the channel information in the overlapping part is directly doubled. If a simple average is used, the weights calculated by the two methods are assumed to be the same.

[0157] The above processing method is obviously unreasonable. Channels in different groups have different information weights because they refer to the information of other channels in the group during the calculation process. Therefore, this paper uses the channel attention and spatial attention parameters learned previously, and uses them as the weights of channel fusion by taking advantage of their ability to reflect channel attention. For example, the schematic diagram of channel weighted summation and channel random permutation is as follows Figure 6 shown.

[0158] With the help of the attention parameters obtained from the previous training as weights, the features are fused. Assume that the adjacent group features are F x and F x+1 , the channel attention parameters are CA x , CA x+1 , the spatial attention parameters are SA x , SA x+1 , their overlapping features The channel attention mechanism parameters corresponding to the overlapping parts are Spatial attention mechanism parameters corresponding to the overlapping part

[0159] The attention weight W of the corresponding feature map is calculated by the channel and spatial attention parameters x_x+1 =CA x_x+1 ·SA x_x+1 , W x+1_x =CA x+1_x ·SA x+1_x , then the final output feature is W x_x+1 and W x+1_x is the weighted sum of weights, which is calculated as follows:

[0160]

[0161] The first and last channels, however, require special processing during the calculation because they participate in only one grouping calculation and have no overlap. Furthermore, to prevent the top and bottom channels from being consistently excluded from overlapping grouping calculations, all channels calculated using the attention mechanism are randomly permuted. This random permutation disrupts the order of the channels, resolving the issue of top and bottom channels not being included in overlapping groupings and further enhancing the network's generalization.

[0162] The beneficial effects of this embodiment include:

[0163] This embodiment proposes a multi-organ segmentation method based on a multi-overlapping grouping attention mechanism. This method introduces a learnable multi-overlapping feature grouping strategy into the segmentation network, and connects the channel and spatial attention mechanisms in series to improve the model's ability to express key areas and recognize complex organ structures. At the same time, the combination of channel disorder and weighted fusion mechanism enhances the generalization performance of the model. Compared with traditional methods, this method can effectively improve the segmentation accuracy in low-contrast images such as abdominal plain CT scans, and has stronger cross-scene adaptability, providing reliable technical support for high-risk medical tasks such as puncture surgery planning.

[0164] Reference Figure 7 The embodiment of the present application further provides a noncontrast CT multi-organ segmentation device based on a multi-overlapping grouping attention mechanism, which can implement the above-mentioned noncontrast CT multi-organ segmentation method based on a multi-overlapping grouping attention mechanism. The device includes:

[0165] A feature extraction unit, configured to extract a feature map from a plain scan CT image;

[0166] A grouping number determining unit, configured to determine the number of groups of the feature map according to the number of channels of the image segmentation model;

[0167] an overlapping grouping unit, configured to perform overlapping grouping on the feature map according to the grouping number so as to overlap the channels of features of adjacent groups and obtain overlapping group features;

[0168] A channel attention processing unit, configured to extract the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features;

[0169] A spatial attention processing unit, configured to extract the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features;

[0170] a feature fusion unit, configured to fuse the channel features and the spatial features of each channel to obtain a fused feature;

[0171] An image segmentation unit is used to perform organ segmentation according to the fusion features using the image segmentation model.

[0172] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0173] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of the present application. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0174] It can be understood that the contents of the above method embodiments are all applicable to the embodiments of the present device, the functions specifically implemented by the embodiments of the present device are the same as those of the method of the present application, and the beneficial effects achieved are also the same as those achieved by the method of the present application.

[0175] See also Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0176] The processor 801 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0177] The memory 802 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 802 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 802 and is called by the processor 801 to execute the methods of the embodiments of this application.

[0178] Input / output interface 803, used to implement information input and output;

[0179] Communication interface 804, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0180] Bus 805 , which transmits information between various components of the device (e.g., processor 801 , memory 802 , input / output interface 803 , and communication interface 804 );

[0181] The processor 801 , the memory 802 , the input / output interface 803 and the communication interface 804 are connected to each other in communication within the device via a bus 805 .

[0182] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method of the present application is implemented.

[0183] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0184] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0185] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0186] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0187] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0188] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0189] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0190] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0191] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0192] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0194] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0195] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A multi-organ segmentation method for plain scan CT based on a multi-overlapping grouping attention mechanism, characterized by: The method comprises the following steps: Extract the feature map from the plain scan CT image; Determining the number of groups of the feature map according to the number of channels of the image segmentation model; Overlapping and grouping the feature maps according to the number of groups so that channels of features of adjacent groups overlap to obtain overlapping group features; Extracting the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features; Extracting the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features; Fusing the channel features and the spatial features of each channel to obtain a fused feature; The image segmentation model is used to perform organ segmentation according to the fusion features.

2. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to claim 1 is characterized in that: Determining the number of groups of the feature map according to the number of channels of the image segmentation model comprises the following steps: Determine the number of channels corresponding to the layer with the least number of channels in the image segmentation model as the number of grouping channels; Determining the number of groups according to the number of grouping channels; The expression of the number of groups is: Wherein, G is the number of groups, C is the number of channels in each group, and c is the number of channels in each group.

3. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to claim 1 is characterized in that: The overlapping division of the feature graph according to the number of groups so as to make channels of features of adjacent groups overlap to obtain overlapping group features comprises the following steps: The feature map is overlapped according to the number of groups so that 1 / 2 channels of features of adjacent groups overlap to obtain the overlapping group features.

4. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to claim 1, characterized in that: The method of extracting the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features includes the following steps: The feature statistics of each channel in each group are averaged to obtain the statistical information of each group; The statistical information is: Wherein, s is the statistical information, H represents the height of the feature map, L represents the length of the feature map, and W represents the width of the feature map; F x is the characteristic of grouping; The statistical information is transformed according to a channel scale factor and a channel bias factor, and then subjected to nonlinear activation to obtain a channel attention parameter; The channel attention parameters are: CA x =δ(F c (s))=δ(S1s+b1)); Among them, CA x is the channel attention parameter, S1 is the channel scale factor, b1 is the channel bias factor, F c (s) is the grouped channel feature; δ is the nonlinear activation; Multiplying the channel attention parameter by the feature of the corresponding group to obtain the channel feature; The channel characteristics are: F′ x =CA x ·F x ; Among them, F′ x is the channel characteristic.

5. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to claim 1, characterized in that: The method of extracting the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features includes the following steps: Calculating the mean and variance of the channel features according to the grouping to perform regularization; The regularized channel features are transformed according to the spatial scale factor and the spatial bias factor, and then the spatial attention parameters are obtained through nonlinear activation; The spatial attention parameters are: SA x =δ(F c (GN(F′ x )))=δ(S2·GN(F′ x )+b2)); Among them, SA x is the spatial attention parameter, S2 is the spatial scale factor, b2 is the spatial bias factor, GN represents the regularization; F c (s) is the channel feature of the group; δ is the nonlinear activation; F′ x is the channel feature; Multiplying the spatial attention parameter by the corresponding channel feature to obtain the spatial feature; The spatial characteristics are: F″ x =SA·F′ x ; Among them, F x is the spatial feature.

6. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to claim 1, characterized in that: The fusing of the channel features and the spatial features of each channel to obtain a fused feature comprises the following steps: Define the adjacent grouping features as F x and F x+1 , the channel attention parameters are CA x , CA x+1 , the spatial attention parameters are SA x , SA x+1 , the overlapping characteristics of adjacent groups are , the channel attention parameters corresponding to the overlapping features are The spatial attention parameters corresponding to overlapping features The attention weight of the feature map is calculated by the channel attention parameter and the spatial attention parameter as W x_x+1 =CA x_x+1 ·SA x_x+1 , W x+1_x =CA x+1_x ·SA x+1_x ; W x_x+1 and W x+1_x Performing weighted summation on the weights to obtain the fusion feature; The fusion features are: Wherein, F is the fusion feature.

7. The method for multi-organ segmentation in plain scan CT based on multi-overlapping grouping attention mechanism according to any one of claims 1 to 6, characterized in that: Before fusing the channel features and the spatial features of the respective channels to obtain a fused feature, the method further includes the following steps: The channel features and the spatial features corresponding to each channel are randomly permuted to adjust the order between channels.

8. A multi-organ segmentation device for plain scan CT based on a multi-overlapping grouping attention mechanism, characterized by: The device comprises: A feature extraction unit, configured to extract a feature map from a plain scan CT image; A grouping number determining unit, configured to determine the number of groups of the feature map according to the number of channels of the image segmentation model; an overlapping grouping unit, configured to perform overlapping grouping on the feature map according to the grouping number so as to overlap the channels of features of adjacent groups and obtain overlapping group features; A channel attention processing unit, configured to extract the overlapping grouping features using the channel attention mechanism in the image segmentation model to obtain channel features; A spatial attention processing unit, configured to extract the channel features using the spatial attention mechanism in the image segmentation model to obtain spatial features; a feature fusion unit, configured to fuse the channel features and the spatial features of each channel to obtain a fused feature; An image segmentation unit is used to perform organ segmentation according to the fusion features using the image segmentation model.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images