Liver echinococcosis CT image intelligent detection method based on domain self-adaption, electronic equipment and storage medium

Abstract

The invention provides a liver echinococcosis CT image intelligent detection method based on domain self-adaption, electronic equipment and a storage medium. The method comprises the following steps: collecting liver CT image data; carrying out image preprocessing on a multi-scale structure preserving enhancement model based on gradient flow field guidance; carrying out target detection labeling; and model training. The method has the beneficial effects that the influence of different CT equipment on liver echinococcosis detection can be effectively reduced by utilizing a field adaptive technology, so that the accuracy and consistency of AI-assisted doctor diagnosis are improved, the medical cost can be saved, and the medical experience and health management of patients in remote areas can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of image detection, and in particular relates to a domain-adaptive method, electronic equipment and storage medium for intelligent detection of hepatic echinococcosis CT images. Background Art

[0002] Hepatic echinococcosis is a zoonotic parasitic disease that mainly occurs in economically underdeveloped pastoral areas and high-altitude areas. These areas usually have harsh climatic conditions and limited medical resources. CT scanning is the main means of diagnosing HE, but the traditional manual segmentation method is time-consuming and labor-intensive, and highly dependent on the doctor's subjective experience and judgment, making it difficult to ensure the repeatability of the results. The existing CT image detection also has the following limitations: When traditional AI models are used for testing, they will encounter the challenge of "domain gap" caused by differences in data distribution. The "domain gap" stems from the differences in data acquisition and processing between different models of CT equipment. This difference is mainly manifested in differences in imaging protocols, reconstruction algorithms, etc., resulting in significant differences in contrast, noise, spatial resolution, etc. in the scanned images of the same patient. Therefore, even if a deep learning model trained on one device performs well on one device, it may experience reduced accuracy or performance collapse on another device. The most direct way to solve this gap is to collect enough medical data from all other fields and have it annotated by professional doctors. But it is obvious that this direct approach is almost impossible to achieve, whether it is the scarcity of data or the expensive expert annotation. Summary of the invention

[0003] In view of this, the present invention aims to propose a domain-adaptive intelligent detection method, electronic device and storage medium for hepatic echinococcosis CT images, so as to solve the low accuracy and high misdiagnosis rate of hepatic echinococcosis caused by differences in equipment and models in traditional CT image detection.

[0004] To achieve the above object, the technical solution of the present invention is achieved as follows: A domain-adaptive intelligent detection method for hepatic echinococcosis CT images comprises the following steps: S1. Collect liver CT imaging data; S2, performing image preprocessing on the liver CT image data collected in step S1 based on a multi-scale structure-preserving enhancement model guided by a gradient flow field; S3, performing target detection and annotation on the image data preprocessed in step S2; S4. Perform model training on the image data annotated with target detection in step S3.

[0005] Furthermore, in step S2, the liver CT image data collected in step S1 are preprocessed based on the multi-scale structure-preserving enhancement model guided by the gradient flow field, including the following steps: S21, Estimation of gradient fields and structure tensors at different scales; S22, introduce structural continuity measurement, and obtain preliminary enhancement results through fusion; S23. Process the image using non-local mean denoising and adaptive sharpening.

[0006] Furthermore, in step S21, the gradient field and the structure tensor are estimated at different scales, including: For the input CT image At different scales Perform Gaussian convolution to get a smooth image: ; For each smoothed image , calculate the gradient component: ; ; This gives the gradient field: ; Constructing the structure tensor : ; in, represents a smoothed image, The variance is The Gaussian kernel of represents the convolution operation, It represents the horizontal spatial coordinate in the image. It represents the vertical coordinate of the horizontal direction in the image. Indicates that the Gaussian kernel is The derivative in the direction, Indicates that the Gaussian kernel is Derivative in direction.

[0007] Furthermore, in step S22, a structural continuity metric is introduced, and a preliminary enhancement result is obtained through fusion, including: Introducing structural continuity metrics for different regions of CT images: ; in, represents the structural continuity measure, is a hyperparameter for adjusting sensitivity; Represents the gradient strength of the local area of ​​the image in the main direction, Represents the gradient strength of the local area of ​​the image in the secondary direction. Indicates that the region lacks significant directionality; when When , it means that the area has obvious line structure or edge information; At every scale Next, through the kernel function For the center pixel Enhancement, kernel function According to all pixels in the neighborhood With the center pixel The weights are assigned based on the similarity between them.

[0008] Furthermore, in step S23, the image is processed using non-local mean denoising and adaptive sharpening, including: At every scale Next, it is a target pixel , define the adaptive enhancement kernel To weight the information of surrounding pixels: ; Spatial proximity : ; in, represents the adaptive enhancement kernel, which is used to weight the information of surrounding pixels. Represents the grayscale difference, represents the directional consistency function, , Both represent the main direction angles of the target pixel and the surrounding pixels. is a hyperparameter that controls the range of spatial smoothing; Attenuation according to the Euclidean distance through a Gaussian function; Grayscale Difference : ; in, A hyperparameter to control the sensitivity of grayscale differences; Direction similarity: ; in, and All are main direction angles; The input image is scaled The smoothed version is obtained by , using the kernel function in a fixed neighborhood Perform weighted summation within to obtain the enhanced pixel value : ; Calculate for all Weight , perform weighted summation on the grayscale value of each pixel in the neighborhood and then divide it by the weighted sum; Each scale σ gets an enhanced result , through structural continuity Weighted fusion of each scale: ; ; in, Indicates enhancing the result weight.

[0009] Furthermore, in step S3, when the object of target detection annotation is cystic echinococcosis data, a cystic echinococcosis data set is established, and the categories include single cystic echinococcosis, multiascus echinococcosis, internal capsule collapse echinococcosis, necrotizing and consolidating echinococcosis, calcified echinococcosis, infiltrative echinococcosis, and liquefied cavitary echinococcosis.

[0010] Furthermore, in step S4, model training is performed on the image data annotated with target detection in step S3, including the following steps: S41. Select features whose similarity with the target domain feature map is higher than a set value from the source domain feature map of the CT image to form a feature set. ; S42, based on feature set The maximum mean difference is used to measure the difference in feature distribution between the source domain and the target domain; S43. By calculating the distance between the source domain features and the target domain features, the model training impact of the source domain features is dynamically adjusted in the loss calculation.

[0011] Further, in step S41, features whose similarity with the target domain feature map is higher than a set value are selected from the source domain feature map of the CT image to form a feature set ,include: Cosine similarity is selected as the similarity measure, and the formula is as follows: ; in, is the feature vector in the source domain, is the feature vector in the target domain; Setting similarity threshold , similarity threshold It is used to filter out features with similarity higher than the set value by setting the similarity threshold Forming a feature set : ; in, is the source domain feature, is the target domain feature.

[0012] Further, in step S42, based on the feature set The maximum mean difference is used to measure the difference in feature distribution between the source domain and the target domain, including: Calculate source domain features and its closest target domain feature The pairwise distance of the local domain alignment loss is taken as the average of all the minimum pairwise Euclidean distances. : ; in, is the batch size of the source domain, is the batch size of the target domain; Introducing global domain alignment loss Quantify the global difference between source and target domain features; by using the first-order statistics of the distribution of source and target domain features, the global difference between domains is approximately expressed as: ; The global distribution difference between the source domain and the target domain is a non-Gaussian distribution, which introduces a parameterized nonlinear function Capture deeper global distribution differences: ; in, , is a learnable parameter, so the global domain difference As shown below: ; in, represents the transpose of the weight matrix, Indicates the global represents transpose, represents the mean of the source domain features, represents the mean of the target domain features, Represents the bias term, which is used to adjust the calculation of the global difference.

[0013] Furthermore, in step S43, by calculating the distance between the source domain features and the target domain features, the model training influence of the source domain features is dynamically adjusted in the loss calculation, including: Introducing dynamic training parameters, which can be automatically adjusted according to the current learning state of the model; ; in, Indicates The source domain feature influence factor in each training iteration, is a preset constant; The source domain features are divided into multiple levels according to their similarity with the target domain features, among which the low-similarity features retain the gradient update: ; ; ; in, is the maximum similarity, It is the level coefficient automatically calculated based on the maximum similarity and is used to adjust the weighting of the loss; is the source domain loss of the yolo model; Yes Control Hyperparameters of .

[0014] An electronic device includes a processor and a memory that is communicatively connected to the processor and is used to store instructions executable by the processor, wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor, and the processor is used to execute the domain-adaptive method for intelligent detection of hepatic echinococcosis CT images.

[0015] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the domain-adaptive-based intelligent detection method for hepatic echinococcosis CT images is implemented.

[0016] Compared with the prior art, the domain-adaptive intelligent detection method, electronic device and storage medium for hepatic echinococcosis CT images described in the present invention have the following advantages: The present invention describes an intelligent detection method for hepatic echinococcosis CT images based on domain adaptation, an electronic device and a storage medium. The method utilizes domain adaptation technology to effectively reduce the impact of different CT devices on the detection of hepatic echinococcosis, thereby improving the accuracy and consistency of AI-assisted doctor diagnosis, which can not only save medical costs, but also improve the medical experience and health management of patients in remote areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 It is a schematic diagram of the overall method flow of an embodiment of the present invention; Figure 2 This is a schematic diagram of the image preprocessing process according to an embodiment of the present invention; Figure 3 The figure is a schematic diagram of the model training process described in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0019] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0020] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0021] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0022] like Figures 1 to 3 As shown, a domain-adaptive method for intelligent detection of hepatic echinococcosis CT images includes the following steps: S1. Data collection: collect high-quality liver CT imaging data to ensure data representativeness and diversity.

[0023] S2. Perform image preprocessing on the liver CT image data collected in step S1 based on a multi-scale structure-preserving enhancement model guided by a gradient flow field.

[0024] S3. Perform target detection and annotation on the image data preprocessed in step S2.

[0025] S4. Perform model training on the image data annotated with target detection in step S3.

[0026] The final experimental results of this method were: cystic echinococcosis: mAP50-95=95.8%, an increase of 3.1%; alveolar echinococcosis: mAP50-95=96.8%, an increase of 2.2%.

[0027] In a preferred embodiment of the present invention, in step S2, the liver CT image data collected in step S1 are preprocessed based on a multi-scale structure-preserving enhancement model guided by a gradient flow field: In the preprocessing of liver CT images, it is necessary to balance noise suppression with the preservation and enhancement of anatomical structures (such as blood vessels, lesion boundaries, etc.). The multi-scale structure-preserving enhancement model proposed in this invention first estimates the gradient flow field and structure tensor at different scales, and obtains preliminary enhancement results through fusion. Subsequently, non-local mean denoising and adaptive sharpening are used to further improve the visualization quality and diagnostic value of the image. Specifically: For the input CT image At different scales Perform Gaussian convolution to obtain a series of smooth images: ; For each smoothed image , calculate the gradient component: ; ; This gives the gradient field: ; Constructing the structure tensor : ; in, represents a smoothed image, The variance is The Gaussian kernel of represents the convolution operation, It represents the horizontal spatial coordinate in the image. It represents the vertical coordinate of the horizontal direction in the image. Indicates that the Gaussian kernel is The derivative in the direction, Indicates that the Gaussian kernel is Derivative in direction.

[0028] In an image, the structural features of different regions vary greatly. In order to properly adjust the enhancement strength, a structural continuity measure is introduced: ; in, represents the structural continuity measure, is a hyperparameter for adjusting sensitivity. Represents the gradient strength of the local area of ​​the image in the main direction, Represents the gradient strength of the local area of ​​the image in the secondary direction. Indicates that the region lacks significant directionality; when , it indicates that the region has obvious line structure or edge information.

[0029] At every scale Next, for the center pixel The core of the enhancement lies in the kernel function , which is based on all pixels in the neighborhood With the center pixel The larger the weight, the more and The closer they are, the greater the proportion they account for in the fusion.

[0030] At every scale Next, it is a target pixel , define the adaptive enhancement kernel To weight the information of surrounding pixels: ; Spatial proximity : ; in, represents the adaptive enhancement kernel, which is used to weight the information of surrounding pixels. Represents the grayscale difference, represents the directional consistency function, , Both represent the main direction angles of the target pixel and the surrounding pixels. is a hyperparameter that controls the range of spatial smoothing.

[0031] The Gaussian function is used to decay according to the Euclidean distance: the farther the distance, the smaller the weight; the closer the distance, the larger the weight.

[0032] Grayscale Difference : ; Taking into account the similarity of local grayscale values: the closer the grayscale values ​​of two pixels are, the easier it is for them to reinforce each other; on the contrary, when the gap is large, the weight decay is obvious. is a hyperparameter that controls the sensitivity of grayscale differences.

[0033] Direction similarity: ; If the main direction angle and If the difference is close, it means that the two positions are similar in structure and should be given a higher weight. (almost opposite direction), then reduce its impact. Add some modular operations.

[0034] The input image is scaled The smoothed version is obtained by , using the above kernel function in a fixed neighborhood Perform weighted summation within to obtain the enhanced pixel value : ; In the discrete implementation, the above integral is replaced by a weighted sum of pixels in the neighborhood: (1) Calculate for all pixels Weight , perform weighted summation on the grayscale value of each pixel in the neighborhood and then divide it by the weighted sum.

[0035] Each scale σ can get an enhanced result In order to strengthen the coarse structure and retain the fine texture in general, the structural continuity Weighted fusion of each scale: ; ; in, Indicates the weight of the enhanced result. If When there are more obvious structural differences in the region, Smaller or larger, depending on the model design, this scale can receive a relatively higher or lower weight. By integrating the enhanced information at each scale, the characteristics of the gradient flow field are also retained.

[0036] In a preferred embodiment of the present invention, in step S3, target detection and annotation are performed on cystic echinococcosis data to establish a cystic echinococcosis data set, which is classified into single cystic echinococcosis, multiascus echinococcosis, internal capsule collapse echinococcosis, necrotizing and consolidative echinococcosis, calcified echinococcosis, infiltrative echinococcosis, and liquefaction and cavitary echinococcosis.

[0037] In a preferred embodiment of the present invention, in step S4, it is proposed in the background technology above that the traditional AI model will encounter the challenge of "domain gap" caused by data distribution differences when performing detection. The "domain gap" originates from the differences in data acquisition and processing between different models of CT equipment. This difference is mainly manifested in differences in imaging protocols, reconstruction algorithms, etc., resulting in significant differences in contrast, noise, spatial resolution, etc. in the scanned images of the same patient. Therefore, even if a deep learning model trained by one device performs well on one device, it may experience a decrease in accuracy or performance collapse on another device. The most direct way to solve this gap is to collect enough medical data from all other fields and have professional doctors annotate it. But it is obvious that this direct method is almost impossible to achieve, whether it is the scarcity of data or the expensive expert annotation. Therefore, the present invention has developed a new model to solve these problems in CT. This domain adaptive model, the intelligent identification and cross-domain fusion model of hepatic echinococcosis (HCE-YOLO), mainly optimizes the loss function.

[0038] Specifically, first, we need to select feature instances with high similarity to the target domain feature map from the source domain feature map. The first step in this process is to calculate the source domain feature and target domain features The similarity between them. Cosine similarity is selected as the similarity measure, and the specific formula is as follows: ; here, is the feature vector in the source domain, is the feature vector in the target domain. Once the similarity between the source domain features and the target domain features is calculated, a similarity threshold needs to be set This threshold is used to filter out features with high similarity and ensure that the selected features are meaningful to the target domain features. By setting this threshold, a feature set can be formed. : ; In practical applications, It is obtained through experimental tuning. is the source domain feature, is the target domain feature.

[0039] Secondly, once you get The feature set can be used to measure the difference in feature distribution between the source domain and the target domain using the maximum mean difference (MMD): Calculate source domain features and its nearest feature in the target domain The pairwise distance of the local domain alignment loss is calculated by averaging all the minimum pairwise distances (pairwise Euclidean distances). : ; in is the batch size of the source domain, is the batch size of the target domain. The batch size is the number of samples processed by the model in one forward propagation and back propagation. Minimizing the pairwise distance can directly optimize the feature alignment of different domains. It can alleviate local feature differences by minimizing the pairwise distance between intermediate features in the source and target domains, but this method alone may not be sufficient to completely eliminate the global distribution shift between the source and target domains. In order to further enhance the cross-domain transferability of features, a global domain alignment loss is introduced To quantify the global difference between source and target domain features. By using the first-order statistics of the source and target domain feature distributions, the global difference between domains can be approximated as: ; The global distribution difference between the source domain and the target domain is a non-Gaussian distribution, which may have complex features such as multiple modes, spikes, and skew. This complex distribution feature is difficult to fully describe with simple mean and variance, and requires more powerful statistics. To this end, a parameterized nonlinear function is introduced To capture deeper global distribution differences: ; in, , is a learnable parameter. Therefore, the global domain difference As shown below: ; in, represents the transpose of the weight matrix, Indicates the global represents transpose, represents the mean of the source domain features, represents the mean of the target domain features, Represents the bias term, which is used to adjust the calculation of the global difference.

[0040] Finally, in order to achieve independence of model functions from target domain features, an adaptive gradient stopping mechanism is introduced to dynamically adjust the training influence of source domain features in the loss calculation by calculating the distance between source and target domain features.

[0041] During the training process, the distribution distance between source domain features and target domain features will change. In order to respond to this change in time, dynamic training parameters can be introduced, which will be automatically adjusted according to the current learning state of the model.

[0042] ; in, Indicates The source domain feature influence factor in each training iteration, is a preset small constant. In the early stage of training, the weight of the source domain features is high, and gradually decreases with the increase of the number of iterations. This method can not only flexibly adjust the contribution of the source domain, but also avoid the performance degradation caused by paying too much attention to the target domain too early.

[0043] By grading the similarity between source domain features and target domain features, different gradient control strengths can be applied to features of different levels. Specifically, source domain features can be divided into multiple levels according to their similarity with target domain features. High-similarity features have less influence on the target domain, while low-similarity features can retain more gradient updates: ; ; ; in, is the maximum similarity, It is a level coefficient automatically calculated based on the maximum similarity and is used to adjust the weighting of the loss. is the source domain loss of the yolo model. Yes Control Hyperparameters of .

[0044] The present invention also proposes an electronic device, comprising a processor and a memory that is communicatively connected to the processor and is used to store instructions executable by the processor, wherein the memory stores instructions that can be executed by the processor, and the instructions are executed by the processor, and the processor is used to execute the domain-adaptive intelligent detection method for hepatic echinococcosis CT images.

[0045] The present invention also proposes a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the domain-adaptive-based intelligent detection method for hepatic echinococcosis CT images.

[0046] This method uses domain adaptation technology to effectively reduce the impact of different CT devices on the detection of hepatic echinococcosis, thereby improving the accuracy and consistency of AI-assisted doctor diagnosis, which can not only save medical costs, but also improve the medical experience and health management of patients in remote areas.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A domain-adaptive intelligent detection method for hepatic echinococcosis CT images, characterized in that: The following steps are involved: S1. Collect liver CT imaging data; S2, performing image preprocessing on the liver CT image data collected in step S1 based on a multi-scale structure-preserving enhancement model guided by a gradient flow field; S3, performing target detection and annotation on the image data preprocessed in step S2; S4, performing model training on the image data annotated by the target detection in step S3; In step S2, the liver CT image data collected in step S1 are preprocessed based on the multi-scale structure-preserving enhancement model guided by the gradient flow field, including the following steps: S21, Estimation of gradient fields and structure tensors at different scales; S22, introduce structural continuity measurement, and obtain preliminary enhancement results through fusion; S23, processing the image using non-local mean denoising and adaptive sharpening; In step S4, model training is performed on the image data annotated with target detection in step S3, including the following steps: S41. Select features whose similarity with the target domain feature map is higher than a set value from the source domain feature map of the CT image to form a feature set. ; S42, based on feature set The maximum mean difference is used to measure the difference in feature distribution between the source domain and the target domain; S43. By calculating the distance between the source domain features and the target domain features, the model training impact of the source domain features is dynamically adjusted in the loss calculation.

2. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 1, characterized in that: In step S21, the gradient field and structure tensor are estimated at different scales, including: For the input CT image At different scales Perform Gaussian convolution to get a smooth image: ； For each smoothed image , calculate the gradient component: ； ； This gives the gradient field: ； Constructing the structure tensor : ； in, represents a smoothed image, The variance is The Gaussian kernel of represents the convolution operation, It represents the horizontal spatial coordinate in the image. It represents the vertical coordinate of the horizontal direction in the image. Indicates that the Gaussian kernel is The derivative in the direction, Indicates that the Gaussian kernel is Derivative in direction.

3. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 2, characterized in that: In step S22, the structural continuity measure is introduced, and a preliminary enhancement result is obtained through fusion, including: Introducing structural continuity metrics for different regions of CT images: ； in, represents the structural continuity measure, is a hyperparameter for adjusting sensitivity; Represents the gradient strength of the local area of ​​the image in the main direction, Represents the gradient strength of the local area of ​​the image in the secondary direction. Indicates that the region lacks significant directionality; when When , it means that the area has obvious line structure or edge information; At every scale Next, through the kernel function For the center pixel Enhancement, kernel function According to all pixels in the neighborhood With the center pixel The weights are assigned based on the similarity between them.

4. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 3 is characterized in that: In step S23, the image is processed using non-local mean denoising and adaptive sharpening, including: At every scale Next, it is a target pixel , define the adaptive enhancement kernel To weight the information of surrounding pixels: ； Spatial proximity : ； in, represents the adaptive enhancement kernel, which is used to weight the information of surrounding pixels. Represents the grayscale difference, represents the directional consistency function, , Both represent the main direction angles of the target pixel and the surrounding pixels. is a hyperparameter that controls the range of spatial smoothing; Attenuation according to the Euclidean distance through a Gaussian function; Grayscale Difference : ； in, A hyperparameter to control the sensitivity of grayscale differences; Direction similarity: ； in, and All are main direction angles; The input image is scaled The smoothed version is obtained by , using the kernel function in a fixed neighborhood Perform weighted summation within to obtain the enhanced pixel value : ； Calculate for all Weight , perform weighted summation on the grayscale value of each pixel in the neighborhood and then divide it by the weighted sum; Each scale σ gets an enhanced result , through structural continuity Weighted fusion of each scale: ； ； in, Indicates enhancing the result weight.

5. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 1, characterized in that: In step S3, when the object of target detection and annotation is cystic echinococcosis data, a cystic echinococcosis data set is established, and the categories include single cystic echinococcosis, multiascus echinococcosis, internal capsule collapse echinococcosis, necrotizing and consolidating echinococcosis, calcified echinococcosis, infiltrative echinococcosis, and liquefied cavitary echinococcosis.

6. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 1, characterized in that: In step S41, features whose similarity with the target domain feature map is higher than a set value are selected from the source domain feature map of the CT image to form a feature set ,include: Cosine similarity is selected as the similarity measure, and the formula is as follows: ； in, is the feature vector in the source domain, is the feature vector in the target domain; Setting similarity threshold , similarity threshold It is used to filter out features with similarity higher than the set value by setting the similarity threshold Forming a feature set : ； in, is the source domain feature, is the target domain feature.

7. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 6, characterized in that: In step S42, based on the feature set The maximum mean difference is used to measure the difference in feature distribution between the source domain and the target domain, including: Calculate source domain features and its closest target domain feature The pairwise distance of the local domain alignment loss is taken as the average of all the minimum pairwise Euclidean distances. : ； in, is the batch size of the source domain, is the batch size of the target domain; Introducing global domain alignment loss Quantify the global difference between source and target domain features; by using the first-order statistics of the distribution of source and target domain features, the global difference between domains is approximately expressed as: ； The global distribution difference between the source domain and the target domain is a non-Gaussian distribution, which introduces a parameterized nonlinear function Capture deeper global distribution differences: ； in, , is a learnable parameter, so the global domain difference As shown below: ； in, represents the transpose of the weight matrix, Indicates the global represents transpose, represents the mean of the source domain features, represents the mean of the target domain features, Represents the bias term, which is used to adjust the calculation of the global difference.

8. The method for intelligent detection of hepatic echinococcosis CT images based on domain adaptation according to claim 7, characterized in that: In step S43, by calculating the distance between the source domain features and the target domain features, the model training influence of the source domain features is dynamically adjusted in the loss calculation, including: Introducing dynamic training parameters, which can be automatically adjusted according to the current learning state of the model; ； in, Indicates The source domain feature influence factor in each training iteration, is a preset constant; The source domain features are divided into multiple levels according to their similarity with the target domain features, among which the low-similarity features retain the gradient update: ； ； ； in, is the maximum similarity, It is the level coefficient automatically calculated based on the maximum similarity and is used to adjust the weighting of the loss; is the source domain loss of the yolo model; Yes Control Hyperparameters of .

9. An electronic device, comprising a processor and a memory connected to the processor for storing instructions executable by the processor, characterized in that: The memory stores instructions that can be executed by the processor, and the instructions are executed by the processor. The processor is used to execute the domain-adaptive intelligent detection method for hepatic echinococcosis CT images described in any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the domain-adaptive intelligent detection method for hepatic echinococcosis CT images described in any one of claims 1 to 8 is implemented.

Images