Image unsupervised anomaly detection method, system, product, medium and equipment

Abstract

The invention belongs to the technical field of image anomaly detection. The invention provides an image unsupervised anomaly detection method and system, a product, a medium and equipment. The method comprises the following steps: extracting a first OCT image feature according to an OCT image and an encoder of a teacher network; extracting a second OCT image feature according to the OCT image and an encoder of a student network; the second OCT image features are sent to a global decoder branch to be compressed to obtain one-dimensional features, and the one-dimensional features are recovered to global student features through continuous multiple times of up-sampling and a hybrid self-task module; the second OCT image features are sent to a local decoder branch to obtain local student features; and calculating a first deviation value between the first OCT image feature and the global student feature, calculating a second deviation value between the first OCT image feature and the local student feature, and taking the sum of the first deviation value and the second deviation value as an anomaly detection score. According to the invention, the accuracy of OCT retina image anomaly detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image anomaly detection, and in particular to an image unsupervised anomaly detection method, system, product, medium and equipment. Background Art

[0002] The statements in this section merely provide background art related to the present invention and do not necessarily constitute prior art.

[0003] In ophthalmology, OCT retinal images are an important means of evaluating retinal health and diagnosing eye diseases, especially in the early detection of diseases such as glaucoma, macular degeneration, and diabetic retinopathy. However, traditional manual diagnostic methods rely on the experience of experts and are affected by factors such as physician subjectivity, fatigue, and time pressure, which may lead to missed or misdiagnoses. In addition, with the advancement of OCT technology, the amount of medical imaging data has increased dramatically, and the efficiency and consistency of manual inspections face severe challenges, making it difficult to meet the needs of modern ophthalmology clinics for efficiency and accuracy. Therefore, automated anomaly detection technology based on computer vision and deep learning has become an effective means to solve these problems. As a high-resolution three-dimensional imaging technology, OCT retinal images provide extremely rich retinal structural information, which is crucial for detecting subtle lesions. Through the introduction of deep learning technologies (such as convolutional neural networks (CNNs) and generative adversarial networks (GANs), computer vision systems can automatically learn the features of retinal images and effectively identify potential anomalies.

[0004] In OCT retinal images, the types of abnormalities are complex and diverse, and the early symptoms of retinal diseases are often not obvious, which makes it difficult to collect enough abnormal samples for supervised learning. Therefore, unsupervised anomaly detection methods have become the focus of research. This type of method relies only on normal retinal images for training and detects potential abnormalities by learning the features of normal images. Existing unsupervised anomaly detection methods based on knowledge distillation achieve anomaly detection by exploiting the differences in feature learning between the teacher network and the student network. However, due to the generalization ability of the network, the features learned by the student network tend to be similar to those of the teacher network, making it difficult for the model to effectively distinguish between normal and abnormal samples, making anomaly detection more difficult. In addition, existing unsupervised anomaly detection methods mainly focus on the detection of global abnormalities, but when dealing with complex pathological abnormalities or early local abnormalities (such as minor lesions, local retinal changes, etc.), there are still major deficiencies, and it is difficult to effectively identify and locate these subtle abnormal changes. Summary of the invention

[0005] In order to address the deficiencies of the prior art, the present invention provides an unsupervised image anomaly detection method, system, product, medium and device, which improve the accuracy of OCT retinal image anomaly detection and can provide important technical support for the early detection and clinical diagnosis of ophthalmic diseases.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides an unsupervised anomaly detection method for images.

[0008] An unsupervised anomaly detection method for an image includes the following processes:

[0009] Acquire an OCT image to be detected;

[0010] Extracting a first OCT image feature based on the OCT image and an encoder of a teacher network;

[0011] Extracting a second OCT image feature according to the OCT image and an encoder of the student network;

[0012] Sending the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restoring the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing with the self-task module;

[0013] Sending the second OCT image feature to the local decoder branch to obtain a local student feature;

[0014] A first deviation value between the first OCT image feature and the global student feature is calculated, and a second deviation value between the first OCT image feature and the local student feature is calculated. The sum of the first deviation value and the second deviation value is used as an abnormality detection score, and abnormality detection of the OCT image is performed according to the abnormality detection score.

[0015] As a further limitation of the first aspect of the present invention, the OCT image is a preprocessed OCT image, and the preprocessing includes: size scaling, normalization and standardization.

[0016] As a further limitation of the first aspect of the present invention, the one-dimensional feature is restored to a global student feature through continuous multiple upsampling and mixing from the task module, comprising:

[0017]

[0018] in, Indicates splicing, F sG is the global student characteristic, F s2is a one-dimensional feature, Mamba is a mixed self-task module, C1 represents the dimension of the encoded image, H1 represents the height of the encoded image, and W1 represents the width of the encoded image.

[0019] As a further limitation of the first aspect of the present invention, sending the second OCT image feature to the local decoder branch to obtain the local student feature includes:

[0020]

[0021] Among them, F sL is a local student feature. DW-Conv3×3(·) is a 3×3 depth convolution. ↑p(·) indicates that the feature is upsampled to the original resolution by the nearest neighbor interpolation at a specific level. Indicates that the input features are pooled to Size, F L′ =DW-Conv3×3(F s1 ∈R C1×H1×W1 )), F s1 is the second OCT image feature.

[0022] As a further limitation of the first aspect of the present invention, a distillation loss is used to guide the student decoder network to learn, and the knowledge distillation loss L A for:

[0023] L A =L G +L L ;

[0024] Among them, L G =||F t ∈R C1×H1×W1 -F sG ∈R C1×H1×W1 || 2 , L L =||F t ∈R C1×H1×W1 -F sL ∈R C1×H1×W1 || 2 , F t is the first OCT image feature, F sL is the local student characteristic, F sG It is a global student characteristic.

[0025] As a further limitation of the first aspect of the present invention, the sum of the first deviation value and the second deviation value is taken as the abnormality detection score S, including:

[0026] S=F t -F sL +F t -F sG;

[0027] Among them, F t is the first OCT image feature, F sL is the local student characteristic, F sG for global student characteristics;

[0028] The image is judged to be abnormal according to the set threshold. When the abnormality detection score S is greater than the set threshold, it is judged to be abnormal, otherwise there is no abnormality.

[0029] In a second aspect, the present invention provides an image unsupervised anomaly detection system.

[0030] An image unsupervised anomaly detection system, comprising:

[0031] An image acquisition unit is configured to: acquire an OCT image to be detected;

[0032] The teacher network feature extraction unit is configured to: extract a first OCT image feature according to the OCT image and an encoder of the teacher network;

[0033] A student network feature extraction unit is configured to: extract a second OCT image feature according to the OCT image and an encoder of the student network;

[0034] The global student feature extraction unit is configured to: send the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restore the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing from the task module;

[0035] The local student feature extraction unit is configured to: send the second OCT image feature to the local decoder branch to obtain the local student feature;

[0036] The abnormality detection score generating unit is configured to: calculate a first deviation value between the first OCT image feature and the global student feature, calculate a second deviation value between the first OCT image feature and the local student feature, take the sum of the first deviation value and the second deviation value as the abnormality detection score, and perform abnormality detection on the OCT image according to the abnormality detection score.

[0037] In a third aspect, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium;

[0038] a processor adapted to execute a computer program;

[0039] A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the image unsupervised anomaly detection method as described in the first aspect of the present invention is implemented.

[0040] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program is suitable for being loaded by a processor and executing the image unsupervised anomaly detection method as described in the first aspect of the present invention.

[0041] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, and when the computer program is executed by a processor, the image unsupervised anomaly detection method as described in the first aspect of the present invention is implemented.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. Aiming at the problem of unsupervised anomaly detection in OCT retinal images, the present invention designs a dual heterogeneous knowledge distillation network, which detects anomalies in images by utilizing the feature extraction differences between the teacher network and the student network. It can not only detect structural abnormalities in the retina (such as retinopathy, macular degeneration, etc.), but also effectively identify subtle pathological abnormalities.

[0044] 2. The teacher network of the present invention adopts an encoder structure, and the student network adopts an encoder plus decoder structure. The encoder part of the student network uses a convolutional neural network (CNN), and the decoder combines CNN and Mamba network models. In order to improve the model's ability to capture long-distance dependencies, a hybrid self-task Mamba module is designed for the student network to capture remote spatial dependencies through the Mamba network.

[0045] 3. The dual heterogeneous knowledge distillation network of the present invention adopts a multi-scale feature generation unit (MFGU) to interact with non-local features to enhance the model's ability to reconstruct details.

[0046] 4. Through the design of dual heterogeneous knowledge distillation, the teacher network and the student network can complement each other, thereby effectively detecting abnormal areas in OCT retinal images and providing accurate positioning information.

[0047] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings in the specification, which constitute 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 improper limitations on the present invention.

[0049] Figure 1 A schematic diagram of the process of the unsupervised anomaly detection method for images provided in Example 1 of the present invention;

[0050] Figure 2 A schematic diagram of an unsupervised anomaly detection system for images provided in Embodiment 2 of the present invention;

[0051] Figure 3 A schematic diagram of a computer device provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0053] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0054] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0055] Embodiment 1:

[0056] This implementation proposes an unsupervised anomaly detection method for images, including the following process:

[0057] S1: Data acquisition and preprocessing.

[0058] First, the OCT device is used to image the object to collect and acquire the original RAW data; then, the original data is processed using an image reconstruction algorithm to generate an OCT image, and the image is subjected to necessary preprocessing operations.

[0059] S1.1: Collect image data and define the original image as X∈R C×H×W , where C represents the original image dimension, H represents the height of the original image, and W represents the width of the original image. The original image data is resized using the following formula:

[0060]

[0061]

[0062] Among them, S(x, y) represents the corresponding pixel point in the scaled image, (x, y) is the pixel coordinate, X represents the original image, and X w and X h Represents the width and height of the original image, S w and S h Represents the width and height of the scaled image, k w and k h Indicates the scaling factor.

[0063] S1.2: Normalize the image data using the following formula:

[0064]

[0065] Among them, S max and S min They represent the maximum and minimum values ​​of the specific values ​​of all pixels in the image, respectively. c represents the image dimension, h represents the height of the image, and w represents the width of the image.

[0066] S1.3: Standardize the image data using the following formula:

[0067]

[0068] Among them, mean represents the mean of each channel, and std represents the standard deviation of each channel.

[0069] S2: Input the image to the teacher network.

[0070] Use the teacher network to train on the ImageNet dataset to extract image features X∈R C×H×W , and use the encoder part of the teacher network to extract the features F of the OCT image t ∈R C1×H1×W1 , where C1 represents the dimension of the encoded image, H1 represents the height of the encoded image, and W1 represents the width of the encoded image.

[0071] S3: Input the image to the student network.

[0072] The OCT images are simultaneously input into the student network. The student network adopts a single encoder and dual decoder structure, and uses distillation loss to train the network model. Specifically, it includes:

[0073] S3.1: OCT is input into the student network. The student network adopts a multi-task network model, in which the encoder adopts a CNN network model. The dual decoders are a global decoder branch and a local decoder branch, which are used to detect global anomalies and local anomalies.

[0074] S3.2: Input image X∈R C×H×W , the feature F of the student network is obtained through the encoding layer of the student network s1 ∈R C1 ×H1×W1 .

[0075] S3.3: The feature F s1 ∈R C1×H1×W1 Send it to the global decoder branch for feature compression and compress it into a one-dimensional feature F s2 ∈R C1×1×1 , then the feature Fs2 ∈R C1×1×1 After four consecutive upsampling and mixing of self-task modules, the global student feature F is restored sG ∈R C1×H1×W1 , the hybrid self-task module adopts CNN+Mamba structure. The specific formula is:

[0076]

[0077] in, Indicates splicing.

[0078] S3.4: Feature F s1 ∈R C1×H1×W1 It is sent to the local decoder branch and passed through the multi-scale feature generation unit (MFGU) to obtain the local student feature F sL ∈R C1×H1×W1 , the specific formula is:

[0079] F L′ =DW-Conv3×3(F s1 ∈R C1×H1×W1 )) (6);

[0080]

[0081] Among them, DW-Conv3×3(·) is a 3×3 depth convolution, ↑p(·) means that the features are upsampled to the original resolution by the nearest neighbor interpolation at a specific level, and Indicates that the input features are pooled to The size of .

[0082] The distillation loss is used to guide the student decoder network to learn. The optimization goal of the knowledge distillation module is:

[0083] L A =L G +L L (8);

[0084] L G =||F t ∈R C1×H1×W1 -F sG ∈R C1×H1×w1 || 2 (9);

[0085] L L =||F t ∈R C1×H1×W1 -F sL ∈R C1×H1×W1 || 2 (10).

[0086] S4: Calculate anomaly detection score.

[0087] After the model training is completed, during testing, the test data can be input into the teacher network and the student network for testing to obtain anomaly detection scores.

[0088] S4.1: Transform the feature map F by bilinear interpolation t ∈R C1×H1×W1 , F sL ∈R C1×H1×W1 and F sG ∈R C1×H1×W1 The image is resized to the original resolution and smoothed with a Gaussian kernel of σ = 4.

[0089] S4.2: The anomaly detection score S is obtained by the difference between the feature map E output by the teacher network and the feature map output by the student network. The specific formula is:

[0090] S = mean(F t -F sL +F t -F sG ) (11).

[0091] S4.3: After obtaining the anomaly detection score S, determine whether the image is abnormal based on the set threshold. For example, when the anomaly detection score S is greater than the set threshold, it is determined to be abnormal, otherwise there is no abnormality.

[0092] Embodiment 2:

[0093] like Figure 2 As shown, this implementation provides an image unsupervised anomaly detection system, including:

[0094] An image acquisition unit is configured to: acquire an OCT image to be detected;

[0095] The teacher network feature extraction unit is configured to: extract a first OCT image feature according to the OCT image and an encoder of the teacher network;

[0096] A student network feature extraction unit is configured to: extract a second OCT image feature according to the OCT image and an encoder of the student network;

[0097] The global student feature extraction unit is configured to: send the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restore the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing from the task module;

[0098] The local student feature extraction unit is configured to: send the second OCT image feature to the local decoder branch to obtain the local student feature;

[0099] The abnormality detection score generating unit is configured to: calculate a first deviation value between the first OCT image feature and the global student feature, calculate a second deviation value between the first OCT image feature and the local student feature, take the sum of the first deviation value and the second deviation value as the abnormality detection score, and perform abnormality detection on the OCT image according to the abnormality detection score.

[0100] The specific working process of each of the above units is described in Example 1 and will not be repeated here.

[0101] It is understandable that the above-mentioned various units can be separately or all combined into one or several other units to constitute, or some (some) units therein can also be split into multiple smaller units in function to constitute, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In practical applications, the function of a unit can also be realized by multiple units, or the function of multiple units can be realized by one unit. In other embodiments of the present application, the system can also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0102] According to another embodiment of the present application, the system described in this embodiment can be constructed, and the method of Example 1 of the present application can be implemented by running a computer program (including program code) capable of executing the steps involved in the corresponding method described in Example 1 on a general computing device such as a computer including processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, and loaded into the above-mentioned computing device through the computer-readable recording medium and run therein.

[0103] Embodiment 3:

[0104] like Figure 3 As shown, this implementation provides an electronic device, which includes a processor 1001, a communication interface 1002, and a computer-readable storage medium 1003. The processor 1001, the communication interface 1002, and the computer-readable storage medium 1003 may be connected via a bus or other means.

[0105] Among them, the communication interface 1002 is used to receive and send data, the computer-readable storage medium 1003 can be stored in the memory of the electronic device, the computer-readable storage medium 1003 is used to store a computer program, the computer program includes program instructions, and the processor 1001 is used to execute the program instructions stored in the computer-readable storage medium 1003.

[0106] Processor 1001 (or CPU (Central Processing Unit)) is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to implement corresponding method processes or corresponding functions.

[0107] The processor 1001 is configured to execute the following process:

[0108] Acquire an OCT image to be detected;

[0109] Extracting a first OCT image feature based on the OCT image and an encoder of a teacher network;

[0110] Extracting a second OCT image feature according to the OCT image and an encoder of the student network;

[0111] Sending the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restoring the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing with the self-task module;

[0112] Sending the second OCT image feature to the local decoder branch to obtain a local student feature;

[0113] A first deviation value between the first OCT image feature and the global student feature is calculated, and a second deviation value between the first OCT image feature and the local student feature is calculated. The sum of the first deviation value and the second deviation value is used as an abnormality detection score, and abnormality detection of the OCT image is performed according to the abnormality detection score.

[0114] The detailed process is described in Example 1 and will not be repeated here.

[0115] Embodiment 4:

[0116] This implementation provides a computer-readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The computer-readable storage medium provides a storage space that stores the processing system of the electronic device.

[0117] In addition, the storage space also stores one or more instructions suitable for being loaded and executed by the processor, and these instructions may be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here may be a high-speed RAM memory, or a non-volatile memory, such as at least one disk memory; optionally, it may also be at least one computer-readable storage medium located away from the aforementioned processor.

[0118] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to implement the following process:

[0119] Acquire an OCT image to be detected;

[0120] Extracting a first OCT image feature based on the OCT image and an encoder of a teacher network;

[0121] Extracting a second OCT image feature according to the OCT image and an encoder of the student network;

[0122] Sending the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restoring the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing with the self-task module;

[0123] Sending the second OCT image feature to the local decoder branch to obtain a local student feature;

[0124] A first deviation value between the first OCT image feature and the global student feature is calculated, and a second deviation value between the first OCT image feature and the local student feature is calculated. The sum of the first deviation value and the second deviation value is used as an abnormality detection score, and abnormality detection of the OCT image is performed according to the abnormality detection score.

[0125] The detailed process is described in Example 1 and will not be repeated here.

[0126] Embodiment 5:

[0127] The present implementation provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device performs the following process:

[0128] Acquire an OCT image to be detected;

[0129] Extracting a first OCT image feature based on the OCT image and an encoder of a teacher network;

[0130] Extracting a second OCT image feature according to the OCT image and an encoder of the student network;

[0131] Sending the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restoring the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing with the self-task module;

[0132] Sending the second OCT image feature to the local decoder branch to obtain a local student feature;

[0133] A first deviation value between the first OCT image feature and the global student feature is calculated, and a second deviation value between the first OCT image feature and the local student feature is calculated. The sum of the first deviation value and the second deviation value is used as an abnormality detection score, and abnormality detection of the OCT image is performed according to the abnormality detection score.

[0134] The detailed process is described in Example 1 and will not be repeated here.

[0135] A person skilled in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0136] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instruction can be stored in a computer-readable storage medium or transmitted by a computer-readable storage medium. The computer instruction can be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (for example, coaxial cable, optical fiber, digital line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server, a data center, etc. that contains one or more available media integrated. Available media can be magnetic media (for example, floppy disk, hard disk, tape), optical media (for example, DVD), or semiconductor media (for example, solid-state drive (Solid State Disk, SSD)) and the like.

[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An unsupervised anomaly detection method for an image, characterized in that: The process includes: Acquire an OCT image to be detected; Extracting a first OCT image feature based on the OCT image and an encoder of a teacher network; Extracting a second OCT image feature according to the OCT image and an encoder of the student network; Sending the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restoring the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing with the self-task module; Sending the second OCT image feature to the local decoder branch to obtain a local student feature; A first deviation value between the first OCT image feature and the global student feature is calculated, and a second deviation value between the first OCT image feature and the local student feature is calculated. The sum of the first deviation value and the second deviation value is used as an abnormality detection score, and abnormality detection of the OCT image is performed according to the abnormality detection score.

2. The unsupervised anomaly detection method for images according to claim 1, characterized in that: The OCT image is a preprocessed OCT image, and the preprocessing includes: size scaling, normalization and standardization.

3. The unsupervised anomaly detection method for images according to claim 1, characterized in that: The one-dimensional features are restored to global student features through continuous multiple upsampling and mixing with the task module, including: in, Indicates splicing, F sG is the global student characteristic, F s2 is a one-dimensional feature, Mamba is a mixed self-task module, C1 represents the dimension of the encoded image, H1 represents the height of the encoded image, and W1 represents the width of the encoded image.

4. The unsupervised anomaly detection method for images according to claim 1, wherein: The second OCT image feature is sent to the local decoder branch and passed through the multi-scale feature generation unit to obtain the local student feature, including: Among them, F sL is a local student feature. DW-Conv3×3(·) is a 3×3 depth convolution. ↑p(·) indicates that the feature is upsampled to the original resolution by the nearest neighbor interpolation at a specific level. Indicates that the input features are pooled to Size, F L′ =DW-Conv3×3(F s1 ∈R C1×H1×W1 )), F s1 is the second OCT image feature, C1 represents the dimension of the encoded image, H1 represents the height of the encoded image, and W1 represents the width of the encoded image.

5. The unsupervised anomaly detection method for images according to any one of claims 1 to 4, characterized in that: The distillation loss is used to guide the student decoder network to learn, and the knowledge distillation loss L A for: L A =L G +L L ; Among them, L G =||F t ∈R C1×H1×W1 -F sG ∈R C1×H1×W1 || 2 , L L =||F t ∈R C1×H1×W1 -F sL ∈R C1×H1×W1 || 2 , F t is the first OCT image feature, F sL is the local student characteristic, F sG It is a global student characteristic.

6. The unsupervised anomaly detection method for images according to any one of claims 1 to 4, characterized in that: The sum of the first deviation value and the second deviation value is taken as the abnormality detection score S, including: S=mean(F t -F sL +F t -F sG ); Among them, F t is the first OCT image feature, F sL is the local student characteristic, F sG for global student characteristics; The image is judged to be abnormal according to the set threshold. When the abnormality detection score S is greater than the set threshold, it is judged to be abnormal, otherwise there is no abnormality.

7. An unsupervised anomaly detection system for images, characterized in that: include: An image acquisition unit is configured to: acquire an OCT image to be detected; The teacher network feature extraction unit is configured to: extract a first OCT image feature according to the OCT image and an encoder of the teacher network; A student network feature extraction unit is configured to: extract a second OCT image feature according to the OCT image and an encoder of the student network; The global student feature extraction unit is configured to: send the second OCT image feature to the global decoder branch for compression to obtain a one-dimensional feature, and restore the one-dimensional feature to a global student feature through continuous multiple upsampling and mixing from the task module; The local student feature extraction unit is configured to: send the second OCT image feature to the local decoder branch to obtain the local student feature; The abnormality detection score generating unit is configured to: calculate a first deviation value between the first OCT image feature and the global student feature, calculate a second deviation value between the first OCT image feature and the local student feature, take the sum of the first deviation value and the second deviation value as the abnormality detection score, and perform abnormality detection on the OCT image according to the abnormality detection score.

8. A computer device, characterized in that: include: a processor and a computer readable storage medium; a processor adapted to execute a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the image unsupervised anomaly detection method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the image unsupervised anomaly detection method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method for unsupervised anomaly detection in an image as claimed in any one of claims 1 to 6 is implemented.

Images