Image detection method, device and computer equipment

Abstract

The application relates to an image detection method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: after an initial medical image is acquired, the initial medical image is segmented to obtain a plurality of segmented image blocks, so that a larger medical image is segmented into a plurality of smaller segmented image blocks, and each segmented image block is subsequently processed in a targeted manner; then, the plurality of segmented image blocks are input into a first model to obtain a lesion positioning probability and a pathological classification result of each segmented image block, and an attention map of the initial medical image is output based on the obtained lesion positioning probability and pathological classification result. Since the attention map of the initial medical image can make the image signs in the initial medical image more easily recognized, so that the medical image is detected in a more targeted manner, the accuracy of a detection result obtained by detecting the medical image can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image data processing, and in particular to an image detection method and device, a computer device, a storage medium, and a computer program product. BACKGROUND

[0002] Molybdenum target and magnetic resonance imaging (MRI) are common image examination methods. At present, a computer-aided examination system is usually used to detect the image type of a medical image according to image signs (such as the properties, shapes, and nipple depression of a lump and calcification) of the medical image. However, the detection result accuracy is poor when the computer-aided examination system is used to detect the medical image. SUMMARY

[0003] The present application provides an image detection method and device, a computer device, a computer readable storage medium, and a computer program product. After processing a medical image, the image signs in the medical image are more easily identified, the medical image is more easily detected in a targeted manner, and the detection of the medical image is more accurate.

[0004] In a first aspect, the present application provides an image detection method, which includes:

[0005] obtaining a plurality of segmentation image blocks of an initial medical image;

[0006] obtaining an attention map of the initial medical image according to the segmentation image blocks and a first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is used to represent a lesion positioning probability of each segmentation image block, and the second attention map is used to represent a pathological classification probability of each segmentation image block;

[0007] obtaining a detection result of the initial medical image based on the attention map.

[0008] In a second aspect, the present application also provides an image detection device, which includes:

[0009] a first module configured to obtain a plurality of segmentation image blocks of an initial medical image;

[0010] a second module configured to obtain an attention map of the initial medical image according to the segmentation image blocks and a first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is used to represent a lesion positioning probability of each segmentation image block, and the second attention map is used to represent a pathological classification probability of each segmentation image block;

[0011] The third module is used to obtain the detection results of the initial medical image based on the attention map.

[0012] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above:

[0013] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0014] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above methods.

[0015] This application provides an image detection method, apparatus, computer device, computer-readable storage medium, and computer program product. The method includes: after acquiring an initial medical image, segmenting the initial medical image to obtain multiple segmented image blocks, thus dividing a large medical image into multiple smaller segmented image blocks, facilitating subsequent targeted processing of each segmented image block; then inputting the multiple segmented image blocks into a first model to obtain the lesion localization probability and pathological classification result of each segmented image block; and outputting an attention map of the initial medical image based on the obtained lesion localization probability and pathological classification result. Since the attention map of the initial medical image makes the image features in the initial medical image easier to identify, enabling more targeted detection of the medical image, it can improve the accuracy of the detection results obtained from medical image detection. Attached Figure Description

[0016] Figure 1 This is an application environment diagram of an image detection method in one embodiment;

[0017] Figure 2 This is a flowchart illustrating an image detection method in one embodiment;

[0018] Figure 3 This is a flowchart illustrating the image detection method in another embodiment;

[0019] Figure 4 This is a flowchart illustrating the image detection method in another embodiment;

[0020] Figure 5 This is a flowchart illustrating the image detection method in another embodiment;

[0021] Figure 6 This is a flowchart illustrating the image detection method in another embodiment;

[0022] Figure 7 Flowchart of the image detection method in another embodiment;

[0023] Figure 8 Block diagram of the image detection device in an embodiment;

[0024] Figure 9 Internal structure diagram of the computer device in an embodiment. DETAILED DESCRIPTION

[0025] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0026] The image detection method provided by the present application can be applied to the application environment as shown in Figure 1 . Wherein, the medical image acquisition device 102 can be that after acquiring the initial medical image, the acquired initial medical image is transmitted to the terminal 104 or the server 106 through the network, the terminal 104 or the server 106 performs segmentation processing on the initial medical image, then the processed segmented image block is input into the first model, the first model is used to analyze the lesion positioning probability and the pathological classification probability of each segmented image block, the lesion positioning probability and the pathological classification probability of each segmented image block are obtained, and the first attention map and the second attention map corresponding to the initial medical image are further obtained, and the detection result of the initial medical image is obtained based on the first attention map and the second attention map. Wherein, the attention map can locate the lesion on the medical image, and at the same time, the pathological classification result used for clinical diagnosis can be obtained, so as to provide more reliable basis for subsequent auxiliary diagnosis of diseases, so as to achieve the effect of improving the accuracy of disease diagnosis. Wherein, the medical image acquisition device 102 can be but not limited to various CT scanning devices, B-ultrasound devices, MR devices, etc. The terminal 104 can be but not limited to personal computers, notebook computers, tablet computers, etc. The server 106 can be realized by an independent server or a server cluster composed of multiple servers.

[0027] In an embodiment, as shown in Figure 2 , an image detection method is provided, which is taken as an example to illustrate the server in Figure 1 , including the following steps:

[0028] Step S202, a plurality of segmented image blocks of an initial medical image are acquired.

[0029] The initial medical image can be collected by a medical image collection device. After the medical image collection device collects the initial medical image, the initial medical image can be sent to a corresponding terminal device for subsequent processing by the terminal device. After the medical image collection device collects the initial medical image, the initial medical image can also be sent to a corresponding terminal device first, and then transmitted to a server by the terminal device through a network for subsequent processing by the server. The present application does not limit this. The medical image collection device may, for example, be a computed tomography device, an ultrasonic diagnostic device, an MRI diagnostic device, an X-ray diagnostic device, a CT scanning device, a magnetic resonance scanning device, a B-ultrasound device, etc.

[0030] The initial medical image can be a plurality of medical images collected by the same medical image collection device, or a plurality of medical images collected by a plurality of medical image collection devices. The initial medical image can be a medical image of the human brain, a medical image of the human chest, a medical image of the human breast, or a medical image of the human abdominal cavity, etc. The initial medical image can belong to one examiner or a plurality of examiners. If the initial medical image belongs to a plurality of examiners, the terminal device can classify the initial medical image according to different examiners after receiving the initial medical image, so as to carry out targeted analysis for different examiners.

[0031] At present, the initial medical image collected by the medical image collection device is generally a large-size image with a large amount of information. If medical analysis is directly performed on the initial medical image, the pressure on the server to process data will increase, and the large amount of information in the image will result in inaccurate analysis results due to the inability to perform targeted analysis. To solve this problem, the present application performs segmentation processing on the initial medical image after the server obtains the initial medical image, and divides the large-size initial medical image into a plurality of segmented image blocks. The segmentation can be performed according to the pixels of the initial medical image according to a predetermined pixel gradient, according to the RGB data of the initial medical image according to a predetermined RGB range, or according to the depth information of the initial medical image according to a predetermined depth range, etc. The present application does not limit this. The image can also be calibrated in the initial medical image by a detection frame to obtain a plurality of image detection frames.

[0032] For example, the initial medical image with a voxel size of 512*512*512 can be segmented into image blocks with a voxel size of 64*64*64, 125 image blocks with a voxel size of 64*64*64 can be randomly selected, the initial medical image with a voxel size of 512*512*512 can be segmented into image blocks with a voxel size of 32*32*32, 125 image blocks with a voxel size of 32*32*32 can be randomly selected, the initial medical image with a voxel size of 512*512*512 can be segmented into image blocks with a voxel size of 16*16*16, 125 image blocks with a voxel size of 16*16*16 can be randomly selected, or the initial medical image with a voxel size of 512*512*512 can be segmented into image blocks with a voxel size of 64*64*64 and image blocks with a voxel size of 32*32*32, 125 image blocks with a voxel size of 64*64*64 and 125 image blocks with a voxel size of 32*32*32 can be randomly selected. It should be noted that a different number or different size of image blocks with a corresponding voxel size can be selected for subsequent detection operations according to different requirements, which is not limited in the present application.

[0033] It should be noted that the smaller the pixel value of each image block after image segmentation is, the more conducive to subsequent analysis. However, the smaller the pixel value is, the more the number of segmented image blocks obtained after segmentation is, which will increase the analysis pressure of the server and reduce the image analysis efficiency. Therefore, the user can set different segmentation strategies according to the analysis requirements and the actual situation of the initial medical image to process different initial medical images, so as to improve the analysis efficiency of the server while ensuring the analysis quality. For example, if the initial medical image is preliminarily identified by a professional diagnostician to have a probability of less than 10% of existing abnormalities, the initial medical image can be segmented by setting a larger pixel gradient, so that the number of segmented image blocks is smaller, which is convenient for the server to process the segmented image blocks quickly.

[0034] In step S204, an attention map of the initial medical image is obtained according to the segmented image blocks and the first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is used to represent the lesion positioning probability of each segmented image block, and the second attention map is used to represent the pathological classification result of each segmented image block.

[0035] The first model can be obtained by unsupervised training of an initial neural network model using image blocks containing lesions and image blocks without lesions, can be obtained by supervised training of the initial neural network model using image blocks containing lesions and image blocks without lesions, or can be obtained by training the initial neural network model using a single algorithm, an integrated algorithm, image blocks containing lesions, and image blocks without lesions, and the application does not limit the same. The lesion positioning probability is used to represent the probability of whether the segmented image block contains a lesion. The pathological classification result is used to represent the image type classification result of the segmented image block, for example, the segmented image block is a normal image type or an abnormal image type. Further, the pathological classification result can also represent a specific classification result of the abnormal image, for example, a first classification result, a second classification result, a third classification result, and the like. Taking a breast mammogram as an example, the first classification result represents that the segmented image block is a breast nipple indentation type image (for example, no nipple indentation, mild nipple indentation, complete nipple indentation, and the like), the second classification result represents that the segmented image block is a breast gland density type image (for example, no gland, high-density gland, equal-density gland, low-density gland, and fat-containing density gland), and the third classification result represents that the segmented image block is a breast BI-RADS type image (for example, BI-RADS level 0, BI-RADS level 1, BI-RADS level 2, and the like).

[0036] Then, based on the above segmentation of the initial medical image, a plurality of segmented image blocks are obtained, and then the plurality of segmented image blocks need to be analyzed one by one. The analysis method can be to input the plurality of segmented image blocks into the first model, and the first model analyzes the plurality of segmented image blocks respectively. It should be noted that the first model can obtain the lesion positioning probability and the pathological classification probability from the same pixel segmented image block, or can obtain the lesion positioning probability and the pathological classification result from different pixel segmented image blocks, and the application does not limit the same.

[0037] When the segmented image blocks are input into the first model, the first model can obtain the lesion positioning probability by comparing the segmented image blocks with the image blocks with lesions and / or the image blocks without lesions. The first model can also obtain the pathological classification result by comparing the segmented image blocks with the image blocks with diseases. For example, 300 segmented image blocks with 512*512 pixel values are input into the first model, the first model compares each segmented image block with the image blocks with lesions and / or the image blocks without lesions, obtains the comparison result, and determines the lesion positioning probability of each segmented image block based on the comparison result. Then, 300 segmented image blocks with 128*128 pixel values are input into the first model, the first model compares each segmented image block with the image blocks with diseases, obtains the comparison result, and determines the pathological classification result of each segmented image block based on the comparison result. The first model in the server can output the corresponding attention map based on the analysis result of each segmented image block. The first attention map represents the lesion positioning probability of each segmented image block, and the second attention map represents the pathological classification result of each segmented image block. In this way, the positioning accuracy and the classification accuracy of the initial medical image can be improved.

[0038] In step S206, the detection result of the initial medical image is obtained based on the attention map.

[0039] In the above method, the attention map output by the first model after the plurality of segmented image blocks are input into the first model can reflect the region where the lesion is located and the classification result of the image. Further, the detection result obtained by further detection based on the attention map is more accurate than the detection result obtained by detection based on only the initial medical image. The detection result can be an image type determination result of the initial medical image, for example, the detection result is that the image type of the initial medical image is the first type, the image type of the initial medical image is the second type, the image type of the initial medical image is the third type, the image type of the initial medical image is the fourth type, etc. The first type represents that the initial medical image is a breast nipple depression type image, the second type represents that the initial medical image is a breast gland density type image, etc. The detection result can be obtained according to the attention map and the corresponding detection model, or can be obtained according to the attention map, the initial medical image, and the corresponding detection model. The detection result can be represented by a specific classification result or a probability, which is not limited in the present application. For example, the first attention map and the second attention map are input into the detection model, and the detection result output by the detection model is that the initial medical image is a first type image.

[0040] The application provides an image detection method, which comprises the following steps: after an initial medical image is acquired, the initial medical image is segmented to obtain a plurality of segmented image blocks, so that the large medical image is segmented into a plurality of small segmented image blocks, and subsequent processing of each segmented image block is targeted; then, the plurality of segmented image blocks are input into a first model to obtain a lesion positioning probability and a pathological classification result of each segmented image block, and an attention map of the initial medical image is output based on the obtained lesion positioning probability and the pathological classification result. Since the attention map of the initial medical image can make the image signs in the initial medical image more easily recognized, so that the medical image is more targeted for detection, the accuracy of the detection result obtained by detecting the medical image can be improved.

[0041] In one embodiment, as shown in the figure, Figure 3 The embodiment is an optional method embodiment for obtaining the attention map of the initial medical image according to the segmented image blocks and the first model, and the method comprises the following steps:

[0042] In step S302, the segmented image blocks are input into the first model to obtain the output results corresponding to the segmented image blocks, wherein the output results comprise the lesion positioning probability and / or the pathological classification result.

[0043] Based on the plurality of segmented image blocks obtained by segmenting the initial medical image by the server, the first model can compare the segmented image blocks with the image blocks with lesions and / or the image blocks without lesions to obtain the lesion positioning probability. The first model can also compare the segmented image blocks with the image blocks with abnormalities to obtain the pathological classification result, and the like. For example, after the first model analyzes the first segmented image block, the output lesion positioning probability of the first segmented image block is 10%, the output lesion positioning probability of the second segmented image block is 10%, the output lesion positioning probability of the third segmented image block is 30%, the output lesion positioning probability of the fourth segmented image block is 20%, the output lesion positioning probability of the fifth segmented image block is 60%, and the like. Further, it can be illustrated that the probability of the first segmented image block having a lesion is 10%, the probability of the second segmented image block having a lesion is 10%, the probability of the third segmented image block having a lesion is 30%, the probability of the fourth segmented image block having a lesion is 20%, and the probability of the fifth segmented image block having a lesion is 60%.

[0044] In step S304, the attention map of the initial medical image is obtained based on the output results and the image parameters of the segmented image blocks.

[0045] For example, the image parameters of the segmented image blocks can be the RGB value of the image, the saturation of the image, the gray value of the image, and the like.

[0046] After the lesion segmentation probability and / or the lesion property classification probability determined based on the first model, the target image parameters corresponding to each segmented image block can be determined based on the lesion positioning probability and the pathological classification result of each segmented image block, and then the initial image parameters of each segmented image block are modified according to the target image parameters corresponding to each segmented image block, and finally the attention map of the initial medical image is obtained according to the modified segmented image blocks. The first model can be to display the corresponding target image parameters in each segmented image block after the target image parameters of each segmented image block are determined, and finally the attention map of the initial medical image is obtained according to each segmented image block after the target image parameters are displayed; it can also be to modify the initial image parameters of the initial medical image according to the average value of the target image parameter values of each segmented image block, and the modified initial medical image is taken as the attention map, etc. The present application does not limit this. It should be noted here that the server can store a plurality of information corresponding tables of lesion positioning probability and image parameters, and / or a plurality of information corresponding tables of pathological classification result and image parameters. When each segmented image block inputs the first model to obtain the lesion positioning probability and / or the pathological classification result corresponding to each segmented image block, the server can obtain the corresponding information table from the corresponding memory address to determine the target image parameters of each segmented image block, so as to quickly determine the target image parameters of each segmented image block and improve the processing efficiency of the initial medical image.

[0047] The present application provides an image detection method, which inputs each segmented image block into a first model, and modifies the image parameters of each segmented image block based on the lesion positioning probability and the pathological classification result of each segmented image block, or modifies the image parameters of the initial medical image. The image obtained after processing is more conducive to subsequent image detection, and improves the accuracy of image detection.

[0048] In one embodiment, as shown in Figure 4 The present embodiment is an optional method embodiment for obtaining the attention map of the initial medical image based on the output result and the image parameters of each segmented image block, which includes the following steps:

[0049] Step S402, determining the first image parameter adjustment strategy of each segmented image block according to the lesion positioning probability corresponding to each segmented image block.

[0050] The server can be configured to continue to obtain the information corresponding table of the lesion positioning probability and the image parameter from the corresponding memory address, and determine the target image parameter of each segmented image block. For example, for the first segmented image block, the first model outputs a lesion positioning probability of 70%, and the server determines the target RGB value of the first segmented image block as (176 23 31) according to the information corresponding table of the lesion positioning probability and the RGB value. For the second segmented image block, the first model outputs a lesion positioning probability of 10%, and the server determines the target RGB value of the second segmented image block as (255 192 203) according to the information corresponding table of the breast hyperplasia probability and the RGB value. For the third segmented image block, the first model outputs a lesion positioning probability of 30%, and the server determines the target RGB value of the third segmented image block as (176 224 230) according to the information corresponding table of the lesion positioning probability and the RGB value. The above examples are not exhaustive.

[0051] In step S404, the image parameter of each segmented image block is modified based on the first image parameter adjustment strategy of each segmented image block.

[0052] After determining the target image parameter of each segmented image block in the above step, the first image parameter modification strategy of each segmented image block is determined according to the initial image parameter and the target image parameter of each segmented image block. The first image parameter modification strategy is to modify the initial image parameter of the segmented image block to the target image parameter. After modifying the image parameter of the segmented image block, the initial medical image is equivalent to being information-labeled, which makes the subsequent image detection more directional, improves the efficiency of image detection, and improves the accuracy of image detection.

[0053] In step S406, the modified segmented image blocks are mapped based on the initial medical image, and a first attention map is obtained according to the mapping result.

[0054] Wherein, because the first model is for the lesion positioning probability and the pathological classification result of each segmented image block, the image detection of the application is to detect the initial medical image to obtain the detection result of the initial medical image, and the scattered lesion positioning probability and pathological classification result of the segmented image block can only be reflected on the initial medical image to obtain the detection result of the initial medical image. Therefore, it is necessary to integrate the above segmented image blocks with modified image parameters, and the integration needs to map the modified segmented image blocks based on the initial medical image, that is, arrange the modified segmented image blocks according to the original arrangement of the initial medical image to obtain the first attention map corresponding to the initial medical image. The first attention map is different from the initial medical image in that the image parameters are different.

[0055] The application provides an image detection method, which modifies the image parameters of each segmented image block based on the lesion segmentation probability of each segmented image block, and maps the modified segmented image blocks based on the initial medical image to obtain a first attention map. The first attention map shows the lesion position in the initial medical image, which facilitates subsequent detection of the initial medical image with more directionality and further obtains more accurate detection results.

[0056] In one embodiment, as shown in FIG. 5, the embodiment is another optional method embodiment for obtaining the attention map of the initial medical image based on the output result and the image parameters of each segmented image block, which includes the following steps: Figure 5

[0057] Step S502, determining the second image parameter modification strategy of each segmented image block according to the pathological classification result corresponding to each segmented image block;

[0058] Step S504, modifying the image parameters of each segmented image block based on the second image parameter modification strategy of each segmented image block;

[0059] Step S506, mapping the modified segmented image blocks based on the initial medical image, and obtaining a second attention map according to the mapping result.

[0060] ​The method for obtaining the second attention map is the same as the method for obtaining the first attention map, and the difference between the two is that the second attention map is determined based on the pathological classification result corresponding to each segmented image block. The pathological classification result has different image parameters from the lesion positioning probability, so the obtained second attention map is different from the first attention map. The second attention map can reflect the pathological classification result of each segmented image block, and in combination with the first attention map, it can provide more directional and more reliable data support for subsequent detection of the initial medical image, thereby improving the detection efficiency and accuracy of the initial medical image. The specific method for determining the second attention map is referred to the determination method of the first attention map, which will not be repeated here.

[0061] The present application provides an image detection method. The method modifies the image parameters of each segmented image block based on the pathological classification result of each segmented image block, and maps the modified each segmented image block based on the initial medical image to obtain a second attention map. The second attention map can represent the classification result of the pathology to which each segmented image block belongs, and provide more reliable data support for subsequent detection of the initial medical image, so as to obtain the detection result of the initial medical image more quickly. At the same time, the accuracy of the detection result can be improved.

[0062] In one embodiment, the present embodiment is based on an attention map to obtain an optional method embodiment of a detection result of an initial medical image. The method embodiment includes the following steps:

[0063] Based on the attention map, the initial medical image, and a second model, a detection result of the initial medical image is obtained.

[0064] Optionally, the first attention map and the initial medical image are input into the second model to obtain a detection result of the initial medical image.

[0065] Optionally, the second attention map and the initial medical image are input into the second model to obtain a detection result of the initial medical image.

[0066] Optionally, the first attention map, the second attention map, and the initial medical image are input into the second model to obtain a detection result of the initial medical image.

[0067] The initial medical image can be a medical image of a tissue of interest, which can be, for example, a breast, a liver, a spleen, a stomach, a heart, a kidney, a large intestine, a small intestine, a large arm, a small arm, a large leg, a small leg, a hand, a foot, a reproductive organ, a neck, a head, etc.

[0068] The second model can be used to classify the initial medical images by image type, thereby achieving a more accurate classification. The second model can be obtained through unsupervised training of the initial neural network model using medical images containing lesions, medical images without lesions, attention map samples corresponding to medical images with lesions, and attention maps corresponding to medical images without lesions; it can also be obtained through supervised training of the initial neural network model using medical images containing lesions, medical images without lesions, attention map samples corresponding to medical images with lesions, and attention maps corresponding to medical images without lesions; or it can be obtained by training the initial neural network model using a single algorithm, an ensemble algorithm, medical images containing lesions, medical images without lesions, attention map samples corresponding to medical images with lesions, and attention maps corresponding to medical images without lesions, etc. The attention maps can include a first attention map corresponding to medical images with lesions and a first attention map corresponding to medical images without lesions; or it can include a first attention map and a second attention map corresponding to medical images with lesions and a first attention map and a second attention map corresponding to medical images without lesions, etc.

[0069] The first attention map and second attention map obtained based on the above steps can be input into the second model along with the initial medical image to obtain the detection result of the initial medical image. Specifically, this can be done by inputting the first attention map and the initial medical image into the second model to obtain the detection result of the initial medical image; it can be done by inputting the second attention map and the initial medical image into the second model to obtain the detection result of the initial medical image; or it can be done by inputting the first attention map, the second attention map, and the initial medical image into the second model to obtain the detection result of the initial medical image. This application does not limit this approach.

[0070] This application provides an image detection method. The method inputs an attention map and an initial medical image into a second model. Since the attention map can locate lesions and classify images pathologically, the second model, combined with the initial medical image and the corresponding attention map, can quickly obtain the detection results of the initial medical image, thereby improving the detection efficiency and the accuracy of the detection results.

[0071] In one embodiment, such as Figure 6 As shown, this embodiment is an optional method embodiment for training the first model, which includes the following steps:

[0072] Step S602: Obtain segmented image block samples of multiple medical image samples of the first tissue of interest, wherein the segmented image block samples include segmented image block samples without lesions and segmented image block samples with lesions.

[0073] Step S604, training the first preset prediction model based on each segmented image block sample to obtain a first model.

[0074] The first interested tissue can be the same tissue as the target interested tissue, and the medical image samples of the first interested tissue and the medical image of the target interested tissue belong to different objects. For example, the target interested tissue is a breast, and the first interested tissue is also a breast, but the initial medical image of the target interested tissue is a breast molybdenum target image from an A object, and the medical image samples of the first interested tissue are breast molybdenum target images from B1, C1, D1, E1, F1, G1, H1, I1 and other objects. The medical image samples can be image samples including lesions and image samples without lesions, such as breast molybdenum target images with mastitis, breast molybdenum target images with breast fibroadenoma, breast molybdenum target images with breast hyperplasia, etc. Then, the medical image samples are segmented to obtain segmented image samples, the segmented image samples are labeled with lesion positioning probabilities, the first image parameters corresponding to each lesion positioning probability are determined, and / or the pathological classification results are labeled to determine the second image parameters corresponding to each pathological classification result. Finally, each segmented image block sample, the lesion positioning probability corresponding to each segmented image block sample, the pathological classification result corresponding to each segmented image block sample, the first image parameter and the second image parameter are input into the first prediction model for model training to obtain the first model. The first prediction model can be, for example, a convolutional neural network model, a recurrent neural network model, a deep belief neural network model, a generative adversarial network model, etc.

[0075] In one embodiment, as shown in FIG. 7, the present embodiment is an optional method embodiment for training the second model, which includes the following steps: Figure 7

[0076] Step S702, obtaining a plurality of medical image samples of a second interested tissue and attention map samples corresponding to the medical image samples, wherein the medical image samples include image samples without lesions and image samples with lesions;

[0077] Step S704, training a second preset prediction model based on each medical image sample and each attention map sample to obtain a second model.

[0078] ​The second interested tissue can be the same tissue as the target interested tissue, and the medical image sample of the second interested tissue and the medical image of the target interested tissue belong to different objects. The medical image sample of the second interested tissue can be a mammogram image from B1 object, C1 object, D1 object, E1 object, F1 object, G1 object, H1 object, I1 object, and the like. The medical image sample can also be a mammogram image from B2 object, C2 object, D2 object, E2 object, F2 object, G2 object, H2 object, I2 object, and the like. The medical image sample can be an image sample including a lesion and an image sample without a lesion, for example, a mammogram image of a breast with mastitis, a mammogram image of a breast with fibroadenoma, a mammogram image of a breast with hyperplasia, and the like. Then, the medical image sample of the second interested tissue is processed to obtain a plurality of attention maps corresponding to the medical image sample according to the method of processing the initial medical image to obtain the attention map described above, and then each medical image sample, each medical image sample corresponding attention map sample and corresponding detection result (first type, second type, third type, etc.) are input into a second prediction model to obtain a second model. The second prediction model can be a convolutional neural network model, a recurrent neural network model, a deep belief neural network model, a generative adversarial network model, and the like.

[0079] In the following, the image detection method provided by the present application is applied to breast cancer screening, and the image detection method of the present application is described in detail as follows:

[0080] Mammogram images of B1 object, C1 object, D1 object, E1 object, F1 object, G1 object, H1 object, I1 object, and the like are obtained. Among the mammogram images of the plurality of objects, 100 mammogram images of the image type of breast cancer and 100 mammogram images of the image type of normal breast are included.

[0081] One hundred mammograms of breast cancer and one hundred mammograms of normal breast were segmented to obtain 400 mammogram blocks of breast cancer and 400 mammogram blocks of normal breast. For each mammogram block of breast cancer, lesion localization probability and pathological classification results were labeled, as were those for the mammogram blocks of normal breast. Then, 400 mammograms of breast cancer images, along with their corresponding lesion localization probabilities and pathological classification results, and 400 mammograms of normal breast images, along with their corresponding lesion localization probabilities (probability of benign breast lesions, probability of malignant breast lesions, etc.) and pathological classification results (BI-RADS classification results, classification results of breast calcification nature, classification results of lesion nature, etc.; BI-RADS classification results, for example, BI-RADS level 0, 1; classification results of breast calcification nature, for example, diffuse calcification, regionally distributed calcification, scattered calcification; classification results of lesion nature, for example, none, mass, calcification, mass with calcification, asymmetric density, structural distortion, etc.), the correspondence between lesion localization probabilities and image parameters, and the correspondence between pathological classification results and image parameters, are input into the first convolutional neural network model for training to obtain the first model;

[0082] 400 mammograms of breast cancer images and 400 mammograms of normal breast images are input into the first model to obtain the corresponding attention map.

[0083] One hundred mammograms of breast cancer and one hundred mammograms of normal breast, along with their corresponding attention maps, are input into the second convolutional neural network model for training to obtain the second model.

[0084] Obtain the target mammogram of object A, and then input the target mammogram of object A into the first model to obtain the first attention map and / or the second attention map corresponding to the target mammogram;

[0085] The first attention map, the second attention map, and the target mammogram are input into the second model to obtain the detection result of the target mammogram; wherein, the detection result of the target mammogram is that the target mammogram is an image of the first type, and the first type represents the image type of the target mammogram as the breast cancer image type.

[0086] It should be understood that although each step in the flowchart involved in the embodiments described above is shown in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, there is no strict order limitation for the execution of these steps, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps.

[0087] Based on the same inventive concept, the embodiments of the present application also provide an image detection device for implementing the above-mentioned image detection method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more image detection device embodiments provided below can refer to the limitations of the image detection method in the above text, which will not be repeated here.

[0088] In one embodiment, as shown in Figure 8 An image detection device is provided, comprising: a first module 802, a second module 804 and a third module 806, wherein:

[0089] The first module 802 is configured to obtain a plurality of segmentation image blocks of an initial medical image.

[0090] The second module 804 is configured to obtain an attention map of the initial medical image according to each segmentation image block and a first model, wherein the attention map comprises a first attention map and / or a second attention map, the first attention map is used to represent a lesion positioning probability of each segmentation image block, and the second attention map is used to represent a pathological classification result of each segmentation image block.

[0091] The third module 806 is configured to obtain a detection result of the initial medical image based on the attention map.

[0092] In one embodiment, the second module 804 is specifically configured to input each segmentation image block into the first model to obtain an output result corresponding to each segmentation image block, wherein the output result comprises a lesion positioning probability and / or a pathological classification result.

[0093] The output result and the image parameters of each segmentation image block are used to obtain the attention map of the initial medical image.

[0094] In one embodiment, the second module 804 is specifically further configured to determine a first image parameter modification strategy of each segmentation image block according to the lesion positioning probability corresponding to each segmentation image block.

[0095] modify the image parameters of each segmented image block based on the first image parameter modification strategy of each segmented image block;

[0096] map the modified segmented image blocks based on the initial medical image, and obtain a first attention map according to a mapping result.

[0097] In an embodiment, the second module 804 is specifically further configured to determine a second image parameter modification strategy of each segmented image block according to the pathological classification result corresponding to each segmented image block;

[0098] modify the image parameters of each segmented image block based on the second image parameter modification strategy of each segmented image block;

[0099] map the modified segmented image blocks based on the initial medical image, and obtain a second attention map according to a mapping result.

[0100] In an embodiment, the third module 806 is specifically configured to obtain a detection result of the initial medical image based on the attention map, the initial medical image, and the second model.

[0101] In an embodiment, the third module 806 is specifically further configured to input the first attention map and the initial medical image into the second model to obtain the detection result of the initial medical image,

[0102] or,

[0103] input the second attention map and the initial medical image into the second model to obtain the detection result of the initial medical image;

[0104] or,

[0105] input the first attention map, the second attention map, and the initial medical image into the second model to obtain the detection result of the initial medical image.

[0106] In an embodiment, the image detection apparatus described above further includes a training module (not shown in the figure),

[0107] The training module is configured to obtain segmented image block samples of a plurality of medical image samples of a first tissue of interest, wherein the segmented image block samples include segmented image block samples without lesions and segmented image block samples with lesions.

[0108] train a preset first prediction model based on each segmented image block sample to obtain a first model.

[0109] In an embodiment, the training module is further configured to obtain a plurality of medical image samples of a second tissue of interest and attention map samples corresponding to the medical image samples, wherein the medical image samples include image samples without lesions and image samples with lesions.

[0110] training the second preset prediction model based on each medical image sample and each attention map sample to obtain a second model.

[0111] Each of the modules in the image detection apparatus can be implemented wholly or partially by software, hardware, and combinations thereof. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in the computer device in software form, so as to be invoked and executed by the processor to perform operations corresponding to the modules.

[0112] In an embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in Figure 9 The computer device includes a processor, a memory, and a network interface connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store medical image data. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement an image detection method.

[0113] Those skilled in the art can understand that Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0114] In an embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the following steps:

[0115] obtaining a plurality of segmentation image blocks of an initial medical image;

[0116] obtaining an attention map of the initial medical image according to each segmentation image block and the first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is configured to represent lesion positioning probabilities of each segmentation image block, and the second attention map is configured to represent pathological classification results of each segmentation image block;

[0117] obtaining a detection result of the initial medical image based on the attention map.

[0118] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0119] The segmented image blocks are input into the first model to obtain output results corresponding to the segmented image blocks, wherein the output results include lesion positioning probabilities and / or pathological classification results;

[0120] Based on the output results and the image parameters of the segmented image blocks, an attention map of the initial medical image is obtained.

[0121] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0122] According to the lesion positioning probabilities corresponding to the segmented image blocks, first image parameter modification strategies of the segmented image blocks are determined;

[0123] Based on the first image parameter modification strategies of the segmented image blocks, the image parameters of the segmented image blocks are modified;

[0124] The modified segmented image blocks are mapped based on the initial medical image, and a first attention map is obtained according to the mapping results.

[0125] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0126] According to the pathological classification results corresponding to the segmented image blocks, second image parameter modification strategies of the segmented image blocks are determined;

[0127] Based on the second image parameter modification strategies of the segmented image blocks, the image parameters of the segmented image blocks are modified;

[0128] The modified segmented image blocks are mapped based on the initial medical image, and a second attention map is obtained according to the mapping results.

[0129] In one embodiment, the processor, when executing the computer program, also implements the following steps: based on the attention map, the initial medical image and the second model, a detection result of the initial medical image is obtained.

[0130] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0131] The first attention map and the initial medical image are input into the second model to obtain a detection result of the initial medical image,

[0132] or,

[0133] The second attention map and the initial medical image are input into the second model to obtain a detection result of the initial medical image;

[0134] or,

[0135] The first attention map, the second attention map and the initial medical image are input into the second model to obtain a detection result of the initial medical image.

[0136] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0137] Obtaining segmentation image block samples of a plurality of first interest tissues of medical image samples, wherein the segmentation image block samples include segmentation image block samples without lesions and segmentation image block samples with lesions;

[0138] Training a preset first prediction model based on each segmentation image block sample to obtain a first model.

[0139] In one embodiment, the processor, when executing the computer program, also implements the following steps:

[0140] Obtaining a plurality of medical image samples of second interest tissues and attention map samples corresponding to the medical image samples, wherein the medical image samples include image samples without lesions and image samples with lesions;

[0141] Training a preset second prediction model based on each medical image sample and each attention map sample to obtain a second model.

[0142] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the following steps:

[0143] Obtaining a plurality of segmentation image blocks of an initial medical image;

[0144] Obtaining an attention map of the initial medical image according to each segmentation image block and the first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is used to represent lesion positioning probabilities of each segmentation image block, and the second attention map is used to represent pathological classification results of each segmentation image block;

[0145] Obtaining a detection result of the initial medical image based on the attention map.

[0146] In one embodiment, the computer program, when executed by the processor, also implements the following steps:

[0147] Inputting each segmentation image block into the first model to obtain an output result corresponding to each segmentation image block, wherein the output result includes a lesion positioning probability and / or a pathological classification result;

[0148] Obtaining the attention map of the initial medical image based on the output result and image parameters of each segmentation image block.

[0149] In one embodiment, the computer program, when executed by the processor, also implements the following steps:

[0150] determine a first image parameter modification strategy for each segmented image block according to the lesion positioning probability corresponding to each segmented image block;

[0151] modify the image parameters of each segmented image block based on the first image parameter modification strategy of each segmented image block;

[0152] map the modified segmented image blocks based on the initial medical image, and obtain a first attention map according to the mapping result.

[0153] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0154] determine a second image parameter modification strategy for each segmented image block according to the pathological classification result corresponding to each segmented image block;

[0155] modify the image parameters of each segmented image block based on the second image parameter modification strategy of each segmented image block;

[0156] map the modified segmented image blocks based on the initial medical image, and obtain a second attention map according to the mapping result.

[0157] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0158] obtain a detection result of the initial medical image based on the attention map, the initial medical image and the second model.

[0159] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0160] input the first attention map and the initial medical image into the second model to obtain a detection result of the initial medical image,

[0161] or,

[0162] input the second attention map and the initial medical image into the second model to obtain a detection result of the initial medical image;

[0163] or,

[0164] input the first attention map, the second attention map and the initial medical image into the second model to obtain a detection result of the initial medical image.

[0165] In one embodiment, the computer program is executed by the processor to further implement the following steps:

[0166] obtain segmented image block samples of medical image samples of a plurality of first tissues of interest, wherein the segmented image block samples include segmented image block samples without lesions and segmented image block samples with lesions;

[0167] training the first preset prediction model based on each segmented image block sample, to obtain a first model.

[0168] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0169] obtaining a plurality of second medical image samples of a tissue of interest and attention map samples corresponding to the medical image samples, wherein the medical image samples include image samples without lesions and image samples with lesions;

[0170] training a second preset prediction model based on each medical image sample and each attention map sample, to obtain a second model.

[0171] In one embodiment, a computer program product is provided, comprising a computer program which, when executed by the processor, implements the following steps:

[0172] obtaining a plurality of segmented image blocks of an initial medical image;

[0173] obtaining an attention map of the initial medical image according to each segmented image block and the first model, wherein the attention map includes a first attention map and / or a second attention map, the first attention map is used to represent lesion positioning probabilities of each segmented image block, and the second attention map is used to represent pathological classification results of each segmented image block;

[0174] obtaining a detection result of the initial medical image based on the attention map.

[0175] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0176] inputting each segmented image block into the first model to obtain an output result corresponding to each segmented image block, wherein the output result includes a lesion positioning probability and / or a pathological classification result;

[0177] obtaining an attention map of the initial medical image based on the output result and image parameters of each segmented image block.

[0178] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0179] determining a first image parameter modification strategy of each segmented image block according to the lesion positioning probability corresponding to each segmented image block;

[0180] modifying image parameters of each segmented image block based on the first image parameter modification strategy of each segmented image block;

[0181] mapping the modified each segmented image block based on the initial medical image, and obtaining the first attention map according to the mapping result.

[0182] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0183] determining a second image parameter modification strategy for each segmented image block according to the pathological classification result corresponding to each segmented image block;

[0184] modifying the image parameters of each segmented image block based on the second image parameter modification strategy of each segmented image block;

[0185] mapping the modified segmented image blocks based on the initial medical image, and obtaining a second attention map according to the mapping result.

[0186] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0187] obtaining a detection result of the initial medical image based on the attention map, the initial medical image and the second model.

[0188] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0189] inputting the first attention map and the initial medical image into the second model to obtain a detection result of the initial medical image,

[0190] or,

[0191] inputting the second attention map and the initial medical image into the second model to obtain a detection result of the initial medical image;

[0192] or,

[0193] inputting the first attention map, the second attention map and the initial medical image into the second model to obtain a detection result of the initial medical image.

[0194] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0195] obtaining segmented image block samples of medical image samples of a plurality of first interest tissues, wherein the segmented image block samples include segmented image block samples without lesions and segmented image block samples with lesions;

[0196] training a preset first prediction model based on each segmented image block sample to obtain a first model.

[0197] In one embodiment, the computer program, when executed by the processor, further implements the following steps:

[0198] obtaining a plurality of medical image samples of a second interest tissue and attention map samples corresponding to the medical image samples, wherein the medical image samples include image samples without lesions and image samples with lesions;

[0199] training the second preset prediction model based on each medical image sample and each attention map sample to obtain a second model.

[0200] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0201] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist, they should be considered as the scope of the present disclosure.

[0202] The above embodiments only express several implementation ways of the present application, and the description is specific and detailed, but it should not be understood as a limitation to the patent scope of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An image detection method characterized by, The method comprises: obtaining a plurality of segmentation image blocks of an initial medical image; inputting each of the segmentation image blocks into a first model to obtain an output result corresponding to each of the segmentation image blocks, wherein the output result comprises a lesion positioning probability and a pathological classification result; determining a first image parameter adjustment strategy for each of the segmentation image blocks according to the lesion positioning probability corresponding to each of the segmentation image blocks; modifying the image parameters of each of the segmentation image blocks based on the first image parameter adjustment strategy of each of the segmentation image blocks; mapping the modified segmentation image blocks based on the initial medical image, and obtaining a first attention map according to the mapping result; the first attention map is used to represent the lesion positioning probability of each of the segmentation image blocks; the image parameters of the segmentation image blocks are any one of the RGB value of the image, the saturation of the image, and the gray value of the image; determining a second image parameter adjustment strategy for each of the segmentation image blocks according to the pathological classification result corresponding to each of the segmentation image blocks; modifying the image parameters of each of the segmentation image blocks based on the second image parameter adjustment strategy of each of the segmentation image blocks; mapping the modified segmentation image blocks based on the initial medical image, and obtaining a second attention map according to the mapping result; the second attention map is used to represent the pathological classification result of each of the segmentation image blocks; obtaining a detection result of the initial medical image based on the first attention map and the second attention map; the detection result is a classification result of the initial medical image, or a classification result probability of the initial medical image.

2. The method of claim 1, wherein, The inputting each of the segmentation image blocks into the first model to obtain the output result corresponding to each of the segmentation image blocks comprises: After the segmentation image block is input into the first model, the first model obtains the lesion positioning probability by comparing the segmentation image block with an image block having a lesion and / or an image block having no lesion; the first model obtains the pathological classification result by comparing the segmentation image block with an image block having a disease.

3. The method of claim 1, wherein, The obtaining the detection result of the initial medical image based on the first attention map and the second attention map comprises: obtaining the detection result of the initial medical image based on the first attention map, the second attention map, the initial medical image, and a second model.

4. The method of claim 3, wherein, The obtaining the detection result of the initial medical image based on the attention map, the initial medical image, and the second model comprises: inputting the first attention map and the initial medical image into the second model to obtain the detection result of the initial medical image; or, inputting the second attention map and the initial medical image into the second model to obtain the detection result of the initial medical image; or, inputting the first attention map, the second attention map, and the initial medical image into the second model to obtain the detection result of the initial medical image.

5. The method of claim 1, wherein, The method further comprises: obtain segmentation image block samples of a plurality of first medical image samples of a tissue of interest, wherein the segmentation image block samples comprise non-lesion segmentation image block samples and lesion segmentation image block samples; train a preset first prediction model based on each of the segmentation image block samples, to obtain the first model.

6. The method of claim 4, wherein, The method further comprises: obtain a plurality of second medical image samples of the tissue of interest and attention map samples corresponding to the medical image samples, wherein the medical image samples comprise non-lesion image samples and lesion image samples; train a preset second prediction model based on each of the medical image samples and each of the attention map samples, to obtain the second model.

7. The method of claim 6, wherein, The second prediction model is any one of a convolutional neural network model, a recurrent neural network model, a deep belief neural network model, and a generative adversarial network model.

8. The method of claim 1, wherein, The initial medical image is a medical image of a tissue of interest, and the tissue of interest comprises any one of a breast, a liver, a spleen, a stomach, a heart, a kidney, a large intestine, a small intestine, a large arm, a small arm, a large leg, a small leg, a hand, a foot, a reproductive organ, a neck, and a head.

9. An image detection apparatus characterized by comprising: The apparatus comprises: a first module configured to obtain a plurality of segmentation image blocks of an initial medical image; a second module configured to determine a first image parameter adjustment strategy for each of the segmentation image blocks according to a lesion positioning probability corresponding to the segmentation image block, modify image parameters of each of the segmentation image blocks based on the first image parameter adjustment strategy for the segmentation image block, map each of the modified segmentation image blocks based on the initial medical image, and obtain a first attention map according to a mapping result, determine a second image parameter adjustment strategy for each of the segmentation image blocks according to a pathological classification result corresponding to the segmentation image block, modify image parameters of each of the segmentation image blocks based on the second image parameter adjustment strategy for the segmentation image block, map each of the modified segmentation image blocks based on the initial medical image, and obtain a second attention map according to a mapping result, wherein the first attention map is used to represent the lesion positioning probability of each of the segmentation image blocks, the second attention map is used to represent the pathological classification result of each of the segmentation image blocks, and the image parameters of the segmentation image block are any one of an RGB value of an image, a saturation of an image, and a grayscale value of an image; a third module configured to obtain a detection result of the initial medical image based on the attention map, wherein the detection result is a classification result of the initial medical image or a classification result probability of the initial medical image. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The processor, when executing the computer program, implements the steps of the method in any one of claims 1 to 8. The processor, when executing the computer program, implements the steps of the method in any one of claims 1 to 8.

Images