A three-dimensional lesion stripping method for medical images

Abstract

The present invention relates to a method for three-dimensional lesion exfoliation from medical images. The method comprises obtaining ultrasound medical images of the patient's ovaries from all directions, pre-processing them, and then inputting them into a three-dimensional image processing system. The ultrasound medical images of the ovaries from all directions are then 3D reconstructed to obtain a three-dimensional image of the ovaries and adjacent tissues. The three-dimensional image is sliced along a certain direction to obtain multiple first slice images for image feature extraction. Features matching preset lesion feature items are obtained through feature comparison and marked, the lesion area is retained, and areas outside the lesion area are deleted. The first three-dimensional image of the lesion is reassembled and reconstructed. If the first three-dimensional image of the lesion is identified, a CT medical image of the ovary is obtained and the lesion is reconstructed again in three dimensions. A second three-dimensional image of the lesion, along with the lesion category and size, is obtained and output. The three-dimensional reconstruction of the ovarian lesion can accurately assist doctors in making accurate diagnoses of patients' conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of image data processing, and in particular relates to a three-dimensional lesion stripping method for medical images. Background Art

[0002] Because the ovaries are located intraperitoneally, have abundant blood flow inside the ovaries, and are closely connected to multiple organs, ovarian lesions in medical images taken of the ovaries will be affected by other adjacent tissues, making it difficult for doctors to analyze the lesions, resulting in low accuracy in doctors' diagnosis of patients' ovarian lesions.

[0003] Therefore, with the development of image processing technology, before doctors diagnose a patient's condition based on medical images of various parts of the body, they generally first use image recognition systems to perform image recognition on the medical images. In existing technologies, medical images are input into image recognition systems, which have pre-set algorithms and built-in image data corresponding to various lesions. Once the medical image is input into the image recognition system, each area in the input image is compared with the pre-set lesion image data to determine the type and size of the lesion, thereby assisting doctors in diagnosing the patient's condition.

[0004] However, due to the special location of the ovaries, located in the ovarian fossa on the side wall of the pelvis, and being blocked by other adjacent organs and having rich blood flow, the imaging equipment has a certain impact on taking ovarian medical images. Therefore, it is necessary to improve the image processing process in the existing technology to highlight ovarian lesions with the help of existing ovarian medical images and assist doctors in making accurate diagnoses of patients' conditions. Summary of the Invention

[0005] The purpose of the present invention is to provide a three-dimensional lesion stripping method for medical images, which is used to improve the image processing process in the existing technology, so as to highlight ovarian lesions with the help of existing ovarian medical images and assist doctors in making accurate diagnosis of patients' diseases.

[0006] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0007] A method for three-dimensional lesion removal from a medical image comprises the following steps:

[0008] S1: Obtaining ovarian ultrasound medical images of the patient in all directions through ultrasound equipment, performing image preprocessing on the ovarian ultrasound medical images, and inputting the preprocessed ovarian ultrasound medical images into a three-dimensional image processing system;

[0009] S2: The three-dimensional image processing system performs three-dimensional reconstruction based on the ovarian ultrasound medical images in various directions to obtain a three-dimensional image of the ovary and adjacent tissues in the medical image;

[0010] S3: performing image slicing processing on the constructed three-dimensional image of the ovary and adjacent tissues according to a specified thickness and along a certain direction of the three-dimensional image to obtain a plurality of first slice images of the specified thickness;

[0011] S4: performing image feature extraction on each of the plurality of first slice images, comparing the extracted image features with preset lesion image features, obtaining features that match the preset lesion feature items, marking the lesion area in each first slice image based on the features that match the preset lesion feature items, retaining the lesion area, and deleting areas other than the lesion area;

[0012] S5: Recombining the retained areas of each first slice image in the original slice order to construct a first three-dimensional image of the lesion, identifying the first three-dimensional image of the lesion, and determining the lesion category and size. If the lesion belongs to a specified category or its size is greater than a preset threshold, executing step S6; otherwise, outputting the three-dimensional lesion data including the lesion category and size.

[0013] S6: Obtain ovarian CT medical images of the patient in all directions through the CT device, and after pre-processing the ovarian CT medical images, repeat steps S2-S5 to obtain and output the second and third-dimensional images of the lesion, as well as the lesion type and size. The doctor assists in diagnosing the patient's condition based on the output data.

[0014] Preferably, the specific process of image preprocessing in step S1 is as follows:

[0015] S11: After performing image denoising processing on the ovarian ultrasound medical images in various directions, performing image scaling processing on the ovarian ultrasound medical images in various directions, and scaling the ovarian ultrasound medical images in various directions to a matching size;

[0016] S12: performing image segmentation on the ovarian ultrasound medical image in each direction after the scaling process, and obtaining multiple image blocks from each ovarian ultrasound medical image;

[0017] S13: Remap the pixel values of each pixel point within each image block according to the specified rules, calculate the grayscale value of each pixel point in each image block, and count the number of pixels in each image block that exceed a preset threshold, and evenly distribute the part that exceeds the threshold to the specified grayscale value.

[0018] Preferably, the specific process of performing three-dimensional reconstruction based on ovarian ultrasound medical images in various directions in step S2 is as follows:

[0019] S21: Based on step S12, the grayscale value of each pixel in each ovarian ultrasound medical image is obtained, and the grayscale value of each pixel is transformed according to a specified rule to obtain a depth value of each pixel;

[0020] S22: generating a set of data points expressing a 3D shape based on the depth value of each pixel point, wherein information of each data point in the set of data points includes coordinates in a three-dimensional space;

[0021] S23: Simplifying the set of data points to obtain a grid composed of a specified shape, mapping each ovarian ultrasound medical image to the grid surface according to the correspondence between the position of the pixel point and the grid, and realizing a three-dimensional image of the tissue in the ovarian ultrasound medical image.

[0022] Preferably, the specific process of performing image feature extraction on the plurality of first slice images in step S3 is as follows:

[0023] S31: creating a convolutional neural network model for extracting image features, training the convolutional neural network model using labeled ovarian ultrasound medical images, and copying the plurality of first slice images into a plurality of groups, where the number of groups of slice images is the same as the category of the image features to be extracted;

[0024] S32: The convolutional neural network model is set to have multiple input channels, the number of the input channels matches the number of slice groups, and slice images of different groups are input into different input channels of the convolutional neural network model in the order of slices;

[0025] S33: Different input channels correspond to different combinations of convolutional layers and pooling layers. Sliding windows of different sizes are set in each convolutional layer. Each sliding window is set to a matrix of a specified size. The matrix of each convolutional layer slides on each slice image. Different convolutional layers extract different local feature maps of the image through different point multiplication operations. The pooling layer downsamples the output of the convolutional layer. The output of the convolutional layer is calculated as the maximum value through the window set in the pooling layer.

[0026] S34: A fully connected layer is set after the combination layer of the convolution layer and the pooling layer to perform feature fusion on the output of the combination layer of the convolution layer and the pooling layer in different channels, and transmit it to the output layer for output.

[0027] Preferably, in step S34, the specific process of performing feature fusion on the combined layer outputs of the convolutional layer and the pooling layer in different channels by the fully connected layer is as follows:

[0028] S341: The fully connected layer converts the local features extracted by different channels into local feature vectors;

[0029] S342: Calculating the feature vectors corresponding to the local features extracted by different channels according to a preset rule to obtain a final global feature vector;

[0030] S343: Mapping the global feature vector to probabilities corresponding to features of different categories through a preset activation function.

[0031] Preferably, in step S4, the extracted image features are compared with the preset lesion image features to obtain the features that match the preset lesion feature items. The specific process is as follows:

[0032] S41: converting the preset lesion image features into corresponding feature vectors, and matching the global feature vector in step S343 with the feature vector corresponding to the preset lesion image feature conversion;

[0033] S42: Verify the matching result based on the corresponding probabilities of features of different categories, obtain the lesion features in the slice image, and obtain the area corresponding to the lesion features in the slice image.

[0034] Preferably, when the lesion features in the slice image acquired in step S42 show the presence of multiple types of lesions, different colors are applied to the corresponding areas according to the different types of lesions.

[0035] Preferably, in step S5, when the retained area based on each first slice image is reassembled according to the original slice order, different three-dimensional reconstructions are performed on different lesions according to different colors.

[0036] The beneficial effects of the present invention include:

[0037] The present invention provides a method for three-dimensional lesion exfoliation from medical images. The method obtains ultrasound medical images of the patient's ovaries from all directions, pre-processes them, and then inputs them into a three-dimensional image processing system. The ultrasound medical images of the ovaries from all directions are then 3D reconstructed to obtain a three-dimensional image of the ovaries and adjacent tissues. The 3D image is sliced along a certain direction to obtain multiple first slice images for image feature extraction. Feature matching is obtained through feature comparison and marked, retaining the lesion area and deleting areas outside the lesion area. The first three-dimensional image of the lesion is reassembled and reconstructed. If the lesion belongs to a specified category or its size exceeds a preset threshold, an ovarian CT medical image is obtained and the lesion is reconstructed again in three dimensions. A second three-dimensional image of the lesion, along with the lesion category and size, is obtained and output. The three-dimensional reconstruction of ovarian lesions can accurately assist doctors in making accurate diagnoses of patients' conditions.

[0038] First, by constructing a three-dimensional image of the ovary and adjacent tissues based on ovarian ultrasound medical imaging, when a lesion is present in the image, a combined three-dimensional image of the lesion and the ovarian tissue can be obtained. However, since the lesion and the tissue are intertwined, even if a three-dimensional image is constructed, it is impossible to clearly obtain the relevant data of the lesion. Therefore, the present invention slices the constructed combined three-dimensional image of the lesion and the ovarian tissue, and performs feature extraction and other processing based on the sliced image, removing the image area except the lesion area, retaining only the lesion area, and combining the sliced images of each lesion area to obtain three-dimensional data of the lesion, providing a data basis for subsequent lesion identification and size calculation, and improving the efficiency and accuracy of auxiliary diagnosis.

[0039] Secondly, for some of the critical and severe lesions, CT images are obtained through CT imaging equipment, and then three-dimensional reconstruction of the lesions is performed based on the CT images, avoiding the limitations of lesion display in ovarian ultrasound medical images obtained by ultrasound equipment, and can further improve the accuracy of the three-dimensional images of the lesions and further improve the efficiency and accuracy of auxiliary diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the architecture of the neural network model of the present invention.

[0041] Figure 2 The figure is a flow chart of the method for three-dimensional lesion stripping from medical images of the present invention. DETAILED DESCRIPTION

[0042] The following is combined with Figure 1~Figure 2 The present invention is described in further detail:

[0043] Example 1

[0044] See attached Figure 2 As shown, a method for three-dimensional lesion stripping of medical images includes the following steps:

[0045] S1: Ultrasound medical images of the patient's ovaries are acquired from all directions using ultrasound equipment. These images are pre-processed and input into a 3D image processing system. This pre-processing includes image denoising to prevent noise from affecting the subsequent 3D exfoliation and reconstruction of the lesion. Subsequent image scaling ensures that all images are at the same size, providing a standardized data foundation for subsequent 3D reconstruction of the ovaries and adjacent tissues, avoiding mismatching issues during the 3D reconstruction process.

[0046] S2: The 3D image processing system performs 3D reconstruction based on the ovarian ultrasound medical images in various directions, obtaining a 3D image of the ovary and adjacent tissues within the medical image. Because different tissues and lesions within them appear differently in the image, and because the mutual influence of adjacent tissues can result in a one-sided ovarian ultrasound medical image in a particular direction, it is necessary to obtain ovarian ultrasound medical images in various directions to ensure more accurate subsequent 3D imaging results.

[0047] S3: The constructed three-dimensional image of the ovary and adjacent tissues is sliced at a specified thickness and along a certain direction of the three-dimensional image to obtain multiple first slice images of the specified thickness. During the image slicing process, a slicing direction must be pre-set. The slicing direction is set perpendicular to a certain axis of the tissue. That is, a section is set perpendicular to the certain axis of the tissue. The three-dimensional image is sliced according to the specified length through the set section.

[0048] S4: Extracting image features from each of the plurality of first slice images, comparing the extracted image features with preset lesion image features, obtaining features that match the preset lesion feature items, and marking the lesion region in each first slice image based on the features that match the preset lesion feature items, retaining the lesion region, and deleting regions outside the lesion region. Deleting regions outside the lesion region involves performing image excision, i.e., excising the image region except the lesion region along the edge of the lesion region.

[0049] S5: Based on the area retained in each first slice image, the image is reassembled in the order of the original slices to construct a first three-dimensional image of the lesion, the first three-dimensional image of the lesion is identified, and the lesion category and size are determined. When it belongs to the specified category or its size is greater than a preset threshold, step S6 is executed, otherwise the three-dimensional data of the lesion including the lesion category and size is output. Since only the lesion area is retained in each image, the three-dimensional image of the lesion can be directly obtained by reassembling the images based on the retained lesion area. When a separate three-dimensional image of the lesion is obtained, it can greatly facilitate the subsequent identification of the lesion type and the calculation of the lesion size. However, due to the inherent reasons of the ultrasound imaging equipment, the obtained ovarian ultrasound medical images have certain limitations, such as low imaging quality, or the display of the lesion may not be clear and accurate, resulting in certain errors in the final constructed three-dimensional image of the lesion. Therefore, if the first 3D image constructed based on ovarian ultrasound medical imaging indicates the presence of a specified lesion or its size exceeds a preset threshold, such as a critical lesion such as an ovarian tumor or a lesion that is too large, further ovarian CT imaging is required to further analyze the lesion's data through 3D imaging to improve the accuracy of the auxiliary diagnosis of the patient's condition. If the lesion is minor or small, no further ovarian CT imaging is required.

[0050] S6: Obtain ovarian CT medical images of the patient in all directions through the CT device, and after pre-processing the ovarian CT medical images, repeat steps S2-S5 to obtain and output the second and third-dimensional images of the lesion, as well as the lesion type and size. The doctor assists in diagnosing the patient's condition based on the output data.

[0051] Although with the development of image processing technology, the existing technology already exists in which doctors use image recognition systems to perform image recognition on medical images before making a diagnosis based on medical images of various parts of the patient. The medical image is input into the image recognition system, which has a preset corresponding algorithm and built-in corresponding image data of various lesions. After the medical image is input into the image recognition system, each area in the input image is compared with the preset lesion image data to determine the type and size of the lesion, thereby assisting the doctor in diagnosing the patient's condition. However, due to the special location of the ovary, which is located in the ovarian fossa on the side wall of the pelvis and is blocked by other adjacent organs and has rich blood flow, it has a certain impact on the imaging equipment's ability to capture ovarian medical images. Therefore, it is necessary to improve the image processing process in the existing technology to highlight ovarian lesions with the help of existing ovarian medical images and assist doctors in accurately diagnosing the patient's condition.

[0052] Therefore, the present application obtains ovarian ultrasound medical images of the patient in all directions and inputs them into a three-dimensional image processing system after pre-processing; performs three-dimensional reconstruction on the ovarian ultrasound medical images in all directions to obtain three-dimensional images of the ovary and adjacent tissues; performs slice processing along a certain direction of the three-dimensional image to obtain multiple first slice images for image feature extraction, obtains features matching preset lesion feature items through feature comparison and marks them, retains the lesion area, and deletes areas outside the lesion area; recombines and constructs the first three-dimensional image of the lesion, identifies the first three-dimensional image of the lesion, and when it belongs to a specified category or its size is greater than a preset threshold, obtains an ovarian CT medical image and performs three-dimensional reconstruction of the lesion again, obtains a second three-dimensional image of the lesion, as well as the lesion category and size, and outputs them. In this process, the three-dimensional construction of the ovarian lesion can accurately assist doctors in making an accurate diagnosis of the patient's condition. On the one hand, it can improve the efficiency of the assisted doctors in making the diagnosis, and on the other hand, it can effectively improve the accuracy of the assisted doctors in making the diagnosis.

[0053] Example 2

[0054] Based on Example 1, the specific process of image preprocessing in step S1 is as follows:

[0055] S11: After performing image denoising on the ovarian ultrasound medical images in all directions, perform image scaling on the ovarian ultrasound medical images in all directions to a matching size. The image denoising process is as follows: deploying a filter and setting the filter size to 5x5, placing the 5x5 filter on each pixel of the image: covering the image with the current pixel as the center, calculating the average value of the surrounding pixels: averaging all pixel values within the filter coverage area to obtain a new pixel value, assigning the new pixel value to the pixel value at the current position, and replacing the original pixel value with the calculated new pixel value.

[0056] S12: performing image segmentation on the ovarian ultrasound medical image in each direction after the scaling process, and obtaining multiple image blocks from each ovarian ultrasound medical image;

[0057] S13: Remap the pixel values of each pixel point within each image block according to the specified rules, calculate the grayscale value of each pixel point in each image block, and count the number of pixels in each image block that exceed a preset threshold, and evenly distribute the part that exceeds the threshold to the specified grayscale value.

[0058] The specific process of performing three-dimensional reconstruction based on ovarian ultrasound medical images in various directions in step S2 is as follows:

[0059] S21: Based on step S12, the grayscale value of each pixel in each ovarian ultrasound medical image is obtained, and the grayscale value of each pixel is transformed according to a specified rule to obtain a depth value of each pixel;

[0060] S22: generating a set of data points expressing a 3D shape based on the depth value of each pixel point, wherein information of each data point in the set of data points includes coordinates in a three-dimensional space;

[0061] S23: Simplifying the set of data points to obtain a grid composed of a specified shape, mapping each ovarian ultrasound medical image to the grid surface according to the correspondence between the position of the pixel point and the grid, and realizing a three-dimensional image of the tissue in the ovarian ultrasound medical image.

[0062] Example 3

[0063] On the basis of Example 1 or Example 2, the specific process of performing image feature extraction on the plurality of first slice images in step S3 is as follows:

[0064] S31: creating a convolutional neural network model for extracting image features, training the convolutional neural network model using labeled ovarian ultrasound medical images, and copying the plurality of first slice images into a plurality of groups, where the number of groups of slice images is the same as the category of the image features to be extracted;

[0065] S32: The convolutional neural network model is set to have multiple input channels, the number of the input channels matches the number of slice groups, and slice images of different groups are input into different input channels of the convolutional neural network model in the order of slices;

[0066] S33: Different input channels correspond to different combinations of convolutional layers and pooling layers. Sliding windows of different sizes are set in each convolutional layer. Each sliding window is set to a matrix of a specified size. The matrix of each convolutional layer slides on each slice image. Different convolutional layers extract different local feature maps of the image through different point multiplication operations. The pooling layer downsamples the output of the convolutional layer. The output of the convolutional layer is calculated as the maximum value through the window set in the pooling layer.

[0067] S34: A fully connected layer is set after the combination layer of the convolution layer and the pooling layer to perform feature fusion on the output of the combination layer of the convolution layer and the pooling layer in different channels, and transmit it to the output layer for output.

[0068] In this embodiment, the specific process of performing feature fusion on the combined layer outputs of the convolutional layer and the pooling layer in different channels in step S34 is as follows:

[0069] S341: The fully connected layer converts the local features extracted by different channels into local feature vectors;

[0070] S342: Calculating the feature vectors corresponding to the local features extracted by different channels according to a preset rule to obtain a final global feature vector;

[0071] S343: Mapping the global feature vector to probabilities corresponding to features of different categories through a preset activation function.

[0072] See also Figure 1 As shown, the convolutional neural network model for extracting image features of the present invention is provided with an input layer, a combination layer of a convolution layer and a pooling layer, a fully connected layer and an output layer. The input layer is provided with multiple input channels, and the number of the input channels is the same as the category of the image features to be extracted. The image features include edge features, corner features, texture features, etc. The input channels are matched with the combination layer of the convolution layer and the pooling layer, that is, each input channel is matched with a combination layer of a specified group of convolution layers and pooling layers for extracting an image feature. That is to say, the edge feature has an input channel input and is extracted through a specific combination layer of a convolution layer and a pooling layer. The corner features, texture features, etc. of other categories are also input through different input channels and then extracted through a specific combination layer of a convolution layer and a pooling layer. Therefore, the output of the combination layer of the convolution layer and the pooling layer is image features of different categories, which are then integrated and processed by the fully connected layer and output through the output layer.

[0073] In step S4, the extracted image features are compared with the preset lesion image features to obtain the features that match the preset lesion feature items. The specific process is as follows:

[0074] S41: converting the preset lesion image features into corresponding feature vectors, and matching the global feature vector in step S343 with the feature vector corresponding to the preset lesion image feature conversion;

[0075] S42: Verify the matching result based on the corresponding probabilities of features of different categories, obtain the lesion features in the slice image, and obtain the area corresponding to the lesion features in the slice image.

[0076] If the lesion features in the slice image obtained in step S42 indicate the presence of multiple types of lesions, different colors are applied to the corresponding areas according to the different types of lesions. In step S5, when the retained areas based on each first slice image are reassembled according to the original slice order, different lesions are reconstructed three-dimensionally according to the different colors.

[0077] In summary, the present invention provides a method for three-dimensional lesion exfoliation from medical images. The method obtains ultrasound medical images of the patient's ovaries from all directions, pre-processes them, and then inputs them into a three-dimensional image processing system. The ultrasound medical images of the ovaries from all directions are then 3D reconstructed to obtain a three-dimensional image of the ovaries and adjacent tissues. Slicing is performed along a certain direction of the three-dimensional image to obtain multiple first slice images for image feature extraction. Features matching preset lesion feature items are obtained through feature comparison and marked, retaining the lesion area and deleting areas outside the lesion area. The first three-dimensional image of the lesion is reassembled and reconstructed. If the first three-dimensional image of the lesion is identified, a CT medical image of the ovary is obtained and the lesion is reconstructed again in three dimensions. A second three-dimensional image of the lesion, along with the lesion category and size, is obtained and output. The three-dimensional reconstruction of ovarian lesions can accurately assist doctors in making accurate diagnoses of patients' conditions.

[0078] By constructing a three-dimensional image of the ovary and adjacent tissues based on ovarian ultrasound medical images, when there is a lesion in the image, a combined three-dimensional image of the lesion and the ovarian tissue can be obtained. However, since the lesion and the tissue are intertwined, even if the three-dimensional image is constructed, it is impossible to clearly obtain the relevant data of the lesion. The constructed combined three-dimensional image of the lesion and the ovarian tissue is further sliced, and feature extraction and other processing are performed based on the sliced image. The image area other than the lesion area is removed, and only the lesion area is retained. The sliced images of each lesion area are combined to obtain three-dimensional data of the lesion, providing a data basis for subsequent lesion identification and size calculation, thereby improving the efficiency and accuracy of auxiliary diagnosis. Due to the limitations of the lesion display in the ovarian ultrasound medical images obtained by ultrasound equipment for some of the critical and serious lesions, it is necessary to obtain CT images again through CT imaging equipment, and then reconstruct the lesion in three dimensions based on the CT images, further improving the accuracy of the lesion three-dimensional image, and further improving the efficiency and accuracy of auxiliary diagnosis.

Claims

1. A method for three-dimensional lesion removal from medical images, characterized in that: The following steps are involved: S1: Obtaining ovarian ultrasound medical images of the patient in all directions through ultrasound equipment, performing image preprocessing on the ovarian ultrasound medical images, and inputting the preprocessed ovarian ultrasound medical images into a three-dimensional image processing system; S2: The three-dimensional image processing system performs three-dimensional reconstruction based on the ovarian ultrasound medical images in various directions to obtain a three-dimensional image of the ovary and adjacent tissues in the medical image; S3: performing image slicing processing on the constructed three-dimensional image of the ovary and adjacent tissues according to a specified thickness and along a certain direction of the three-dimensional image to obtain a plurality of first slice images of the specified thickness; S4: performing image feature extraction on each of the plurality of first slice images, comparing the extracted image features with preset lesion image features, obtaining features that match the preset lesion feature items, marking the lesion area in each first slice image based on the features that match the preset lesion feature items, retaining the lesion area, and deleting areas other than the lesion area; S5: Recombining the retained areas of each first slice image in the original slice order to construct a first three-dimensional image of the lesion, identifying the first three-dimensional image of the lesion, and determining the lesion category and size. If the lesion belongs to a specified category or its size is greater than a preset threshold, executing step S6; otherwise, outputting the three-dimensional lesion data including the lesion category and size. S6: Obtain ovarian CT medical images of the patient in all directions through the CT device, and after pre-processing the ovarian CT medical images, repeat steps S2-S5 to obtain and output the second and third-dimensional images of the lesion, as well as the lesion type and size. The doctor assists in diagnosing the patient's condition based on the output data.

2. The method for three-dimensional lesion removal from medical images according to claim 1, characterized in that: The specific process of image preprocessing in step S1 is as follows: S11: After performing image denoising processing on the ovarian ultrasound medical images in various directions, performing image scaling processing on the ovarian ultrasound medical images in various directions, and scaling the ovarian ultrasound medical images in various directions to a matching size; S12: performing image segmentation on the ovarian ultrasound medical image in each direction after the scaling process, and obtaining multiple image blocks from each ovarian ultrasound medical image; S13: Remap the pixel values of each pixel point within each image block according to the specified rules, calculate the grayscale value of each pixel point in each image block, and count the number of pixels in each image block that exceed a preset threshold, and evenly distribute the part that exceeds the threshold to the specified grayscale value.

3. The method for three-dimensional lesion removal from medical images according to claim 2, characterized in that: The specific process of performing three-dimensional reconstruction based on ovarian ultrasound medical images in various directions in step S2 is as follows: S21: Based on step S12, the grayscale value of each pixel in each ovarian ultrasound medical image is obtained, and the grayscale value of each pixel is transformed according to a specified rule to obtain a depth value of each pixel; S22: generating a set of data points expressing a 3D shape based on the depth value of each pixel point, wherein information of each data point in the set of data points includes coordinates in a three-dimensional space; S23: Simplifying the set of data points to obtain a grid composed of a specified shape, mapping each ovarian ultrasound medical image to the grid surface according to the correspondence between the position of the pixel point and the grid, and realizing a three-dimensional image of the tissue in the ovarian ultrasound medical image.

4. The method for three-dimensional lesion removal from medical images according to claim 1, characterized in that: The specific process of performing image feature extraction on the plurality of first slice images in step S3 is as follows: S31: creating a convolutional neural network model for extracting image features, training the convolutional neural network model using labeled ovarian ultrasound medical images, and copying the plurality of first slice images into a plurality of groups, where the number of groups of slice images is the same as the category of the image features to be extracted; S32: The convolutional neural network model is set to have multiple input channels, the number of the input channels matches the number of slice groups, and slice images of different groups are input into different input channels of the convolutional neural network model in the order of slices; S33: Different input channels correspond to different combinations of convolutional layers and pooling layers. Sliding windows of different sizes are set in each convolutional layer. Each sliding window is set to a matrix of a specified size. The matrix of each convolutional layer slides on each slice image. Different convolutional layers extract different local feature maps of the image through different point multiplication operations. The pooling layer downsamples the output of the convolutional layer. The output of the convolutional layer is calculated as the maximum value through the window set in the pooling layer. S34: A fully connected layer is set after the combination layer of the convolution layer and the pooling layer to perform feature fusion on the output of the combination layer of the convolution layer and the pooling layer in different channels, and transmit it to the output layer for output.

5. The method for three-dimensional lesion removal from medical images according to claim 4, characterized in that: The specific process of feature fusion of the combined layer outputs of the convolutional layer and the pooling layer in different channels by the fully connected layer in step S34 is as follows: S341: The fully connected layer converts the local features extracted by different channels into local feature vectors; S342: Calculating the feature vectors corresponding to the local features extracted by different channels according to a preset rule to obtain a final global feature vector; S343: Mapping the global feature vector to probabilities corresponding to features of different categories through a preset activation function.

6. The method for three-dimensional lesion removal from medical images according to claim 5, characterized in that: In step S4, the extracted image features are compared with the preset lesion image features to obtain the features that match the preset lesion feature items. The specific process is as follows: S41: converting the preset lesion image features into corresponding feature vectors, and matching the global feature vector in step S343 with the feature vector corresponding to the preset lesion image feature conversion; S42: Verify the matching result based on the corresponding probabilities of features of different categories, obtain the lesion features in the slice image, and obtain the area corresponding to the lesion features in the slice image.

7. The method for three-dimensional lesion removal from medical images according to claim 6, characterized in that: When the lesion features in the slice image acquired in step S42 show that there are multiple types of lesions, different colors are applied to the corresponding areas according to the different types of lesions.

8. The method for three-dimensional lesion removal from medical images according to claim 7, characterized in that: In step S5 , when the retained area based on each first slice image is reassembled according to the original slice sequence, different three-dimensional reconstructions are performed on different lesions according to different colors.

Images