Dual-view digital breast tomosynthesis-based architectural distortion detection method and system

Abstract

The present invention relates to the technical field of image analysis, and provides a dual-view digital breast tomosynthesis-based architectural distortion detection method and system. The method comprises the following steps: collecting three-dimensional breast tomosynthesis images of two projection position views and processing same into two-dimensional tomographic slices; using bounding boxes to label positions of architectural distortions in the two-dimensional tomographic slices; pairing the labeled two-dimensional tomographic slices of the two projection position views of the same breast; using a computer-aided detection model to predict two-dimensional detection boxes corresponding to the positions of the architectural distortions in the paired two-dimensional tomographic slices and to extract from the two-dimensional detection boxes representation vectors of breast tissue structures; using the representation vectors of the breast tissue structures of the paired two-dimensional tomographic slices to match the two-dimensional detection boxes in the tomographic slices of the two projection position views; fusing the two-dimensional detection boxes into three-dimensional detection boxes; and, on the basis of single-view and dual-view information, removing some of the three-dimensional detection boxes to obtain finally kept three-dimensional detection boxes.

Description

Structural distortion detection method and system based on dual-view digital breast tomosynthesis Technical Field

[0001] The present invention relates to the technical field of image analysis, and more particularly to a method and system for detecting structural distortion based on dual-view digital breast tomosynthesis. Background Art

[0002] Traditional two-dimensional full-field digital mammography (FFDM) has a high probability of false detection in architectural distortion (AD) screening.

[0003] Digital Breast Tomosynthesis (DBT), as an emerging mammography technology, can provide richer internal information and effectively reduce the probability of misdiagnosis caused by overlapping breast glands through tomographic imaging.

[0004] Architectural distortion detection models based on single-view digital breast tomosynthesis images perform poorly in detecting atypical architectural distortions because these architectural distortions are often obscured by breast tissue without architectural distortion or are difficult to correctly identify due to overlapping parenchymal tissue. This results in low overall sensitivity and accuracy of architectural distortion detection models based on single-view digital breast tomosynthesis images.

[0005] Summary of the Invention

[0006] The present invention aims to overcome the defects of the above-mentioned prior art in the low overall sensitivity and accuracy of the structural distortion detection model based on single-view digital breast tomosynthesis images, and provides a structural distortion detection method and system based on dual-view digital breast tomosynthesis with higher overall sensitivity and accuracy.

[0007] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0008] Acquire three-dimensional breast tomosynthesis images containing two projection positions and view angles, and process them into two-dimensional image slices;

[0009] Use bounding boxes to mark the locations of structural distortions in two-dimensional image slices;

[0010] Pairing the annotated two-dimensional image slices of the same breast from two different projection positions and perspectives;

[0011] A computer-aided detection model is used to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extract the representation vector of the breast tissue structure within the two-dimensional detection frame;

[0012] Using the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices, the two-dimensional detection frames in the image slices of the two projection positions are matched;

[0013] fusing the paired two-dimensional detection frames into a three-dimensional detection frame;

[0014] Calculating the predicted probability deviation measure of the three-dimensional detection frame, the three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and the breast parenchyma ratio of the three-dimensional detection frame respectively, and eliminating the three-dimensional detection frame according to the calculation results to obtain a retained three-dimensional detection frame;

[0015] For the retained 3D detection frames, the proportion of the 2D detection frames paired with them that have been eliminated is counted, and the 3D detection frames with a proportion greater than a preset threshold are eliminated to obtain the final retained 3D detection frames.

[0016] The present invention also proposes a structural distortion prediction system based on dual-view digital breast tomosynthesis for implementing the above-mentioned structural distortion detection method based on dual-view digital breast tomosynthesis. The system comprises:

[0017] An image processing module is used to acquire three-dimensional breast tomography images including two projection position viewing angles and process them into two-dimensional image slices;

[0018] A labeling module for labeling the location of structural distortion in a two-dimensional image slice using a bounding box;

[0019] A pairing module is used to pair two annotated two-dimensional image slices belonging to the same breast from two projection positions and perspectives;

[0020] A two-dimensional detection frame prediction module is used to use a computer-aided detection model to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extract the representation vector of the breast tissue structure within the two-dimensional detection frame; and use the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices to match the two-dimensional detection frames in the image slices of the two projection positions and perspectives;

[0021] The three-dimensional detection frame prediction module is used to fuse the paired two-dimensional detection frames output by the two-dimensional detection frame prediction module into a three-dimensional detection frame, and respectively calculate the prediction probability deviation measure of the three-dimensional detection frame, the three-dimensional anatomical coordinate space position of the three-dimensional detection frame, and the breast parenchyma ratio of the three-dimensional detection frame. According to the calculation results, the three-dimensional detection frame is eliminated to obtain a retained three-dimensional detection frame; for the retained three-dimensional detection frame, the ratio of the two-dimensional detection frame paired with the eliminated two-dimensional detection frame is counted, and the three-dimensional detection frame with a ratio greater than a preset threshold is eliminated, and the final retained three-dimensional detection frame is output as the structural distortion prediction result.

[0022] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0023] Three-dimensional breast tomography images containing two projection positions and perspectives are collected and processed into two-dimensional image slices. The two labeled two-dimensional image slices belonging to the same breast are paired. The locations of structural distortions in the paired two-dimensional image slices are used to predict paired two-dimensional detection frames. The paired two-dimensional detection frames are used to predict three-dimensional detection frames, and multiple methods are used to remove incorrectly marked three-dimensional detection frames. The dual-view breast tomography images are used to improve the probability of accurate detection of structurally distorted breast tissue occluded by breast tissue without structural distortion. The breast parenchyma ratio is used to decide whether to remove the three-dimensional detection frame to improve the probability of accurate detection of structural distortion in overlapping parenchymal tissue. Therefore, this scheme has high sensitivity and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] FIG1 is a flow chart of a method for detecting structural distortion based on dual-view digital breast tomosynthesis proposed in Example 1;

[0025] FIG2 is a schematic diagram of the structure of the computer-aided detection model proposed in Example 1;

[0026] FIG3 is a schematic diagram of the detection process of the computer-aided detection model proposed in Example 1;

[0027] FIG4 is a diagram showing the effectiveness of the computer-aided detection model proposed in Example 1 under different structural parameters;

[0028] FIG5 is a schematic diagram of the prediction probability deviation measurement threshold proposed in Example 1;

[0029] FIG6 is a schematic diagram of the pectoralis major voxel distance proposed in Example 1;

[0030] FIG7 is a schematic diagram of the breast parenchyma ratio proposed in Example 1;

[0031] FIG8 is a schematic diagram of position quantiles proposed in Example 1;

[0032] FIG9 is a performance comparison example diagram proposed in Example 2;

[0033] FIG10 is a schematic diagram of the ablation experiment results proposed in Example 2;

[0034] FIG11 is an overall framework diagram of the structural distortion prediction system based on dual-view digital breast tomosynthesis proposed in Example 3. DETAILED DESCRIPTION

[0035] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting the present embodiment;

[0036] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0037] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0038] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0039] Example 1

[0040] This embodiment proposes a structural distortion detection method based on dual-view digital breast tomosynthesis. FIG1 is a flow chart of the structural distortion detection method based on dual-view digital breast tomosynthesis according to this embodiment.

[0041] The structural distortion detection method based on dual-view digital breast tomosynthesis proposed in this embodiment includes the following steps:

[0042] S1: Acquire three-dimensional breast tomosynthesis images containing two projection positions and view angles, and process them into two-dimensional image slices;

[0043] S2: Use a bounding box to mark the location of structural distortion in the two-dimensional image slice;

[0044] S3: Pairing the annotated two-dimensional image slices of the same breast from two different projection positions and viewing angles;

[0045] S4: using a computer-aided detection model to predict a two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extracting a representation vector of the breast tissue structure within the two-dimensional detection frame;

[0046] S5: using the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices, matching the two-dimensional detection frames in the image slices at the two projection positions and viewing angles;

[0047] S6: Fusing the paired two-dimensional detection frames into a three-dimensional detection frame;

[0048] S7: Calculate the prediction probability deviation measure of the 3D detection frame, the 3D anatomical coordinate space position of the 3D detection frame, and the breast parenchyma ratio of the 3D detection frame, and remove the 3D detection frame according to the calculation results to obtain a retained 3D detection frame;

[0049] S8: Counting the proportion of the two-dimensional detection frames that are paired with the retained three-dimensional detection frames and being eliminated, and eliminating the three-dimensional detection frames whose proportion is greater than a preset threshold to obtain the final retained three-dimensional detection frames.

[0050] In the specific implementation process, three-dimensional breast tomography images containing two projection position perspectives are collected and processed into two-dimensional image slices. The two-dimensional image slices with marked two-dimensional image slices from the two projection position perspectives belonging to the same breast are paired. The paired two-dimensional detection frames are used to predict the positions of structural distortions in the paired two-dimensional image slices. The three-dimensional detection frames are used to predict the three-dimensional detection frames, and multiple methods are used to remove incorrectly marked three-dimensional detection frames. The dual-view breast tomography images are used to improve the accurate detection probability of structurally distorted breast tissue occluded by breast tissue without structural distortion, and the breast parenchyma ratio is used to decide whether to remove the three-dimensional detection frame to improve the accurate detection probability of structural distortion in overlapping parenchymal tissue. Therefore, this scheme has high sensitivity and accuracy.

[0051] In an optional embodiment, the two projection position viewing angles include: a head-foot viewing angle and a medial-lateral oblique viewing angle;

[0052] The three-dimensional breast tomosynthesis images corresponding to the head-foot view angle include: three-dimensional breast tomosynthesis images corresponding to the left head-foot view angle and the right head-foot view angle;

[0053] The three-dimensional breast tomosynthesis images corresponding to the internal and external oblique lateral views include: three-dimensional breast tomosynthesis images corresponding to the left internal and external oblique lateral view and the right internal and external oblique lateral view;

[0054] When acquiring 3D breast tomosynthesis images, the images were screened based on the inclusion criteria and the gold standard, and only those that met both the inclusion criteria and the gold standard were acquired;

[0055] The inclusion criteria include: 3D breast tomosynthesis images of patients who participated in breast cancer screening for the first time and had not undergone breast surgery or biopsy before participating in the screening;

[0056] The gold standard includes: 3D breast tomosynthesis imaging of patients with architectural distortion in at least one breast tissue confirmed by biopsy, surgery, or follow-up;

[0057] After standardization and desensitization processing of the acquired 3D breast tomosynthesis images, the 3D breast tomosynthesis images are split into 2D tomographic images;

[0058] When using a bounding box to mark the location of structural distortion in a two-dimensional image slice, at each location of structural distortion, the location of the clearest breast slice, the two slices above it, and the two slices below it are marked with a bounding box;

[0059] When pairing two annotated two-dimensional image slices belonging to the same breast from two projection positions and perspectives, the two-dimensional image slice corresponding to the left head-foot position within the bounding box is paired with the two-dimensional image slice corresponding to the left medial-lateral oblique position, and the two-dimensional image slice corresponding to the right head-foot position within the bounding box is paired with the two-dimensional image slice corresponding to the right medial-lateral oblique position.

[0060] In this optional embodiment, cross-analysis of the ipsilateral views of the Left Cranio-Caudal (LCC)-Left Mediolateral Oblique (LMLO) or the Right Cranio-Caudal (RCC)-Right Mediolateral Oblique (RMLO) can reveal structural distortions that may be overlooked in a single view, thereby reducing the uncertainty of structural distortions. Studies have shown that dual-view screening can significantly reduce the number of missed cancer cases compared to single-view screening. As an example, the acquired breast tomosynthesis images are stored in the original DICOM medical image format. The acquisition device used is Hologic's Selenia Dimension mammography system; image data is stored in 16-bit grayscale, with pixel sizes within slices ranging from 0.086 mm to 0.108 mm (mean ± standard deviation: 0.089 ± 0.005), and a physical spacing of 1 mm between adjacent slices. As an example, a radiologist with more than three years of experience, using imaging reports as a reference, annotates the locations of each structural distortion lesion with bounding boxes on the breast slice with the clearest image, as well as on the two slices above and below it, using this as the 3D gold standard for detection. In this optional embodiment, annotation information such as the diseased area, nipple location, areola location, and axillary tail location of the breast is not required. Only the locations of each structural distortion lesion with bounding boxes on the breast slice with the clearest image, as well as on the two slices above and below it, are needed to pair annotated 2D image slices from two projection positions and perspectives belonging to the same breast.

[0061] In an optional embodiment, the computer-aided detection model includes: a feature map pyramid network, a task sharing subnetwork, a structural distortion score subnetwork, a detection box subnetwork, and an image representation subnetwork;

[0062] The task sharing sub-network includes m1 convolutional layers;

[0063] The detection box sub-network and the structural distortion score sub-network include m2 3×3 convolutional layers;

[0064] The image representation subnetwork includes m3 1×1 convolutional layers;

[0065] The steps of using a computer-aided detection model to predict a two-dimensional detection frame corresponding to a structural distortion position in a paired two-dimensional image slice and extracting a representation vector of the breast tissue structure within the two-dimensional detection frame include:

[0066] The paired two-dimensional image slices are simultaneously input into a feature map pyramid network. The feature map pyramid network is based on the feature maps of the paired two-dimensional image slices. The feature maps are input into a task sharing subnetwork. The task sharing subnetwork outputs features extracted from the feature maps, and the features are input into a structural distortion score subnetwork, a detection frame subnetwork, and an image representation subnetwork. The detection frame subnetwork is used to locate and output the location of the structural distortion, the structural distortion score subnetwork is used to calculate and output the structural distortion score corresponding to the location of the structural distortion, and the image representation subnetwork is used to extract and output a representation vector of the breast structural tissue within the location of the structural distortion. The output results of the structural distortion score subnetwork, the detection frame subnetwork, and the image representation subnetwork are used to form a two-dimensional detection candidate frame corresponding to the structural distortion location in the two-dimensional image slice, which includes the structural distortion location and the structural distortion score, and a representation vector of the breast tissue structure within the two-dimensional detection candidate frame.

[0067] In this optional embodiment, Figure 2 is a structural schematic diagram of the computer-aided detection model proposed in this embodiment. As shown in Figure 2, a feature pyramid network (FPN) is used to extract multi-scale features of the image, and a single-view two-dimensional deep learning detection network is formed together with the task sharing subnetwork, structural distortion score subnetwork, detection box subnetwork and image representation subnetwork constructed by the fully convolutional network; as an exemplary illustration, Figure 3 is a detection process schematic diagram of the computer-aided detection model proposed in this embodiment. As shown in Figure 3, a Siamese network is constructed using a single-view two-dimensional detection network to simultaneously extract image features of dual-view digital breast tomosynthesis images and obtain candidate detection boxes for dual views, and the characterization of tissue structure within the detection box is achieved by contrastive learning using triplet loss.

[0068] In an optional embodiment, before using the computer-aided detection model to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extracting the representation vector of the breast tissue structure in the two-dimensional detection frame, the paired two-dimensional image slices are divided into a training set X according to a preset ratio. train and validation set X valuation Using the training set X train The computer-aided detection model is trained for a preset number of rounds. In each round of training, based on the loss function The computer-aided detection model is trained until the loss function When the number of iterations is minimized or reaches a preset threshold, the training is stopped to obtain the trained computer-aided detection model corresponding to each round;

[0069] Using the validation set X valuation , verify the reliability test index of the trained computer-aided detection model corresponding to each round, and select the computer-aided detection model with the highest reliability test index as the final computer-aided detection model;

[0070] When using a computer-aided detection model to predict a two-dimensional detection frame corresponding to a structural distortion position in a paired two-dimensional image slice and extracting a representation vector of the breast tissue structure within the two-dimensional detection frame, the final computer-aided detection model is used for prediction and extraction;

[0071] Among them, the loss function The expression is: Softplus(r)=ln(1+exp(r))

[0072] Where, represents the loss function used to train the structural distortion score subnetwork, C represents the number of predicted categories, k represents the current predicted category, and F(x) (k) represents the predicted probability function for category k, γ represents the adjustment factor of the focal loss, represents the loss function used to train the detection frame positioning, x, y, w, h represent the horizontal coordinate of the detection frame center point, the vertical coordinate of the detection frame center point, the width of the detection frame and the height of the detection frame respectively, t i represents the detection box parameters after scale transformation, v i Represents the scale-transformed annotation box parameters, δ represents the smoothing function parameters, and α represents the loss scale adjustment factor, which is used to balance the scale of the triplet loss and other losses. represents the triplet loss function, N A represents the number of positive samples selected, A represents the positive sample set, j represents a positive sample, m represents the distance parameter between different categories, D jp Indicates the distance between the current sample and the positive sample, D jn Represents the distance between the current sample and the negative sample, and B represents the negative sample set.

[0073] As an example, when the paired two-dimensional image slices are divided according to a preset ratio, they are divided into a training set X train , validation set X valuation and the test set X test , the preset ratio is: 5:2:3; the test set X testThe reliability test index for evaluating the final computer-aided detection model is used to obtain the evaluation index for the validation set, and the evaluation index for the validation set is used as the final evaluation index for the computer-aided detection model; as an exemplary illustration, the data is randomly divided into a training set, a validation set, and a test set at the patient level; the preset rounds are 25 rounds, and the reliability test index is MTPF (Mean Time Between Failure). Each round uses the training set for training in a random order, and at the end of each training round, the MTPF index of the model is verified using the validation set; at the end of the entire training, the model with the highest MTPF index in the validation set is selected as the optimal model, and the test set is used for final verification; as an exemplary illustration, is the Focal Loss function, is the Huber Loss (smooth L1 loss) function, The triplet loss function is used. The computer-aided detection model learns the feature expression of the tissue structure in the detection frame through dual-view image comparison based on the triplet loss function. As an example, Figure 4 shows the performance of the computer-aided detection model proposed in this embodiment under different structural parameters. In Figure 4, R@k represents the maximum sensitivity R that can be achieved while tolerating k false positives in each image. "-" indicates that the sensitivity does not reach 80%, and the bold value is the highest value corresponding to the R@k item. As an example, the structural parameters m1, m2, and m3 of the final computer-aided detection model are 3, 1, and 2 respectively. In this optional embodiment, a twin network architecture is used to extract and compare information from the two views, and a triplet loss module is introduced. The anatomical structure relationship between the ipsilateral views is used as supervision information to guide the fusion of multi-view feature information. This method not only completes the detection and structural feature extraction tasks simultaneously during training, but also effectively integrates multi-view information for detection and removes false positives, thereby improving the overall performance of the model.

[0074] In an optional embodiment, the step of matching the two-dimensional detection frames in the paired two-dimensional image slices using the breast tissue structure representation vector of the two-dimensional detection frames in the paired two-dimensional image slices includes:

[0075] Obtain the two-dimensional detection candidate frames corresponding to the head-foot position and the two-dimensional detection candidate frames corresponding to the medial and lateral positions in the paired two-dimensional image tomography, calculate the correlation between the two-dimensional detection candidate frames using the breast tissue structure representation vectors corresponding to the two-dimensional detection candidate frames, and based on the correlation, use the Hungarian matching algorithm to determine the one-to-one matching relationship between the two-dimensional detection candidate frames to obtain the paired two-dimensional detection frames, and use the formula P 1_refineAdjust the structural distortion score within the paired 2D detection candidate box;

[0076] Wherein, the formula P 1_refine The expression is: P 1_refine =ω*P1+(1-ω)P2

[0077] Where ω represents the weight parameter, P1 represents the structural distortion score of the current two-dimensional candidate box, and P2 represents the structural distortion score of the two-dimensional candidate box paired with P1.

[0078] In an optional embodiment, the step of fusing the paired two-dimensional detection frames into a three-dimensional detection frame includes:

[0079] Based on the paired two-dimensional detection frames, the centroids of the two-dimensional detection frames on each slice of the paired two-dimensional image slices are calculated, and all the calculated centroids are projected onto the same plane. All the centroids projected onto the same plane are clustered using a density-based spatial clustering method to obtain a plurality of cluster centers.

[0080] All 2D detection frames belonging to the same cluster center are stacked in the upper and lower layers of the breast slice to form a series of 3D connected regions.

[0081] Use 3D morphological erosion and dilation operations to connect adjacent connected branches in the same cluster and remove connected branches that only appear in one or two faults;

[0082] Calculate the bounding box of each remaining three-dimensional connected branch, take the bounding box as the ROI in the three-dimensional space, take the maximum two-dimensional prediction probability in the ROI as the prediction probability of the entire ROI, and the three-dimensional detection box is the ROI.

[0083] In this optional embodiment, the three-dimensional structural distortion candidate region (ROI) is determined by the steps of two-dimensional centroid clustering, constructing three-dimensional connected regions, filling gaps and eliminating isolated regions, wherein in the two-dimensional centroid clustering step, the density-based spatial clustering method (Density Based Spatial Clustering of Applications with Noise, DBSCAN) is used for clustering.

[0084] In an optional embodiment, the steps of respectively calculating a prediction probability deviation measure of the three-dimensional detection frame, a three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and a breast parenchyma ratio of the three-dimensional detection frame, and removing the three-dimensional detection frame based on the calculation results to obtain a retained three-dimensional detection frame include:

[0085] Calculating the prediction probability deviation metric of the three-dimensional detection frame and removing the three-dimensional detection frame whose prediction probability deviation metric is greater than a preset threshold, the steps include:

[0086] For any 3D detection frame, assuming that it contains K faults, the prediction probability of the computer-aided detection model on each fault is arranged from top to bottom according to the spatial position, forming a prediction probability sequence, which is recorded as Among them, P i represents the predicted probability value of the i-th fault;

[0087] Using a sliding window of width W, calculate the sequence The sliding mean and sliding standard deviation of , where W is the preset value;

[0088] The sliding standard deviation corresponding to the maximum sliding mean is selected as the prediction probability deviation metric of the 3D detection frame;

[0089] Eliminate the 3D detection frames whose predicted probability deviation measure is greater than the preset threshold;

[0090] Obtaining pectoralis major muscle position information and nipple position information, and constructing a three-dimensional anatomical coordinate space based on the pectoralis major muscle position information and nipple position information using the pectoralis major muscle axis, the nipple axis, and the upper and lower fault direction axes;

[0091] The pectoralis major axis includes: a boundary line between the pectoralis major region and the breast region;

[0092] The nipple axis includes: a vertical line starting from the nipple position and perpendicular to the pectoralis major axis;

[0093] The upper and lower fault direction axes include: the direction axes of the upper and lower layers between the faults;

[0094] The pectoralis major axis, the nipple axis and the upper and lower fault direction axes are perpendicular to each other;

[0095] Calculating the 3D anatomical coordinate space position of the 3D detection frame and removing the 3D detection frame whose 3D anatomical coordinate space position is outside a preset acceptance range, the steps include:

[0096] Based on the three-dimensional anatomical coordinate space, the statistical training set X train The voxel distance frequency between the geometric center of the structural distortion position of the paired two-dimensional image tomography and the pectoralis major muscle line is selected from the voxel distance frequency, and the lower limit value and the upper limit value of the voxel distance acceptance range are selected. The expression of the voxel distance acceptance range is: [v low ,v high ], where v low Indicates the lower limit of the voxel distance acceptance range, v high Indicates the upper limit of the voxel distance acceptance range;

[0097] Calculating the 3D anatomical coordinate space position of the 3D detection frame, and removing the 3D detection frame whose 3D anatomical coordinate space position is outside a preset acceptance range;

[0098] The three-dimensional anatomical coordinate space position of the three-dimensional detection frame includes: the voxel distance frequency from the geometric center of the three-dimensional detection frame to the pectoralis major line; the preset acceptance range includes: voxel distance acceptance range [v low ,v high ].

[0099] As an exemplary illustration, the width W of the sliding window is an odd value; as an exemplary illustration, W=5; as an exemplary illustration, FIG5 is a schematic diagram of the predicted probability deviation metric threshold proposed in this embodiment; FIG5 shows the relationship between the threshold of the MTPF calculated using the training set and the predicted probability deviation metric. As an exemplary illustration, according to FIG5, the present application determines that the three-dimensional candidate box with a predicted probability deviation metric greater than 0.19 is a misjudgment result and removes it; FIG6 is a schematic diagram of the pectoralis major voxel distance proposed in this embodiment, and FIG6 shows the cumulative distribution histogram of the distance from the pectoralis major voxel at each threshold calculated using the training set. As an exemplary illustration, according to FIG6, the present application selects v low =10,v high =1000.

[0100] In an optional embodiment, the step of respectively calculating a prediction probability deviation measure of the three-dimensional detection frame, a three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and a breast parenchyma ratio of the three-dimensional detection frame, and removing the three-dimensional detection frame based on the calculation results to obtain a retained three-dimensional detection frame further includes:

[0101] Calculating the breast parenchyma ratio of the 3D detection frame and removing the 3D detection frame whose breast parenchyma ratio is less than a preset threshold, the steps include:

[0102] The 2D fault where the 3D geometric center of the 3D prediction frame is located is selected as the central fault. The OTSU adaptive threshold segmentation algorithm is used on the central fault to distinguish the solid tissue and fat tissue in the breast area, and the breast solid proportion S of the 3D detection frame is calculated. p , the S p The expression is:

[0103] Where, P c represents the number of solid tissue pixels in the breast area of ​​the central section of the 3D prediction box, C bs Indicates the size of the 3D prediction box of the central fault;

[0104] The proportion of breast parenchyma S pThe 3D detection frames smaller than the preset threshold are eliminated, and the position quantile acceptance range [t low ,t high ], where t low Indicates the lower limit of the acceptance range of the position quantile, t high It represents the upper limit of the position quantile acceptance range. The three-dimensional prediction frame within the central fault range that exceeds the position quantile acceptance range is eliminated. The expression of the position quantile is:

[0105] Where, t pua represents the position quantile, p num Indicates the ordinal number of the layer where the 3D prediction box is located, p all Indicates the total number of fault layers.

[0106] As an exemplary illustration, FIG7 is a schematic diagram of the breast parenchyma ratio proposed in this embodiment, and FIG7 shows the frequency distribution of the breast parenchyma ratio in each interval calculated using the training set; FIG8 is a schematic diagram of the position quantile proposed in this embodiment, and FIG8 shows the frequency distribution of the position quantile metric in each interval calculated using the training set; As an exemplary illustration, according to FIG8, the t selected in this application low =0.15, t high =0.9.

[0107] In an optional embodiment, the steps of obtaining the structural distortion detection result include counting the proportion of the 2D detection frames paired with the 2D detection frames of the remaining 3D detection frames and removing the 3D detection frames whose proportion is greater than a preset threshold.

[0108] Perform pairwise inner product on the representation vectors of the breast tissue structure of the two-dimensional detection frame of the retained three-dimensional detection frame and its paired two-dimensional detection frame to obtain the correlation between the paired two-dimensional detection frames. Based on the correlation, the Hungarian matching algorithm is used to determine the one-to-one pairing relationship between the paired two-dimensional detection frames. The mapping expression of the pairing relationship is: P(a) = b

[0109] Where a represents the two-dimensional detection frame corresponding to one projection position angle in the paired two-dimensional detection frame, and b represents the two-dimensional detection frame corresponding to another projection position angle in the paired two-dimensional detection frame;

[0110] Combine the two-dimensional detection frames corresponding to all the remaining three-dimensional detection frames into a set A, and calculate the proportion P of the two-dimensional detection frames paired with the two-dimensional detection frames of the remaining three-dimensional detection frames that are eliminated based on the set A. m , the proportion is measured by pairwise false positive P m to measure;

[0111] The pairwise false positive metric P mThe 3D detection frames larger than the preset threshold are eliminated to obtain the final remaining 3D detection frames;

[0112] Among them, the P m The calculation expression is: B={x1,x2,x3,...x i ,…}

[0113] Where B represents any remaining 3D detection box, x i Represents the i-th two-dimensional prediction box that makes up the three-dimensional detection box B.

[0114] As an example, let the pairwise false positive metric P m 3D detection boxes with values ​​greater than 1 are removed.

[0115] Example 2

[0116] This embodiment is based on the structural distortion detection method based on dual-view digital breast tomosynthesis proposed in Example 1 and is compared with other common models using the following indicators:

[0117] (1) Mean True Positive Fraction (MTPF): The average correct detection rate within the range of 0.05 to 2.0 false positives (FPs) per volume. This metric measures the model's ability to correctly identify true architectural distortions (AD) at different false positive thresholds. (2) Sensitivity at x False Positives (FPs) per volume (R@x): The highest sensitivity that can be achieved when the average number of false positives in each group of breast tomosynthesis is no higher than x. This metric reflects the sensitivity of the model when the number of false positives is limited. (3) Number of FPs at 80% Sensitivity (FPs@0.8): The number of false positive results generated by the model while maintaining 80% sensitivity (i.e., correctly identifying 80% of true positives). This metric is used to evaluate the number of false positive results generated by the model while maintaining high sensitivity.

[0118] Figure 9 is an example diagram of the performance comparison proposed in this embodiment. Figure 9 shows the performance comparison results of this method and other models. R@k in Figure 9 means that the highest sensitivity R can be achieved while tolerating k false alarm results for each image, "-" means that the sensitivity does not reach 80%, and the bold value is the highest value corresponding to the R@k item; this application removes false alarm results of structural distortion positions based on dual perspectives, three-dimensional continuity of breast tissue and prior knowledge of breast anatomy. In order to compare the impact of these three methods of removing false alarm results on this method, this method conducts ablation experiments on these three methods of removing false alarm results. Figure 10 is a schematic diagram of the ablation experiment results proposed in this embodiment. R@k in Figure 10 means that the highest sensitivity R can be achieved while tolerating k false alarm results for each image, "-" means that the sensitivity does not reach 80%, and the bold value is the highest value corresponding to the R@k item.

[0119] Example 3

[0120] This embodiment proposes a structural distortion prediction system based on dual-view digital breast tomosynthesis, which is used to implement the structural distortion detection method based on dual-view digital breast tomosynthesis proposed in Example 1. Figure 11 is an overall framework diagram of the structural distortion prediction system based on dual-view digital breast tomosynthesis in this embodiment.

[0121] The structural distortion prediction system based on dual-view digital breast tomosynthesis includes:

[0122] An image processing module is used to acquire three-dimensional breast tomography images including two projection position viewing angles and process them into two-dimensional image slices;

[0123] A labeling module for labeling the location of structural distortion in a two-dimensional image slice using a bounding box;

[0124] A pairing module is used to pair two annotated two-dimensional image slices belonging to the same breast from two projection positions and perspectives;

[0125] A two-dimensional detection frame prediction module is used to use a computer-aided detection model to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extract the representation vector of the breast tissue structure within the two-dimensional detection frame; and use the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices to match the two-dimensional detection frames in the image slices of the two projection positions and perspectives;

[0126] The three-dimensional detection frame prediction module is used to fuse the paired two-dimensional detection frames output by the two-dimensional detection frame prediction module into a three-dimensional detection frame, and respectively calculate the prediction probability deviation measure of the three-dimensional detection frame, the three-dimensional anatomical coordinate space position of the three-dimensional detection frame, and the breast parenchyma ratio of the three-dimensional detection frame. According to the calculation results, the three-dimensional detection frame is eliminated to obtain the retained three-dimensional detection frame. The proportion of the two-dimensional detection frame paired with the retained three-dimensional detection frame that has been eliminated is counted for the retained three-dimensional detection frame, and the three-dimensional detection frame whose proportion is greater than a preset threshold is eliminated, and the final retained three-dimensional detection frame is output as the structural distortion position prediction result.

[0127] In this embodiment, a three-dimensional breast tomosynthesis image containing two projection positions and perspectives is input into a structural distortion prediction system based on dual-view digital breast tomosynthesis. The system uses bounding boxes to mark the locations of structural distortions in the two-dimensional image slices. Based on this marked information, the final retained three-dimensional detection box (the specific location of the structural distortion in the two-dimensional image slices) is predicted to obtain a structural distortion location prediction result.

[0128] It can be understood that the structural distortion prediction system based on dual-view digital breast tomosynthesis in this embodiment improves the method of Example 1. The optional options in the above-mentioned Example 1 are also applicable to this embodiment, so they will not be repeated here. The same or similar reference numerals correspond to the same or similar components; the terms describing the positional relationships in the accompanying drawings are only for illustrative purposes and should not be understood as limiting this embodiment; it is obvious that the above-mentioned embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for detecting structural distortion based on dual-view digital breast tomosynthesis, characterized in that: The following steps are involved: Acquire three-dimensional breast tomosynthesis images containing two projection positions and view angles, and process them into two-dimensional image slices; Use bounding boxes to mark the locations of structural distortions in two-dimensional image slices; Pairing the annotated two-dimensional image slices of the same breast from two different projection positions and perspectives; A computer-aided detection model is used to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extract the representation vector of the breast tissue structure within the two-dimensional detection frame; Using the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices, the two-dimensional detection frames in the image slices of the two projection positions are matched; fusing the paired two-dimensional detection frames into a three-dimensional detection frame; Calculating the predicted probability deviation measure of the three-dimensional detection frame, the three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and the breast parenchyma ratio of the three-dimensional detection frame respectively, and eliminating the three-dimensional detection frame according to the calculation results to obtain a retained three-dimensional detection frame; For the retained 3D detection frames, the proportion of the 2D detection frames paired with them that have been eliminated is counted, and the 3D detection frames with a proportion greater than a preset threshold are eliminated to obtain the final retained 3D detection frames.

2. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 1, wherein: The two projection position angles include: head-foot angle and medial-lateral angle; The three-dimensional breast tomosynthesis images corresponding to the head-foot view angle include: three-dimensional breast tomosynthesis images corresponding to the left head-foot view angle and the right head-foot view angle; The three-dimensional breast tomosynthesis images corresponding to the internal and external oblique lateral views include: three-dimensional breast tomosynthesis images corresponding to the left internal and external oblique lateral view and the right internal and external oblique lateral view; When acquiring 3D breast tomosynthesis images, the images were screened based on the inclusion criteria and the gold standard, and only those that met both the inclusion criteria and the gold standard were acquired; The inclusion criteria include: 3D breast tomosynthesis images of patients who participated in breast cancer screening for the first time and had not undergone breast surgery or biopsy before participating in the screening; The gold standard includes: 3D breast tomosynthesis imaging of patients with architectural distortion in at least one breast tissue confirmed by biopsy, surgery, or follow-up; After standardization and desensitization processing of the acquired 3D breast tomosynthesis images, the 3D breast tomosynthesis images are split into 2D tomographic images; When using a bounding box to mark the location of structural distortion in a two-dimensional image slice, at each location of structural distortion, the location of the clearest breast slice, the two slices above it, and the two slices below it are marked with a bounding box; When pairing two annotated two-dimensional image slices belonging to the same breast from two projection positions and perspectives, the two-dimensional image slice corresponding to the left head-foot position within the bounding box is paired with the two-dimensional image slice corresponding to the left medial-lateral oblique position, and the two-dimensional image slice corresponding to the right head-foot position within the bounding box is paired with the two-dimensional image slice corresponding to the right medial-lateral oblique position.

3. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 2, wherein: The computer-aided detection model includes: a feature map pyramid network, a task sharing subnetwork, a structural distortion score subnetwork, a detection box subnetwork and an image representation subnetwork; The task sharing sub-network includes m1 convolutional layers; The detection box sub-network and the structural distortion score sub-network include m w 3×3 convolutional layers; The image representation subnetwork includes m3 1×1 convolutional layers; The steps of using a computer-aided detection model to predict a two-dimensional detection frame corresponding to a structural distortion position in a paired two-dimensional image slice and extracting a representation vector of the breast tissue structure within the two-dimensional detection frame include: The paired two-dimensional image slices are simultaneously input into a feature map pyramid network. The feature map pyramid network is based on the feature maps of the paired two-dimensional image slices. The feature maps are input into a task sharing subnetwork. The task sharing subnetwork outputs features extracted from the feature maps, and the features are input into a structural distortion score subnetwork, a detection frame subnetwork, and an image representation subnetwork. The detection frame subnetwork is used to locate and output the location of the structural distortion, the structural distortion score subnetwork is used to calculate and output the structural distortion score corresponding to the location of the structural distortion, and the image representation subnetwork is used to extract and output a representation vector of the breast structural tissue within the location of the structural distortion. The output results of the structural distortion score subnetwork, the detection frame subnetwork, and the image representation subnetwork are used to form a two-dimensional detection candidate frame corresponding to the structural distortion location in the two-dimensional image slice, which includes the structural distortion location and the structural distortion score, and a representation vector of the breast tissue structure within the two-dimensional detection candidate frame.

4. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 3, wherein: Before using the computer-aided detection model to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image tomography and extracting the representation vector of the breast tissue structure in the two-dimensional detection frame, the paired two-dimensional image tomography is divided into a training set X according to a preset ratio. train and validation set X valuation Using the training set X train The computer-aided detection model is trained for a preset number of rounds. In each round of training, based on the loss function The computer-aided detection model is trained until the loss function When the number of iterations is minimized or reaches a preset threshold, the training is stopped to obtain the trained computer-aided detection model corresponding to each round; Using the validation set X valuation , verify the reliability test index of the trained computer-aided detection model corresponding to each round, and select the computer-aided detection model with the highest reliability test index as the final computer-aided detection model; When using a computer-aided detection model to predict a two-dimensional detection frame corresponding to a structural distortion position in a paired two-dimensional image slice and extracting a representation vector of the breast tissue structure within the two-dimensional detection frame, the final computer-aided detection model is used for prediction and extraction; Among them, the loss function The expression is: Softplus(r)=ln(1+exp(r)) Where, represents the loss function used to train the structural distortion score subnetwork, C represents the number of predicted categories, k represents the current predicted category, and F(x) (k) represents the predicted probability function for category k, γ represents the adjustment factor of the focal loss, represents the loss function used to train the detection frame positioning, x, y, w, h represent the horizontal coordinate of the detection frame center point, the vertical coordinate of the detection frame center point, the width of the detection frame and the height of the detection frame respectively, t i represents the detection box parameters after scale transformation, v i Represents the scale-transformed annotation box parameters, δ represents the smoothing function parameters, and α represents the loss scale adjustment factor, which is used to balance the scale of the triplet loss and other losses. represents the triplet loss function, N A represents the number of positive samples selected, A represents the positive sample set, j represents a positive sample, m represents the distance parameter between different categories, D jp Indicates the distance between the current sample and the positive sample, D jn Represents the distance between the current sample and the negative sample, and B represents the negative sample set.

5. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 4, characterized in that: The steps of matching the two-dimensional detection frames in the paired two-dimensional image slices using the breast tissue structure representation vector of the two-dimensional detection frames in the paired two-dimensional image slices include: Obtain the two-dimensional detection candidate frames corresponding to the head-foot position and the two-dimensional detection candidate frames corresponding to the medial and lateral positions in the paired two-dimensional image tomography, calculate the correlation between the two-dimensional detection candidate frames using the breast tissue structure representation vectors corresponding to the two-dimensional detection candidate frames, and based on the correlation, use the Hungarian matching algorithm to determine the one-to-one matching relationship between the two-dimensional detection candidate frames to obtain the paired two-dimensional detection frames, and use the formula P 1_refine Adjust the structural distortion score within the paired 2D detection candidate box; Wherein, the formula P 1_refine The expression is: P 1_refine =ω*P1+(1-ω)P2 Where ω represents the weight parameter, P1 represents the structural distortion score of the current two-dimensional candidate box, and P2 represents the structural distortion score of the two-dimensional candidate box paired with P1.

6. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to any one of claims 1 to 5, characterized in that: The step of fusing the paired two-dimensional detection frames into a three-dimensional detection frame includes: Based on the paired two-dimensional detection frames, the centroids of the two-dimensional detection frames on each slice of the paired two-dimensional image slices are calculated, and all the calculated centroids are projected onto the same plane. All the centroids projected onto the same plane are clustered using a density-based spatial clustering method to obtain a plurality of cluster centers. All 2D detection frames belonging to the same cluster center are stacked in the upper and lower layers of the breast slice to form a series of 3D connected regions. Use 3D morphological erosion and dilation operations to connect adjacent connected branches in the same cluster and remove connected branches that only appear in one or two faults; Calculate the bounding box of each remaining three-dimensional connected branch, take the bounding box as the ROI in the three-dimensional space, take the maximum two-dimensional prediction probability in the ROI as the prediction probability of the entire ROI, and the three-dimensional detection box is the ROI.

7. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 6, wherein: The steps of respectively calculating a prediction probability deviation measure of the three-dimensional detection frame, a three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and a breast parenchyma ratio of the three-dimensional detection frame, and eliminating the three-dimensional detection frame according to the calculation results to obtain a retained three-dimensional detection frame include: Calculating the prediction probability deviation metric of the three-dimensional detection frame and removing the three-dimensional detection frame whose prediction probability deviation metric is greater than a preset threshold, the steps include: For any 3D detection frame, assuming that it contains K faults, the prediction probability of the computer-aided detection model on each fault is arranged from top to bottom according to the spatial position, forming a prediction probability sequence, which is recorded as Among them, P i represents the predicted probability value of the i-th fault; Using a sliding window of width W, calculate the sequence The sliding mean and sliding standard deviation of , where W is the preset value; The sliding standard deviation corresponding to the maximum sliding mean is selected as the prediction probability deviation metric of the 3D detection frame; Eliminate the 3D detection frames whose predicted probability deviation measure is greater than the preset threshold; Obtaining pectoralis major muscle position information and nipple position information, and constructing a three-dimensional anatomical coordinate space based on the pectoralis major muscle position information and nipple position information using the pectoralis major muscle axis, the nipple axis, and the upper and lower fault direction axes; The pectoralis major axis includes: a boundary line between the pectoralis major region and the breast region; The nipple axis includes: a vertical line starting from the nipple position and perpendicular to the pectoralis major axis; The upper and lower fault direction axes include: the direction axes of the upper and lower layers between the faults; The pectoralis major axis, the nipple axis and the upper and lower fault direction axes are perpendicular to each other; Calculating the 3D anatomical coordinate space position of the 3D detection frame and removing the 3D detection frame whose 3D anatomical coordinate space position is outside a preset acceptance range, the steps include: Based on the three-dimensional anatomical coordinate space, the statistical training set X train The voxel distance frequency between the geometric center of the structural distortion position of the paired two-dimensional image tomography and the pectoralis major muscle line is selected from the voxel distance frequency, and the lower limit value and the upper limit value of the voxel distance acceptance range are selected. The expression of the voxel distance acceptance range is: [v low ,v high ], where v low Indicates the lower limit of the voxel distance acceptance range, v high Indicates the upper limit of the voxel distance acceptance range; Calculating the 3D anatomical coordinate space position of the 3D detection frame, and removing the 3D detection frame whose 3D anatomical coordinate space position is outside a preset acceptance range; The three-dimensional anatomical coordinate space position of the three-dimensional detection frame includes: the voxel distance frequency from the geometric center of the three-dimensional detection frame to the pectoralis major line; the preset acceptance range includes: voxel distance acceptance range [v low ,v high ].

8. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 7, wherein: The steps of respectively calculating a prediction probability deviation measure of the three-dimensional detection frame, a three-dimensional anatomical coordinate spatial position of the three-dimensional detection frame, and a breast parenchyma ratio of the three-dimensional detection frame, and removing the three-dimensional detection frame according to the calculation results to obtain a retained three-dimensional detection frame further include: Calculating the breast parenchyma ratio of the 3D detection frame and removing the 3D detection frame whose breast parenchyma ratio is less than a preset threshold, the steps include: The 2D fault where the 3D geometric center of the 3D prediction frame is located is selected as the central fault. The OTSU adaptive threshold segmentation algorithm is used on the central fault to distinguish the solid tissue and fat tissue in the breast area, and the breast solid proportion S of the 3D detection frame is calculated. p , the S p The expression is: Where, P c represents the number of solid tissue pixels in the breast area of ​​the central section of the 3D prediction box, C bs Indicates the size of the 3D prediction box of the central fault; The proportion of breast parenchyma S p The 3D detection frames smaller than the preset threshold are eliminated, and the position quantile acceptance range [t low ,t high ], where t low Indicates the lower limit of the acceptance range of the position quantile, t high It represents the upper limit of the position quantile acceptance range. The three-dimensional prediction frame within the central fault range that exceeds the position quantile acceptance range is eliminated. The expression of the position quantile is: Where, t pua represents the position quantile, p num Indicates the ordinal number of the layer where the 3D prediction box is located, p all Indicates the total number of fault layers.

9. The method for detecting structural distortion based on dual-view digital breast tomosynthesis according to claim 8, wherein: The steps of calculating the ratio of the 2D detection frames that are paired with the remaining 3D detection frames and discarding them, and discarding the 3D detection frames whose ratio is greater than a preset threshold, to obtain the structural distortion detection results include: The inner product of the breast tissue structure representation vectors of the 2D detection frame of the retained 3D detection frame and its paired 2D detection frame is performed to obtain the correlation between the paired 2D detection frames. Based on the correlation, the Hungarian matching algorithm is used to determine the one-to-one pairing relationship between the paired 2D detection frames. The mapping expression of the pairing relationship is: P(a)=b Where a represents the two-dimensional detection frame corresponding to one projection position angle in the paired two-dimensional detection frame, and b represents the two-dimensional detection frame corresponding to another projection position angle in the paired two-dimensional detection frame; Combine the two-dimensional detection frames corresponding to all the remaining three-dimensional detection frames into a set A, and calculate the proportion P of the two-dimensional detection frames paired with the two-dimensional detection frames of the remaining three-dimensional detection frames that are eliminated based on the set A. m , the proportion is measured by pairwise false positive P m to measure; The pairwise false positive metric P m The 3D detection frames larger than the preset threshold are eliminated to obtain the final remaining 3D detection frames; Among them, the P m The calculation expression is: B={x1,x2,x3,...x i ,…} Where B represents any remaining 3D detection box, x i Represents the i-th two-dimensional prediction box that makes up the three-dimensional detection box B.

10. A structural distortion prediction system based on dual-view digital breast tomosynthesis, used to implement the structural distortion detection method based on dual-view digital breast tomosynthesis according to any one of claims 1 to 9, characterized in that: include: An image processing module is used to acquire three-dimensional breast tomography images including two projection position viewing angles and process them into two-dimensional image slices; A labeling module for labeling the location of structural distortion in a two-dimensional image slice using a bounding box; A pairing module is used to pair two annotated two-dimensional image slices belonging to the same breast from two projection positions and perspectives; A two-dimensional detection frame prediction module is used to use a computer-aided detection model to predict the two-dimensional detection frame corresponding to the structural distortion position in the paired two-dimensional image slices and extract the representation vector of the breast tissue structure within the two-dimensional detection frame; and use the breast tissue structure representation vector of the two-dimensional detection frame in the paired two-dimensional image slices to match the two-dimensional detection frames in the image slices of the two projection positions and perspectives; The three-dimensional detection frame prediction module is used to fuse the paired two-dimensional detection frames output by the two-dimensional detection frame prediction module into a three-dimensional detection frame, and respectively calculate the prediction probability deviation measurement of the three-dimensional detection frame, the three-dimensional anatomical coordinate space position of the three-dimensional detection frame, and the breast substance proportion of the three-dimensional detection frame. According to the calculation results, the three-dimensional detection frame is eliminated to obtain a retained three-dimensional detection frame; for the retained three-dimensional detection frame, the proportion of the two-dimensional detection frame paired with the two-dimensional detection frame that has been eliminated is counted, and the three-dimensional detection frame with a proportion greater than a preset threshold is eliminated, and the final retained three-dimensional detection frame is output as the structural distortion position prediction result.

Images