Image processing model training method and image processing method

Abstract

The embodiment of the invention provides an image processing model training method and an image processing method, and the image processing model training method comprises the steps: determining target training data and enhanced training data; performing object detection on the sample image and the flipped sample image by using the initial image processing model to obtain a detection result of the sample image and a flipped detection result of the flipped sample image; determining predicted object region information of the target object in the sample image based on the detection result and the overturning detection result; and carrying out model training on the initial image processing model based on the object region information, the predicted object region information, the detection result, the sample label, the overturning detection result and the overturning sample label to obtain a trained target image processing model, thereby obtaining the target image processing model capable of more accurately carrying out object detection.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of artificial intelligence technology, and in particular, to an image processing model training method; one or more embodiments of this specification also relate to another image processing model training method, an image processing method, another image processing method, an information processing method, a computing device, a computer-readable storage medium, and a computer program product. Background Art

[0002] With the continuous development of computer technology and artificial intelligence technology, data processing operations can be performed through neural network models, thereby improving the efficiency of data processing; for example, object detection operations can be performed on images through neural network models.

[0003] The current neural network model has the problem of inaccurate detection results when performing object detection on images. Therefore, how to obtain accurate detection results in the process of performing object detection through the neural network model has become an urgent problem that needs to be solved. Summary of the Invention

[0004] In view of this, embodiments of this specification provide an image processing model training method. One or more embodiments of this specification also relate to another image processing model training method, an image processing method, another image processing method, an information processing method, a computer-assisted cancer diagnosis method, a computer-assisted nasopharyngeal carcinoma diagnosis method, a computer-assisted breast cancer diagnosis method, an image processing model training device, another image processing model training device, an image processing device, another image processing device, an information processing device, a computing device, a computer-readable storage medium, and a computer program product to address technical deficiencies in the prior art.

[0005] According to a first aspect of an embodiment of this specification, there is provided an image processing model training method, comprising:

[0006] Determining target training data and enhanced training data, wherein the target training data includes a sample image of a target object located at a target location, a sample label of the sample image, and object region information of the target object, and the enhanced training data includes a flipped sample image corresponding to the sample image and a flipped sample label of the flipped sample image;

[0007] Using the initial image processing model, perform object detection on the sample image and the flipped sample image respectively to obtain a detection result of the sample image and a flip detection result of the flipped sample image;

[0008] Determining predicted object region information of the target object in the sample image based on the detection result and the flip detection result;

[0009] Based on the object area information, the predicted object area information, the detection result, the sample label, the flip detection result and the flip sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0010] According to a second aspect of the embodiments of this specification, there is provided an image processing model training device, comprising:

[0011] a training data determination module configured to determine target training data and enhanced training data, wherein the target training data includes a sample image of a target object located at a target location, a sample label of the sample image, and object region information of the target object; and the enhanced training data includes a flipped sample image corresponding to the sample image and a flipped sample label of the flipped sample image;

[0012] an object detection module configured to perform object detection on the sample image and the flipped sample image using an initial image processing model, and obtain a detection result of the sample image and a flip detection result of the flipped sample image;

[0013] a region information determination module configured to determine predicted object region information of the target object in the sample image based on the detection result and the flip detection result;

[0014] The model training module is configured to perform model training on the initial image processing model based on the object area information, the predicted object area information, the detection results, the sample labels, the flip detection results and the flip sample labels to obtain a trained target image processing model.

[0015] According to a third aspect of the embodiments of this specification, another image processing model training method is provided, including:

[0016] Determining target training data and enhanced training data, wherein the target training data includes a sample CT image of a target tumor located in the nasopharynx, a sample label of the sample CT image, and tumor region information of the target tumor; and the enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image;

[0017] Using the initial image processing model, respectively perform tumor detection on the sample CT image and the flipped sample CT image to obtain a tumor detection result of the sample CT image and a flipped tumor detection result of the flipped sample CT image;

[0018] Determining predicted tumor region information of the target tumor in the sample CT image based on the tumor detection result and the flipped tumor detection result;

[0019] Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0020] According to a fourth aspect of the embodiments of this specification, another image processing model training device is provided, comprising:

[0021] a training data determination module configured to determine target training data and enhanced training data, wherein the target training data includes a sample CT image of a target tumor located in the nasopharynx, a sample label of the sample CT image, and tumor region information of the target tumor; and the enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image;

[0022] a tumor detection module configured to perform tumor detection on the sample CT image and the flipped sample CT image using an initial image processing model, and obtain a tumor detection result of the sample CT image and a flipped tumor detection result of the flipped sample CT image;

[0023] a region information determination module, configured to determine predicted tumor region information of the target tumor in the sample CT image based on the tumor detection result and the flipped tumor detection result;

[0024] The model training module is configured to perform model training on the initial image processing model based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label to obtain a trained target image processing model.

[0025] According to a fifth aspect of the embodiments of this specification, there is provided an image processing method, including:

[0026] Determine an image of a target object to be detected located at a target location;

[0027] A target image processing model is used to perform object detection on the image to be detected to obtain a detection result of the image to be detected, wherein the target image processing model is obtained by performing model training on an initial image processing model based on the object area information determined by the target training data and the enhanced training data, the predicted object area information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label. The target training data includes the sample image of the target object located at the target position, the sample label of the sample image, and the object area information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object area information is determined based on the detection result and the flipped detection result.

[0028] According to a sixth aspect of the embodiments of this specification, there is provided an image processing apparatus, including:

[0029] An image determination module is configured to determine an image of a target object to be detected located at a target location;

[0030] The object detection module is configured to perform object detection on the image to be detected using a target image processing model to obtain a detection result of the image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on the object area information determined by target training data and enhanced training data, the predicted object area information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label. The target training data includes the sample image of the target object located at the target position, the sample label of the sample image, and the object area information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model. The flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object area information is determined based on the detection result and the flipped detection result.

[0031] According to a seventh aspect of the embodiments of this specification, another image processing method is provided, which is applied to a client of a medical system, including:

[0032] In response to a user's clicking operation on a display interface of the client, determining a medical image to be detected;

[0033] The medical image to be detected is sent to the server of the medical system, and the detection result returned by the server is received, wherein the detection result is the detection result output after the medical image to be detected is subjected to object detection by a target image processing model, and the target image processing model is obtained by training an initial image processing model based on the object region information determined by target training data and enhanced training data, the predicted object region information of the target object in the sample image, the detection result, the sample label, the flip detection result, and the flip sample label, the target training data includes the sample image of the target object located at the target site, the sample label of the sample image, and the object region information of the target object, and the enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image, the detection result is determined by performing object detection on the sample image using the initial image processing model, the flip detection result is determined by performing object detection on the flipped sample image using the initial image processing model, and the predicted object region information is determined based on the detection result and the flip detection result;

[0034] The detection result is presented to the user through the presentation interface.

[0035] According to an eighth aspect of the embodiments of this specification, another image processing device is provided, which is applied to a client of a medical system, including:

[0036] an image determination module, configured to determine a medical image to be detected in response to a user's clicking operation on the display interface of the client;

[0037] An object detection module is configured to send the medical image to be detected to the server of the medical system and receive the detection result returned by the server, wherein the detection result is the detection result output after object detection is performed on the medical image to be detected by a target image processing model, and the target image processing model is obtained by training an initial image processing model based on the object region information determined by target training data and enhanced training data, the predicted object region information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label. The target training data includes the sample image of the target object located at the target site, the sample label of the sample image, and the object region information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object region information is determined based on the detection result and the flipped detection result.

[0038] The result display module is configured to display the detection result to the user through the display interface.

[0039] According to a ninth aspect of the embodiments of this specification, there is provided an information processing method, including:

[0040] Determine a CT image of a target tumor to be detected located in the nasopharynx;

[0041] Performing tumor detection on the CT image to be detected using a target image processing model to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0042] An information processing result is determined according to the tumor detection result of the CT image to be detected.

[0043] According to a tenth aspect of the embodiments of this specification, there is provided an information processing device, including:

[0044] an image determination module, configured to determine a CT image to be detected of a target tumor located in the nasopharynx;

[0045] The object detection module is configured to perform tumor detection on the CT image to be detected using a target image processing model to obtain a tumor detection result of the CT image to be detected.

[0046] Wherein, the target image processing model is obtained by training the initial image processing model based on the tumor region information determined by the target training data and the enhanced training data, the predicted tumor region information of the target tumor in the sample CT image, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0047] The information processing module is configured to determine an information processing result according to the tumor detection result of the CT image to be detected.

[0048] According to an eleventh aspect of the embodiments of this specification, a method for computer-aided diagnosis of cancer is provided, comprising:

[0049] Determine a CT image of a target tumor to be detected located at a target site;

[0050] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located at the target site, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0051] According to a twelfth aspect of the embodiments of this specification, a computer-aided diagnosis method for nasopharyngeal carcinoma is provided, comprising:

[0052] Determine the CT image of the target tumor located in the nasopharynx to be detected:

[0053] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by model training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0054] According to a thirteenth aspect of the embodiments of this specification, a computer-aided diagnosis method for breast cancer is provided, comprising:

[0055] Determine a CT image of a target tumor to be detected located in a breast area;

[0056] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by model training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the breast area, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0057] According to a fourteenth aspect of the embodiments of this specification, there is provided a computing device, including:

[0058] memory and processor;

[0059] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of any one of the above methods are implemented.

[0060] According to the fifteenth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores a computer program / instruction, and the computer program / instruction implements the steps of any of the above methods when executed by a processor.

[0061] According to the sixteenth aspect of the embodiments of this specification, a computer program product is provided, comprising a computer program / instruction, which implements the steps of any one of the above methods when executed by a processor.

[0062] One or more embodiments of the present specification provide an image processing model training method, which can use a large number of target training data and enhanced training data to train the initial image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to train the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained through the object area information, predicted object area information, detection results, sample labels, flipped detection results and flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; subsequently, based on the target image processing model, accurate detection results can be obtained in the process of performing object detection on the image. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a schematic diagram of an application of an image processing method provided by an embodiment of this specification;

[0064] Figure 2 This is a flowchart of an image processing model training method provided by one embodiment of this specification;

[0065] Figure 3 is a schematic diagram of a CT scan in an image processing model training method provided in one embodiment of this specification;

[0066] Figure 4 This is a flowchart of a processing process of an image processing model training method provided by one embodiment of this specification;

[0067] Figure 5 is a flowchart of another image processing model training method provided by one embodiment of this specification;

[0068] Figure 6 is a flowchart of an image processing method provided by one embodiment of this specification;

[0069] Figure 7 is a flowchart of another image processing method provided by an embodiment of this specification;

[0070] Figure 8 is a flow chart of an information processing method provided by one embodiment of this specification;

[0071] Figure 9 This is a structural diagram of an image processing model training device provided by one embodiment of this specification;

[0072] Figure 10 This is a schematic diagram of the structure of another image processing model training device provided by one embodiment of this specification;

[0073] Figure 11 This is a schematic diagram of the structure of an image processing device provided by one embodiment of this specification;

[0074] Figure 12 is a structural diagram of another image processing device provided by an embodiment of this specification;

[0075] Figure 13 This is a schematic diagram of the structure of an information processing device provided by one embodiment of this specification;

[0076] Figure 14 This is a structural block diagram of a computing device provided by one embodiment of this specification. DETAILED DESCRIPTION

[0077] The following description sets forth many specific details to facilitate a thorough understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0078] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a," "the," and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0079] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0080] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0081] In one or more embodiments of this specification, a large model refers to a deep learning model with large-scale model parameters, typically containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. A large model can also be called a cornerstone model / foundation model. By pre-training a large model with large-scale unlabeled corpus, a pre-trained model with more than 100 million parameters is produced. This model can adapt to a wide range of downstream tasks and has good generalization capabilities, such as a large language model (LLM) and a multi-modal pre-training model.

[0082] When large models are used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. Large models can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design.

[0083] First, the terms involved in one or more embodiments of this specification are explained.

[0084] Nasopharyngeal carcinoma (NPC) is an epithelial malignancy that originates in the nasopharyngeal mucosa. NPC is one of the most common cancers of the nasopharynx, and timely diagnosis and treatment significantly improve its prognosis. Compared with other head and neck cancers, NPC has unique clinical features and is extremely sensitive to radiotherapy (RT). Therefore, most NPC patients are curable with radiotherapy. One of the most important tasks in radiotherapy planning is delineating the gross tumor volume (GTV) on the planning CT (pCT) scan (without the injection of iodinated contrast agent).

[0085] pCT (planning computed tomography): refers to positron emission tomography, an advanced imaging technology widely used in the medical field.

[0086] GTV (Gross Tumor Volume): refers to the gross tumor volume.

[0087] RT (Radiation therapy): refers to radiation therapy.

[0088] MRI (Magnetic Resonance Imaging): refers to magnetic resonance imaging.

[0089] Dice coefficient: It is a way to evaluate the segmentation effect.

[0090] Projection head: A projection head is a network structure used to extract features or representations. It is typically composed of one or more fully connected layers, or convolutional layers, that map high-dimensional input features into a low-dimensional vector space to facilitate subsequent tasks.

[0091] DSC (Dice Similarity Coefficient): An indicator used to evaluate the similarity of binary image segmentation results. The DSC loss function is often used in fields such as medical image segmentation because it is sensitive to the size of the overlapping part and can well measure the degree of overlap between the predicted results and the true label.

[0092] CE Loss (Cross-Entropy Loss): refers to the cross entropy loss function.

[0093] Affine registration: An image processing technique commonly used to globally align medical images from different time points or different modalities.

[0094] Semantic Symmetry,Human anatomical structures usually exhibit bilateral symmetry,,such as the brain, breast, lungs, and pelvis. Exploiting this,symmetry is crucial for various medical image analyses.

[0095] Margin Loss: The Chinese name is boundary loss. Margin Loss is a loss function. Margin Loss defines the loss by setting an "interval" or "boundary" in order to create a certain gap between the predicted value of the correct category and the wrong category.

[0096] SAT (Semantic Asymmetry Tumor): refers to semantic asymmetry tumor.

[0097] SATs (Semantic Asymmetry Tumor segmentation): refers to semantic asymmetric tumor segmentation.

[0098] Siamese: refers to Siamese Net, which is two identical CNNs (Convolutional Neural Networks) with shared weights.

[0099] LeakyReLU: is an improved ReLU function, which is a nonlinear activation function with parameters.

[0100] ReLU (Rectified Linear Unit): The activation function is a mathematical function commonly used in artificial neural networks to introduce nonlinearity and help the network learn complex patterns and relationships in the data.

[0101] InstanceNorm: The full name is Instance Normalization, which is a technique used in deep learning to accelerate training and improve model performance.

[0102] Epoch: refers to the process in which the entire dataset is completely traversed and learned by the model once.

[0103] With the continuous development of computer technology and artificial intelligence technology, data processing operations can be performed through neural network models, thereby improving the efficiency of data processing; for example, object detection operations can be performed on images through neural network models; current neural network models have the problem of inaccurate detection results in the process of performing object detection on images.

[0104] For example, in radiotherapy for nasopharyngeal carcinoma (NPC), clinicians typically precisely outline the primary gross tumor volume (GTV) on non-contrast planning computed tomography (pCT) to accurately deliver radiation doses. This operation is a challenging task for even experienced radiation oncologists. This is because NPC tumors can invade different types of surrounding tissues, which form poor contrast with the NPC tumor in pCT (especially soft tissues such as mucosa, muscle, nerves, etc.).

[0105] In clinical practice, diagnostic MRI is often used to guide manual GTV delineation in pCT. However, cross-modality registration between MRI and pCT is not easy due to large modality gaps and scan range variations. Radiation oncologists typically use rigid transformations to roughly align MRI and pCT by matching bony structures such as the spine, or they simply open two viewers on the computer to simultaneously observe the pCT and MRI and virtually fuse them in their minds to facilitate NPC GTV delineation. This process is very time-consuming (at least 1-2 hours per case) and prone to errors. Therefore, accurate automatic segmentation of NPC GTV in pCT is very necessary. That is, due to the poor contrast between NPC tumors and many adjacent normal tissues, radiation oncologists need to manually delineate the GTV with the help of diagnostic MRI, and virtually align and fuse the MRI in their minds; but this process is both time-consuming and error-prone.

[0106] To address the above problems, this specification provides a method for automatic segmentation of NPC GTV based on deep learning. However, most MRIs do not provide direct measurements of electron density required for RT planning. Therefore, MRI-derived tumor masks still need to be spatially converted to pCT through image registration, which leads to potential alignment errors (because large errors usually occur in cross-modal registration). Some studies have explored segmenting NPC GTV only in enhanced CT scans (UNETR, SwinUNETR and its variant SwinUNETR-v2). However, they generally produce relatively low performance, such as Dice scores (DSC) below 70%. Only a few methods have addressed the problem of automatic segmentation of NPC GTV using CT and MRI. It has been observed that if there is a large registration error between paired CT and MRI, the multimodal segmentation performance may be inferior to the single-modal segmentation performance.

[0107] Based on this, this specification provides an image processing model training method. One or more embodiments of this specification also involve another image processing model training method, an image processing method, another image processing method, an information processing method, a computer-assisted diagnosis method for cancer, a computer-assisted diagnosis method for nasopharyngeal carcinoma, a computer-assisted diagnosis method for breast cancer, an image processing model training device, another image processing model training device, an image processing device, another image processing device, an information processing device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0108] See also Figure 1 , Figure 1 The following diagram illustrates an application of an image processing method according to one embodiment of the present disclosure. The client 102 can determine a CT image of a target tumor located in the nasopharynx, and then send the CT image to the server 104. The server 104 then uses an image segmentation model trained based on target training data and enhanced training data to perform tumor segmentation on the CT image, thereby obtaining a tumor segmentation map. The server 104 then sends the tumor segmentation map to the client 102. This method significantly improves GTV segmentation in nasopharyngeal carcinoma by leveraging semantic symmetry in anatomical structures. The tumor segmentation map can potentially be used in radiotherapy practice to standardize nasopharyngeal carcinoma GTV delineation and reduce the workload and variation of radiation oncologists.

[0109] See also Figure 2 , Figure 2 A flowchart of an image processing model training method provided according to an embodiment of this specification is shown, which specifically includes the following steps.

[0110] Step 202: Determine target training data and enhanced training data, wherein the target training data includes a sample image of a target object located at a target location, a sample label of the sample image, and object area information of the target object; and the enhanced training data includes a flipped sample image corresponding to the sample image and a flipped sample label of the flipped sample image.

[0111] The target training data and the enhanced training data are training data used to perform model training on the initial image processing model, and the enhanced training data are training samples obtained by performing sample enhancement processing on the target training data.

[0112] The target part can be a specific part of the target organism, for example, the target part can be the nasopharynx, chest, lung, breast, pelvic area, brain, etc. The target organism can be a human, an animal, etc. It should be noted that the target part can be a part that satisfies semantic symmetry. The target object can be understood as a target tumor, diseased tissue, diseased area, cancer area, etc. For details, please refer to Figure 3 , Figure 3 is a schematic diagram of a CT scan in an image processing model training method provided in one embodiment of this specification. Figure 3 Three CT scan images of the nasopharynx are shown, where the area framed by white dotted lines is the tumor area. Each CT scan image has a central sagittal axis, and the nasopharynx is bilaterally symmetrical along the central sagittal axis.

[0113] It should be noted that in one or more embodiments of this specification, the image processing model training method is explained by taking its application in the nasopharyngeal carcinoma scenario as an example. The image processing model training method provided in this specification can be applied to cancer scenarios, breast cancer scenarios, or pelvic cancer scenarios. For specific execution methods for cancer, breast cancer, and pelvic cancer scenarios, please refer to the corresponding or corresponding descriptions in one or more embodiments of this specification, and no further details will be given here.

[0114] The sample image can be understood as a CT image, plain CT image, enhanced CT image, pCT and / or MRI used as a training sample. The sample label can be understood as a segmented image used to segment the target object in the sample image; for example, the sample label can be a tumor segmentation image; the sample label can be a semantic segmentation mask. The object region information can be understood as information characterizing the location region of the target object in the sample image, for example, the object region information can be coordinates, a mask, or an anchor box; the flipped sample image can be understood as an image obtained by flipping the sample image, and the flipped sample label can be understood as the sample label corresponding to the flipped sample image.

[0115] Taking the application of the image processing model training method provided in this specification in the precise tumor segmentation scenario as an example, the image processing model training method is explained, wherein the sample image of the target object located at the target location is a CT image of a tumor located in the nasopharynx, the sample label is the semantic segmentation mask corresponding to the sample image, the object area information can be an asymmetric mask for locating the tumor in the CT image, and the flipped sample image can be a CT image obtained by flipping the CT image.

[0116] Based on this, the method determines a CT image of a tumor located in the nasopharynx, the CT image I∈RD×H×W, where D, H, and W are the spatial dimensions of the CT image; and determines a flipped CT image I′ of the CT image I.

[0117] Determine a semantic segmentation mask for the CT image I. The semantic segmentation mask is represented as s∈{0,1}. The shape of the semantic segmentation mask is D×H×W, where 0 represents the background and 1 represents the tumor foreground. Based on the semantic segmentation mask s, determine the flipped semantic segmentation mask s′ of the flipped CT image I′. s and s′ are used as the sample label and flipped sample label.

[0118] Then, an asymmetric mask m (i.e., object region information) is defined to locate the asymmetric region (i.e., tumor region) in the CT image I, i.e., m = ss ∩ s′, where m∈{0, 1}D×H×W. Note that 1 and 0 represent the asymmetric and symmetric regions in the CT image I, respectively.

[0119] In one or more embodiments provided in this specification, in order to improve the accuracy of the image processing model, the training data can be expanded to obtain a larger amount of training data; the model can be trained with more training data to improve the accuracy of the detection results output by the target image processing model; the specific implementation method is as follows.

[0120] The determining of target training data and enhanced training data includes:

[0121] determining the target training data;

[0122] horizontally flipping the sample image to obtain the flipped sample image, and horizontally flipping the sample label to obtain the flipped sample label of the flipped sample image;

[0123] The flipped sample image and the flipped sample label are used as the enhanced training sample.

[0124] Continuing with the above example, after determining the original CT image I, the CT image I is horizontally flipped to obtain a flipped CT image I′. After determining the semantic segmentation mask s of the CT image I, the flipped semantic mask s′ of the flipped CT image I′ can be obtained by performing a horizontal flip operation on the semantic segmentation mask s.

[0125] In one or more embodiments provided in this specification, the target part in the sample image and the flipped sample image provided by this method can be a part that satisfies semantic symmetry; since the target part satisfies semantic symmetry, this method can accurately identify the target object existing in the target part based on semantic asymmetry; and in order to improve the learning efficiency and effect of the image processing model, the sample image and the flipped sample image provided by this method include a central axis, which can symmetrically divide the target part; the specific method for determining the central axis of the sample image and the flipped sample image is as follows.

[0126] The determining of the target training data includes:

[0127] Determining an object image of the target object located at the target site, and determining an object image to be processed from the object image, wherein the object image to be processed is any image in the object image;

[0128] Based on the target part, adding a central axis to the image of the object to be processed to obtain an initial object image, wherein the central axis symmetrically divides the target part;

[0129] Based on the initial object image, adding the central axis to other object images in the object image except the object image to be processed to obtain a target object image;

[0130] Taking the initial object image and the target object image as the sample images, and determining a sample label corresponding to the sample image and object region information of the target object;

[0131] The sample image, the sample label, and the object region information are used as the target training data.

[0132] The central axis may be a central sagittal axis or a sagittal axis.

[0133] Based on the initial object image, the central axis is added to the other object images in the object image except the object image to be processed to obtain the target object image. The initial object image can be used as an atlas. Based on the atlas, affine alignment is used to add the central axis to the other object images in the object image except the object image to be processed to obtain the target object image.

[0134] The manner of determining the sample label corresponding to the sample image and the object region information of the target object may refer to the corresponding content in one or more embodiments.

[0135] Continuing with the above example, the SATs model (image processing model) of this method focuses on asymmetric abnormal areas in CT images. Considering that image asymmetry may originate from pathological or non-pathological sources, such as changes in imaging angle and patient posture, the CT scan images can be pre-processed to ensure that the scan is symmetrical along the central sagittal axis. Specifically, this method selects patient CT images that are bilaterally symmetrical along the central sagittal plane as the atlas. Then, based on this atlas, the other patient CT images are aligned to the atlas space through affine registration, thereby determining the central sagittal axis for the other patient CT images. This step helps to mitigate the influence of other asymmetric anatomical structures of the head and neck that may mislead the model.

[0136] Step 204: using the initial image processing model, perform object detection on the sample image and the flipped sample image respectively to obtain a detection result of the sample image and a flip detection result of the flipped sample image.

[0137] Among them, the image processing model can be understood as a model that performs object detection processing on the target object in the sample image or the flipped sample image. The image processing model can be an image segmentation model or an object recognition model; the initial image processing model can be understood as an image processing model of any round in the process of iterative model training; correspondingly, the target image processing model can be a trained image processing model.

[0138] The detection result can be understood as the result obtained after the initial image processing model performs object detection on the sample image. The detection result can be a tumor segmentation map, a segmentation map, a probability map, a heat map, a boundary map or a mask. The flip detection result can be understood as the detection result obtained after the initial image processing model performs object detection on the flipped sample image.

[0139] Continuing with the above example, when using an image segmentation model to identify tumors in CT images, the image segmentation model can assign each pixel in the input CT image to a specific category, thereby identifying the tumor area and non-tumor area (background area) in the CT. Based on this, after the CT image I and the flipped CT image I′ are input into the image segmentation model, the image segmentation model can assign each pixel in the input CT image to a specific category and output the segmentation maps corresponding to the CT image I and the flipped CT image I′.

[0140] In one or more embodiments provided in this specification, the performing of object detection on the sample image and the flipped sample image using the initial image processing model to obtain a detection result of the sample image and a flip detection result of the flipped sample image includes:

[0141] Encoding the sample image and the flipped sample image respectively using an encoder in the initial image processing model to obtain a sample image encoding of the sample image and a flipped sample image encoding of the flipped sample image;

[0142] The sample image code and the flipped sample image code are decoded respectively by using the decoder in the initial image processing model to obtain the detection result of the sample image and the flip detection result of the flipped sample image.

[0143] Following the previous example, the SATs architecture provided by this method consists of an encoder-decoder module and a projection head. The encoder-decoder is a U-shaped encoder-decoder architecture with shared weights, where the encoder uses the repeated application of 3D residual blocks, each containing two convolutional layers with 3×3×3 kernels. Each convolutional layer is followed by InstanceNorm normalization and LeakyReLU activation. For downsampling, the resolution of the input feature map is halved using convolution operations with a stride of 2. The initial number of filters is 32 and is doubled after each downsampling step to maintain constant time complexity, except for the last layer; in total, the encoder performs four downsampling operations.

[0144] Based on this, the SATs model of this method uses a Siamese segmentation network with an encoder-decoder with shared weights to process the original CT image I and the flipped CT image I′.

[0145] It should be noted that although both components, the encoder-decoder module and the projection head, are involved during training, during inference, the encoder-decoder module can be used.

[0146] Step 206: Determine predicted object region information of the target object in the sample image based on the detection result and the flip detection result.

[0147] The predicted object region information may be understood as object region information predicted using the detection results and the flipped detection results. For example, the object region information may be a tumor region feature code, tumor region coordinates, and the like.

[0148] In one or more embodiments provided in this specification, the image processing model training method in this specification proposes a semantic asymmetry learning method that utilizes the asymmetric characteristics of pharyngeal tumors to achieve accurate nasopharyngeal carcinoma GTV segmentation. The image processing model uses an encoder-decoder with shared weights to process the original CT image I and the flipped CT image I', and then uses a nonlinear projection module and distance metric learning to maximize the difference between abnormal and normal locations. Based on this, GTV segmentation in nasopharyngeal carcinoma can be significantly improved by utilizing semantic symmetry in anatomical structures; the image processing model can be used in radiotherapy practice to standardize nasopharyngeal carcinoma GTV delineation and reduce the workload and variation of radiation oncologists, and the specific implementation method is as follows.

[0149] The determining, based on the detection result and the flip detection result, predicted object region information of the target object in the sample image includes:

[0150] Using a nonlinear processing unit, performing data enhancement on the detection result to obtain an enhanced detection result, and performing data enhancement on the flip detection result to obtain an enhanced flip detection result;

[0151] Determining an object feature corresponding to the target object included in the enhanced detection feature, and determining a flipped object feature of the target object included in the enhanced flip detection result;

[0152] Comparing the object feature and the flipped object feature to determine an enhanced image feature corresponding to the sample image;

[0153] Predicted object region information of the target object in the sample image is determined from the enhanced image features.

[0154] The nonlinear processing unit may be understood as a nonlinear projection module, the enhanced image feature may be understood as an enhanced image feature, and the nonlinear processing unit may be a projection head.

[0155] Continuing with the previous example, the SATs architecture provided by this method includes an encoder-decoder module and a projection head. The projection head uses a nonlinear projection module to transform features (i.e., detection results and flipped detection results) before calculating the distance of the marginal loss, aiming to improve the quality of learned features. It consists of three 1×1×1 convolutional layers, each with 16 channels, followed by a unit normalization layer.

[0156] Specifically, the projection head consists of three convolutional layers (Conv). The first two layers of the projection head can use the ReLU activation function. The specific implementation is as follows:

[0157] 1. The segmentation map of CT image I is input into the projection head (i.e., nonlinear processing unit), and the segmentation map of CT image I (i.e., detection result) is nonlinearly enhanced using three convolutional layers to obtain the CT enhanced image feature E of CT image I (i.e., enhanced detection result);

[0158] 2. The segmentation map of the flipped CT image I′ (i.e., the flip detection result) is input into the projection head (i.e., the nonlinear processing unit). The segmentation map of the flipped CT image I′ is nonlinearly enhanced using three convolutional layers to obtain the CT enhanced image features Ef (i.e., the flipped object features) of the flipped CT image I′.

[0159] 3. Tumor features of the tumor determined from CT enhanced image feature E, and tumor features of the tumor determined from CT enhanced image feature Ef;

[0160] 4. Compare the two tumor features to determine the CT image feature D (i.e., enhanced image feature) corresponding to the CT image I;

[0161] 5. Determine the tumor region feature code corresponding to the tumor from the CT image feature D.

[0162] As can be seen from the above examples, this method assumes that directly applying metric learning to segmentation features may result in information loss and reduce model effectiveness. For example, some asymmetries in CT images are non-pathological, perhaps arising from variations in the patient's head positioning and posture, but they are beneficial for segmentation. Using nonlinear projection, this irrelevant information can be filtered out of the metric learning process, ensuring that it is retained in the features used for segmentation.

[0163] In one or more embodiments provided in this specification, comparing the object feature and the flipped object feature to determine the enhanced image feature corresponding to the sample image includes:

[0164] Using a nonlinear processing unit, determining that the object feature includes a plurality of object feature points, and the flipped object feature includes a plurality of flipped object feature points;

[0165] Determining corresponding feature points of each object feature point from the plurality of flipped object feature points, and determining similarities between each object feature point and the corresponding feature point;

[0166] When the similarity is greater than or equal to a first threshold, the similarity weight between the object feature point and the corresponding feature point is increased, and when the similarity is greater than or equal to the first threshold, the similarity weight between the object feature point and the corresponding feature point is reduced to obtain enhanced image features corresponding to the sample image.

[0167] Continuing with the above example, 1. From the pixel points of the CT enhanced image feature Ef (i.e., the flipped object feature points), determine the pixel points (i.e., corresponding feature points) corresponding to the pixel points of the CT enhanced image feature E (i.e., the object feature points); the pixel points can be understood as information such as features, feature codes, and masks.

[0168] 2. Calculate the similarity of the corresponding pixels in E and Ef to determine the distance between each pixel, and then obtain the pixel-by-pixel distance between E and Ef.

[0169] 3. Based on the pixel-by-pixel distance, the similarity between each pixel in E and each pixel in Ef is pulled closer or further away to obtain the adjusted CT image feature D;

[0170] When the pixel-by-pixel distance is greater than or equal to the preset similarity threshold, the similarity weight between the two pixels is increased, thereby shortening the similarity between the two pixels.

[0171] When the pixel-by-pixel distance is less than or equal to a preset dissimilarity threshold, the similarity weight between the two pixels is reduced, thereby deviating the similarity between the two pixels.

[0172] Based on the above embodiments, the present method introduces a nonlinear projection module and a distance metric learning strategy into the SATs model to refine the generated feature map, thereby maximizing the difference between E and Ef at the corresponding anatomical positions of abnormal and normal anatomical locations.

[0173] Step 208: Based on the object area information, the predicted object area information, the detection result, the sample label, the flip detection result and the flip sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0174] In one or more embodiments provided in this specification, the target site is the nasopharyngeal region, the target object is a target tumor, the sample image is a sample CT image, and the flipped sample image is a flipped sample CT image;

[0175] The performing model training on the initial image processing model based on the object region information, the predicted object region information, the detection result, the sample label, the flip detection result, and the flip sample label to obtain a trained target image processing model includes:

[0176] Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0177] For the explanation of the above embodiment, reference may be made to the corresponding or corresponding explanations in other embodiments, which will not be elaborated here.

[0178] In one or more embodiments provided in this specification, the performing model training on the initial image processing model based on the object region information, the predicted object region information, the detection result, the sample label, the flip detection result, and the flip sample label to obtain a trained target image processing model includes:

[0179] calculating a first loss function based on the object region information and the predicted object region information, wherein the first loss function is used to evaluate a difference between the object region information and the predicted object region information;

[0180] Calculating a second loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the second loss function is used to evaluate the difference between the detection result and the sample label, and the difference between the flip detection result and the flip sample label;

[0181] Calculating a third loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the third loss function is used to evaluate the similarity between the detection result and the sample label, and the similarity between the flip detection result and the flip sample label;

[0182] Based on the first loss function, the second loss function and the third loss function, the model parameters of the initial image processing model are adjusted to obtain the trained target image processing model.

[0183] Continuing with the above example, this method adopts an asymmetric learning strategy. The segmentation loss function for the STA model consists of two parts: a combination of the Dice loss function (the second loss function) and the cross-entropy loss function (the second loss function) for traditional segmentation purposes, and a voxel margin loss (the first loss function) specifically designed for asymmetric abnormal regions.

[0184] For the Dice loss function and the cross entropy loss function, the normalization layer can be used to normalize the two segmentation maps (i.e., the detection result and the flip detection result) to obtain the two normalized segmentation maps (i.e., the normalized detection result and the flip detection result); then, the DSC loss and CE loss are calculated based on the two normalized segmentation maps, sample labels, and flipped sample labels.

[0185] For the metric-based margin loss, this method adopts pixel-level margin loss. In the asymmetric abnormal region, the goal of this method is to minimize the similarity between the feature of any point and its corresponding point on the central sagittal axis; the margin loss between the feature E and the flipped feature E′ is:

[0186]

[0187] The feature E∈RH×W×D×C in the formula, where H, W, and D are spatial dimensions, C is the number of output features, 1 is an indicator function, t defines a boundary for adjusting the degree of semantic asymmetry, (i, j, k) represents the position of voxel (pixel in the tumor area), which can be understood as a three-dimensional coordinate; m is a mask, which represents a specific area.

[0188] One or more embodiments of the present specification provide an image processing model training method, which can use a large number of target training data and enhanced training data to train the initial image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to train the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained through the object area information, predicted object area information, detection results, sample labels, flipped detection results and flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; subsequently, based on the target image processing model, accurate detection results can be obtained in the process of performing object detection on the image.

[0189] The following combined Figure 4 , taking the application of the image processing model training method provided in this specification in the nasopharyngeal carcinoma scenario as an example, the image processing model training method is further explained. Figure 4 A flowchart of a method for training an image processing model according to an embodiment of the present invention is shown. Figure 4 It can be seen that the SATs architecture provided by this method includes an encoder-decoder module and a projection head;

[0190] The encoder-decoder module is a U-shaped encoder-decoder architecture with shared weights. The encoder uses repeated application of 3D residual blocks, each of which contains two convolutional layers with a 3×3×3 kernel. Each convolutional layer is followed by InstanceNorm normalization and LeakyReLU activation.

[0191] For downsampling, a convolution operation with a stride of 2 is used to halve the resolution of the input feature map; the initial number of filters is 32 and is doubled after each downsampling step to maintain constant time complexity, except for the last layer. In total, the encoder performs four downsampling operations.

[0192] The projection head also refers to nonlinear projection. Before calculating the distance of the marginal loss, this scheme uses a nonlinear projection module to transform the image features output by the encoder-decoder module (i.e., the model output of the SATs architecture), aiming to improve the quality of the learned features.

[0193] The projection head consists of three 1×1×1 convolutional layers, each with 16 channels, followed by a unit normalization layer. The first two layers of the projection head use the ReLU activation function. This scheme assumes that directly applying metric learning to segmentation features may lead to information loss and reduce the effectiveness of the model. For example, some asymmetries in CT images are non-pathological and may arise from changes in the patient's head positioning and posture, but they are beneficial for segmentation. Therefore, using nonlinear projection can filter out this irrelevant information from the metric learning process, ensuring that it is retained in the features used for segmentation.

[0194] It should be noted that the following steps take the application of the image processing model training method provided in this specification in the nasopharyngeal carcinoma scenario as an example to explain the image processing model training method. The image processing model training method provided in this specification can be applied to cancer scenarios for computer-assisted diagnosis of cancer, can be applied to breast cancer scenarios for computer-assisted diagnosis of breast cancer, or can be applied to pelvic cancer scenarios for computer-assisted diagnosis of pelvic cancer. For the specific execution methods of cancer, breast cancer, and pelvic cancer scenarios, please refer to the corresponding or corresponding descriptions in the following steps, which will not be elaborated here.

[0195] Based on this, the image processing model training method specifically includes the following steps.

[0196] Step 402: Construct training data.

[0197] The training data of this solution includes original CT images and flipped CT images. The specific method of determining the training data is as follows.

[0198] 1. Acquire a CT image I as a sample image.

[0199] The CT image I refers to a CT image of a tumor located in the nasopharynx, and the CT image I∈RD×H×W, where D, H, and W are the spatial dimensions of the CT image.

[0200] It should be noted that the CT image (i.e., the CT image to be detected in the embodiment of this specification) can be a CT image to be detected of a target tumor located at a target site, a CT image to be detected of a target tumor located at the nasopharynx, a CT image to be detected of a target tumor located at the breast, or a CT image to be detected of a target tumor located at the pelvic site.

[0201] 2. Selecting the corresponding CT images on both sides along the central sagittal plane from CT image I as the atlas;

[0202] 3. Based on the atlas, affine registration is used to align other patients’ CT images to the atlas space, thereby determining the central sagittal axis for other patients’ CT images.

[0203] This step helps mitigate the effects of other asymmetric anatomy of the head and neck that can mislead the model.

[0204] 4. Mark the lesion area in the CT image to obtain the sample label.

[0205] Specifically, this solution can determine a semantic segmentation mask for the CT image I, where the semantic segmentation mask is represented as s∈{0, 1}. The shape of the semantic segmentation mask is D×H×W, where 0 represents the background and 1 represents the tumor foreground.

[0206] 5. Flip the CT image I to obtain the flipped CT image I′ ( Figure 4 I in f ) and the corresponding sample labels.

[0207] Specifically, after determining the original CT image I, the CT image I is horizontally flipped to obtain a flipped CT image I′. After determining the semantic segmentation mask s of the CT image I, a flipped semantic mask s′ of the flipped CT image I′ can be obtained by performing a horizontal flip operation on the semantic segmentation mask s.

[0208] 6. Mark the asymmetric region in the CT image I to obtain an asymmetric mask.

[0209] Specifically, this method can define an asymmetric mask m to locate the asymmetric region (i.e., tumor region) in the CT image I, i.e., m = ss ∩ s′, where m∈{0,1}D×H×W. It should be noted that 1 and 0 represent the asymmetric and symmetric regions in the CT image I, respectively.

[0210] Step 404: Input the training sample into the SATs model to obtain the segmentation map output by the SATs model.

[0211] Specifically, the SATs model in this solution includes an encoder and a decoder, which respectively encode and decode the CT image I and the flipped CT image I′ to obtain the segmentation maps corresponding to the CT image I and the flipped CT image I′. The specific implementation method is as follows:

[0212] In the process of identifying tumors in CT images using image segmentation models, the image segmentation model can assign each pixel in the input CT image to a specific category, thereby identifying the tumor area and non-tumor area (background area) in the CT;

[0213] Based on this, after the CT image I and the flipped CT image I′ are input into the image segmentation model, the image segmentation model can use the encoder to encode the CT image I and the flipped CT image I′, thereby obtaining the image codes corresponding to the CT image I and the flipped CT image I′.

[0214] Then, the image segmentation model can use a decoder to decode the image codes corresponding to the CT image I and the flipped CT image I′, thereby assigning each pixel in the CT image I and the flipped CT image I′ to a specific category and outputting the segmentation maps corresponding to the CT image I and the flipped CT image I′.

[0215] The specific category may be (0, 1), where 0 represents a non-tumor pixel and 1 represents a tumor pixel; and the segmentation map may be a semantic segmentation mask consisting of (0, 1).

[0216] Step 406: Use a normalization layer to normalize the segmentation map and calculate DSC loss and CE Loss.

[0217] Step 408: Utilize the CT enhanced image features E and Ef obtained by the projection head; obtain the adjusted CT image features D based on the metric learning strategy, and calculate the marginal loss function based on the adjusted CT image features D and the asymmetric mask m.

[0218] Specifically, the projection head consists of three convolutional layers (Conv), which are implemented as follows:

[0219] 1. The segmentation map of CT image I is input into the projection head (i.e., nonlinear projection module), and the segmentation map of CT image I is nonlinearly enhanced using three convolutional layers to obtain the CT enhanced image features E of CT image I;

[0220] 2. Input the segmentation map of the flipped CT image I′ into the projection head (i.e., nonlinear projection module), and use three layers of convolutional layers to perform nonlinear enhancement on the segmentation map of the flipped CT image I′ to obtain the CT enhanced image feature Ef of the flipped CT image I′ Figure 4 E in f );

[0221] 3. From the pixel points of the CT enhanced image feature Ef, determine the pixel points corresponding to the pixel points of the CT enhanced image feature E, and calculate the similarity of the corresponding pixel points in E and Ef to determine the distance between each pixel point, and then obtain the pixel-by-pixel distance between E and Ef.

[0222] 4. Based on the pixel-by-pixel distance, the similarity between each pixel in E and each pixel in Ef is pulled closer or further away to obtain the adjusted CT image feature D.

[0223] When the pixel-by-pixel distance is greater than or equal to the preset similarity threshold, the similarity weight between the two pixels is increased, thereby shortening the similarity between the two pixels.

[0224] When the pixel-by-pixel distance is less than or equal to a preset dissimilarity threshold, the similarity weight between the two pixels is reduced, thereby deviating the similarity between the two pixels.

[0225] 5. MarginLoss is calculated based on the tumor region feature encoding in the CT image feature D and the asymmetric mask m.

[0226] It should be noted that after obtaining Margin Loss, DSC loss, and CE loss, the three loss functions are added together to obtain the overall loss function, and the overall loss function is used to adjust the model parameters of the SATs model until the model training stops and the trained SATs model is obtained.

[0227] Based on the above steps, the image processing model training method described in this specification proposes a method for accurate tumor segmentation of NPC in non-contrast CT images using semantic asymmetry. This method, based on the semantic asymmetry characteristics of gross tumors in the nasopharyngeal region, uses a 3D semantically asymmetric tumor segmentation (SATs) method to achieve accurate NPC GTV segmentation even with limited data. This method can directly segment the NPC GTV in non-contrast pCT, thereby avoiding registration errors when aligning MRI or MRI-derived tumor masks with pCT. This solves the problem of poor contrast between the GTV and adjacent normal structures in pCT.

[0228] In this method, the goal is to segment the NPC GTV in non-contrast pCT (the most conventional RT imaging), which can avoid registration errors when aligning MRI or MRI-derived tumor masks with pCT. Direct segmentation of the NPC GTV in non-contrast pCT is challenging because the boundaries between NPC tumors and adjacent soft tissues (such as membranes, muscles, and blood vessels) are very unclear in non-contrast pCT. To alleviate this problem, this method proposes a 3D semantically asymmetric tumor segmentation (SATs) method based on the observation that the healthy nasopharyngeal region is usually bilaterally symmetric (see Figure 3 ), but the presence of NPC tumors will destroy this symmetry. Specifically, in order to exploit anatomical symmetry cues, the pCT is first wrapped with automatically segmented head and neck organs so that the pCT is bilaterally symmetric along the midsagittal plane. This helps to reduce the impact of asymmetric anatomical structures caused by changes in patient head position and posture during CT scanning. Then, a Siamese contrastive learning segmentation framework is developed based on traditional segmentation losses (Dice and / or cross entropy loss) and an additional voxel-level edge loss. The edge loss is applied to the deep features extracted from the original and flipped pCT scans, aiming to minimize the voxel distance between the original and flipped tumor-free regions and encourage a larger distance between the original and flipped tumor-bearing regions, making the GTV features more sensitive to semantic asymmetry.

[0229] Based on the above, it can be seen that the method in this specification can effectively improve the performance of NPC GTV segmentation in non-contrast pCT by utilizing the inherent semantic symmetry of anatomical structures.

[0230] During training, this method uses the stochastic gradient descent algorithm with a momentum of 0.99 as the optimizer. Model training is divided into two stages. In the first stage, only the Siamese encoder is trained for 800 epochs with a learning rate of 1e-2, decayed by a polynomial schedule. In the second stage, the projection head is trained for 200 epochs with a learning rate of 1e-2 for the projection head and a learning rate of 1e-5 for the Siamese encoder, both decayed by a polynomial schedule. The patch size is 56×192×192, and the batch size is 2. In testing, a sliding window strategy is used. For the voxel contrast loss, this method uses a margin hyperparameter t=20.

[0231] During data preparation, an internal dataset was collected and curated from the hospital for the development of the segmentation model (SATs model). This dataset included pCT, contrast-enhanced diagnostic CT, and diagnostic MRI images of 163 NPC patients with T1 and T2 disease stages. The diagnostic CT and MRI images were registered. First, a rigid transformation was applied to the MRI images to approximately align them with the CT images, ensuring similar anatomical localization. Subsequently, a cross-modality deformable registration algorithm (DEEDs) was used to achieve accurate local alignment. The contrast-enhanced CT and MRI images were used to guide radiation oncologists in generating ground-truth GTVs in the pCT. In addition, a publicly available dataset containing 118 non-contrast pCT and contrast-enhanced CT images was collected and curated as an external dataset. The original tumor targets in the publicly available datasets were often larger than those described by the GTV guidelines. Therefore, their labels were curated. Annotations for all datasets were reviewed and edited by two experienced radiation oncologists according to international consensus guidelines for GTV delineation. For evaluation, we randomly select 20% of the internal dataset as the internal test set and use a curated publicly available dataset as the external test dataset. The asymmetric regions in the external data are larger than those in the internal data, making the segmentation task more challenging.

[0232] It is important to note that regarding the comparison methods and evaluation metrics, our method was comprehensively compared with various state-of-the-art methods, including CNN-based methods, transformer-based methods, and Mambaba (a medical image segmentation method), to evaluate its performance. To ensure a fair comparison, all competing models were trained for the same number of epochs, 1000. To avoid model collapse during training, different learning rates were used for the different comparison methods. Generally, CNN-based methods use larger learning rates, while transformer-based methods use smaller ones. For example, for various state-of-the-art methods, the learning rate was set to 1e-2, and stochastic gradient descent with momentum of 0.99 was used as the optimizer. For TransUNet (a CNN-based method), the learning rate was set to 1e-3, and stochastic gradient descent with momentum of 0.99 was used as the optimizer. For UNETR (a transformer-based method), the learning rate was set to 1e-5, and the AdamW optimizer was used. Furthermore, this method uses the Dice similarity coefficient (DSC) (%) to assess the overlap of the segmentation maps and calculates the average surface distance (ASD) (mm) for all cases. Furthermore, this method measures the 95th percentile of the Hausdorff distance (HD95) (mm) to measure the registration accuracy relative to the structure boundary. Generally, better methods should produce higher DSC scores and lower ASD and HD95 values.

[0233] In comparison with state-of-the-art methods, we compared our SATs model with state-of-the-art segmentation methods using an internal dataset to validate GTV segmentation for NPC in non-contrast CT images. With relatively few parameters, our SATs model significantly improved over existing state-of-the-art methods. Specifically, our SATs model achieved improvements of 1.0%, 2.6%, and 3.6% in DSC, ASD, and HD95, respectively, compared to the suboptimal method. Our SATs model also achieved higher accuracy in boundary segmentation, such as the nasal septum.

[0234] During the cross-domain transfer evaluation, two experiments were conducted to verify the generalization performance of our method across different domains: first, our method was trained using the training set of the internal dataset and validated on the test set of the external dataset; second, our method was trained using the training set of the external dataset and validated on the test set of the internal dataset. Our method demonstrated excellent performance in both experiments, particularly in the first, confirming the effectiveness of the learned features associated with asymmetric tumors. Notably, our method achieved the highest mean and lowest variability across both demanding tasks, demonstrating superior and consistent performance. Furthermore, it is clear that our method significantly outperforms when the internally trained model is applied to external data with regions of greater asymmetry.

[0235] In the process of ablation studies, the effects of the projection head (Proj.Head) and the marginal loss (Marg.Loss) on the segmentation model are studied. Table 1 shows the results of the external data. Performance indicators of different segmentation models.

[0236] Table 1

[0237]

[0238] Based on Table 1, we can see that despite the addition of margin loss, this model achieves lower DSC and higher ASD / HD95 compared to the initial model, indicating that margin loss alone is insufficient to improve performance. Furthermore, when the projection head module and margin loss are attached to the baseline, there is a significant performance improvement (+4.98% DSC, -0.60mm ASD, and -1.15mm HD95). Based on Table 1, we can see that semantic asymmetry learning, consisting of the projection head and margin loss, effectively improves tumor segmentation accuracy; this is the impact of semantic asymmetry learning. Our method was compared to a model without the projection head module and / or the margin loss baseline. Our method achieved higher Dice scores and consistently outperformed other models in the majority of the 117 test scans.

[0239] It should be noted that in the embodiments of this specification, a method for semantic asymmetry learning is used using a large amount of unlabeled data. This method can use the proposed unsupervised asymmetric region selection method (through the inverse consistency constraint of bidirectional deformable image registration) to collect additional unlabeled data and detect asymmetric abnormal regions. For unlabeled data, this method calculates marginal loss during training. This method can capture useful GTV segmentation representations from asymmetric regions of unlabeled data, and the performance is improved compared to the benchmark SAT; based on this, the method proposed in this specification has the potential to solve the scarcity of labeled data in the field of medical image analysis.

[0240] In summary, one or more embodiments of the present specification propose a semantic asymmetry learning method that utilizes the asymmetric characteristics of tumors in the pharyngeal duct region to achieve accurate segmentation of nasopharyngeal carcinoma GTV. The SATs model of this method uses a Siamese segmentation network with a shared encoder-decoder to process original and flipped CT images, followed by a nonlinear projection module and distance metric learning to maximize the distinction between abnormal and normal locations. By utilizing the semantic symmetry of anatomical structures, this method can significantly improve the GTV segmentation of NPC, outperforming 9 state-of-the-art methods, which were evaluated on an internal test set and an independent external dataset. In radiotherapy practice, standardizing the delineation of nasopharyngeal carcinoma GTV has potential application value in reducing the workload and variation of radiation oncologists.

[0241] See also Figure 5 , Figure 5 A flowchart of another image processing model training method provided according to an embodiment of this specification is shown, which specifically includes the following steps.

[0242] Step 502: Determine target training data and enhanced training data, wherein the target training data includes a sample CT image of a target tumor located in the nasopharynx, a sample label of the sample CT image, and tumor region information of the target tumor; and the enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image.

[0243] Step 504: Using the initial image processing model, perform tumor detection on the sample CT image and the flipped sample CT image, respectively, to obtain a tumor detection result of the sample CT image and a flipped tumor detection result of the flipped sample CT image:

[0244] Step 506: Determine predicted tumor region information of the target tumor in the sample CT image based on the tumor detection result and the flipped tumor detection result;

[0245] Step 508: Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0246] Another image processing model training method provided by one or more embodiments of this specification can utilize a large amount of target training data and enhanced training data to train an initial image processing model, thereby improving the accuracy of tumor detection results output by a target CT image processing model. Furthermore, during the model training process, to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on a sample CT image and a flipped sample CT image, respectively, to obtain a tumor detection result for the sample CT image and a flipped tumor detection result for the flipped sample CT image. Based on the tumor detection result and the flipped tumor detection result, predicted tumor region information of a target tumor in the sample CT image is determined. The tumor region information and the predicted tumor region information are then used to train the initial image processing model, thereby enabling the target image processing model to accurately detect the target tumor. Therefore, the initial image processing model is trained using the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label, thereby obtaining a target image processing model capable of more accurately performing tumor detection. Subsequently, based on the target image processing model, accurate tumor detection results can be obtained during tumor detection on CT images.

[0247] The above is a schematic scheme of another image processing model training method of this embodiment. It should be noted that the technical scheme of this another image processing model training method and the technical scheme of the above-mentioned one image processing model training method are based on the same concept. For details not described in detail in the technical scheme of the another image processing model training method, please refer to the description of the technical scheme of the above-mentioned one image processing model training method.

[0248] See also Figure 6 , Figure 6 A flowchart of an image processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0249] Step 602: Determine an image of a target object to be detected located at a target location;

[0250] Step 604: Use the target image processing model to perform object detection on the image to be detected to obtain a detection result of the image to be detected, wherein the target image processing model is obtained by performing model training on the initial image processing model based on the object area information determined by the target training data and the enhanced training data, the predicted object area information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label. The target training data includes the sample image of the target object located at the target position, the sample label of the sample image, and the object area information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object area information is determined based on the detection result and the flipped detection result.

[0251] One or more embodiments of the present specification provide an image processing method. In order to obtain accurate detection results using a target image processing model, a large number of target training data and enhanced training data can be used to perform model training on the initial image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained through the object area information, predicted object area information, detection results, sample labels, flipped detection results and flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; then, object detection is performed on the image to be detected based on the target image processing model, thereby obtaining accurate detection results.

[0252] The above is a schematic scheme of an image processing method of this embodiment. It should be noted that the technical scheme of this image processing method and the technical scheme of the above-mentioned image processing model training method are of the same concept. For details not described in detail in the technical scheme of the image processing method, please refer to the description of the technical scheme of the above-mentioned image processing model training method.

[0253] See also Figure 7, Figure 7 A flowchart of another image processing method provided according to an embodiment of the present specification is shown. The another image processing method is applied to a client of a medical system and specifically includes the following steps.

[0254] Step 702: In response to a user's clicking operation on the display interface of the client, determining a medical image to be detected;

[0255] Step 704: Send the medical image to be detected to the server of the medical system, and receive the detection result returned by the server, wherein the detection result is the detection result output after object detection is performed on the medical image to be detected by the target image processing model, and the target image processing model is obtained by training the initial image processing model based on the object region information determined by the target training data and the enhanced training data, the predicted object region information of the target object in the sample image, the detection result, the sample label, the flip detection result, and the flip sample label. The target training data includes the sample image of the target object located at the target part, the sample label of the sample image, and the object region information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flip detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object region information is determined based on the detection result and the flip detection result.

[0256] Step 706: Display the detection result to the user through the display interface.

[0257] Another image processing method provided by one or more embodiments of the present specification, in order to achieve accurate detection results using a target image processing model, can use a large amount of target training data and enhanced training data to train the initial image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to train the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained using the object area information, the predicted object area information, the detection results, the sample labels, the flipped detection results and the flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; then, based on the target image processing model, object detection is performed on the image to be detected sent by the client, thereby obtaining an accurate detection result, and the detection result is returned to the client for display.

[0258] The above is a schematic scheme of another image processing method of this embodiment. It should be noted that the technical scheme of this another image processing method and the technical scheme of the above-mentioned one image processing model training method are based on the same concept. For details not described in detail in the technical scheme of the another image processing method, please refer to the description of the technical scheme of the above-mentioned one image processing model training method.

[0259] See also Figure 8 , Figure 8 A flowchart of an information processing method provided according to an embodiment of this specification is shown, which specifically includes the following steps.

[0260] Step 802: Determine the CT image to be detected of the target tumor located in the nasopharynx:

[0261] Step 804: Perform tumor detection on the CT image to be detected using a target image processing model to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined by target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, a tumor detection result, a sample label, a flipped tumor detection result, and a flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0262] Step 806: Determine an information processing result based on the tumor detection result of the CT image to be detected.

[0263] Among them, the information processing result can be understood as the information processing based on the tumor detection result. The obtained result, for example, the information processing result can be the regional information of the tumor, the tumor type, the radiotherapy strategy for the tumor, etc., which is not specifically limited here.

[0264] One or more embodiments of the present specification provide an information processing method. During information processing, a large amount of target training data and enhanced training data can be used to train an initial image processing model, thereby improving the accuracy of tumor detection results output by the target CT image processing model. Furthermore, during model training, to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on a sample CT image and a flipped sample CT image, respectively, to obtain a tumor detection result for the sample CT image and a flipped tumor detection result for the flipped sample CT image. Based on the tumor detection result and the flipped tumor detection result, predicted tumor region information of a target tumor in the sample CT image is determined. The tumor region information and the predicted tumor region information are then used to train the initial image processing model, thereby enabling the target image processing model to accurately detect the target tumor. Therefore, the initial image processing model is trained using the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label, thereby obtaining a target image processing model capable of more accurately detecting tumors. Subsequently, information processing is performed on the accurate tumor detection result output by the target image processing model, thereby obtaining an accurate information processing result.

[0265] The above is a schematic scheme of an information processing method of this embodiment. It should be noted that the technical scheme of this information processing method and the technical scheme of the above-mentioned image processing model training method are based on the same concept. For details not described in detail in the technical scheme of the information processing method, please refer to the description of the technical scheme of the above-mentioned image processing model training method.

[0266] One or more embodiments of this specification provide a computer-aided cancer diagnosis method, including:

[0267] Determine a CT image of a target tumor to be detected located at a target site;

[0268] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located at the target site, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0269] One or more embodiments of the present specification provide a method for computer-aided diagnosis of cancer. In the process of computer-aided diagnosis of cancer, a large number of target training data and enhanced training data can be used to train the initial image processing model, thereby improving the accuracy of the tumor detection results output by the target CT image processing model. In addition, in the process of model training, in order to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on the sample CT image and the flipped sample CT image respectively, to obtain the tumor detection results of the sample CT image and the flipped tumor detection results of the flipped sample CT image; and based on the tumor detection results and the flipped tumor detection results, determine The predicted tumor region information of the target tumor in the sample CT image is determined; then, the tumor region information and the predicted tumor region information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target tumor by the target image processing model; therefore, the initial image processing model is trained through the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, thereby obtaining a target image processing model that can perform tumor detection more accurately; subsequently, the target image processing model can be used to perform tumor detection on the CT image to be detected of the target tumor located at the target site, thereby outputting accurate tumor detection results for cancer.

[0270] The above is a schematic diagram of a computer-assisted cancer diagnosis method according to this embodiment. It should be noted that the technical solution of this computer-assisted cancer diagnosis method and the technical solution of the aforementioned image processing model training method share the same concept. For details not described in detail in the technical solution of the computer-assisted cancer diagnosis method, please refer to the description of the technical solution of the aforementioned image processing model training method.

[0271] One embodiment of this specification provides a computer-aided diagnosis method for nasopharyngeal carcinoma, comprising:

[0272] Determine the CT image of the target tumor located in the nasopharynx to be detected:

[0273] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by model training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0274] One or more embodiments of the present specification provide a computer-aided diagnosis method for nasopharyngeal carcinoma. During the process of computer-aided diagnosis of nasopharyngeal carcinoma, a large number of target training data and enhanced training data can be used to train the initial image processing model, thereby improving the accuracy of the tumor detection results output by the target CT image processing model. Moreover, during the model training process, in order to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on the sample CT image and the flipped sample CT image, respectively, to obtain the tumor detection results of the sample CT image and the flipped tumor detection results of the flipped sample CT image; and based on the tumor detection results and the flipped tumor detection results, determine The predicted tumor region information of the target tumor in the sample CT image is determined; then, the tumor region information and the predicted tumor region information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target tumor by the target image processing model; therefore, the initial image processing model is trained through the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, thereby obtaining a target image processing model that can perform tumor detection more accurately; subsequently, the target image processing model can be used to perform tumor detection on the CT image to be detected of the target tumor located in the nasopharynx, thereby outputting accurate tumor detection results for nasopharyngeal carcinoma.

[0275] The above is a schematic diagram of a computer-aided diagnosis method for nasopharyngeal carcinoma according to this embodiment. It should be noted that the technical solution of this computer-aided diagnosis method for nasopharyngeal carcinoma and the technical solution of the aforementioned image processing model training method are based on the same concept. For details not described in detail in the technical solution of the computer-aided diagnosis method for nasopharyngeal carcinoma, please refer to the description of the technical solution of the aforementioned computer-aided diagnosis model training method for nasopharyngeal carcinoma.

[0276] One embodiment of this specification provides a computer-aided diagnosis method for breast cancer, comprising:

[0277] Determine a CT image of a target tumor to be detected located in a breast area;

[0278] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by model training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the breast area, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0279] One or more embodiments of the present specification provide a method for computer-aided diagnosis of breast cancer. In the process of computer-aided diagnosis of breast cancer, a large amount of target training data and enhanced training data can be used to train the initial image processing model, thereby improving the accuracy of the tumor detection results output by the target CT image processing model. In addition, in the process of model training, in order to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on the sample CT image and the flipped sample CT image, respectively, to obtain the tumor detection result of the sample CT image and the flipped tumor detection result of the flipped sample CT image; and based on the tumor detection result and the flipped tumor detection result, determine The predicted tumor region information of the target tumor in the sample CT image is determined; then, the tumor region information and the predicted tumor region information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target tumor by the target image processing model; therefore, the initial image processing model is trained through the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, thereby obtaining a target image processing model that can perform tumor detection more accurately; subsequently, the target image processing model can be used to perform tumor detection on the CT image to be detected of the target tumor located in the breast area, thereby outputting accurate tumor detection results for breast cancer.

[0280] The above is a schematic diagram of a computer-assisted diagnosis method for breast cancer according to this embodiment. It should be noted that the technical solution of this computer-assisted diagnosis method for breast cancer shares the same concept as the technical solution of the aforementioned image processing model training method. For details not described in detail in the technical solution of the computer-assisted diagnosis method for breast cancer, please refer to the description of the technical solution of the aforementioned image processing model training method.

[0281] One embodiment of this specification provides a computer-aided diagnosis method for pelvic cancer, comprising:

[0282] Determine a CT image of a target tumor to be detected located in the pelvic region;

[0283] A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the pelvic region, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0284] One or more embodiments of the present specification provide a computer-aided diagnosis method for pelvic cancer. During the process of computer-aided diagnosis of pelvic cancer, a large number of target training data and enhanced training data can be used to train the initial image processing model, thereby improving the accuracy of the tumor detection results output by the target CT image processing model. Moreover, during the model training process, in order to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on the sample CT image and the flipped sample CT image, respectively, to obtain the tumor detection results of the sample CT image and the flipped tumor detection results of the flipped sample CT image; and based on the tumor detection results and the flipped tumor detection results, determine The predicted tumor region information of the target tumor in the sample CT image is determined; then, the tumor region information and the predicted tumor region information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target tumor by the target image processing model; therefore, the initial image processing model is trained through the tumor region information, the predicted tumor region information, the tumor detection results, the sample labels, the flipped tumor detection results and the flipped sample labels, thereby obtaining a target image processing model that can perform tumor detection more accurately; subsequently, the target image processing model can be used to perform tumor detection on the CT image to be detected of the target tumor located in the breast area, thereby outputting accurate tumor detection results for pelvic cancer.

[0285] The above is a schematic diagram of a computer-assisted diagnosis method for pelvic cancer according to this embodiment. It should be noted that the technical solution of this computer-assisted diagnosis method for pelvic cancer and the technical solution of the aforementioned image processing model training method are based on the same concept. For details not described in detail in the technical solution of the computer-assisted diagnosis method for pelvic cancer, please refer to the description of the technical solution of the aforementioned image processing model training method.

[0286] Corresponding to the above method embodiment, this specification also provides an embodiment of an image processing model training device, Figure 9 FIG. 1 shows a schematic diagram of the structure of an image processing model training device provided by an embodiment of this specification. Figure 9 As shown, the device includes:

[0287] The training data determination module 902 is configured to determine target training data and enhanced training data, wherein the target training data includes a sample image of a target object located at a target location, a sample label of the sample image, and object region information of the target object; and the enhanced training data includes a flipped sample image corresponding to the sample image and a flipped sample label of the flipped sample image.

[0288] The object detection module 904 is configured to perform object detection on the sample image and the flipped sample image using the initial image processing model, and obtain a detection result of the sample image and a flip detection result of the flipped sample image;

[0289] a region information determination module 906 configured to determine predicted object region information of the target object in the sample image based on the detection result and the flip detection result;

[0290] The model training module 908 is configured to perform model training on the initial image processing model based on the object area information, the predicted object area information, the detection results, the sample labels, the flip detection results and the flip sample labels to obtain a trained target image processing model.

[0291] Optionally, the model training module 908 is further configured to:

[0292] calculating a first loss function based on the object region information and the predicted object region information, wherein the first loss function is used to evaluate a difference between the object region information and the predicted object region information;

[0293] Calculating a second loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the second loss function is used to evaluate the difference between the detection result and the sample label, and the difference between the flip detection result and the flip sample label;

[0294] Calculating a third loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the third loss function is used to evaluate the similarity between the detection result and the sample label, and the similarity between the flip detection result and the flip sample label;

[0295] Based on the first loss function, the second loss function and the third loss function, the model parameters of the initial image processing model are adjusted to obtain the trained target image processing model.

[0296] Optionally, the area information determination module 906 is further configured to:

[0297] Using a nonlinear processing unit, performing data enhancement on the detection result to obtain an enhanced detection result, and performing data enhancement on the flip detection result to obtain an enhanced flip detection result;

[0298] Determining object region feature information corresponding to the target object contained in the enhanced detection feature, and determining flipped object region feature information of the target object contained in the enhanced flip detection result;

[0299] The object region feature information and the flipped object region feature information are compared to determine the predicted object region information of the target object in the sample image.

[0300] Optionally, the area information determination module 906 is further configured to:

[0301] Determining, using a nonlinear processing unit, an object feature of the target object in the detection result and a flipped object feature of the target object in the flip detection result, wherein the object feature includes a plurality of object feature points and the flipped object feature includes a plurality of flipped object feature points;

[0302] Determining corresponding feature points of each object feature point from the plurality of flipped object feature points, and determining similarities between each object feature point and the corresponding feature point;

[0303] When the similarity is greater than or equal to a first threshold, the similarity weight between the object feature point and the corresponding feature point is increased, and when the similarity is greater than or equal to the first threshold, the similarity weight between the object feature point and the corresponding feature point is reduced to obtain the enhanced detection result and the enhanced flip detection result.

[0304] Optionally, the training data determination module 902 is further configured to:

[0305] Determine the target training data:

[0306] horizontally flipping the sample image to obtain the flipped sample image, and horizontally flipping the sample label to obtain the flipped sample label of the flipped sample image;

[0307] The flipped sample image and the flipped sample label are used as the enhanced training sample.

[0308] Optionally, the object detection module 904 is further configured to:

[0309] Encoding the sample image and the flipped sample image respectively using an encoder in the initial image processing model to obtain a sample image encoding of the sample image and a flipped sample image encoding of the flipped sample image;

[0310] The sample image code and the flipped sample image code are decoded respectively by using the decoder in the initial image processing model to obtain the detection result of the sample image and the flip detection result of the flipped sample image.

[0311] Optionally, the training data determination module 902 is further configured to:

[0312] Determining an object image of the target object located at the target site, and determining an object image to be processed from the object image, wherein the object image to be processed is any image in the object image;

[0313] Based on the target part, a central axis is added to the image of the object to be processed to obtain an initial object image, wherein the central axis symmetrically divides the target part:

[0314] Based on the initial object image, adding the central axis to other object images in the object image except the object image to be processed to obtain a target object image;

[0315] Taking the initial object image and the target object image as the sample images, and determining a sample label corresponding to the sample image and object region information of the target object;

[0316] The sample image, the sample label, and the object region information are used as the target training data.

[0317] Optionally, the target site is a nasopharyngeal region, the target object is a target tumor, the sample image is a sample CT image, and the flipped sample image is a flipped sample CT image;

[0318] The model training module 908 is further configured to:

[0319] Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

[0320] One or more embodiments of the present specification provide an image processing model training device, which can use a large number of target training data and enhanced training data to perform model training on an initial image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on a sample image and a flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained through the object area information, predicted object area information, detection results, sample labels, flipped detection results and flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; subsequently, based on the target image processing model, accurate detection results can be obtained in the process of performing object detection on the image.

[0321] The above is a schematic diagram of an image processing model training device according to this embodiment. It should be noted that the technical solution of this image processing model training device and the technical solution of the aforementioned image processing model training method are based on the same concept. For details not described in detail in the technical solution of the image processing model training device, please refer to the description of the technical solution of the aforementioned image processing model training method.

[0322] Corresponding to the above method embodiment, this specification also provides another image processing model training device embodiment, Figure 10 FIG. 1 shows a structural diagram of another image processing model training device provided by an embodiment of this specification. Figure 10 As shown, the device includes:

[0323] The training data determination module 1002 is configured to determine target training data and enhanced training data, wherein the target training data includes a sample CT image of a target tumor located in the nasopharynx, a sample label of the sample CT image, and tumor region information of the target tumor; and the enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image.

[0324] The tumor detection module 1004 is configured to perform tumor detection on the sample CT image and the flipped sample CT image using the initial image processing model, and obtain a tumor detection result of the sample CT image and a flipped tumor detection result of the flipped sample CT image;

[0325] a region information determination module 1006 configured to determine predicted tumor region information of the target tumor in the sample CT image based on the tumor detection result and the flipped tumor detection result;

[0326] The model training module 1008 is configured to perform model training on the initial image processing model based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label to obtain a trained target image processing model.

[0327] Another image processing model training device provided by one or more embodiments of this specification can use a large amount of target training data and enhanced training data to train an initial image processing model, thereby improving the accuracy of tumor detection results output by the target CT image processing model. In addition, during the model training process, in order to further improve the accuracy of tumor detection, the initial image processing model can be used to perform tumor detection on a sample CT image and a flipped sample CT image, respectively, to obtain a tumor detection result for the sample CT image and a flipped tumor detection result for the flipped sample CT image; and based on the tumor detection result and the flipped tumor detection result, predicted tumor region information of a target tumor in the sample CT image is determined; then, the tumor region information and the predicted tumor region information are used to train the initial image processing model, thereby achieving accurate detection of the target tumor by the target image processing model; therefore, the initial image processing model is trained using the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label, thereby obtaining a target image processing model that can more accurately perform tumor detection; subsequently, based on the target image processing model, accurate tumor detection results can be obtained during the process of performing tumor detection on CT images.

[0328] The above is a schematic diagram of another image processing model training device according to this embodiment. It should be noted that the technical solution of this alternative image processing model training device and the technical solution of the aforementioned alternative image processing model training method are based on the same concept. For details not described in detail in the technical solution of the alternative image processing model training device, please refer to the description of the technical solution of the aforementioned alternative image processing model training method.

[0329] Corresponding to the above method embodiment, this specification also provides an image processing device embodiment, Figure 11 FIG. 1 shows a schematic diagram of the structure of an image processing device provided by an embodiment of this specification. Figure 11 As shown, the device includes:

[0330] An image determination module 1102 is configured to determine an image to be detected of a target object located at a target location;

[0331] The object detection module 1104 is configured to perform object detection on the image to be detected using a target image processing model to obtain a detection result of the image to be detected, wherein the target image processing model is obtained by performing model training on an initial image processing model based on the object area information determined by the target training data and the enhanced training data, the predicted object area information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label. The target training data includes the sample image of the target object located at the target position, the sample label of the sample image, and the object area information of the target object. The enhanced training data includes the flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object area information is determined based on the detection result and the flipped detection result.

[0332] One or more embodiments of the present specification provide an image processing device, which can use a large number of target training data and enhanced training data to perform model training on the initial image processing model in order to obtain accurate detection results using a target image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to perform model training on the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained through the object area information, predicted object area information, detection results, sample labels, flipped detection results and flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; then, object detection is performed on the image to be detected based on the target image processing model, thereby obtaining accurate detection results.

[0333] The above is a schematic diagram of an image processing device according to this embodiment. It should be noted that the technical solution of this image processing device and the technical solution of the above-mentioned image processing method are based on the same concept. For details not described in detail in the technical solution of the image processing device, please refer to the description of the technical solution of the above-mentioned image processing method.

[0334] Corresponding to the above method embodiment, this specification also provides another image processing device embodiment, Figure 12 FIG. 1 shows a schematic diagram of the structure of another image processing device provided by an embodiment of this specification. Figure 12 As shown, another image processing device is applied to a client of a medical system, and the device includes:

[0335] An image determination module 1202 is configured to determine a medical image to be detected in response to a user's clicking operation on the display interface of the client;

[0336] The object detection module 1204 is configured to send the medical image to be detected to the server of the medical system and receive the detection result returned by the server, wherein the detection result is the detection result output after the medical image to be detected is subjected to object detection by a target image processing model, and the target image processing model is obtained by training an initial image processing model based on the object region information determined by target training data and enhanced training data, the predicted object region information of the target object in the sample image, the detection result, the sample label, the flipped detection result, and the flipped sample label, the target training data includes the sample image of the target object located at the target site, the sample label of the sample image, and the object region information of the target object, the enhanced training data includes the flipped sample image corresponding to the sample image, and the flipped sample label of the flipped sample image, the detection result is determined by performing object detection on the sample image using the initial image processing model, the flipped detection result is determined by performing object detection on the flipped sample image using the initial image processing model, and the predicted object region information is determined based on the detection result and the flipped detection result;

[0337] The result display module 1206 is configured to display the detection result to the user through the display interface.

[0338] Another image processing device provided by one or more embodiments of the present specification can use a large number of target training data and enhanced training data to train the initial image processing model in order to obtain accurate detection results using a target image processing model, thereby improving the accuracy of the detection results output by the target image processing model. In addition, during the model training process, in order to further improve the accuracy of object detection, the initial image processing model can be used to perform object detection on the sample image and the flipped sample image respectively to obtain the detection results of the sample image and the flipped detection results of the flipped sample image; and based on the detection results and the flipped detection results, the predicted object area information of the target object in the sample image is determined; then, the object area information and the predicted object area information are used to train the initial image processing model, thereby achieving accurate detection of the target object by the target image processing model; therefore, the initial image processing model is trained by the object area information, the predicted object area information, the detection results, the sample labels, the flipped detection results and the flipped sample labels, thereby obtaining a target image processing model that can perform object detection more accurately; then, based on the target image processing model, object detection is performed on the image to be detected sent by the client, thereby obtaining an accurate detection result, and the detection result is returned to the client for display.

[0339] The above is a schematic diagram of another image processing device according to this embodiment. It should be noted that the technical solution of the alternative image processing device and the technical solution of the aforementioned alternative image processing method share the same concept. For details not described in detail in the technical solution of the alternative image processing device, please refer to the description of the technical solution of the aforementioned alternative image processing method.

[0340] Corresponding to the above method embodiment, this specification also provides an information processing device embodiment, Figure 13 FIG1 shows a schematic diagram of the structure of an information processing device provided by an embodiment of this specification. Figure 13 As shown, the device includes:

[0341] An image determination module 1302 is configured to determine a CT image to be detected of a target tumor located in the nasopharynx;

[0342] The object detection module 1304 is configured to perform tumor detection on the CT image to be detected using the target image processing model to obtain a tumor detection result of the CT image to be detected.

[0343] Wherein, the target image processing model is obtained by training the initial image processing model based on the tumor region information determined by the target training data and the enhanced training data, the predicted tumor region information of the target tumor in the sample CT image, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes the flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

[0344] The information processing module 1306 is configured to determine an information processing result according to the tumor detection result of the CT image to be detected.

[0345] One or more embodiments of the present specification provide an information processing device that, during information processing, can utilize a large amount of target training data and enhanced training data to train an initial image processing model, thereby improving the accuracy of tumor detection results output by the target CT image processing model. Furthermore, during model training, to further improve the accuracy of tumor detection, the initial image processing model can be utilized to perform tumor detection on a sample CT image and a flipped sample CT image, respectively, to obtain a tumor detection result for the sample CT image and a flipped tumor detection result for the flipped sample CT image. Based on the tumor detection result and the flipped tumor detection result, predicted tumor region information of a target tumor in the sample CT image is determined. The tumor region information and the predicted tumor region information are then used to train the initial image processing model, thereby enabling the target image processing model to accurately detect the target tumor. Therefore, the initial image processing model is trained using the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result, and the flipped sample label, thereby obtaining a target image processing model capable of more accurately performing tumor detection. Subsequently, information processing is performed on the accurate tumor detection result output by the target image processing model, thereby obtaining an accurate information processing result.

[0346] The above is a schematic diagram of an information processing device according to this embodiment. It should be noted that the technical solution of the information processing device and the technical solution of the above-mentioned information processing method are based on the same concept. For details not described in detail in the technical solution of the information processing device, please refer to the description of the technical solution of the above-mentioned information processing method.

[0347] Figure 14 14 shows a block diagram of a computing device 1400 according to one embodiment of the present disclosure. Components of the computing device 1400 include, but are not limited to, a memory 1410 and a processor 1420. The processor 1420 is connected to the memory 1410 via a bus 1430, and a database 1450 is used to store data.

[0348] The computing device 1400 also includes an access device 1440 that enables the computing device 1400 to communicate via one or more networks 1460. Examples of such networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 1440 may include one or more of any type of network interface (e.g., a network interface card (NIC)) whether wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, or a near field communication (NFC) interface.

[0349] In one embodiment of the present specification, the above components of the computing device 1400 and Figure 14 Other components not shown in the figure may also be connected to each other, for example, via a bus. Figure 14 The computing device structure block diagram shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art may add or replace other components as needed.

[0350] Computing device 1400 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, personal digital assistant, laptop computer, notebook computer, netbook computer, etc.), a mobile phone (e.g., a smartphone), a wearable computing device (e.g., a smartwatch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or personal computer (PC). Computing device 1400 may also be a mobile or stationary server.

[0351] The processor 1420 is configured to execute the following computer program / instruction, which implements the steps of any one of the above methods when executed by the processor.

[0352] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the computing device embodiment is generally similar to any of the aforementioned method embodiments, so its description is relatively simple. For relevant portions, refer to the partial description of any of the aforementioned method embodiments.

[0353] An embodiment of the present specification further provides a computer-readable storage medium storing a computer program / instruction, which implements the steps of any of the above methods when executed by a processor.

[0354] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the computer-readable storage medium embodiment is generally similar to any of the aforementioned method embodiments, so its description is relatively simple. For relevant portions, refer to the partial description of any of the aforementioned method embodiments.

[0355] An embodiment of the present specification further provides a computer program product, including a computer program / instruction, which implements the steps of any one of the above methods when executed by a processor.

[0356] The above is an illustrative solution of a computer program product of this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of any of the above methods are based on the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of any of the above methods.

[0357] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0358] The computer instructions include computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0359] It should be noted that for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0360] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0361] The preferred embodiments disclosed above are intended only to help illustrate this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made based on the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that those skilled in the art can better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for training an image processing model, comprising: Determine target training data and enhanced training data, wherein the target training data includes a sample image of a target object located at a target location, a sample label of the sample image, and object region information of the target object, and the enhanced training data includes a flipped sample image corresponding to the sample image and a flipped sample label of the flipped sample image; Using the initial image processing model, respectively perform object detection on the sample image and the flipped sample image to obtain a detection result of the sample image and a flip detection result of the flipped sample image; Determining predicted object region information of the target object in the sample image based on the detection result and the flip detection result; Based on the object area information, the predicted object area information, the detection result, the sample label, the flip detection result and the flip sample label, the initial image processing model is trained to obtain a trained target image processing model.

2. The image processing model training method according to claim 1, wherein the initial image processing model is trained based on the object region information, the predicted object region information, the detection result, the sample label, the flip detection result and the flip sample label to obtain a trained target image processing model, comprising: Calculating a first loss function based on the object region information and the predicted object region information, wherein the first loss function is used to evaluate the difference between the object region information and the predicted object region information; Calculating a second loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the second loss function is used to evaluate the difference between the detection result and the sample label, and the difference between the flip detection result and the flip sample label; Calculating a third loss function based on the detection result, the sample label, the flip detection result, and the flip sample label, wherein the third loss function is used to evaluate the similarity between the detection result and the sample label, and the similarity between the flip detection result and the flip sample label; Based on the first loss function, the second loss function and the third loss function, the model parameters of the initial image processing model are adjusted to obtain the trained target image processing model.

3. The image processing model training method according to claim 1, wherein the step of determining the predicted object region information of the target object in the sample image based on the detection result and the flip detection result comprises: Using a nonlinear processing unit, data enhancement is performed on the detection result to obtain an enhanced detection result, and data enhancement is performed on the flip detection result to obtain an enhanced flip detection result; Determining an object feature corresponding to the target object included in the enhanced detection feature, and determining a flip object feature of the target object included in the enhanced flip detection result; Comparing the object feature and the flipped object feature to determine an enhanced image feature corresponding to the sample image; The predicted object region information of the target object in the sample image is determined from the enhanced image features.

4. According to the image processing model training method of claim 3, the step of comparing the object features and the flipped object features to determine the enhanced image features corresponding to the sample image comprises: Using a nonlinear processing unit, determining that the object feature includes a plurality of object feature points, and the flipped object feature includes a plurality of flipped object feature points; Determine corresponding feature points of each object feature point from the plurality of flipped object feature points, and determine similarities between each object feature point and the corresponding feature point; When the similarity is greater than or equal to a first threshold, the similarity weight between the object feature point and the corresponding feature point is increased, and when the similarity is greater than or equal to the first threshold, the similarity weight between the object feature point and the corresponding feature point is reduced to obtain enhanced image features corresponding to the sample image.

5. The image processing model training method according to claim 1, wherein the step of determining target training data and enhancing training data comprises: Determining the target training data; Flipping the sample image horizontally to obtain the flipped sample image, and flipping the sample label horizontally to obtain the flipped sample label of the flipped sample image; The flipped sample image and the flipped sample label are used as the enhanced training sample.

6. The image processing model training method according to claim 1, wherein the initial image processing model is used to perform object detection on the sample image and the flipped sample image to obtain the detection result of the sample image and the flip detection result of the flipped sample image, comprising: Encoding the sample image and the flipped sample image respectively by using the encoder in the initial image processing model to obtain a sample image encoding of the sample image and a flipped sample image encoding of the flipped sample image; The sample image code and the flipped sample image code are decoded respectively by using the decoder in the initial image processing model to obtain the detection result of the sample image and the flip detection result of the flipped sample image.

7. According to the image processing model training method of claim 1 or 5, the determining the target training data comprises: Determine an object image of the target object located at the target part, and determine an object image to be processed from the object image, wherein the object image to be processed is any image in the object image; Based on the target part, adding a central axis to the object image to be processed to obtain an initial object image, wherein the central axis symmetrically divides the target part; Based on the initial object image, adding the central axis to other object images in the object image except the object image to be processed to obtain a target object image; Taking the initial object image and the target object image as the sample images, and determining a sample label corresponding to the sample image and object region information of the target object; The sample image, the sample label, and the object region information are used as the target training data.

8. The image processing model training method according to any one of claims 1 to 6, wherein the target site is the nasopharyngeal region, the target object is a target tumor, the sample image is a sample CT image, and the flipped sample image is a flipped sample CT image; The method of performing model training on the initial image processing model based on the object region information, the predicted object region information, the detection result, the sample label, the flip detection result, and the flip sample label to obtain a trained target image processing model includes: Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

9. A method for training an image processing model, comprising: Determining target training data and enhanced training data, wherein the target training data includes a sample CT image of a target tumor located in the nasopharynx, a sample label of the sample CT image, and tumor region information of the target tumor, and the enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image; Using the initial image processing model, respectively performing tumor detection on the sample CT image and the flipped sample CT image to obtain a tumor detection result of the sample CT image and a flipped tumor detection result of the flipped sample CT image; Determining predicted tumor region information of the target tumor in the sample CT image based on the tumor detection result and the flipped tumor detection result; Based on the tumor region information, the predicted tumor region information, the tumor detection result, the sample label, the flipped tumor detection result and the flipped sample label, the initial image processing model is trained to obtain a trained target image processing model.

10. An image processing method, comprising: Determine an image of a target object to be detected located at a target location; Perform object detection on the image to be detected using a target image processing model to obtain a detection result of the image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on object area information determined based on target training data and enhanced training data, predicted object area information of the target object in the sample image, detection results, sample labels, flip detection results, and flip sample labels. The target training data includes the sample image of the target object located at the target position, the sample label of the sample image, and the object area information of the target object. The enhanced training data includes a flipped sample image corresponding to the sample image and the flipped sample label of the flipped sample image. The detection result is determined by performing object detection on the sample image using the initial image processing model, and the flip detection result is determined by performing object detection on the flipped sample image using the initial image processing model. The predicted object area information is determined based on the detection result and the flip detection result.

11. An image processing method, applied to a client of a medical system, comprising: In response to a user's clicking operation on the display interface of the client, determining a medical image to be detected; Send the medical image to be detected to the server of the medical system, and receive the detection result returned by the server, wherein the detection result is the detection result output after object detection is performed on the medical image to be detected by a target image processing model, and the target image processing model is obtained by training an initial image processing model based on the object region information determined by target training data and enhanced training data, the predicted object region information of the target object in the sample image, the detection result, the sample label, the flip detection result, and the flip sample label, the target training data includes the sample image of the target object located at the target part, the sample label of the sample image, and the object region information of the target object, and the enhanced training data includes the flip sample image corresponding to the sample image and the flip sample label of the flip sample image, the detection result is determined by performing object detection on the sample image using the initial image processing model, the flip detection result is determined by performing object detection on the flip sample image using the initial image processing model, and the predicted object region information is determined based on the detection result and the flip detection result; The detection result is presented to the user through the presentation interface.

12. A method for computer-aided diagnosis of cancer, comprising: Determine a CT image to be detected of a target tumor located at a target site; A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located at the target site, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

13. A computer-aided diagnosis method for nasopharyngeal carcinoma, comprising: Determine the CT image of the target tumor to be detected located in the nasopharynx; A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the nasopharynx, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

14. A computer-aided diagnosis method for breast cancer, comprising: Determine a CT image of a target tumor to be detected located in a breast area; A target image processing model is used to perform tumor detection on the CT image to be detected to obtain a tumor detection result of the CT image to be detected, wherein the target image processing model is obtained by training an initial image processing model based on tumor region information determined based on target training data and enhanced training data, predicted tumor region information of the target tumor in the sample CT image, tumor detection result, sample label, flipped tumor detection result, and flipped sample label. The target training data includes the sample CT image of the target tumor located in the breast area, the sample label of the sample CT image, and the tumor region information of the target tumor. The enhanced training data includes a flipped sample CT image corresponding to the sample CT image and the flipped sample label of the flipped sample CT image. The tumor detection result is determined by performing tumor detection on the sample CT image using the initial image processing model. The flipped tumor detection result is determined by performing tumor detection on the flipped sample CT image using the initial image processing model. The predicted tumor region information is determined based on the tumor detection result and the flipped tumor detection result.

15. A computing device comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method described in any one of claims 1 to 14 are implemented.

16. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 14.

17. A computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 14.

Images