CT gastric cancer prediction method and system based on adaptive diffusion algorithm

Abstract

The invention belongs to the technical field of medical image analysis, and particularly relates to a CT gastric cancer prediction method and system based on an adaptive diffusion algorithm. CT image data of a patient with the advanced gastric cancer treated by the PD1 drug and a focus area mask corresponding to CT are obtained and serve as training data; designing a self-adaptive mask algorithm based on a diffusion model; carrying out denoising and data matching on the training data by utilizing an algorithm from the perspective of image quality; accurately extracting image feature information of a focus region from the denoised training data; screening the extracted image feature information based on the feature information, and calculating influence scores of all features based on a Lasso method; on the basis of the score of the influence, the score corresponds to each feature, and top features are used for prediction, so that CT gastric cancer prediction based on the adaptive diffusion algorithm is realized. The method is used for solving the problems that in the prior art, mask information and a CT image cannot be deeply fused, in addition, interference of artifacts, environments and other factors exists, and most models lack flexible adjustment mechanisms and self-adaptive capacity.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical image analysis, and in particular relates to a CT gastric cancer prediction method and system based on an adaptive diffusion algorithm. Background Art

[0002] Gastric cancer (GC) is the second leading cause of cancer-related death. Most patients already have progressive disease or metastasis when first diagnosed. Most patients in China are diagnosed at an advanced stage, and surgical resection is the mainstay of treatment for GC. For patients with advanced GC, even after resection, the prognosis remains poor, with approximately 20% of patients experiencing recurrence within one year after resection. In recent years, research on immunotherapy has advanced traditional concepts and approaches. Treatment targeting PD1 has demonstrated breakthrough efficacy, and anti-PD1 monoclonal antibodies have been approved as a first-line treatment for advanced GC in the entire population.

[0003] Programmed cell death (PD1) inhibitors have shown breakthrough efficacy in the treatment of gastric cancer. Currently, PD1 drugs have been approved as first-line treatment for advanced gastric cancer. Our goal is to develop a deep learning model to predict and evaluate the therapeutic effect and prognosis of PD1 inhibitors in patients with advanced gastric cancer. The data in this study come from patients with advanced gastric cancer treated with PD1 inhibitors. Existing CT image resources for advanced gastric cancer patients treated with PD1 drugs are scarce. Most existing methods for denoising CT images are based on convolutional neural networks (CNNs), residual networks, etc. Although some progress has been made, existing methods fail to deeply fuse mask information and CT images. In addition, most models lack flexible adjustment mechanisms and adaptive capabilities due to interference from artifacts and environmental factors. Summary of the invention

[0004] The present invention provides a CT gastric cancer prediction method and system based on an adaptive diffusion algorithm, which are used to solve the problems in the prior art that mask information and CT images cannot be deeply integrated, and most models lack flexible adjustment mechanisms and adaptive capabilities due to interference from artifacts and environmental factors.

[0005] The present invention is achieved through the following technical solutions:

[0006] A CT gastric cancer prediction method based on an adaptive diffusion algorithm, the method comprising the following steps:

[0007] Step 1: Obtain CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data;

[0008] Step 2: Design an adaptive masking algorithm based on the diffusion model;

[0009] Step 3: For the training data in step 1, use the algorithm in step 2 to perform denoising and data matching from the perspective of image quality;

[0010] Step 4: De-noise the training data in step 3 and accurately extract the image feature information of the lesion area;

[0011] Step 5: Filter the extracted image feature information based on the feature information in step 4, and calculate the influence scores of all features based on the Lasso method;

[0012] Step 6: Based on the scores of step 5, make them correspond to each feature, use the top features for prediction, and realize CT gastric cancer prediction based on the adaptive diffusion algorithm.

[0013] Furthermore, in step 1, the training data is the maximum cross-section of the patient's stomach three-dimensional data, and the mask is the lesion area manually marked by the doctor.

[0014] Furthermore, the step 2 specifically includes the following steps:

[0015] Step 2.1: Use the convolution algorithm to extract the features of the image and mask;

[0016] Step 2.2: The algorithm sets the diffusion intensity according to the current diffusion step;

[0017] Step 2.3: Output the current noise intensity according to the diffusion step and merge it into the generated noise;

[0018] Step 2.4: Use the noise intensity in step 2.3 and the generated noise to form a new noise, and use it to set the diffusion intensity in step 2.2;

[0019] Step 2.5: Update the mask for the next stage by setting the diffusion intensity of the new noise set in step 2.4.

[0020] Furthermore, the process of step 2.1 is:

[0021]

[0022] Among them, F(i, j) is the value of the output feature map at position (i, j), X is the image with a size of 512*512, K is the convolution kernel, and b is the bias term.

[0023] Step 2.2: The diffusion intensity is:

[0024]

[0025] Among them, σ max is the maximum value of the noise intensity, σ min is the minimum value of the noise intensity, t is the current diffusion step, and T is the total diffusion steps;

[0026] Step 2.3: The current noise intensity is:

[0027]

[0028] Step 2.4: New noise:

[0029] M t '=M t +n t

[0030] The step 2.5: the next stage mask is specifically,

[0031] y t '=X t ·M t '+X t ·(1-M′ t )·α

[0032] Among them, α is a parameter, X t It is CT image data.

[0033] Furthermore, the step 4 specifically includes inputting the training data into a ternary domain attention module embedded behind the extractor to accurately extract feature information;

[0034] The ternary domain attention module is based on the Resnet50 framework, which contains a 50-layer deep structure and extracts multi-level features of the image through residual learning. It is first mapped to the new feature space through the convolutional layer, and then nonlinearity is introduced through the ReLU activation function. Then, the channel feature values ​​are compressed through the global average pooling layer, and the attention weights are calculated through the fully connected layer. At the end of the network, the KQV three domains are used to suppress unimportant information in the original feature map and emphasize the image feature information of the lesion area, thereby extracting image information with salient features.

[0035] Furthermore, the influence score of step 5 is specifically:

[0036]

[0037] Among them, Y is the target variable, X is the feature matrix, δ is the feature influence score, and θ is the regularization parameter that controls the speed of regularization.

[0038] Furthermore, the step 6 is specifically to rank the scores and select the top_k features for prediction:

[0039]

[0040] Among them, Xi is the image feature, β0 is the initial matrix, p is the number of features after screening, is the influence score and b is the bias.

[0041] A CT gastric cancer prediction system based on an adaptive diffusion algorithm, the system uses the CT gastric cancer prediction method based on an adaptive diffusion algorithm as described above, the system comprising:

[0042] Data acquisition module: obtains CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data;

[0043] Algorithm design module: Design of adaptive mask algorithm based on diffusion model;

[0044] Training module: for the training data in step 1, use the algorithm in step 2 to perform denoising and data matching from the perspective of image quality;

[0045] Feature extraction module: Use the denoised training data to accurately extract the texture features and directional features of the lesion area;

[0046] Influence score calculation module: based on the feature information in step 4, the extracted features are screened again, and the influence scores of all features are calculated based on the Lasso method;

[0047] Prediction module: Based on the scores in step 5, the scores are matched to each feature, and the top features are used for prediction to achieve CT gastric cancer prediction based on the adaptive diffusion algorithm.

[0048] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the above method is implemented.

[0049] A computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0050] The beneficial effects of the present invention are:

[0051] The present invention enhances the visibility of key areas, making tumor features clearer, helping doctors make accurate diagnoses, and providing a more reliable basis for formulating treatment plans.

[0052] The optimized feature extraction and sparse processing of the present invention enables the model to capture key information more accurately, thereby more accurately predicting the survival of patients with advanced gastric cancer and providing strong support for clinical decision-making.

[0053] The present invention provides an efficient and accurate tool for predicting the survival of patients with advanced gastric cancer, which helps doctors better assess the condition, optimize the allocation of treatment resources, and improve patient prognosis. It has important clinical significance and social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a structural schematic diagram of the present invention.

[0055] Figure 2 The present invention uses a scatter plot to perform importance analysis and display, and a schematic diagram of importance impact ranking of the screened features.

[0056] Figure 3 It is a schematic diagram of the change of the survival probability predicted by SBF with the survival time (OS) of the two groups of patients in the present invention.

[0057] Figure 4 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0058] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0059] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0060] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0061] The following is attached to this application specification Figure 1-4 , the technical solutions in the embodiments of the present application are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0063] The present invention fuses the mask and the CT image, thereby greatly improving the quality of the original CT image in terms of resolution, and performs high-precision denoising in the lesion area, thereby improving image visualization in medical CT diagnosis, and accurately extracting image features in model prediction analysis of downstream tasks to perform downstream prediction tasks.

[0064] Implementation Method 1

[0065] This embodiment provides a CT gastric cancer prediction method based on an adaptive diffusion algorithm, which fuses the mask and the CT image, thereby greatly improving the quality of the original CT image in terms of resolution, and performing high-precision denoising in the lesion area, thereby improving image visualization in medical CT diagnosis, and accurately extracting image features in model prediction analysis of downstream tasks, and performing downstream prediction tasks. The method includes the following steps:

[0066] Step 1: Obtain CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data;

[0067] Step 2: Design an adaptive masking algorithm based on the diffusion model;

[0068] Step 3: For the training data in step 1, use the algorithm in step 2 to perform denoising and data matching from the perspective of image quality;

[0069] From the perspective of image quality, the algorithm is embedded into the diffusion model, and the mask information is matched and fused with the CT image at different intensities to achieve denoising;

[0070] Step 4: The training data after denoising in step 3 is used to accurately extract the image feature information of the lesion area; the denoised CT image needs to be subjected to accurate feature extraction methods, such as texture features and directional features, to extract feature information;

[0071] Step 5: Filter the extracted image feature information based on the feature information in step 4, and calculate the influence scores of all features based on the Lasso method;

[0072] Step 6: Based on the scores of step 5, make them correspond to each feature, use the top features for prediction, and realize CT gastric cancer prediction based on the adaptive diffusion algorithm.

[0073] Furthermore, in step 1, the training data is the maximum cross-section of the patient's stomach three-dimensional data, and the mask is the lesion area manually marked by the doctor.

[0074] Furthermore, the step 2 specifically includes the following steps:

[0075] Step 2.1: Use the convolution algorithm to extract the features of the image and mask;

[0076] Step 2.2: The algorithm sets the diffusion intensity according to the current diffusion step;

[0077] Step 2.3: Output the current noise intensity according to the diffusion step and merge it into the generated noise;

[0078] Step 2.4: Use the noise intensity in step 2.3 and the generated noise to form a new noise, and use it to set the diffusion intensity in step 2.2;

[0079] Step 2.5: Update the mask for the next stage by setting the diffusion intensity of the new noise set in step 2.4.

[0080] Furthermore, the process of step 2.1 is:

[0081]

[0082] Among them, F(i, j) is the value of the output feature map at position (i, j), X is the image with a size of 512*512, K is the convolution kernel, and b is the bias term.

[0083] Step 2.2: The diffusion intensity is:

[0084]

[0085] Among them, σ max is the maximum value of the noise intensity, σ min is the minimum value of the noise intensity, t is the current diffusion step, and T is the total diffusion steps;

[0086] Step 2.3: The current noise intensity is:

[0087]

[0088] Step 2.4: The noise intensity and the generated noise are fused into new noise:

[0089] M t '=M t +n t

[0090] The step 2.5: the next stage mask is specifically,

[0091] y t'=X t ·M t '+X t ·(1-M′ t )·α

[0092] Among them, α is a parameter, X t It is CT image data.

[0093] Furthermore, the step 4 specifically includes inputting the training data into a ternary domain attention module embedded behind the extractor to accurately extract feature information;

[0094] The ternary domain attention module is based on the Resnet50 framework and contains a 50-layer deep structure. It solves the gradient problem in deep network training through residual learning and extracts multi-level features of the image. It is first mapped to the new feature space through the convolution layer, and then nonlinearity is introduced through the ReLU activation function. Then, the channel feature values ​​are compressed through the global average pooling layer, and the attention weights are calculated through the fully connected layer. At the end of the network, the KQV three domains are used to suppress unimportant information in the original feature map and emphasize the image feature information of the lesion area, thereby extracting image information with salient features.

[0095] Furthermore, the influence score of step 5 is specifically:

[0096]

[0097] Among them, Y is the target variable, X is the feature matrix, δ is the feature influence score, and θ is the regularization parameter that controls the speed of regularization.

[0098] Furthermore, the step 6 is specifically to rank the scores and select the top_k features for prediction:

[0099]

[0100] Among them, Xi is the image feature, β0 is the initial matrix, p is the number of features after screening, is the influence score and b is the bias.

[0101] In recent years, the survival prediction of patients receiving PD-1 inhibitors has received widespread attention. Existing diffusion models generally only blur the key lesion areas, and the mask and CT image are poorly matched during the sampling process, resulting in a high risk of misdiagnosis. The method for predicting survival in advanced gastric cancer proposed in this study has many beneficial effects and functions. First, the adaptive mask diffusion model can better guide the CT diffusion process under different noise intensities by adaptively adding noise to the mask according to the diffusion intensity, thereby improving the quality of CT images. The algorithm module further enhances the visibility of key areas (such as gastric cancer lesion areas) and effectively highlights the edge and texture information of the tumor, which is crucial for the accurate diagnosis and evaluation of gastric cancer.

[0102] In terms of feature extraction, this study introduced a mechanism based on dot product operations and used a scaling factor to normalize the similarity scores. This mechanism not only avoids the problem of gradient vanishing or exploding, but also improves the stability of the model and the training efficiency. In addition, the innovative method based on the sparsification principle guides model optimization through regularization technology and forces the model parameters to develop in the direction of sparsification. This method not only reduces the complexity of the model and prevents overfitting, but also improves the generalization ability and computational efficiency of the model.

[0103] Overall, this method provides an efficient, accurate and robust solution for survival prediction of patients with advanced gastric cancer by combining algorithms, three-domain feature mapping, sparse optimization and survival prediction. This not only helps doctors to more accurately assess the patient's condition and prognosis, but also provides strong support for clinical treatment decisions, thereby improving the patient's treatment effect and quality of life.

[0104] Specifically, the method was verified in a real data set of patients with advanced PD1 gastric cancer from Harbin Cancer Hospital. The accuracy of this method is relatively high, and the C-index indicator has exceeded that of traditional methods. Experiments in visualization and importance scores have demonstrated the superiority of this method.

[0105]

[0106] This table shows the C-index (consistency index) results of different models in predicting the survival period of gastric cancer patients. C-index is an important indicator to measure the prediction accuracy of the model. The higher the value, the better the prediction performance of the model. The values ​​in the table represent the C-index of each model under different experimental settings. The RSF indicator C-index is between 0.5749 and 0.6135, which is relatively low, indicating that the accuracy of the model in predicting the survival period of gastric cancer is limited. This article uses deep learning to extract features from the CT of special patients after processing them with a mask-based diffusion model, and then performs the best performance of the model with a C-index between 0.9403 and 0.9539, combining the advantages of our model and showing excellent performance in both feature selection and prediction accuracy.

[0107] Figure 1 The image shows a series of images in the process of gradual denoising, from the initial high-noise image to the final high-quality and clear image. The image on the far left is the image with the highest noise level, where most of the information is masked by noise and it is difficult to discern the details of the original image. As the image changes to the right, the noise gradually decreases, and the middle images show the different stages of the noise removal process. In these middle images, the image details gradually emerge, the originally blurred structures become clearer, and the contours and features gradually emerge. These images represent different time steps in the denoising process, and each step is closer to the final high-quality image than the previous step. The image on the far right is the final output image, with clear details, higher image quality, and more fully displayed details. This series of images shows the transformation effect from high-noise images to high-quality images, reflecting the significant effect of diffusion models in image restoration. By comparing these images, you can clearly see how the model gradually improves the image quality and finally produces a clear and detailed image.

[0108] Figure 2 As shown in the figure, the importance analysis is performed using a scatter plot to rank the selected features by importance. The model evaluates the distribution of important features, and the importance of the features is measured by the average reduction. It can be seen that the selected important features have a key impact on the model's prediction and may also contain the most valuable information for survival prediction. The top k features are relatively important, indicating that these features play an important role in the process of survival prediction.

[0109] In order to evaluate the survival difference between the high-risk group and the low-risk group, the present invention conducted 5 rounds of cross-validation, and selected one round to demonstrate the predictive ability of our proposed model using the Kaplan-Meier (KM) survival analysis method. First, all patients were divided into a high-risk group and a low-risk group, and then the changes in the survival probability predicted by SBF with the survival time (OS) of the two groups of patients were demonstrated. Figure 3 The red line in the figure represents the high-risk group, and the blue line represents the low-risk group. The table below shows the number of people at risk in the two groups at each time point. Over time, the survival probability of the low-risk group (blue line) was significantly higher than that of the high-risk group (red line). This indicates that the survival of patients in the low-risk group was significantly better than that of patients in the high-risk group during the entire follow-up period. The survival difference between the two groups was statistically significant (usually a p value less than 0.05 is considered statistically significant). This result shows that risk grouping has a significant effect on patient survival, and the prognosis of the high-risk group is significantly worse. The Kaplan-Meier survival curve clearly shows the significant survival difference between the high-risk group and the low-risk group, further proving the effectiveness of our risk score-based grouping method in prognostic assessment. This result provides strong support for the application of this risk scoring system in clinical practice in the future.

[0110] Implementation Method 2

[0111] This embodiment provides a CT gastric cancer prediction system based on an adaptive diffusion algorithm. The system uses the CT gastric cancer prediction method based on an adaptive diffusion algorithm as described in Embodiment 1. The system includes:

[0112] Data acquisition module: obtains CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data;

[0113] Algorithm design module: Design of adaptive mask algorithm based on diffusion model;

[0114] Training module: for the training data in step 1, use the algorithm in step 2 to perform denoising and data matching from the perspective of image quality;

[0115] From the perspective of image quality, the algorithm is embedded into the diffusion model, and the mask information is matched and fused with the CT image at different intensities to achieve denoising;

[0116] Feature extraction module: Use the denoised training data to accurately extract the texture features and directional features of the lesion area; the denoised CT image needs to be accurately extracted using a feature extraction method to extract feature information;

[0117] Influence score calculation module: based on the feature information in step 4, the extracted features are screened again, and the influence scores of all features are calculated based on the Lasso method;

[0118] Prediction module: Based on the scores in step 5, the scores are matched to each feature, and the top features are used for prediction to achieve CT gastric cancer prediction based on the adaptive diffusion algorithm.

[0119] Implementation Method 3

[0120] An embodiment of the present invention provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the memory is used to store software programs and modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory and the processor are connected via a bus. Specifically, the processor implements any step in the first embodiment above by running the computer program stored in the memory.

[0121] It should be understood that in the embodiments of the present invention, the processor referred to may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0122] The memory may include a read-only memory, a flash memory, and a random access memory, and provides instructions and data to the processor. A part or all of the memory may also include a nonvolatile random access memory.

[0123] As can be seen from the above, the electronic device provided by the embodiment of the present invention can implement the CT gastric cancer prediction method based on the adaptive diffusion algorithm as described in the first embodiment by running a computer program, and the mask and CT image are fused and denoised at different intensities on the diffusion model, because the mask auxiliary information indicates the lesion area of ​​CT, so that high-precision denoising can be achieved on the lesion area. In the feature extraction stage, the ternary domain is embedded behind the feature extractor to accurately extract the features of the lesion area, and then the feature screening technology is used to screen the extracted features for influential features, so as to screen out high-quality features for downstream prediction tasks.

[0124] It should be understood that if the above-mentioned integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned implementation method, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method implementations when executed by the processor. Among them, the above-mentioned computer program includes computer program code, and the above-mentioned computer program code can be in source code form, object code form, executable file or some intermediate form. The above-mentioned computer-readable medium may include: any entity or device capable of carrying the above-mentioned computer program code, recording medium, U disk, mobile hard disk, 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 above-mentioned computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0125] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest range consistent with the principles and novel features disclosed herein.

[0126] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the above-mentioned device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the implementation method can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method implementation method, which will not be repeated here.

[0127] It should be noted that the methods and detailed examples provided in the above embodiments can be combined with the devices and equipment provided in the embodiments, and references can be made to each other, and no further details will be given.

[0128] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0129] In the embodiments provided by the present invention, it should be understood that the disclosed device / terminal equipment and method can be implemented in other ways. For example, the device / equipment implementation described above is only illustrative, for example, the division of the above modules or units is only a logical function division, and in actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0130] The above-mentioned embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above-mentioned embodiments, or replace some of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A CT gastric cancer prediction method based on an adaptive diffusion algorithm, characterized in that: The method comprises the following steps: Step 1: Obtain CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data; Step 2: Design an adaptive masking algorithm based on the diffusion model; Step 3: For the training data in step 1, use the algorithm in step 2 to perform denoising and data matching from the perspective of image quality; Step 4: De-noise the training data in step 3 and accurately extract the image feature information of the lesion area; Step 5: Filter the extracted image feature information based on the feature information in step 4, and calculate the influence scores of all features based on the Lasso method; Step 6: Based on the scores of step 5, make them correspond to each feature, use the top features for prediction, and realize CT gastric cancer prediction based on the adaptive diffusion algorithm.

2. The CT gastric cancer prediction method according to claim 1, characterized in that: In step 1, the training data is the maximum cross-section of the patient's stomach three-dimensional data, and the mask is the lesion area manually marked by the doctor.

3. The CT gastric cancer prediction method according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1: Use the convolution algorithm to extract the features of the image and mask; Step 2.2: The algorithm sets the diffusion intensity according to the current diffusion step; Step 2.3: Output the current noise intensity according to the diffusion step and merge it into the generated noise; Step 2.4: Use the noise intensity in step 2.3 and the generated noise to form a new noise, and use it to set the diffusion intensity in step 2.2; Step 2.5: Update the mask for the next stage by setting the diffusion intensity of the new noise set in step 2.

4.

4. The CT gastric cancer prediction method according to claim 3, characterized in that: The process of step 2.1 is: Among them, F(i, j) is the value of the output feature map at position (i, j), X is the image with a size of 512*512, K is the convolution kernel, and b is the bias term. Step 2.2: The diffusion intensity is: Among them, σ max is the maximum value of the noise intensity, σ min is the minimum value of the noise intensity, t is the current diffusion step, and T is the total diffusion steps; Step 2.3: The current noise intensity is: Step 2.4: New noise: M’ t =M t +n t The step 2.5: the next stage mask is specifically, y t ’=X t ·M t ‘+X t ·(1-M′t)·α Among them, α is a parameter, x t It is CT image data.

5. The CT gastric cancer prediction method according to claim 1, characterized in that: Specifically, step 4 includes inputting the training data into the ternary domain attention module embedded behind the extractor to accurately extract feature information; The ternary domain attention module is based on the Resnet50 framework, which contains a 50-layer deep structure and extracts multi-level features of the image through residual learning. It is first mapped to the new feature space through the convolutional layer, and then nonlinearity is introduced through the ReLU activation function. Then, the channel feature values ​​are compressed through the global average pooling layer, and the attention weights are calculated through the fully connected layer. At the end of the network, the KQV three domains are used to suppress unimportant information in the original feature map and emphasize the image feature information of the lesion area, thereby extracting image information with salient features.

6. The CT gastric cancer prediction method according to claim 1, characterized in that: The influence score of step 5 is specifically: Among them, Y is the target variable, X is the feature matrix, δ is the feature influence score, and θ is the regularization parameter that controls the speed of regularization.

7. The CT gastric cancer prediction method according to claim 6, characterized in that: Specifically, step 6 is to rank the scores and select the top k features for prediction: Among them, Xi is the image feature, β0 is the initial matrix, p is the number of features after screening, is the influence score and b is the bias.

8. A CT gastric cancer prediction system based on an adaptive diffusion algorithm, characterized in that: The system uses the CT gastric cancer prediction method based on the adaptive diffusion algorithm as described in any one of claims 1 to 7, and the system includes: Data acquisition module: obtains CT image data of patients with advanced gastric cancer treated with PD1 drugs and the lesion area mask corresponding to CT as training data; Algorithm design module: Design of adaptive mask algorithm based on diffusion model; Training module: For the training data in step 1, use the algorithm in step 2 to perform denoising and data alignment from the perspective of image quality; Feature extraction module: Use the denoised training data to accurately extract the texture features and directional features of the lesion area; Influence score calculation module: based on the feature information in step 4, the extracted features are screened again, and the influence scores of all features are calculated based on the Lasso method; Prediction module: Based on the scores in step 5, the scores are matched to each feature, and the top features are used for prediction to achieve CT gastric cancer prediction based on the adaptive diffusion algorithm.

9. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images