Tumor prediction method and apparatus based on bacterial data

Abstract

The present application relates to the technical field of artificial intelligence, and discloses a tumor prediction method and device based on bacterial data, which comprises the following steps: obtaining a historical tumor image set, performing stage classification on the historical tumor image set based on a fuzzy clustering algorithm to obtain a tumor image classification result, obtaining a bacterial data set corresponding to the historical tumor image set, performing joint labeling on the historical tumor image set based on the tumor image classification result and the bacterial data set to obtain a tumor training set, constructing a tumor prediction network comprising a mixed attention mechanism and a capsule network, performing attention mechanism joint training on the tumor prediction network by using the tumor training set to obtain a standard tumor prediction network, obtaining a to-be-detected tumor image containing bacterial labeling, and outputting a tumor prediction result of the to-be-detected tumor image by using the standard tumor prediction network. The present application can solve the problem of low tumor prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for tumor prediction based on bacterial data. Background Technology

[0002] With the development of artificial intelligence, technologies such as image recognition have been applied to various fields. For example, deep learning has been successfully applied to the field of medical diagnosis, and predicting tumor images using deep learning can effectively improve diagnostic accuracy. However, because tumor images have rich features, traditional convolutional neural networks tend to ignore key locations of lesions in the image. At the same time, tumor prediction relies solely on tumor images, ignoring other tumor-related data (such as bacterial data), resulting in low accuracy in tumor prediction. Summary of the Invention

[0003] This invention provides a tumor prediction method, device, electronic device, and computer-readable storage medium based on bacterial data, with the main purpose of solving the problem of low accuracy in tumor prediction.

[0004] To achieve the above objectives, the present invention provides a tumor prediction method based on bacterial data, comprising: A set of historical tumor images is obtained, and the set of historical tumor images is classified in stages based on a fuzzy clustering algorithm to obtain the tumor image classification results. Obtain the bacterial data set corresponding to the historical tumor image set, and perform joint annotation on the historical tumor image set based on the tumor image classification results and the bacterial data set to obtain a tumor training set; Construct a tumor prediction network that incorporates a hybrid attention mechanism and a capsule network; The tumor prediction network is jointly trained using the tumor training set with an attention mechanism to obtain a standard tumor prediction network. Obtain an image of the tumor to be detected that includes bacterial annotations, and output the tumor prediction result of the tumor image using the standard tumor prediction network.

[0005] Optionally, the step of performing stage classification on the historical tumor image set based on the fuzzy clustering algorithm to obtain tumor image classification results includes: The images in the historical tumor image set are subjected to spatial geometric transformation and color conversion to obtain the original image set; Each image in the original image set is divided into regions to obtain a preset number of sub-regions; Calculate the membership degree of pixels in the sub-region, and calculate the target value of the image in the original image set based on the membership degree; If the target value is greater than the preset target threshold, then each image in the original image set is re-divided into regions, and the step of calculating the membership degree of pixels in the sub-region is returned until the target value is less than or equal to the target threshold. Then, the images in the original image set are divided into different stages according to the cluster centers in the sub-regions to obtain the tumor image classification result.

[0006] Optionally, calculating the membership degree of pixels in the sub-region and calculating the target value of the image in the original image set based on the membership degree includes: Calculate the cluster center of the sub-region, and calculate the membership degree of the pixels in the sub-region based on the cluster center; The target value of the image in the original image set is calculated based on the preset objective function and the membership degree.

[0007] Optionally, the joint annotation of the historical tumor image set based on the tumor image classification results and the bacterial dataset includes: The tumor image classification results are used as the initial annotations for the images in the historical tumor image set; The bacterial data in the bacterial dataset is used as the associated annotation for the images in the historical tumor image dataset; Calculate the initial probability distribution of the initial labels and the association probability distribution of the associated labels; The initial probability distribution and the associated probability distribution are labeled and fused to obtain a joint label set, and a preset number of joint labels are selected from the joint label set as image labels for the historical tumor image set.

[0008] Optionally, the construction of the tumor prediction network, which includes a hybrid attention mechanism and a capsule network, includes: By splicing together a preset spatial attention module and a preset channel attention module, a hybrid attention mechanism unit is obtained; Obtain a convolutional network including multiple convolutional layers, and add the hybrid attention mechanism unit after each convolutional layer of the convolutional network to obtain the original prediction network; A capsule layer and a fully connected layer are added after the original prediction network to obtain the tumor prediction network.

[0009] Optionally, the splicing of a preset spatial attention module and a preset channel attention module yields a hybrid attention mechanism unit, comprising: The spatial attention module (SAM) and the channel attention module (CAM) are connected in series to obtain the first concatenated network. The channel attention module CAM and the channel attention module GCT are connected in series to obtain the second series network; The first serial network and the second serial network are connected in parallel to obtain the hybrid attention mechanism unit.

[0010] Optionally, the step of jointly training the tumor prediction network using the tumor training set with an attention mechanism to obtain a standard tumor prediction network includes: Features of images in the tumor training set are extracted using the convolutional layers in the tumor prediction network. The features are fused using the hybrid attention mechanism unit in the tumor prediction network to obtain hybrid features; The weighted probabilities of the final mixed features are output using the capsule layer in the tumor prediction network. The predicted labels of the weight probabilities are output using the fully connected layers in the tumor prediction network; Construct a confusion matrix based on the predicted annotations and the initial annotations, and calculate the prediction accuracy based on the confusion matrix. If the prediction accuracy is less than or equal to a preset accuracy threshold, adjust the parameters of each layer in the tumor prediction network, return to the step of extracting features of the images in the tumor training set using the convolutional layers in the tumor prediction network, until the prediction accuracy is greater than the accuracy threshold, stop training, and obtain the standard tumor prediction network.

[0011] Optionally, calculating the membership degree of pixels in the sub-region based on the cluster center includes: The membership degree of pixels in the sub-region is calculated using the following formula:

[0012] in, Represents pixels For sub-regions membership degree Indicates the number of sub-regions. Indicates the first Cluster centers, Indicates the first Cluster centers, This is the preset weighted index.

[0013] Optionally, the preset objective function is as follows:

[0014] in, For the target value, These represent the number of rows and columns of the image, respectively. Representation element With cluster center The distance.

[0015] To address the above problems, the present invention also provides a tumor prediction device based on bacterial data, the device comprising: The image classification module is used to acquire a set of historical tumor images, perform stage classification on the set of historical tumor images based on a fuzzy clustering algorithm, and obtain tumor image classification results. An image annotation module is used to obtain a bacterial data set corresponding to the historical tumor image set, and to jointly annotate the historical tumor image set based on the tumor image classification results and the bacterial data set to obtain a tumor training set. A prediction network building module is used to construct a tumor prediction network that includes a hybrid attention mechanism and a capsule network. The prediction network training module is used to perform joint training of the tumor prediction network using the tumor training set to obtain a standard tumor prediction network. The tumor prediction module is used to acquire images of tumors to be detected that contain bacterial annotations, and to output tumor prediction results of the tumor images to be detected using the standard tumor prediction network.

[0016] To address the above problems, the present invention also provides an electronic device, the electronic device comprising: Memory, storing at least one instruction; and The processor executes the instructions stored in the memory to implement the tumor prediction method based on bacterial data described above.

[0017] To address the aforementioned problems, this invention also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the aforementioned tumor prediction method based on bacterial data. Compared to the background art, traditional convolutional neural networks tend to overlook key locations of lesion regions in images, and tumor prediction relies solely on tumor images, resulting in low accuracy. To improve the accuracy of tumor prediction, this invention first performs stage classification on a historical tumor image set based on a fuzzy clustering algorithm to obtain tumor image classification results. Then, based on the tumor image classification results and the bacterial data set, the historical tumor image set is jointly labeled, resulting in a richer tumor training set without manual labeling. Secondly, the tumor training set is used to train a tumor prediction network including a hybrid attention mechanism and a capsule network. The hybrid attention mechanism allows for further in-depth feature extraction from the images, improving the accuracy of tumor prediction. Meanwhile, since the training data uses tumor image classification results and bacterial datasets to jointly annotate historical tumor image sets, taking into account tumor-related bacterial data, the accuracy of tumor prediction is further improved. Therefore, the tumor prediction method, device, electronic device and computer-readable storage medium based on bacterial data proposed in this invention can solve the problem of low accuracy in tumor prediction. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a tumor prediction method based on bacterial data provided in an embodiment of the present invention. Figure 2 for Figure 1 A flowchart illustrating one step of the embodiment; Figure 3 for Figure 1 A flowchart illustrating another step in the embodiment; Figure 4 This is a functional block diagram of a tumor prediction device based on bacterial data provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device for implementing the bacterial data-based tumor prediction method according to an embodiment of the present invention. The realization of the objectives, functional characteristics, and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a tumor prediction method based on bacterial data. The execution entity of the bacterial data-based tumor prediction method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the bacterial data-based tumor prediction method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0021] Example 1:

[0022] Reference Figure 1 The diagram shown is a flowchart illustrating a tumor prediction method based on bacterial data according to an embodiment of the present invention. In this embodiment, the tumor prediction method based on bacterial data includes: S1. Obtain a set of historical tumor images, and perform stage classification on the set of historical tumor images based on a fuzzy clustering algorithm to obtain the tumor image classification result.

[0023] In this embodiment of the invention, the historical tumor image set can be medical images of the gastrointestinal tract. For example, in the clinical medical field, the historical tumor image set can be MRI (magnetic resonance imaging) images or CT images of tumors. For instance, since the treatment methods for tumors differ at different stages, fuzzy clustering algorithms can be used to cluster and classify tumors, improving the accuracy of tumor prediction.

[0024] In an optional embodiment of the present invention, the fuzzy clustering algorithm includes the FCM algorithm. The FCM algorithm clusters pixels in an image into c classes, then calculates the squared metric distance between each pixel and each cluster center, and uses an iterative algorithm to minimize it, thereby clustering the pixels with the highest similarity into one class.

[0025] For details, refer to Figure 2 As shown, the stage classification of the historical tumor image set based on the fuzzy clustering algorithm to obtain the tumor image classification result includes: S10. Perform spatial geometric transformation and color conversion on the images in the historical tumor image set to obtain the original image set; S11. Divide each image in the original image set into regions to obtain a preset number of sub-regions; S12. Calculate the membership degree of the pixels in the sub-region, and calculate the target value of the image in the original image set based on the membership degree; S13. If the target value is greater than the preset target threshold, then each image in the original image set is re-divided into regions, and the step of calculating the membership degree of the pixels in the sub-region is returned until the target value is less than or equal to the target threshold. Then, the images in the original image set are divided into different stages according to the cluster centers in the sub-regions to obtain the tumor image classification result.

[0026] In this embodiment of the invention, since tumor images are small sample data, spatial geometric transformations and color transformations are used to enhance the data and improve image classification performance. Spatial geometric transformations include flipping, random cropping, translation, and scaling of the original image; color transformations include operations on the colors of the original image, such as brightness, saturation, contrast, grayscale conversion, and binarization.

[0027] It should be explained that fuzzy clustering algorithms require pre-setting the number of regions to be divided, for example, dividing each image into three sub-regions.

[0028] Further, calculating the membership degree of pixels in the sub-region and calculating the target value of the image in the original image set based on the membership degree includes: Calculate the cluster center of the sub-region, and calculate the membership degree of the pixels in the sub-region based on the cluster center; The target value of the image in the original image set is calculated based on the preset objective function and the membership degree.

[0029] In an optional embodiment of the present invention, calculating the membership degree of pixels in the sub-region based on the cluster center includes: The membership degree of pixels in the sub-region is calculated using the following formula:

[0030] in, Represents pixels For sub-regions membership degree Indicates the number of sub-regions. Indicates the first Cluster centers, Indicates the first Cluster centers, This is the preset weighted index. In this embodiment of the invention, cluster centers can be calculated based on the K-means clustering method.

[0031] In an optional embodiment of the present invention, the preset objective function is as follows:

[0032] in, For the target value, These represent the number of rows and columns of the image, respectively. Representation element With cluster center The distance.

[0033] In this embodiment of the invention, tumors can be classified into different categories based on the fuzzy clustering results. For example, gastrointestinal tumors can be classified into early-stage, mid-stage, and late-stage tumors.

[0034] S2. Obtain the bacterial data set corresponding to the historical tumor image set, and perform joint annotation on the historical tumor image set based on the tumor image classification results and the bacterial data set to obtain the tumor training set.

[0035] In this embodiment of the invention, the bacterial data set corresponding to the historical tumor image set refers to bacterial colony data from the patient's gastrointestinal tract, such as the number of bacteria like Bifidobacterium, Lactobacillus, and Helicobacter pylori. By associating bacterial data with tumor images, the accuracy of tumor prediction can be further improved.

[0036] For details, refer to Figure 3 As shown, the joint annotation of the historical tumor image set based on the tumor image classification results and the bacterial dataset includes: S20. Use the tumor image classification results as the initial annotations for the images in the historical tumor image set; S21. Use the bacterial data in the bacterial dataset as the associated annotation for the images in the historical tumor image set; S22. Calculate the initial probability distribution of the initial label and the association probability distribution of the associated label; S23. The initial probability distribution and the associated probability distribution are labeled and fused to obtain a joint label set, and a preset number of joint labels are selected from the joint label set as image labels for the historical tumor image set.

[0037] In an optional embodiment of the present invention, the initial probability distribution is as follows: , For initial labeling, For tumor images, the associated probability distribution is as follows: , For related annotations, Image of a tumor.

[0038] In an optional embodiment of the present invention, the annotation and fusion of the initial probability distribution and the associated probability distribution includes: The initial probability distribution and the associated probability distribution are labeled and fused using the following joint probability formula:

[0039] in, The probabilities after annotation and fusion are given. The number of images in the historical tumor image collection. For the initial probability distribution, For the correlation probability distribution, For initial labeling, For related annotations, Image of a tumor.

[0040] In this embodiment of the invention, the initial annotation and associated annotation are selected as the image annotation based on the probability of the annotation fusion from high to low. For example, the annotation of a certain tumor image includes: mid-stage tumor, number of Bifidobacteria a, and number of Lactobacillus b.

[0041] S3. Construct a tumor prediction network that includes a hybrid attention mechanism and a capsule network.

[0042] In this embodiment of the invention, the capsule network is composed of capsules rather than neurons. Compared to the pooling layers of convolutional neural networks, which cause a large amount of useful information to be lost when the feature map passes through, the capsule network learns specific features at specific locations in the image and then generates a vector composed of a set of neurons. This can preserve the pose and spatial information between target objects and improve the accuracy of target object prediction.

[0043] Specifically, the construction of the tumor prediction network, which includes a hybrid attention mechanism and a capsule network, includes: By splicing together a preset spatial attention module and a preset channel attention module, a hybrid attention mechanism unit is obtained; Obtain a convolutional network including multiple convolutional layers, and add the hybrid attention mechanism unit after each convolutional layer of the convolutional network to obtain the original prediction network; A capsule layer and a fully connected layer are added after the original prediction network to obtain the tumor prediction network.

[0044] In an optional embodiment of the present invention, the hybrid attention mechanism unit can be a Feature Attention Unit (FAU), using a mixture of Spatial Attention Module (SAM), Channel Attention Module (CAM), and Channel Attention Module (GCT). The capsule layer includes a main capsule layer and digital capsule layers. The main capsule layer is used to output vectors from the features of the previous layer. The number of capsules in the digital capsule layer is determined by the number of categories to be predicted. For example, if the prediction is for early-stage, mid-stage, and late-stage tumors, then three capsules are used to output the weights for each category. The fully connected layer is used to reconstruct the tumor image, thereby further optimizing the network model.

[0045] In an optional embodiment of the present invention, the hybrid attention mechanism unit obtained by splicing a preset spatial attention module and a preset channel attention module includes: The spatial attention module (SAM) and the channel attention module (CAM) are connected in series to obtain the first concatenated network. The channel attention module CAM and the channel attention module GCT are connected in series to obtain the second series network; The first serial network and the second serial network are connected in parallel to obtain the hybrid attention mechanism unit.

[0046] Meanwhile, the tumor prediction network may include 12 layers, wherein layers 1, 3, and 5 are convolutional layers, layers 2, 4, and 6 are FAU layers, layers 7 and 8 are master capsule layers and digital capsule layers, and layers 9, 10, 11, and 12 are fully connected layers.

[0047] S4. The tumor prediction network is jointly trained using the tumor training set to obtain a standard tumor prediction network.

[0048] In this embodiment of the invention, since the tumor prediction network includes multi-layered hybrid attention mechanism units, it can effectively acquire the global and local connections of image data, thereby effectively improving the training effect.

[0049] Specifically, the step of jointly training the tumor prediction network using the tumor training set with an attention mechanism to obtain a standard tumor prediction network includes: Features of images in the tumor training set are extracted using the convolutional layers in the tumor prediction network. The features are fused using the hybrid attention mechanism unit in the tumor prediction network to obtain hybrid features; The weighted probabilities of the final mixed features are output using the capsule layer in the tumor prediction network. The predicted labels of the weight probabilities are output using the fully connected layers in the tumor prediction network; Construct a confusion matrix based on the predicted annotations and the initial annotations, and calculate the prediction accuracy based on the confusion matrix. If the prediction accuracy is less than or equal to a preset accuracy threshold, adjust the parameters of each layer in the tumor prediction network, return to the step of extracting features of the images in the tumor training set using the convolutional layers in the tumor prediction network, until the prediction accuracy is greater than the accuracy threshold, stop training, and obtain the standard tumor prediction network.

[0050] In one optional embodiment of the present invention, image features are extracted through convolutional layers, and the features are spliced ​​and fused based on multiple attention mechanisms in the hybrid attention mechanism unit, which can continuously deepen the extracted features and improve the accuracy of tumor prediction.

[0051] It's important to explain that the confusion matrix essentially displays the number of correctly and incorrectly predicted images. The confusion matrix includes four basic metrics: 1. The number of images that were actually positive but were predicted as positive (TP); 2. The number of images that were actually positive but were predicted as negative (FN); 3. The number of images that were actually negative but were predicted as positive (FP); 4. The number of images that were actually negative but were predicted as negative (TN). For example, if there are 100 images of early-stage tumors, 80 of them are predicted as early-stage tumors (TP) and 20 are predicted as non-early-stage tumors (FN); if there are 100 images of non-early-stage tumors, 30 of them are predicted as early-stage tumors (FP) and 70 are predicted as non-early-stage tumors (TN).

[0052] In an optional embodiment of the present invention, the prediction accuracy can be the F1-Score value, that is: ,in, , .

[0053] S5. Obtain a tumor image to be detected containing bacterial annotations, and output the tumor prediction result of the tumor image to be detected using the standard tumor prediction network.

[0054] In this embodiment of the invention, the patient's tumor image (CT image, etc.) is labeled with the patient's gastrointestinal bacteria data to obtain a tumor image to be detected, and the tumor image to be detected is predicted using the standard tumor prediction network to obtain the tumor prediction result for the patient.

[0055] Compared to the background technology, traditional convolutional neural networks tend to overlook key locations in lesion areas within images, and tumor prediction relies solely on tumor images, resulting in low accuracy. To improve tumor prediction accuracy, this invention first performs stage classification on a historical tumor image set using a fuzzy clustering algorithm to obtain tumor image classification results. Then, based on these classification results and a bacterial dataset, the historical tumor image set is jointly annotated, providing a richer tumor training set without manual annotation. Next, the tumor training set is used to train a tumor prediction network incorporating a hybrid attention mechanism and capsule networking. The hybrid attention mechanism further extracts image features, improving prediction accuracy. Furthermore, since the training data incorporates tumor-related bacterial data through joint annotation of the historical tumor image set using tumor image classification results and a bacterial dataset, the accuracy of tumor prediction is further enhanced. Therefore, the tumor prediction method based on bacterial data proposed in this invention can solve the problem of low tumor prediction accuracy.

[0056] Example 2:

[0057] like Figure 4 The diagram shown is a functional block diagram of a tumor prediction device based on bacterial data provided in an embodiment of the present invention, which can implement the monitoring method in Embodiment 1.

[0058] The tumor prediction device 100 based on bacterial data described in this invention can be installed in an electronic device. Depending on the functions implemented, the tumor prediction device 100 based on bacterial data may include an image classification module 101, an image annotation module 102, a prediction network construction module 103, a prediction network training module 104, and a tumor prediction module 105. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0059] The image classification module 101 is used to acquire a set of historical tumor images, perform stage classification on the set of historical tumor images based on a fuzzy clustering algorithm, and obtain tumor image classification results. The image annotation module 102 is used to obtain the bacterial data set corresponding to the historical tumor image set, and to jointly annotate the historical tumor image set based on the tumor image classification result and the bacterial data set to obtain a tumor training set. The prediction network construction module 103 is used to construct a tumor prediction network that includes a hybrid attention mechanism and a capsule network; The prediction network training module 104 is used to perform joint training of the tumor prediction network using the tumor training set to obtain a standard tumor prediction network. The tumor prediction module 105 is used to acquire a tumor image to be detected containing bacterial annotations, and to output the tumor prediction result of the tumor image to be detected using the standard tumor prediction network.

[0060] In detail, the modules in the tumor prediction device 100 based on bacterial data described in this embodiment of the invention employ the same methods as described above during use. Figure 1 The method uses the same techniques as the tumor prediction method based on bacterial data described in the previous section and can produce the same technical effects, so it will not be repeated here.

[0061] Example 3:

[0062] like Figure 5 The diagram shown is a schematic representation of an electronic device for implementing the tumor prediction method based on bacterial data, according to an embodiment of the present invention.

[0063] The electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a tumor prediction method program 12 based on bacterial data.

[0064] The memory 11 includes at least one type of readable storage medium, such as flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 1. Furthermore, the memory 11 can include both internal and external storage units of the electronic device 1. The memory 11 can be used not only to store application software and various types of data installed on the electronic device 1, such as the code of the tumor prediction method program 12 based on bacterial data, but also to temporarily store data that has been output or will be output.

[0065] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device via various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., tumor prediction methods based on bacterial data) and calls data stored in the memory 11 to perform various functions of the electronic device 1 and process data.

[0066] The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0067] Figure 5 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 5 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0068] For example, although not shown, the electronic device 1 may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0069] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is typically used to establish communication connections between the electronic device 1 and other electronic devices.

[0070] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), and optionally, a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device 1 and to display a visual user interface.

[0071] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0072] The tumor prediction method program 12 based on bacterial data stored in the memory 11 of the electronic device 1 is a combination of multiple instructions, which, when run in the processor 10, can achieve the following: A set of historical tumor images is obtained, and the set of historical tumor images is classified in stages based on a fuzzy clustering algorithm to obtain the tumor image classification results. Obtain the bacterial data set corresponding to the historical tumor image set, and perform joint annotation on the historical tumor image set based on the tumor image classification results and the bacterial data set to obtain a tumor training set; Construct a tumor prediction network that incorporates a hybrid attention mechanism and a capsule network; The tumor prediction network is jointly trained using the tumor training set with an attention mechanism to obtain a standard tumor prediction network. Obtain an image of the tumor to be detected that includes bacterial annotations, and output the tumor prediction result of the tumor image using the standard tumor prediction network.

[0073] Specifically, the processor 10's implementation method for the above instructions can be found in [reference needed]. Figures 1 to 5 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0074] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0075] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following: A set of historical tumor images is obtained, and the set of historical tumor images is classified in stages based on a fuzzy clustering algorithm to obtain the tumor image classification results. Obtain the bacterial data set corresponding to the historical tumor image set, and perform joint annotation on the historical tumor image set based on the tumor image classification results and the bacterial data set to obtain a tumor training set; Construct a tumor prediction network that incorporates a hybrid attention mechanism and a capsule network; The tumor prediction network is jointly trained using the tumor training set with an attention mechanism to obtain a standard tumor prediction network. Obtain an image of the tumor to be detected that includes bacterial annotations, and output the tumor prediction result of the tumor image using the standard tumor prediction network.

[0076] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0077] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0078] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention and not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention.

Claims

1. A tumor prediction method based on bacterial data, characterized in that, The method includes: A set of historical tumor images is obtained, and the set of historical tumor images is classified in stages based on a fuzzy clustering algorithm to obtain the tumor image classification results. Obtain the bacterial dataset corresponding to the historical tumor image set, and perform joint annotation on the historical tumor image set based on the tumor image classification results and the bacterial dataset, including: The tumor image classification results are used as the initial annotations for the images in the historical tumor image set; The bacterial data in the bacterial dataset is used as the associated annotation for the images in the historical tumor image dataset; Calculate the initial probability distribution of the initial labels and the association probability distribution of the associated labels; The initial probability distribution and the associated probability distribution are labeled and fused to obtain a joint label set, and a preset number of joint labels are selected from the joint label set as image labels for the historical tumor image set to obtain a tumor training set; Constructing a tumor prediction network that incorporates a hybrid attention mechanism and capsule networks, including: By splicing together a preset spatial attention module and a preset channel attention module, a hybrid attention mechanism unit is obtained; a convolutional network including multiple convolutional layers is obtained, and the hybrid attention mechanism unit is added after each convolutional layer of the convolutional network to obtain the original prediction network; a capsule layer and a fully connected layer are added after the original prediction network to obtain the tumor prediction network. The method of splicing a preset spatial attention module and a preset channel attention module to obtain a hybrid attention mechanism unit includes: connecting the spatial attention module SAM and the channel attention module CAM in series to obtain a first series network; connecting the channel attention module CAM and the channel attention module GCT in series to obtain a second series network; and connecting the first series network and the second series network in parallel to obtain the hybrid attention mechanism unit. The tumor prediction network is jointly trained using the tumor training set to obtain a standard tumor prediction network. Obtain an image of the tumor to be detected that includes bacterial annotations, and output the tumor prediction result of the tumor image using the standard tumor prediction network.

2. The tumor prediction method based on bacterial data as described in claim 1, characterized in that, The stage classification of the historical tumor image set based on the fuzzy clustering algorithm to obtain tumor image classification results includes: The images in the historical tumor image set are subjected to spatial geometric transformation and color conversion to obtain the original image set; Each image in the original image set is divided into regions to obtain a preset number of sub-regions; Calculate the membership degree of pixels in the sub-region, and calculate the target value of the image in the original image set based on the membership degree; If the target value is greater than the preset target threshold, then each image in the original image set is re-divided into regions, and the step of calculating the membership degree of pixels in the sub-region is returned until the target value is less than or equal to the target threshold. Then, the images in the original image set are divided into different stages according to the cluster centers in the sub-regions to obtain the tumor image classification result.

3. The tumor prediction method based on bacterial data as described in claim 2, characterized in that, The calculation of the membership degree of pixels in the sub-region, and the calculation of the target value of the image in the original image set based on the membership degree, includes: Calculate the cluster center of the sub-region, and calculate the membership degree of the pixels in the sub-region based on the cluster center; The target value of the image in the original image set is calculated based on the preset objective function and the membership degree.

4. The tumor prediction method based on bacterial data as described in claim 1, characterized in that, The step of using the tumor training set to jointly train the tumor prediction network using an attention mechanism to obtain a standard tumor prediction network includes: Features of images in the tumor training set are extracted using the convolutional layers in the tumor prediction network. The features are fused using the hybrid attention mechanism unit in the tumor prediction network to obtain hybrid features; The weighted probabilities of the final mixed features are output using the capsule layer in the tumor prediction network. The predicted labels of the weight probabilities are output using the fully connected layers in the tumor prediction network; Construct a confusion matrix based on the predicted annotations and the initial annotations, and calculate the prediction accuracy based on the confusion matrix. If the prediction accuracy is less than or equal to a preset accuracy threshold, adjust the parameters of each layer in the tumor prediction network, return to the step of extracting features of the images in the tumor training set using the convolutional layers in the tumor prediction network, until the prediction accuracy is greater than the accuracy threshold, stop training, and obtain the standard tumor prediction network.

5. The tumor prediction method based on bacterial data as described in claim 3, characterized in that, The step of calculating the membership degree of pixels in the sub-region based on the cluster center includes: calculating the membership degree of pixels in the sub-region using the following formula: in, Represents pixels For sub-regions membership degree Indicates the number of sub-regions. Indicates the first Cluster centers, Indicates the first Cluster centers, This is the preset weighted index.

6. A tumor prediction device based on bacterial data, characterized in that, The device includes: The image classification module is used to acquire a set of historical tumor images, perform stage classification on the set of historical tumor images based on a fuzzy clustering algorithm, and obtain tumor image classification results. An image annotation module is used to obtain a bacterial dataset corresponding to the historical tumor image set, and to jointly annotate the historical tumor image set based on the tumor image classification result and the bacterial dataset, including: using the tumor image classification result as the initial annotation for the images in the historical tumor image set; using the bacterial data in the bacterial dataset as the associated annotation for the images in the historical tumor image set; calculating the initial probability distribution of the initial annotation and the associated probability distribution; performing annotation fusion on the initial probability distribution and the associated probability distribution to obtain a joint annotation set, and selecting a preset number of joint annotations from the joint annotation set as image annotations for the historical tumor image set to obtain a tumor training set; A prediction network construction module is used to construct a tumor prediction network including a hybrid attention mechanism and a capsule network. The module includes: splicing a preset spatial attention module and a preset channel attention module to obtain a hybrid attention mechanism unit; obtaining a convolutional network including multiple convolutional layers, adding the hybrid attention mechanism unit after each convolutional layer to obtain an original prediction network; adding a capsule layer and a fully connected layer after the original prediction network to obtain the tumor prediction network; the splicing of the preset spatial attention module and the preset channel attention module to obtain the hybrid attention mechanism unit includes: concatenating the spatial attention module (SAM) and the channel attention module (CAM) to obtain a first concatenated network; concatenating the channel attention module (CAM) and the channel attention module (GCT) to obtain a second concatenated network; and concatenating the first and second concatenated networks in parallel to obtain the hybrid attention mechanism unit. A prediction network training module is used to jointly train the tumor prediction network using the tumor training set to obtain a standard tumor prediction network. The tumor prediction module is used to acquire images of tumors to be detected that contain bacterial annotations, and to output tumor prediction results of the tumor images to be detected using the standard tumor prediction network.

Images