Esophageal tumor auxiliary diagnosis method and device based on machine learning

Abstract

The invention relates to the technical field of medical image processing, discloses an esophageal image processing method, and particularly relates to an esophageal tumor auxiliary diagnosis method and device based on machine learning. According to the embodiment of the invention, the acquired CT image is subjected to noise reduction processing, feature extraction processing and classification processing in sequence, so that tumor segmentation and tumor staging identification in the CT image are realized. Moreover, the feature extraction process and the classification processing process in the embodiment are constructed based on fusion of the neural network and the classifier, so that the accuracy of segmentation recognition under the condition of the small sample of the medical image is realized.

Description

Technical Field

[0001] The present application relates to the technical field of medical image processing, and is an esophageal image processing method, specifically an esophageal tumor auxiliary diagnosis method and device based on machine learning. Background Art

[0002] At present, the treatment of esophageal cancer usually adopts surgical intervention, radiotherapy, chemotherapy, and the combination of traditional Chinese and Western medicine. Among them, tumor resection surgery and radiotherapy are local treatments, which only target the tumor site; while chemotherapy refers to the use of chemical drugs in the body to kill tumor cells. The combination of radiotherapy and surgery can improve the success rate of surgical resection and significantly improve long-term survival. Although the prognosis of patients with middle and late stage esophageal cancer is not good, if it can be treated as soon as possible, the prognosis will be much better. Therefore, in clinical work, if it can help doctors quickly and accurately obtain the segmentation results of the esophageal cancer tumor area, and determine the specific stage of esophageal cancer, thereby reducing the misdiagnosis rate and missed diagnosis rate, and prescribing the right medicine, it will be of great help to the treatment of patients. Summary of the invention

[0003] In order to achieve the above technical goals, the embodiments of the present application provide an esophageal tumor auxiliary diagnosis method and device based on machine learning, which can realize the automatic segmentation and identification of esophageal tumors.

[0004] In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0005] In a first aspect, a method for auxiliary diagnosis of esophageal tumors based on machine learning is provided, the method is applied to a server, the server is configured with patient label information and corresponding detection data, the method comprises: determining the patient label information based on the patient's unique identification code, and retrieving a CT image in the corresponding detection data based on the patient label information; the CT image is obtained by performing a computer tomography scan on the patient's abdominal area; performing feature extraction on the CT image, and segmenting the background of the CT image based on the extracted features to obtain a target feature map containing only a region of interest; classifying and identifying the target feature map based on a classification model to obtain a classification result; the classification result is a tumor staging combination result, and the tumor staging combination result includes a combination result of a first classification, a second classification, and a third classification; the first classification includes four sub-levels, the second classification includes four sub-levels, and the third classification includes two sub-levels.

[0006] Furthermore, the unique identification code includes a unique virtual identification code and a unique entity identification code, and determining the patient label information based on the patient's unique identification code includes: determining a corresponding identification device based on the type of code to be identified, and identifying the unique virtual identification code or the unique entity identification code based on the identification device, and retrieving an identification result, wherein the identification result includes the patient label information.

[0007] Furthermore, the feature extraction of the CT image and the segmentation of the background of the CT image based on the extracted features include: segmenting the CT image based on a segmentation model; the segmentation model includes an encoder and a decoder, the encoder is used to receive the CT image, and the decoder is used to output the target feature map; the encoder is composed of multiple downsampling layers, and the decoder is composed of multiple upsampling layers, each layer of the encoder is connected to each layer of the decoder by a jump connection, and an image information entropy module is provided at the end of each layer of the decoder, and the image information entropy module is used to enhance the edge features of the segmented target.

[0008] Furthermore, a feature intensive module is provided in each layer of the encoder and each layer of the decoder, and the feature intensive module includes a first convolutional layer, a second convolutional layer, a batch normalization layer and an activation function layer connected in sequence, and the input of the current layer of the feature intensive module is the feature map of all previous layers, and the feature map output in the current layer is the input in the subsequent layer.

[0009] Furthermore, the first convolutional layer and the second convolutional layer are deformable convolutions.

[0010] Furthermore, an attention module is provided in each layer of the decoder, and the attention module includes a channel attention submodule and a spatial attention submodule; two multi-layer perceptrons are provided between the input and output of the spatial attention submodule, and two convolution blocks are provided in the channel attention submodule.

[0011] Furthermore, the classification model includes a global average pooling layer connected to the output end of the decoder, and a classifier connected to the global average pooling layer; the classifier is a support vector machine constructed based on a polynomial kernel function.

[0012] Furthermore, the method also includes denoising the CT image, specifically including: performing wavelet transform on the CT image to obtain high-frequency features and low-frequency features of the CT image, and performing guided filtering on the high-frequency features and the low-frequency features respectively to obtain corresponding updated high frequencies and updated low frequencies, and performing image reconstruction based on inverse wavelet transform to obtain the denoised CT image.

[0013] In a second aspect, a machine learning-based auxiliary diagnosis device for esophageal tumors is provided, the device being applied to a server, the server being configured with patient label information and corresponding detection data, the device comprising: an image acquisition module, for determining the patient label information based on a unique identification code of the patient, and retrieving a CT image in the corresponding detection data based on the patient label information; an image segmentation module, for extracting features from the CT image, and segmenting the background of the CT image based on the extracted features to obtain a target feature map containing only an area of ​​interest; and a recognition module, for classifying and identifying the target feature map based on a classification model to obtain a classification result.

[0014] Furthermore, the device also includes: an image noise reduction module, used to perform noise reduction processing on the CT image.

[0015] In the technical solution provided in the embodiments of the present application,. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] The methods, systems and / or programs in the accompanying drawings will be further described according to exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments, wherein example numbers represent similar mechanisms in the various views of the accompanying drawings.

[0018] Figure 1 It is a schematic diagram of the structure of the auxiliary diagnosis system provided in the embodiment of the present application.

[0019] Figure 2 This is a flow chart of the auxiliary diagnosis method for esophageal tumors based on machine learning provided in an embodiment of the present application.

[0020] Figure 3 It is a schematic diagram of the segmentation model structure provided in an embodiment of the present application.

[0021] Figure 4 It is a schematic diagram of the esophageal tumor segmentation result provided in the embodiment of the present application.

[0022] Figure 5 It is a schematic diagram of the structure of the auxiliary diagnosis device for esophageal tumors provided in an embodiment of the present application.

[0023] Figure 6 It is a schematic diagram of the terminal device structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0025] In the following detailed description, numerous specific details are set forth by way of example in order to provide a comprehensive understanding of the relevant guidance. However, it will be apparent to those skilled in the art that the present application may be practiced without these details. In other cases, well-known methods, procedures, systems, compositions and / or circuits have been described at a relatively high level, without detail, in order to avoid unnecessarily obscuring aspects of the present application.

[0026] Flowcharts are used in the present application to illustrate the execution process performed by the system according to the embodiment of the present application. It should be clearly understood that the execution process of the flowchart may not be performed in order. On the contrary, these execution processes may be performed in reverse order or simultaneously. In addition, at least one other execution process may be added to the flowchart. One or more execution processes may be deleted from the flowchart.

[0027] Before further describing the embodiments of the present invention in detail, the nouns and terms involved in the embodiments of the present invention are described. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations.

[0028] (1) In response, it is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0029] (2) Based on is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or have a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0030] Esophageal cancer is a major global health challenge, with approximately 400,000 people dying from it each year. Esophageal squamous cell carcinoma and esophageal adenocarcinoma are two different histological types. Currently, the incidence of esophageal cancer is on the rise, and early detection of esophageal cancer can significantly improve patient prognosis and quality of life. The 5-year survival rate of early esophageal cancer after treatment is over 90%, while the 5-year survival rate of advanced esophageal cancer is less than 20%. The survival outcome of esophageal cancer diagnosed early is better than that of esophageal cancer diagnosed late. Early detection of esophageal cancer can significantly increase the resection rate of endoscopic submucosal dissection (ESD) and reduce the surgical procedure rate and mortality rate of patients. Therefore, early detection of lesions is crucial to reducing the mortality rate and overall burden of esophageal cancer.

[0031] However, since the definition of early esophageal cancer is limited to the mucosal layer and submucosal layer, its lesion characteristics are often not very obvious. This will bring more workload to endoscopists and easily lead to missed diagnosis. In recent years, with the continuous development of computer technology, computer-aided diagnosis systems are increasingly used in clinical diagnosis and treatment. The embodiment of this application aims to use the most commonly used CT images for diagnosing early esophageal cancer, and use machine learning and deep learning methods to help doctors make diagnoses faster and more accurately.

[0032] For the above technical background, see Figure 1 The present embodiment provides an auxiliary diagnosis system 10, which at least includes a doctor end 11, an image acquisition end 12 and a server 13, wherein the image acquisition end is used to acquire the patient's CT image and upload the CT image to the server. A storage unit is provided in the server, and the storage unit stores the CT image correspondingly based on the patient's unique identification code and the image tag. The doctor end retrieves the CT image through the patient's unique identification code, and sends a command to the processing unit in the server. The processing unit is configured with the auxiliary diagnosis method provided in the present embodiment, and processes the retrieved CT image based on the command to obtain a processing result.

[0033] Among them, the command issued by the doctor side in this embodiment can be not only auxiliary diagnosis, but also other commands. However, since the solution provided in this embodiment is mainly used for auxiliary diagnosis of esophageal tumors, the command issued by the doctor side is this command.

[0034] For the auxiliary diagnosis method executed in the processing unit, please refer to Figure 2 , an esophageal tumor auxiliary diagnosis method based on machine learning, comprising the following steps:

[0035] Step S21. Determine the patient label information based on the patient's unique identification code, and retrieve the CT image in the corresponding detection data based on the patient label information.

[0036] In this embodiment, the unique identification code includes a unique virtual identification code and a unique entity identification code. The unique virtual identification code can be a real-time updated QR code and a real-time updated barcode, and can also include any other virtual form of identifiable coding; the first entity identification code can be a coding form that stores information through physical media such as a medical card. The acquisition of the above virtual identification code and entity identification code is identified by the device terminal corresponding to the doctor's end, the unique virtual identification code is obtained by a barcode scanning gun, a barcode scanning terminal, etc., and the unique entity identification code is identified by a terminal device such as a card reader.

[0037] With regard to determining the patient label information through the patient's unique identification code, the corresponding identification device is determined by the type of code to be identified, i.e., the virtual code or the physical code, and the unique virtual identification code or the unique physical identification code is identified based on the identification device to retrieve the identification result, which includes the patient label information.

[0038] Among them, the patient tag information is the tag information stored in the server, which is set based on the data storage logic of the server, and each specific patient has corresponding tag information, which is edited based on a specific coding logic. This content will not be repeated in this embodiment.

[0039] Step S22: Perform noise reduction processing on the CT image.

[0040] In this embodiment, in order to avoid the scanning imaging noise problem caused by equipment reasons and low-dose scanning, the CT image needs to be subjected to noise reduction processing before being processed.

[0041] Among them, the denoising process is mainly a wavelet transform denoising method. Specifically, the CT image is first transformed by wavelet transform to obtain the high-frequency component and low-frequency component of the CT image, and the high-frequency component and the low-frequency component are guided filtered respectively to obtain the corresponding updated high-frequency component and updated low-frequency component, and finally the image is reconstructed by inverse wavelet transform to obtain the denoised CT image.

[0042] Specifically, the wavelet transform can be implemented using the wavelet transform technology in the prior art, by adding a group of random noises to the CT image, and then performing a wavelet transform on the CT image with the random noise added based on the wavelet function to obtain the high-frequency component and the low-frequency component of the image.

[0043] Then the high frequency component and the low frequency component are guided and filtered respectively, where the guide filtering is shown as follows: i =a k I i +b k , Where q is the output image, I is the guide image, i is the index of the pixel, and q and I are in the window w with pixel k as the center and r as the radius. k In the example above, P is the input image, and minimizing the error between p and q can be obtained: where μ k and σ k For i in window w k The mean and variance in ; ε is a positive parameter that controls the smoothness. The formula for calculating the filter output is: in, and are the average values ​​of a and b over the window centered on i, respectively.

[0044] The high-frequency component and the low-frequency component are filtered by the guided filtering, and then the filtered high-frequency component and the low-frequency component are reconstructed by inverse wavelet transform to obtain the final denoised CT image.

[0045] In this embodiment, the above-mentioned noise reduction processing can achieve noise reduction processing for CT images. Compared with the existing wavelet filtering, the guided filtering method can better retain the internal details of the image, making the denoised image closer to the original image and having a better denoising effect.

[0046] Step S23: extract features from the CT image, and segment the background of the CT image based on the extracted features to obtain a target feature map containing only the region of interest.

[0047] In this embodiment, the above processing is an image segmentation process, which is used to segment the CT image without a region of interest, that is, the image that does not contain a tumor, and only segment the region of interest with the tumor to obtain the target feature map required for subsequent identification.

[0048] The above processing is implemented through a segmentation model. For the structure of this segmentation model, please refer to Figure 3 , using U-Net network as the basic structure. It includes an encoder and a decoder, where the encoder is used to receive the denoised CT image and the decoder is used to output the target feature map.

[0049] The encoder is composed of multiple downsampling layers, the decoder is composed of multiple upsampling layers, and each layer of the encoder is connected to each layer of the decoder by a jump connection, and an image information entropy module is provided at the end of each layer of the decoder. In this embodiment, the image information entropy module is used to enhance the edge features of the segmentation target.

[0050] Among them, a feature intensive module is set in each layer of the encoder and the decoder, and the feature intensive module is composed of a first convolutional layer, a second convolutional layer, a batch normalization layer and an activation function layer connected in sequence.

[0051] The feature intensive module set in the encoder in this embodiment is used to extract the features of the esophageal GTV, especially the edge information. Among them, the first convolution layer is a 1×1×1 convolution. The number of feature channels is reduced by using 1×1×1 convolution. This layer is called the bottleneck layer. In addition to reducing parameters and speeding up the training speed of the network, the bottleneck layer can also increase the depth of the network. In each layer of the feature intensive module, the input of the current layer is the feature map of all previous layers, and the feature map learned in this layer is the input in the subsequent layer, thus forming a complete interconnected structure. In this embodiment, the problem of gradient vanishing is alleviated by dense connection, the propagation between features is strengthened, and a regularization effect is achieved.

[0052] The setting of the second convolutional layer is different from that of the first convolutional layer. Since the boundary of esophageal cancer tumors is complex, changeable and irregular, conventional convolution kernels cannot play a good role in the segmentation task of esophageal tumors. In this embodiment, the key layer of the segmentation model is replaced with a deformable convolution, which can better adapt to the complex curvature changes of the cancer lesion boundary, so that the network can better extract the features hidden in the details on the original basis. Among them, the formula for deformable convolution is as follows: Where Δp i is the offset of each sampling point, which can be obtained by bilinear interpolation or other methods.

[0053] Among them, the image information entropy module can enhance the feature map information in the encoding stage of the encoder. The processing method is to calculate the information entropy of the feature map and re-weight it through the image information entropy module to obtain a new enhanced feature map.

[0054] Specifically, for a K-level grayscale image, the K value is set to 255, and the i (i∈1,...,k) level grayscale is represented by pi. Then, the entropy of the corresponding grayscale image is calculated based on the following formula: Among them, 0≤i≤k, the accumulation of information entropy of different gray levels is defined as information entropy, then the information entropy of the entire image is realized based on the following formula: k = 255; where pi refers to the probability of each level of grayscale pixels in the entire image. When pi = 0, pilog(pi) = 0. pi is calculated from the grayscale histogram, which is the ratio of the number of pixels of grayscale i to the total number of pixels in the image.

[0055] For each layer of the decoder in this embodiment, an attention module is also provided, wherein the attention module includes a channel attention submodule and a spatial attention submodule.

[0056] Specifically, two multi-layer perceptrons are arranged between the input and output of the spatial attention submodule, and two convolution blocks are arranged in the channel attention submodule.

[0057] For the segmentation model in the embodiment of the present application, a hybrid loss function composed of Focal and Dice is used in the training stage. The hybrid loss function is as follows: Loss = 0.6 Loss Dice +0.4Loss Focal ; The Dice loss function is based on the following formula: Where x and y are the first result and the second result respectively, |X| and |Y| are the number of the first result and the second result, where the first result is a positive sample and the second result is a negative sample; the Focal loss function is based on the following expression: Loss Focal =-α t (1-p t ) γ ×log(p t )-(1-α)p γ ×(1-y)log(1-p), where y is the label value of the sample, p is the probability that the model predicts that a sample is a positive sample, and α and γ are both hyperparameters, which are 0.25 and 2 respectively in this embodiment.

[0058] Step S24: classify and identify the target feature map based on the classification model to obtain a classification result.

[0059] In this embodiment, the classification model includes a global average pooling layer connected to the output end of the decoder, and a classifier connected to the global average pooling layer.

[0060] Among them, a support vector machine based on a polynomial kernel function is used as the classifier.

[0061] Generally speaking, neural networks are usually used to implement classification for machine learning, and specifically, a fully connected layer is used to classify and identify the feature map output by the decoder. Although neural networks have strong classification capabilities, due to the small sample size and very similar characteristics of images of different categories in medical images, in this embodiment, classification is achieved by using a support vector machine. The support vector machine can find the best compromise between the complexity of the model and the learning ability, thereby obtaining the best classification. Compared with traditional neural network classifiers, it can use kernel functions to map samples to high-dimensional space, and construct an optimal hyperplane to divide the feature space, thereby maximizing the classification boundary, and performs well in nonlinear situations.

[0062] In this embodiment, the classification result is a tumor staging combination result, and the tumor staging combination result includes the combination result of the first classification, the second classification and the third classification; the first classification includes four sub-levels, the second classification includes four sub-levels, and the third classification includes two sub-levels. Specifically, the above classification is established based on the clinical classification method of malignant tumors, "TNM staging", which is used to judge the degree of lesions of various cancers. Among them, TNM is the first classification, the second classification and the third classification, and the meanings of T, N and M are different. T staging is the size and distribution of the primary tumor, which is divided into four sub-levels of T1, T2, T3 and T4 according to the degree. The larger the value, the larger the mass and the wider the infiltration area; in addition, it can also be divided into two levels of Tis and T0. These two levels represent that the tumor is carcinoma in situ, that is, the tumor only occurs in the epithelium, and no tumor lesions are found at the location of the examination. N staging is divided into four sub-levels of N0, N1, N2 and N3, which measures the number of regional lymph node metastases related to the tumor. N0 means no lymph node metastasis. The larger the value from N1 to N3, the more local lymph node metastasis there is. If it is not possible to determine whether there is lymph node metastasis, NX is used to represent it. The M stage indicates the situation of distant metastasis of regional lymph nodes. M0 means no distant metastasis of regional lymph nodes; and M1 means distant metastasis of regional lymph nodes.

[0063] In this embodiment, different stages of cancer can be divided by different combinations of the three stages T, N, and M, namely, esophageal cancer stages I, II, III, and IV.

[0064] The embodiment of the present application realizes tumor segmentation and tumor staging recognition in CT images by performing noise reduction, feature extraction and classification processing on the collected CT images in sequence. In addition, the feature extraction process and classification process in this embodiment are constructed based on neural network and classifier fusion, which realizes the accuracy of segmentation and recognition under small sample conditions of medical images.

[0065] See also Figure 5Based on step S110-step S140, a virtual device for execution is also provided, which is an esophageal tumor auxiliary diagnosis device 500 based on machine learning. This device can be configured in any software product to execute the process of step S110-step S140. This device can be a separate plug-in module or a complete computer product. This device includes:

[0066] An image noise reduction module 51, used for performing noise reduction processing on the CT image;

[0067] An image acquisition module 52, configured to determine the patient tag information based on the patient's unique identification code, and retrieve a CT image in the corresponding detection data based on the patient tag information;

[0068] An image segmentation module 53 is used to extract features from the CT image and segment the background of the CT image based on the extracted features to obtain a target feature map containing only a region of interest;

[0069] The recognition module 54 classifies and recognizes the target feature map based on the classification model to obtain a classification result.

[0070] See also Figure 6 , the above method can also be integrated into the provided terminal device 300, and the device may have relatively large differences due to different configurations or performances, and may include one or more processors 601 and memory 602, and the memory 602 may store one or more storage applications or data. Among them, the memory 602 can be a temporary storage or a permanent storage. The application stored in the memory 302 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the terminal device. Furthermore, the processor 601 can be configured to communicate with the memory 602, and the terminal device executes a series of computer executable instructions in the memory 602. The terminal device may also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input or output interfaces 605, one or more keyboards 606, etc.

[0071] In a specific embodiment, the terminal device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions in the terminal device, and the one or more programs are configured to be executed by one or more processors, including computer executable instructions for performing the following:

[0072] Performing noise reduction processing on the CT image;

[0073] and screening the image features to obtain the best features;

[0074] Determine the patient label information based on the patient's unique identification code, and retrieve the CT image in the corresponding detection data based on the patient label information;

[0075] Extracting features from the CT image, and segmenting the background of the CT image based on the extracted features to obtain a target feature map containing only a region of interest;

[0076] The target feature map is classified and identified based on the classification model to obtain a classification result.

[0077] The following is a detailed introduction to the various components of the processor:

[0078] In this embodiment, the processor is an application specific integrated circuit (ASIC), or is configured to implement one or more integrated circuits of the embodiments of the present application, such as one or more microprocessors (digital signal processors, DSP), or one or more field programmable gate arrays (field programmable gate arrays, FPGA).

[0079] Optionally, the processor may execute various functions by running or executing a software program stored in the memory and calling data stored in the memory, such as executing the above-mentioned Figure 2 The method shown.

[0080] In a specific implementation, as an embodiment, the processor may include one or more microprocessors.

[0081] Among them, the memory is used to store the software program that executes the solution of the present application, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0082] Optionally, the memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor, or may exist independently and be coupled to the processing unit through the interface circuit of the processor, and the embodiments of the present application do not specifically limit this.

[0083] It should be noted that the structure of the processor shown in this embodiment does not constitute a limitation on the device, and the actual device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0084] In addition, the technical effects of the processor can refer to the technical effects of the method described in the above method embodiment, which will not be repeated here.

[0085] It should be understood that the processor in the embodiments of the present application may 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. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0086] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0087] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0088] In this application, "at least one" means one or more, and "more" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0089] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0090] Those of ordinary skill 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 this application.

[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0092] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0094] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0095] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks.

[0096] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for auxiliary diagnosis of esophageal tumors based on machine learning, characterized in that: The method is applied to a server, wherein the server is configured with patient label information and corresponding detection data, and the method comprises: Determine the patient tag information based on the patient's unique identification code, and retrieve a CT image in the corresponding detection data based on the patient tag information; the CT image is obtained by performing a computerized tomography scan on the patient's abdominal area; Extracting features from the CT image, and segmenting the background of the CT image based on the extracted features to obtain a target feature map containing only a region of interest; The target feature map is classified and identified based on the classification model to obtain a classification result; the classification result is a tumor staging combination result, and the tumor staging combination result includes a combination result of a first classification, a second classification and a third classification; the first classification includes four sub-levels, the second classification includes four sub-levels, and the third classification includes two sub-levels.

2. The machine learning-based auxiliary diagnosis method for esophageal tumors according to claim 1, characterized in that: The unique identification code includes a unique virtual identification code and a unique entity identification code. The determination of the patient label information based on the patient's unique identification code includes: determining a corresponding identification device based on the type of code to be identified, and identifying the unique virtual identification code or the unique entity identification code based on the identification device, and retrieving an identification result, wherein the identification result includes the patient label information.

3. The machine learning-based auxiliary diagnosis method for esophageal tumors according to claim 1, characterized in that: The method of extracting features from the CT image and segmenting the background of the CT image based on the extracted features includes: segmenting the CT image based on a segmentation model; the segmentation model includes an encoder and a decoder, the encoder is used to receive the CT image, and the decoder is used to output the target feature map; the encoder is composed of multiple downsampling layers, and the decoder is composed of multiple upsampling layers, each layer of the encoder is connected to each layer of the decoder by a jump connection, and an image information entropy module is provided at the end of each layer of the decoder, and the image information entropy module is used to enhance the edge features of the segmented target.

4. The machine learning-based auxiliary diagnosis method for esophageal tumors according to claim 3, characterized in that: A feature intensive module is provided in each layer of the encoder and each layer of the decoder. The feature intensive module includes a first convolutional layer, a second convolutional layer, a batch normalization layer and an activation function layer connected in sequence, and the input of the current layer of the feature intensive module is the feature map of all previous layers, and the feature map output in the current layer is the input in the subsequent layer.

5. The method for auxiliary diagnosis of esophageal tumors based on machine learning according to claim 4, characterized in that: The first convolution layer is a 1×1×1 convolution, and the second convolution layer is a deformable convolution.

6. The machine learning-based auxiliary diagnosis method for esophageal tumors according to claim 3, characterized in that: An attention module is also provided in each layer of the decoder, and the attention module includes a channel attention submodule and a spatial attention submodule; two multi-layer perceptrons are provided between the input and output of the spatial attention submodule, and two convolution blocks are provided in the channel attention submodule.

7. The method for auxiliary diagnosis of esophageal tumors based on machine learning according to claim 3, characterized in that: The classification model includes a global average pooling layer connected to the output end of the decoder, and a classifier connected to the global average pooling layer; the classifier is a support vector machine constructed based on a polynomial kernel function.

8. The method for auxiliary diagnosis of esophageal tumors based on machine learning according to claim 1, characterized in that: The method also includes performing denoising on the CT image, specifically including: performing wavelet transform on the CT image to obtain high-frequency components and low-frequency components of the CT image, and performing guided filtering on the high-frequency components and the low-frequency components respectively to obtain corresponding updated high-frequency components and updated low-frequency components, and performing image reconstruction based on inverse wavelet transform to obtain the denoised CT image.

9. An esophageal tumor auxiliary diagnosis device based on machine learning, characterized in that: The device is applied to a server, and the server is configured with patient label information and corresponding detection data. The device includes: An image acquisition module, used to determine the patient label information based on the patient's unique identification code, and retrieve the CT image in the corresponding detection data based on the patient label information; An image segmentation module, used for extracting features from the CT image, and segmenting the background of the CT image based on the extracted features to obtain a target feature map containing only a region of interest; The recognition module classifies and recognizes the target feature map based on the classification model to obtain a classification result.

10. The esophageal tumor auxiliary diagnosis device based on machine learning according to claim 9, characterized in that: The device also includes: The image noise reduction module is used to perform noise reduction processing on the CT image.

Images