Pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model

Abstract

The present invention provides a pan-organ lymph node metastatic cancer analysis system based on a joint deep learning model, which includes: acquiring and performing image segmentation and image annotation on lymphatic images to obtain domain images and image display responses corresponding to each image domain in the lymphatic images, then performing classification processing to obtain corresponding image domain classes, analyzing each class feature based on the class features corresponding to each image domain class to determine the abnormal information of the corresponding image domain class, determining the cancer cell information contained in the lymphatic images according to the abnormal information, establishing a cancer variable corresponding to each image domain in the lymphatic images based on the cancer cell information, establishing the degree of cancer cell metastasis of the corresponding patient based on the cancer variable, marking the cancer cell information and cancer variable corresponding to each image domain in the lymphatic images, generating and displaying the case information of the corresponding patient, and analyzing the basic situation of the patient in a timely manner after the patient has completed the scan to generate a valuable case for the convenience of doctors.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model. Background Art

[0002] Lymph nodes (LNs) are the core of the immune system and are secondary lymphoid organs distributed throughout the human body. Cancer in lymph nodes either originates from the lymph nodes or, more often, from metastases from the primary organ. Lymph nodes are usually the first metastatic sites of various cancers and are crucial for tumor staging and prognosis. The degree of tumor metastasis is considered a powerful predictor of recurrence and survival. Pan-organ lymph node metastatic cancer refers to cancer cells metastasizing from the primary focus to the lymph nodes of multiple organs in the body through the lymphatic circulation or blood circulation, forming metastatic cancer. This usually indicates that the cancer has progressed to an advanced stage and requires timely diagnosis and treatment.

[0003] Generally speaking, the analysis steps for pan-organ lymph node metastatic cancer are as follows: Doctors understand the patient's medical history and clinical manifestations and combine them with the medical imaging examinations performed by the patient to evaluate the severity of the cancer. When necessary, pathological examinations are required to determine whether there is cancer cell metastasis in the lymph nodes. Obviously, although the traditional method requires extremely high professional skills from doctors, doctors need to go through years of practice and learning to accurately determine whether there is cancer cell metastasis in the lymph nodes, and the work efficiency is greatly improved. Now, a system that can assist doctors in judging the condition is needed.

[0004] Therefore, the present invention provides a pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model. Summary of the Invention

[0005] The pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model of the present invention analyzes the basic situation of the patient in a timely manner after the patient has completed the scan, generates a reference-worthy case, and provides convenience for doctors.

[0006] The present invention provides a pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model, including:

[0007] An image segmentation and annotation module, which is used to obtain and perform image segmentation and image annotation on lymphatic images, and obtain the domain images and image display responses corresponding to each image domain in the lymphatic images;

[0008] A domain image classification module, which is used to perform classification processing on the image domains based on the image display responses, obtain the corresponding image domain classes, and based on the class features corresponding to each image domain class;

[0009] Anomaly analysis and localization module, which is used to analyze each of the class features to determine the anomaly information of the corresponding image domain class, and determine the cancer cell information contained in the lymphatic image according to the anomaly information;

[0010] Anomaly level determination module, which is used to establish the cancer variable corresponding to each image domain in the lymphatic image based on the cancer cell information, and establish the degree of cancer cell metastasis of the corresponding patient based on the cancer variable;

[0011] Information sorting and display module, which is used to mark the cancer cell information and cancer variable corresponding to each image domain in the lymphatic image, generate the case information of the corresponding patient and display it.

[0012] In an implementable manner,

[0013] It further includes:

[0014] Metastasis analysis and estimation module, which is used to establish the historical metastasis information and current health information of the corresponding patient according to the case information;

[0015] Estimate the metastasis of the current health information according to the historical metastasis information to obtain the cancer metastasis estimation information of the corresponding patient, and transmit it to the information sorting and display module for display.

[0016] In an implementable manner,

[0017] The image segmentation and annotation module includes:

[0018] Lymphatic image capture unit, which is used to scan the patient using a specified medical imaging technology to generate the lymphatic image of the patient;

[0019] Lymphatic image segmentation unit, which is used to perform sliding window segmentation on the lymphatic image to generate a number of image domains;

[0020] Lymphatic image annotation unit, which is used to establish an image comparison sample according to the preset cortical features, compare each of the image domains with the image comparison sample respectively, and establish the domain image display response corresponding to each image domain according to the comparison result.

[0021] In an implementable manner,

[0022] It further includes:

[0023] Lesion pre-analysis module, which is used to obtain the cortical information of the corresponding image domain according to the domain image display response;

[0024] Establish the abnormal cortical level of the corresponding patient according to the cortical information;

[0025] When the cortical abnormality level is greater than the preset level, draw the abnormal cortical contour contained in the corresponding image domain according to the cortical information;

[0026] Locate each of the abnormal cortical contours in the lymphatic image, and obtain the immune function corresponding to each organ in the lymphatic image, so as to obtain the immune abnormal function corresponding to each of the abnormal cortical contours;

[0027] Establish the manifestation lesion information of the corresponding patient according to the immune abnormal function.

[0028] In an implementable manner,

[0029] The domain image classification module includes:

[0030] An image pooling unit, which is used to perform image pooling processing on each of the image domains respectively by using a convolutional neural network to obtain an abstract image domain corresponding to each of the image domains;

[0031] A feature generation unit, which is used to establish abstract description information corresponding to the image domain based on the abstract image domain, establish a pooling correction rate based on the size ratio between the abstract image domain and the corresponding image domain, and establish image description information corresponding to the image domain according to the abstract description information and the pooling correction rate;

[0032] An image classification unit, which is used to establish description features corresponding to the image domain according to the image description information, classify the image domains with consistent description features into one category, and generate corresponding image domain classes;

[0033] A class feature establishment unit, which is used to establish basic features corresponding to the image domain class based on the description features, establish an image sequence according to the image domain class, generate a difference feature between each sequence graph and the basic features in the image sequence, and use each difference feature to correct the corresponding basic features to generate class features corresponding to the image domain class.

[0034] In an implementable manner,

[0035] The abnormal analysis and location module includes:

[0036] An image analysis and mapping unit, which is used to establish the image composition structure of each image domain class according to the class features, establish a mapping background corresponding to the image domain class according to the image composition structure, and map the images in the class in the same image domain class to the corresponding mapping background to obtain a corresponding mapping result;

[0037] An analysis starting point determination unit, which is used to determine the feature prominent contour of each image in the corresponding image domain class according to the class features, locate the feature prominent contour in the corresponding mapping result, and establish an analysis starting point position in the mapping result according to the location result;

[0038] Anomaly step-by-step analysis unit, which is used to retrieve corresponding normal information samples based on the class features, establish a search step size according to the normal information samples, and perform normal information screening and abnormal information screening on the mapping results starting from the analysis starting position to obtain the normal information and abnormal information corresponding to each mapping result;

[0039] Cancer cell determination unit, which is used to adjust the information parameters of the corresponding abnormal information according to the class features, generate abnormal information corresponding to the image domain class, establish a cancer cell distribution list corresponding to the image domain class and cancer cell actual situation information corresponding to the image domain class according to the abnormal information, find the positions of cancer cells contained in each lymph image according to the cancer cell distribution list, and establish cancer cell information corresponding to the patient in combination with the corresponding cancer cell actual situation information.

[0040] In an implementable manner,

[0041] The abnormal level determination module includes:

[0042] Model establishment and analysis unit, which is used to establish lymph information corresponding to the patient according to the lymph image, establish a cell model corresponding to the patient according to the lymph information and the cancer cell information, and mark the cancer cell model domain in the cell model;

[0043] Cancer variable analysis unit, which is used to determine the cancerous organs corresponding to each cancer cell model domain, establish an organ model domain according to the normal organ data corresponding to each cancerous organ, integrate the organ model domain into the cell model, and determine the cancerous area corresponding to each cancerous organ according to the integration result;

[0044] Organ function analysis unit, which is used to determine the influence parameter of cancer cells on the corresponding cancerous organ according to the cancerous area, and adjust the corresponding organ model domain by using the influence parameter to obtain the organ function characteristics corresponding to the patient;

[0045] Cancer cell information analysis unit, which is used to determine the cancerous depth of the corresponding cancerous organ according to the organ function characteristics, and combine the corresponding cancerous area to obtain the cancer variable corresponding to each image domain.

[0046] In an implementable manner,

[0047] The abnormal level determination module further includes:

[0048] Cancer metastasis analysis unit, which is used to retrieve the case information of the patient and combine the cancer variable corresponding to each image domain to obtain the cancer increase or decrease amount corresponding to each image domain;

[0049] Generate the cancer cell metastasis degree corresponding to the patient according to the cancer increase or decrease amount.

[0050] In an implementable manner,

[0051] The information sorting and display module includes:

[0052] An ordinary marking unit, used to mark the cancer cell information and cancer variable corresponding to each image domain in the lymph image, and generate a diagnostic image;

[0053] A lymph node screening unit, used to establish a grading screening mechanism according to a preset number of lymph node grading information, and use the grading screening mechanism to screen the diagnostic image;

[0054] A lymph node grading unit, used to obtain the lymph nodes included in the diagnostic image and the lymph node grade corresponding to each lymph node when the diagnostic image contains lymph nodes;

[0055] A case formulation generation unit, used to generate case information corresponding to the patient and display it according to the cancer cell information, cancer variable, and cancer variable when the diagnostic image contains lymph nodes.

[0056] In an implementable manner,

[0057] It further includes:

[0058] A suggestion generation module, used to generate a suggested treatment method for doctors to refer to according to the case information.

[0059] The beneficial effects that the present invention can achieve are as follows: In order to better assist doctors in diagnosis and treatment and improve doctors' work efficiency, first, the lymph image is segmented and labeled, so as to classify the obtained domain images and determine the class features corresponding to each domain class. By analyzing the class features, the abnormal information of the image domain class is further determined. Thus, the cancer cell information contained in the lymph image can be basically determined according to the abnormal information. Further, through deep learning analysis, the cancer variable and the degree of cancer cell metastasis corresponding to each image domain are obtained. Finally, a case information is generated for doctors to refer to. In this way, the cancer cells can be preliminarily screened automatically after the patient is scanned, reducing the workload of doctors. And the case information is generated shortly after the patient finishes the imaging scan, effectively avoiding the phenomena of misdiagnosis or taking the wrong case, facilitating the doctor to give symptomatic and effective treatment to the patient, enabling the doctor to better engage in the treatment work process, and reducing the waiting time of the patient.

[0060] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification or be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained through the structures specifically pointed out in the written specification, claims, and drawings.

[0061] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0062] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention. In the accompanying drawings:

[0063] Figure 1 It is a schematic diagram of the composition of the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model in the embodiment of the present invention;

[0064] Figure 2 It is a schematic diagram of the composition of the domain image classification module in the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model in the embodiment of the present invention;

[0065] Figure 3 It is a schematic diagram of the process of sliding window segmentation in the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model in the embodiment of the present invention;

[0066] Figure 4 It is a schematic diagram of the sliding window segmentation marking process in the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model in the embodiment of the present invention. Detailed Embodiments

[0067] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0068] Embodiment 1

[0069] This embodiment provides a pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model, as Figure 1 shown, including:

[0070] An image segmentation and annotation module, which is used to obtain and perform image segmentation and image annotation on lymphatic images to obtain domain images and image display responses corresponding to each image domain in the lymphatic images;

[0071] A domain image classification module, which is used to classify the image domains based on the image display responses to obtain corresponding image domain classes, and based on the class features corresponding to each image domain class;

[0072] An abnormal analysis and localization module, which is used to analyze each class feature to determine the abnormal information of the corresponding image domain class, and determine the cancer cell information contained in the lymphatic image according to the abnormal information;

[0073] An abnormal level determination module, which is used to establish cancer variables corresponding to each image domain in the lymphatic image based on the cancer cell information, and establish the degree of cancer cell metastasis of the corresponding patient based on the cancer variables;

[0074] An information sorting and display module, which is used to mark the cancer cell information and cancer variables corresponding to each image domain in the lymphatic image, generate the case information of the corresponding patient and display it.

[0075] In this example, the processes of image segmentation and image annotation are semantic segmentation and semantic annotation respectively;

[0076] In this example, the domain image is a part of the lymphatic image and is the result of image segmentation of the lymphatic image;

[0077] In this example, the process of classifying the image domain is to perform clustering analysis;

[0078] In this example, several sub-images included in the same image domain class have the same class characteristics, and each image domain class corresponds to one class characteristic;

[0079] In this example, the abnormal information represents the information that does not belong to normal lymph nodes in the image domain class;

[0080] In this example, the cancer cell information includes information such as the distribution and area of cancer cells in the lymphatic image;

[0081] In this example, the degree of cancer cell metastasis has 3 levels, namely: T stage (primary tumor): cancer cells are limited to the primary tumor site, N stage (lymph node metastasis): cancer cells have metastasized to nearby lymph nodes, M stage (distant metastasis): cancer cells have spread to other parts of the body, such as the liver, lungs, etc., staging stage:

[0082] In this example, each patient corresponds to one case information, and the case information is updated in real time as the patient undergoes examinations.

[0083] Working principle and beneficial effects of the above technical solution: To better assist doctors in diagnosis and treatment and improve doctors' work efficiency, first, lymph images are segmented and labeled, then the obtained domain images are classified, and the class features corresponding to each domain class are determined. By analyzing the class features, the abnormal information of the image domain class is further determined. Thus, the cancer cell information contained in the lymph image can be basically determined according to the abnormal information. Further, through deep learning analysis, the cancer variable and the degree of cancer cell metastasis corresponding to each image domain are obtained. Finally, a case information is generated for doctors' reference. In this way, the cancer cells can be preliminarily screened automatically after the patient is scanned, reducing the workload of doctors. And the case information is generated shortly after the patient finishes the imaging scan, effectively avoiding the phenomena of misdiagnosis or taking the wrong case, facilitating doctors to carry out symptomatic and effective treatment for patients, enabling doctors to better engage in the treatment process, and reducing the waiting time of patients.

[0084] Example 2

[0085] Based on Example 1, the pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model further includes:

[0086] A metastasis analysis and estimation module, used to establish the historical metastasis information and current health information of the corresponding patient according to the case information;

[0087] According to the historical metastasis information, perform metastasis estimation on the current health information to obtain the cancer metastasis estimation information of the corresponding patient, and transmit it to the information sorting and display module for display.

[0088] Working principle and beneficial effects of the above technical solution: To improve the practicability of the system and further reduce the workload of doctors, perform cancer metastasis estimation on the patient according to the patient's case information and current health information, so that doctors can have a clear understanding and better implement effective treatment methods.

[0089] Example 3

[0090] Based on Example 1, in the pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model, the image segmentation and labeling module includes:

[0091] A lymph image capturing unit, used to scan the patient using a specified medical imaging technology to generate the lymph image of the patient;

[0092] A lymph image segmentation unit, used to perform sliding window segmentation on the lymph image to generate a number of image domains;

[0093] The lymphatic image annotation unit is used to establish image comparison samples according to preset cortical features, compare each of the image domains respectively using the image comparison samples, and establish the domain image display response corresponding to each of the image domains according to the comparison results.

[0094] In this example, the specified medical imaging technology can be CT scan, MRI, PET-CT scan, ultrasound examination;

[0095] In this example, sliding window segmentation represents the process of dividing the lymphatic image into image domains with a specification of 3*3 pixels;

[0096] In this example, the preset cortical features represent the features presented when the cortical region of the lymph node unfolds an immune response;

[0097] In this example, the domain image display response represents the display response of the cortical region in the domain image, including: lymph node enlargement, lymphocyte proliferation, inflammatory response, leukocyte aggregation, lymphoid follicle dilation;

[0098] In this example, the process of performing sliding window segmentation is as Figure 3 shown.

[0099] The working principle and beneficial effects of the above technical solution: In order to analyze the features of lymphatic images simply and efficiently, after scanning, the lymphatic image of the patient is generated, and then it is segmented into several image domains. By comparing the cortical features contained in each image domain, the display response of the patient is determined, which is convenient for subsequent phase domain image classification work.

[0100] Example 4

[0101] Based on Example 3, the pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model further includes:

[0102] A lesion pre-analysis module, which is used to obtain the cortical information of the corresponding image domain according to the domain image display response;

[0103] Establish an abnormal cortical grade corresponding to the patient according to the cortical information;

[0104] When the cortical abnormality grade is greater than the preset grade, draw the abnormal cortical contour contained in the corresponding image domain according to the cortical information;

[0105] Locate each of the abnormal cortical contours in the lymphatic image, and obtain the immune function corresponding to each organ in the lymphatic image, and obtain the immune abnormal function corresponding to each of the abnormal cortical contours;

[0106] Establish the manifestation lesion information corresponding to the patient according to the immune abnormal function.

[0107] In this example, the cortical information represents the responses of the immune work carried out in the cortical region of the lymph node, including: antigen presentation and recognition, T cell activation, B cell activation, cytokine release, proliferation and diffusion;

[0108] In this example, the abnormal cortical grades include: unstimulated grade, activated grade, tumor infiltrated grade, metastatic grade;

[0109] In this example, the abnormal cortical contour represents the contour formed by the abnormal cortex in the lymphatic image;

[0110] In this example, each organ corresponds to one or more immune functions;

[0111] In this example, the dominant lesion information includes lymph node enlargement, lymphocyte proliferation, inflammatory response, leukocyte aggregation, and lymph follicle dilation.

[0112] The working principle and beneficial effects of the above technical solution: By analyzing the cortical information to judge the patient's condition, it can effectively further analyze the patient's immune situation and facilitate doctors and patients to understand the actual condition.

[0113] Example 5

[0114] Based on Example 1, the pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model is characterized in that the domain image classification module, as Figure 2 shown, includes:

[0115] An image pooling unit, which is used to perform image pooling processing on each of the image domains respectively by using a convolutional neural network to obtain an abstract image domain corresponding to each of the image domains;

[0116] A feature generation unit, which is used to establish abstract description information of the corresponding image domain based on the abstract image domain, establish a pooling correction rate based on the size ratio between the abstract image domain and the corresponding image domain, and establish image description information of the corresponding image domain according to the abstract description information and the pooling correction rate;

[0117] An image classification unit, which is used to establish description features of the corresponding image domain according to the image description information, classify the image domains with consistent description features into one category, and generate a corresponding image domain class;

[0118] A class feature establishment unit, which is used to establish basic features of the corresponding image domain class based on the description features, establish an image sequence according to the image domain class, generate a difference feature between each sequence diagram and the basic feature in the image sequence, and use each difference feature to correct the corresponding basic feature to generate class features of the corresponding image domain class.

[0119] In this example, image pooling processing represents the process of reducing the size of the image domain;

[0120] In this example, the abstract image domain represents the result of pooling the image domain, and its essence is also the image domain;

[0121] In this example, the abstract description information represents an abstract and error-prone description of the image domain;

[0122] In this example, the image description information represents the information used to represent the basic situation of the image domain;

[0123] In this example, the size ratio represents the area ratio between the abstract image domain and the corresponding image domain;

[0124] In this example, the pooling correction rate is related to the size ratio and is positively correlated;

[0125] In this example, the description feature represents the appearance feature of the image domain;

[0126] In this example, the basic feature represents the common feature of multiple image domains in the image domain class;

[0127] In this example, the image sequence is a sequence generated by arranging the images in the image domain class;

[0128] In this example, the difference feature represents the difference between the basic features of the sequence diagram domains;

[0129] In this example, the number of class features is 1, and one image domain class corresponds to one class feature.

[0130] The working principle and beneficial effects of the above technical solution: In order to better analyze the features contained in each image domain, first perform image pooling processing on the image domain to obtain its corresponding abstract image domain, then establish the abstract description features of the abstract image domain, and combine the size ratio between the two image domains after pooling to establish the description features of the image domain. Furthermore, classify the image domains with consistent description features, and finally use the difference features between the basic features of different sequence diagram domains to establish the class features of the classified image domains. Use the pooling technology to simplify the structure of the image domain, so that the image domains with similar features can be classified, reducing the classification error and improving the efficiency of subsequent image analysis.

[0131] Example 6

[0132] Based on Example 1, the abnormal analysis and positioning module of the pan-organ lymph node metastatic cancer analysis system based on the joint deep learning model includes:

[0133] An image analysis mapping unit, which is used to establish an image composition structure for each image domain class according to the class features, establish a mapping background corresponding to the image domain class according to the image composition structure, and map the in-class images in the same image domain class into the corresponding mapping background to obtain corresponding mapping results;

[0134] An analysis starting point determination unit, which is used to determine the characteristic prominent contour of each in-class image in the corresponding image domain class according to the class features, locate the characteristic prominent contour in the corresponding mapping result, and establish an analysis starting point position in the mapping result according to the positioning result;

[0135] An abnormal step-by-step analysis unit, which is used to retrieve the corresponding normal information sample based on the class features, establish a search step size according to the normal information sample, and perform normal information screening and abnormal information screening on the mapping result starting from the analysis starting point position to obtain the normal information and abnormal information corresponding to each mapping result;

[0136] A cancer cell determination unit, which is used to adjust the information parameters of the corresponding abnormal information according to the class features, generate abnormal information for the corresponding image domain class, establish a cancer cell distribution list for the corresponding image domain class and cancer cell live information for the corresponding image domain class according to the abnormal information, find the positions of cancer cells contained in each lymph image according to the cancer cell distribution list, and establish cancer cell information for the corresponding patient in combination with the corresponding cancer cell live information.

[0137] In this example, the image composition structure represents the composition of the average pixel values corresponding to different regions in the image domain class;

[0138] In this example, the composition structure of the multiple mapping regions included in the mapping background is consistent with the image composition structure;

[0139] In this example, the in-class image represents the image domain in the image domain class;

[0140] In this example, the characteristic prominent contour represents the contour corresponding to the image that can be used to express the class features in the in-class image;

[0141] In this example, the analysis starting point position represents the starting point for analyzing abnormal information;

[0142] In this example, the normal information sample represents the sample presented by the normal information in the image domain class;

[0143] In this example, the cancer cell distribution list represents a list that statistically counts the distribution quantities of cancer cells in different image domains.

[0144] Working principle and beneficial effects of the above technical solution: In order to further determine the basic information of cancer cells in a patient, first determine the image composition of each image domain class according to class features, so as to establish a mapping background for mapping. Then, the normal information and abnormal information can be searched in the mapping structure, and the abnormal information can be established. In this way, the cancer cell distribution list and cancer cell actual situation information in the patient's body can be determined, and the location and basic information of cancer cells can be further determined in detail. Moreover, during the analysis process, the same type of images are analyzed simultaneously, improving the analysis efficiency.

[0145] Example 7

[0146] Based on the system for analyzing metastatic cancer of pan-organ lymph nodes based on a joint deep learning model in Example 1, the abnormal level determination module includes:

[0147] A model establishment and analysis unit, used to establish lymph information corresponding to a patient according to the lymph image, establish a cell model corresponding to the patient according to the lymph information and the cancer cell information, and mark the cancer cell model domain in the cell model;

[0148] A cancer variable analysis unit, used to determine the cancerous organ corresponding to each cancer cell model domain, establish an organ model domain according to the normal organ data corresponding to each cancerous organ, and integrate the organ model domain into the cell model, and determine the cancerous area corresponding to each cancerous organ according to the integration result;

[0149] An organ function analysis unit, used to determine the influence parameter of cancer cells on the corresponding cancerous organ according to the cancerous area, and adjust the corresponding organ model domain by using the influence parameter to obtain the organ function characteristics of the corresponding patient;

[0150] A cancer cell information analysis unit, used to determine the cancer depth of the corresponding cancerous organ according to the organ function characteristics, and combine the corresponding cancerous area to obtain the cancer variable corresponding to each image domain.

[0151] In this example, the cancer cell model domain represents the area in the cell model that belongs to the cancer cell model. That is to say, the cancer cell model domain is a part of the cancer cell model.

[0152] Working principle and beneficial effects of the above technical solution: In order to further determine the degree of cancer cell lesions in a patient, first establish a cell model that can express the lymphatic information of the patient based on the lymphatic images, then mark the area of cancer cells in the cell model, determine the corresponding cancerous organs, and establish an organ model domain, so as to know the cancerous area of the cancerous organs, and combine the organ function characteristics of the organs to determine the corresponding cancerous depth. In this way, the amount of cancer in each image domain can be determined, and the cancer results of each organ can be accurately analyzed through this method, enabling doctors to master the existing functions of each organ and facilitating symptomatic treatment.

[0153] Example 8

[0154] Based on Example 7, for the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model, the abnormal level determination module further includes:

[0155] Cancer metastasis analysis unit, used to retrieve the case information of the patient and obtain the cancer increase or decrease amount corresponding to each image domain in combination with the amount of cancer in each image domain;

[0156] Generate the degree of cancer cell metastasis corresponding to the patient according to the cancer increase or decrease amount.

[0157] Working principle and beneficial effects of the above technical solution: In order to facilitate doctors to further analyze the physical health status of patients, determine the cancer increase or decrease amount of each image domain according to the case information of the patient, so as to determine the degree of cancer cell metastasis, enabling doctors to master the fluctuations in the physical health of patients in a short time.

[0158] Example 9

[0159] Based on Example 1, for the pan-organ lymph node metastatic cancer analysis system based on the combined deep learning model, the information sorting and display module includes:

[0160] Ordinary marking unit, used to mark the cancer cell information and the amount of cancer corresponding to each image domain in the lymphatic image to generate a diagnostic image;

[0161] Lymph node screening unit, used to establish a grade screening mechanism according to a preset number of lymph node grade information, and use the grade screening mechanism to screen the diagnostic image;

[0162] Lymph node grading unit, used to obtain the lymph nodes included in the diagnostic image and the corresponding lymph node grades of each lymph node when the diagnostic image contains lymph nodes;

[0163] A case formulation generation unit is configured to, when lymph nodes are included in the diagnostic image, generate and display case information corresponding to a patient based on the cancer cell information, cancer variables, and cancer variables.

[0164] In this example, the lymph node grades include: Grade 0 (lymph nodes not involved, no swelling, normal texture), Grade 1 (lymph nodes slightly involved, slightly palpable swelling, may have slight discomfort), Grade 2 (lymph nodes moderately involved, significantly enlarged, harder texture, may be accompanied by pain or tenderness), Grade 3 (lymph nodes severely involved, significantly enlarged, hard texture, obvious pain, may affect the function of surrounding tissues);

[0165] In this example, the marking process is as Figure 4 shown.

[0166] The working principle and beneficial effects of the above technical solution: To facilitate patients and doctors to understand the patient's physical condition, the lymph node grade corresponding to each lymph node is displayed in a concise and clear manner, and the case information of the patient is established based on the cancer cell information, cancer information, and cancer variables, improving the case and enhancing the authenticity of the case.

[0167] Example 10

[0168] Based on Example 1, the pan-organ lymph node metastatic cancer analysis system based on a combined deep learning model further includes:

[0169] A treatment suggestion generation module is configured to generate a suggested treatment method for doctors' reference based on the case information.

[0170] The working principle and beneficial effects of the above technical solution: Doctors can make appropriate adjustments according to the suggested treatment method and then treat the patient, reducing the efficiency of doctors issuing medical orders.

[0171] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A pan-organ lymph node metastasis cancer analysis system based on a joint deep learning model, characterized by: include: An image segmentation and annotation module is used to obtain and perform image segmentation and image annotation on the lymphatic image, and obtain a domain image and image display response corresponding to each image domain in the lymphatic image; A domain image classification module, used to classify the image domain based on the image display response, obtain the corresponding image domain class, and obtain the class feature corresponding to each image domain class; An abnormality analysis and positioning module, used to analyze each of the class features to determine abnormal information of the corresponding image domain class, and determine cancer cell information contained in the lymphatic image based on the abnormal information; an abnormality level determination module, used to establish a cancer number corresponding to each image domain in the lymphatic image based on the cancer cell information, and to establish a cancer cell metastasis degree of the corresponding patient based on the cancer number; An information sorting and display module is used to mark the cancer cell information and cancer variable corresponding to each image domain in the lymphatic image, generate case information of the corresponding patient and display it; The image segmentation and annotation module comprises: A lymphatic image capturing unit, used for scanning a patient using a specified medical imaging technology to generate a lymphatic image of the patient; A lymphatic image segmentation unit, used for performing sliding window segmentation on the lymphatic image to generate a plurality of image domains; A lymphatic image annotation unit, used to establish image comparison samples according to preset cortical features, use the image comparison samples to compare each of the image domains respectively, and establish a domain image display response corresponding to each of the image domains according to the comparison results; The domain image classification module comprises: An image pooling unit, used to perform image pooling processing on each of the image domains using a convolutional neural network to obtain an abstract image domain corresponding to each of the image domains; a feature generation unit, used to establish abstract description information of a corresponding image domain based on the abstract image domain, establish a pooling correction rate based on a size ratio between the abstract image domain and the corresponding image domain, and establish image description information of the corresponding image domain according to the abstract description information and the pooling correction rate; An image classification unit, used for establishing description features of a corresponding image domain according to the image description information, classifying image domains with consistent description features into one category, and generating a corresponding image domain class; A class feature establishment unit is used to establish basic features of the corresponding image domain class based on the descriptive features, establish an image sequence according to the image domain class, generate difference features between each sequence image and the basic features in the image sequence, use each of the difference features to correct the corresponding basic features, and generate class features of the corresponding image domain class.

2. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: Also includes: A metastasis analysis and estimation module, used to establish historical metastasis information and current health information of the corresponding patient according to the case information; The current health information is estimated to metastasize according to the historical metastasis information to obtain cancer metastasis estimation information of the corresponding patient, and the estimated information is transmitted to the information sorting and display module for display.

3. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: Also includes: A lesion pre-analysis module, used for obtaining cortical information of a corresponding image domain according to the domain image display response; establishing an abnormal cortical grade for the corresponding patient according to the cortical information; When the cortical abnormality level is greater than a preset level, drawing an abnormal cortical contour contained in a corresponding image domain according to the cortical information; Locating each of the abnormal cortical contours in the lymphatic image, and acquiring the immune function corresponding to each organ in the lymphatic image, to obtain the abnormal immune function corresponding to each of the abnormal cortical contours; The manifested lesion information of the corresponding patient is established according to the abnormal immune function.

4. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: Abnormal analysis and positioning module, including: An image analysis mapping unit, used to establish an image composition structure of each image domain class according to the class features, establish a mapping background of the corresponding image domain class according to the image composition structure, and map the class images in the same image domain class to the corresponding mapping background to obtain a corresponding mapping result; An analysis starting point establishing unit, used to determine the characteristic prominent contour of the image in each class in the corresponding image domain class according to the class feature, locate the characteristic prominent contour in the corresponding mapping result, and establish the analysis starting point position in the mapping result according to the positioning result; an abnormal step-by-step analysis unit, used to retrieve corresponding normal information samples based on the class features, establish a search step length according to the normal information samples, and filter the mapping results for normal information and abnormal information starting from the analysis starting point to obtain normal information and abnormal information corresponding to each mapping result; The cancer cell determination unit is used to adjust the information parameters of the corresponding abnormal information according to the class characteristics, generate abnormal information of the corresponding image domain class, establish a cancer cell distribution list of the corresponding image domain class and cancer cell actual information of the corresponding image domain class according to the abnormal information, search for the position of the cancer cells contained in each of the lymphatic images according to the cancer cell distribution list, and establish the cancer cell information of the corresponding patient in combination with the corresponding cancer cell actual information.

5. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: The abnormal level determination module includes: A model building and analysis unit, used to build lymphatic information of a corresponding patient according to the lymphatic image, build a cell model of the corresponding patient according to the lymphatic information and the cancer cell information, and mark a cancer cell model domain in the cell model; a cancer variable analysis unit, used to determine the cancerous organ corresponding to each cancer cell model domain, establish an organ model domain according to the normal organ data corresponding to each cancerous organ, integrate the organ model domain into the cell model, and determine the cancerous area corresponding to each cancerous organ according to the fusion result; an organ function analysis unit, used to determine the influence parameters of cancer cells on the corresponding cancerous organ according to the cancerous area, and to adjust the corresponding organ model domain using the influence parameters to obtain the organ function characteristics of the corresponding patient; The cancer cell information analysis unit is used to determine the cancer depth of the corresponding cancerous organ according to the organ function characteristics, and obtain the cancer volume corresponding to each image domain in combination with the corresponding cancer area.

6. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 5, characterized in that: The abnormal level determination module further includes: A cancer metastasis analysis unit is used to retrieve the case information of the patient, combine the cancer amount corresponding to each image domain, and obtain the cancer increase or decrease amount corresponding to each image domain; The degree of cancer cell metastasis of the corresponding patient is generated according to the increase or decrease amount of canceration.

7. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: The information sorting and display module comprises: A common marking unit, used for marking cancer cell information and cancer quantity corresponding to each image domain in the lymphatic image to generate a diagnostic image; A lymph node screening unit, used to establish a level screening mechanism according to a plurality of preset lymph node level information, and to screen the diagnostic image using the level screening mechanism; a lymph node grading unit, used for obtaining the lymph nodes contained in the diagnostic image and the lymph node grade corresponding to each of the lymph nodes when the diagnostic image contains lymph nodes; The case generation unit is used to generate case information of the corresponding patient according to the cancer cell information, cancer quantity and cancer quantity and display the case information when the diagnostic image contains lymph nodes.

8. The pan-organ lymph node metastasis cancer analysis system based on the joint deep learning model according to claim 1, characterized in that: Also includes: The suggestion generation module is used to generate suggested treatment methods based on the case information for the doctor's reference.

Images