A method and system for detecting endometrial cancer based on deep learning

Abstract

The application provides a kind of detection method and system of endometrial carcinoma based on deep learning, it is related to cancer detection technical field, the method comprises: obtaining preliminary survey data;Determine the preliminary survey result;Determine whether to need to carry out histopathological examination;If it needs to carry out histopathological examination, obtain the hysteroscopy guided biopsy data of patient to be measured, determine the histopathological diagnosis result;Determine whether to need to carry out further detection;If it needs to carry out further detection, obtain the patient physiological state data, medical image data and laboratory test data of patient to be measured;Determine patient physiological state influence coefficient;The medical image data, laboratory test data and patient physiological state influence coefficient are handled by the trained tumor deterioration time prediction model, determine the predicted tumor deterioration time;Determine endometrial carcinoma risk coefficient;Generation detection report.According to the application, the accuracy of the detection result of endometrial carcinoma can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cancer detection, and particularly relates to a deep learning-based endometrial carcinoma detection method and system. BACKGROUND

[0002] In the related art, endometrial carcinoma can be detected by in-vivo pathological detection of a patient to be detected, which can improve the detection accuracy, but the related art is difficult to predict the tumor deterioration speed, that is, it is difficult to predict the tumor deterioration time according to the ADC value and histological type and other data.

[0003] The information disclosed in the background section of this application is only intended to deepen the understanding of the general background of the application and should not be regarded as acknowledging or implying in any form that this information constitutes prior art known to those skilled in the art. SUMMARY

[0004] The present application provides a deep learning-based endometrial carcinoma detection method and system, which can solve the technical problem that the related art is difficult to predict the tumor deterioration time.

[0005] According to a first aspect of the present application, a deep learning-based endometrial carcinoma detection method is provided, comprising:

[0006] Obtaining preliminary investigation data of a patient to be detected;

[0007] Determining a preliminary investigation result according to the preliminary investigation data;

[0008] Determining whether histopathological detection is needed according to the preliminary investigation result;

[0009] If histopathological detection is needed, obtaining hysteroscopy-guided biopsy data of the patient to be detected, and determining a histopathological diagnosis result according to the hysteroscopy-guided biopsy data;

[0010] Determining whether further detection is needed according to the histopathological diagnosis result;

[0011] If further detection is needed, obtaining patient physiological state information, medical image information and laboratory test information of the patient to be detected;

[0012] Determining a patient physiological state influence coefficient according to the patient physiological state information;

[0013] Processing the medical image information, the laboratory test information and the patient physiological state influence coefficient by a trained tumor deterioration time prediction model to determine a predicted tumor deterioration time;

[0014] determining an endometrial cancer risk coefficient according to the predicted tumor progression time;

[0015] generating a detection report according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient.

[0016] According to the present application, a preliminary investigation result is determined according to the preliminary investigation data, comprising:

[0017] According to the preliminary investigation data, an abnormal bleeding identification result, an abnormal discharge identification result and an abnormal pain identification result are determined;

[0018] According to the abnormal bleeding identification result, the abnormal discharge identification result and the abnormal pain identification result, the preliminary investigation result is determined.

[0019] According to the present application, a patient physiological state influence coefficient is determined according to the patient physiological state information, comprising:

[0020] According to the patient physiological state information, a patient age, a patient BMI, a patient disease history, a toxic T cell density, a helper regulatory T cell ratio and a macrophage ratio are determined;

[0021] According to the patient disease history, a diabetes identification result and a cardiovascular disease identification result are obtained;

[0022] According to the patient age, a patient age risk identification result is determined;

[0023] According to the patient age risk identification result, the patient BMI, the toxic T cell density, the helper regulatory T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, the patient physiological state influence coefficient is determined.

[0024] According to the present application, a patient physiological state influence coefficient is determined according to the patient age risk identification result, the patient BMI, the toxic T cell density, the helper regulatory T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, comprising:

[0025] According to the formula

[0026]

[0027] determining a patient physiological state influence coefficient of the i-th to-be-tested patient , wherein if is a conditional function, and is a logical operator of "and", is a diabetes identification result of the i-th to-be-tested patient, , is a patient BMI of the i-th to-be-tested patient, The patient age risk identification result for the i-th patient to be tested. , For the cardiovascular disease identification result of the i-th patient, , The density of toxic T cells in the i-th patient being tested. To preset the cytotoxic T cell density threshold, The ratio of helper regulatory T cells for the i-th patient being tested. To preset the threshold for the auxiliary regulation of the T cell ratio, The ratio of macrophages in the i-th patient being tested. The preset macrophage ratio threshold is used.

[0028] According to the present invention, the training steps of the tumor progression time prediction model include:

[0029] Obtain historical physiological data, historical medical imaging data, and historical laboratory test data from multiple historical patients;

[0030] Based on the historical patient physiological status data, the influence coefficient of historical patient physiological status was determined;

[0031] Based on the historical medical imaging data, determine the historical degree of myometrial infiltration, historical ADC value, and historical depth of cervical stroma invasion.

[0032] Based on the historical laboratory test data, determine the risk level of historical histological types and the identification results of historical lymphovascular space infiltration;

[0033] The historical patient physiological status influence coefficient, historical myometrial invasion degree, historical ADC value, historical cervical stromal invasion depth, historical histological type risk level, and historical lymphovascular space invasion identification results are processed by the tumor deterioration time prediction model to obtain the historical training deterioration time.

[0034] Obtain the actual historical deterioration time of multiple historical patients;

[0035] Based on the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical myometrial infiltration degree, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type risk level, and the historical lymphovascular space infiltration identification results, the training loss function of the tumor deterioration time prediction model is determined.

[0036] The tumor progression time prediction model is trained according to the training loss function to obtain the trained tumor progression time prediction model.

[0037] According to the application, the training loss function of the tumor deterioration time prediction model is determined according to the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical muscle layer infiltration degree, the historical ADC value, the historical depth of cervical stroma invasion, the historical histological type risk level and the historical lymphatic vessel space infiltration identification result, and includes:

[0038] According to the formula

[0039]

[0040] The training loss function of the tumor deterioration time prediction model is determined , wherein, is the historical training deterioration time of the kth historical patient, is the historical actual deterioration time of the kth historical patient, is the historical ADC value of the kth historical patient, is a preset ADC value threshold, is the historical muscle layer infiltration degree of the kth historical patient, is a preset muscle layer infiltration degree threshold, is the historical depth of cervical stroma invasion of the kth historical patient, is a preset historical depth of cervical stroma invasion threshold, is the historical histological type risk level of the kth historical patient, is a preset histological type risk level threshold, is the historical lymphatic vessel space infiltration identification result of the kth historical patient, is the historical patient physiological state influence coefficient of the kth historical patient, is a preset patient physiological state influence coefficient threshold, K is the number of historical patients, k≤K, k and K are both positive integers.

[0041] According to the application, the medical image data, the laboratory test data and the patient physiological state influence coefficient are processed by the trained tumor deterioration time prediction model to determine the predicted tumor deterioration time, including:

[0042] According to the medical image data, the muscle layer infiltration degree, the ADC value and the depth of cervical stroma invasion are determined;

[0043] According to the laboratory test data, the histological type risk level and the lymphatic vessel space infiltration identification result are determined;

[0044] ​The tumor deterioration time prediction model is trained, and the patient physiological state influence coefficient, the muscle layer infiltration degree, the ADC value, the cervical interstitial invasion depth, the histological type danger level and the lymphatic vessel space infiltration recognition result are processed to obtain a predicted tumor deterioration time.

[0045] According to the application, according to the predicted tumor deterioration time, the endometrial cancer danger coefficient is determined, comprising:

[0046] According to the predicted tumor deterioration time and the set tumor deterioration time threshold, a first difference value is determined.

[0047] According to the first difference value and the set tumor deterioration time threshold, a first ratio is determined.

[0048] According to the first ratio, the endometrial cancer danger coefficient is determined.

[0049] According to the second aspect of the application, a deep learning-based endometrial cancer detection system is provided, comprising:

[0050] The preliminary data module obtains the preliminary investigation data of the patient to be tested.

[0051] The preliminary result module determines the preliminary investigation result according to the preliminary investigation data.

[0052] The pathological detection module determines whether histopathological detection is needed according to the preliminary investigation result.

[0053] The pathological result module obtains the hysteroscopy-guided biopsy data of the patient to be tested if histopathological detection is needed, and determines the histopathological diagnosis result according to the hysteroscopy-guided biopsy data.

[0054] The deep detection module determines whether further detection is needed according to the histopathological diagnosis result.

[0055] The detailed information module obtains the patient physiological state information, medical image information and laboratory test information of the patient to be tested if further detection is needed.

[0056] The influence coefficient module determines the patient physiological state influence coefficient according to the patient physiological state information.

[0057] The prediction time module processes the medical image information, the laboratory test information and the patient physiological state influence coefficient through a trained tumor deterioration time prediction model to determine a predicted tumor deterioration time.

[0058] The danger coefficient module determines the endometrial cancer danger coefficient according to the predicted tumor deterioration time.

[0059] a detection report module configured to generate a detection report according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient.

[0060] Technical effects: According to the present application, the preliminary investigation of the to-be-tested patient is performed to determine whether histopathological detection is needed, the accuracy of the detection result is improved through histopathological detection, further, in the case of confirmed endometrial cancer, the tumor deterioration time is predicted according to the patient physiological state data, medical image data and laboratory test data of the to-be-tested patient, the risk degree of the tumor is evaluated according to the predicted tumor deterioration time, the endometrial cancer risk coefficient is determined, and the detection report is generated, which improves the comprehensiveness and accuracy of the detection result of endometrial cancer. When determining the patient physiological state influence coefficient, the patient physiological state influence coefficient can be determined according to the patient age risk identification result, patient BMI, toxic T cell density, helper T cell ratio, macrophage ratio, diabetes identification result and cardiovascular disease identification result. In the calculation process, the health status of the to-be-tested patient can be evaluated from four aspects of metabolic status, age status, cardiovascular disease status and immune status, which improves the comprehensiveness and accuracy of the patient physiological state influence coefficient. When determining the training loss function, the training loss function of the tumor deterioration time prediction model can be determined according to the historical patient physiological state influence coefficient, historical training deterioration time, historical actual deterioration time, historical myometrial invasion degree, historical ADC value, historical depth of cervical stroma invasion, historical histological type risk level and lymphatic vessel space invasion identification result. In the calculation process, the influence of the historical patient physiological state influence coefficient, historical training deterioration time, historical actual deterioration time, historical myometrial invasion degree, historical ADC value, historical depth of cervical stroma invasion, historical histological type risk level and lymphatic vessel space invasion identification result on the tumor deterioration speed is determined, and the influence of the above data on the error of the historical training deterioration time is determined, and the training loss function is set based on the influence and the relative error of the historical training deterioration time, so that the tumor deterioration time prediction model reduces the training loss function in the training process, and more accurately improves the precision of the tumor deterioration time prediction model.

[0061] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, rather than limiting the present application. Other features and aspects of the present application will be more apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other embodiments can also be obtained from these drawings without creative labor.

[0063] Figure 1 An exemplary flowchart of a method for detecting endometrial cancer based on deep learning according to an embodiment of the present application is shown.

[0064] Figure 2 An exemplary diagram for determining a preliminary investigation result according to an embodiment of the present application is shown.

[0065] Figure 3 An exemplary diagram for determining a patient physiological state influence coefficient according to an embodiment of the present application is shown.

[0066] Figure 4 An exemplary diagram for determining a predicted tumor malignancy time according to an embodiment of the present application is shown.

[0067] Figure 5 An exemplary diagram for determining an endometrial cancer risk coefficient according to an embodiment of the present application is shown.

[0068] Figure 6 An exemplary block diagram of a system for detecting endometrial cancer based on deep learning according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0070] The technical solutions of the present application will be described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and some embodiments may not be described again for the same or similar concepts or processes.

[0071] Figure 1 An exemplary flowchart of a method for detecting endometrial cancer based on deep learning according to an embodiment of the present application is shown, which comprises:

[0072] Step S1, obtaining preliminary investigation data of a patient to be tested;

[0073] Step S2, determining a preliminary investigation result according to the preliminary investigation data;

[0074] Step S3, determining whether histopathological detection is needed according to the preliminary investigation result;

[0075] Step S4, if histopathological detection is needed, obtaining hysteroscopy-guided biopsy data of the patient to be tested, and determining a histopathological diagnosis result according to the hysteroscopy-guided biopsy data;

[0076] Step S5, determining whether further detection is needed according to the histopathological diagnosis result;

[0077] Step S6, if further detection is needed, obtaining patient physiological state information, medical image information and laboratory test information of the patient to be tested;

[0078] Step S7, determining a patient physiological state influence coefficient according to the patient physiological state information;

[0079] Step S8, processing the medical image information, the laboratory test information and the patient physiological state influence coefficient by using a trained tumor deterioration time prediction model to determine a predicted tumor deterioration time;

[0080] Step S9, determining an endometrial cancer risk coefficient according to the predicted tumor deterioration time;

[0081] Step S10, generating a detection report according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient.

[0082] The detection method of endometrial cancer based on deep learning according to the embodiments of the present application can preliminarily investigate the patient to be tested, determine whether histopathological detection is needed, improve the accuracy of the detection result through histopathological detection, further predict the tumor deterioration time according to the patient physiological state information, the medical image information and the laboratory test information of the patient to be tested in the case of diagnosing endometrial cancer, evaluate the risk degree of the tumor according to the predicted tumor deterioration time, determine an endometrial cancer risk coefficient, and generate a detection report, thereby improving the comprehensiveness and accuracy of the detection result of endometrial cancer.

[0083] According to an embodiment of the present application, in step S1, preliminary investigation data of the patient to be tested is obtained.

[0084] For example, the patient is preliminarily investigated through oral inquiry or questionnaire investigation to investigate whether the patient has abnormal symptoms and obtain preliminary investigation data.

[0085] According to one embodiment of the present application, in step S2, a preliminary investigation result is determined according to the preliminary investigation data.

[0086] Figure 2 A schematic diagram of determining a preliminary investigation result according to an embodiment of the present application is exemplarily shown.

[0087] According to one embodiment of the present application, step S2 comprises:

[0088] In step S21, abnormal bleeding identification result, abnormal discharge identification result and abnormal pain identification result are determined according to the preliminary investigation data.

[0089] In step S22, a preliminary investigation result is determined according to the abnormal bleeding identification result, the abnormal discharge identification result and the abnormal pain identification result.

[0090] For example, whether the patient to be tested has abnormal uterine bleeding (such as postmenopausal bleeding, premenopausal abnormal bleeding), abnormal vaginal discharge (such as discharge being serous or bloody) and abnormal lower abdominal pain is determined by oral inquiry or questionnaire, if the patient to be tested has abnormal uterine bleeding, the abnormal bleeding identification result is 1, otherwise, the abnormal bleeding identification result is 0, if the patient to be tested has abnormal vaginal discharge, the abnormal discharge identification result is 1, otherwise, the abnormal discharge identification result is 0, if the patient to be tested has abnormal lower abdominal pain, the abnormal pain identification result is 1, otherwise, the abnormal pain identification result is 0; the preliminary investigation result is determined according to the sum of the abnormal bleeding identification result, the abnormal discharge identification result and the abnormal pain identification result.

[0091] According to one embodiment of the present application, in step S3, whether histopathological detection is needed is determined according to the preliminary investigation result.

[0092] For example, if the preliminary investigation result is equal to 0, it means that no warning symptoms occur, and no histopathological detection is needed, if the preliminary investigation result is greater than 0, it means that abnormal uterine bleeding or abnormal vaginal discharge or abnormal lower abdominal pain occurs, and histopathological detection is needed.

[0093] According to one embodiment of the present application, in step S4, if histopathological detection is needed, hysteroscopy-guided biopsy data of the patient to be tested is obtained, and a histopathological diagnosis result is determined according to the hysteroscopy-guided biopsy data.

[0094] For example, if histopathological detection is needed, hysteroscopy-guided biopsy is performed on the patient to be tested, hysteroscopy-guided biopsy data is obtained, histopathological examination is performed on the hysteroscopy-guided biopsy data obtained by sampling, and a histopathological diagnosis result is determined, if it is determined that the patient to be tested has endometrial cancer, the histopathological diagnosis result is 1, otherwise, the histopathological diagnosis result is 0.

[0095] According to an embodiment of the present application, in step S5, it is determined whether further detection is needed according to the histopathological diagnosis result.

[0096] For example, if the histopathological diagnosis result is 1, indicating that endometrial cancer is determined, further detection is needed, and if the histopathological diagnosis result is 0, indicating that endometrial cancer is not determined, further detection is not needed.

[0097] According to an embodiment of the present application, in step S6, if further detection is needed, patient physiological state data, medical image data and laboratory test data of the patient to be tested are obtained.

[0098] For example, according to the medical history and normal physiological condition test of the patient to be tested, patient physiological state data is obtained, transvaginal ultrasound, pelvic MRI and other tests are performed on the patient to be tested to obtain medical image data, laboratory tests are performed on the samples obtained by hysteroscopy guided biopsy to obtain laboratory test data.

[0099] According to an embodiment of the present application, in step S7, a patient physiological state influence coefficient is determined according to the patient physiological state data.

[0100] Figure 3 An exemplary schematic diagram of determining a patient physiological state influence coefficient according to an embodiment of the present application is shown.

[0101] According to an embodiment of the present application, step S7 comprises:

[0102] Step S71, according to the patient physiological state data, determining patient age, patient BMI, patient disease history, toxic T cell density, helper T cell ratio and macrophage ratio;

[0103] Step S72, according to the patient disease history, obtaining diabetes identification result and cardiovascular disease identification result;

[0104] Step S73, according to the patient age, determining patient age risk identification result;

[0105] Step S74, according to the patient age risk identification result, the patient BMI, the toxic T cell density, the helper T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, determining the patient physiological state influence coefficient.

[0106] For example, according to the medical record of the to-be-tested patient, the patient's age, patient's height, patient's weight and patient's disease history are obtained, the patient's BMI is determined according to the patient's height and the patient's weight, the CD8+ cytotoxic T cell density, the CD4+ helper T cell / regulatory T cell ratio and the CD68+ / CD163+ macrophage ratio are detected by immunohistochemical staining, multicolor immunofluorescence and IHC double staining, that is, the cytotoxic T cell density, the helper T cell / regulatory T cell ratio and the macrophage ratio; according to the patient's disease history, the diabetes identification result and the cardiovascular disease identification result are determined, if the to-be-tested patient has a history of diabetes, the diabetes identification result is 1, otherwise, the diabetes identification result is 0, if the to-be-tested patient has a history of cardiovascular disease, the cardiovascular disease identification result is 0.5, otherwise, the cardiovascular disease identification result is 1; if the age of the to-be-tested patient is greater than or equal to 70 years old, the patient's age risk identification result is 0.5, otherwise, the patient's age risk identification result is 1; according to the patient's age risk identification result, the patient's BMI, the cytotoxic T cell density, the helper T cell / regulatory T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, the patient's immunity and health status are evaluated, and the patient's physiological state influence coefficient is determined.

[0107] According to one embodiment of the present application, step S74 comprises: determining the patient's physiological state influence coefficient of the i-th to-be-tested patient according to formula (1) ,

[0108] (1)

[0109] Wherein, if is a conditional function, and is a logical operator of "and", is the diabetes identification result of the i-th to-be-tested patient, , is the patient's BMI of the i-th to-be-tested patient, is the patient's age risk identification result of the i-th to-be-tested patient, , is the cardiovascular disease identification result of the i-th to-be-tested patient, , is the cytotoxic T cell density of the i-th to-be-tested patient, is a preset cytotoxic T cell density threshold, is the helper T cell / regulatory T cell ratio of the i-th to-be-tested patient, is a preset helper T cell / regulatory T cell ratio threshold, is the macrophage ratio of the i-th to-be-tested patient, is a preset macrophage ratio threshold.

[0110] According to one embodiment of the present application, in formula (1), the conditional function The value of the condition function includes the following two cases: when the condition of is met, it indicates that the i-th to-be-tested patient has a history of diabetes and is overweight, when the patient has diabetes and the BMI is greater than or equal to 35, the to-be-tested patient is likely to have metabolic syndrome, metabolic abnormalities can cause chronic inflammation, chronic inflammation promotes tumor growth, and the to-be-tested patient is in poor physical condition in terms of metabolism, the value of the condition function is 0.5, and when the condition of is not met, it indicates that the i-th to-be-tested patient does not have a history of diabetes and is overweight, the to-be-tested patient is less likely to have metabolic syndrome, and the to-be-tested patient is in normal physical condition in terms of metabolism, the value of the condition function is 1.

[0111] According to one embodiment of the present application, is the patient age risk identification result of the i-th to-be-tested patient, when the age of the i-th to-be-tested patient is greater than or equal to 70 years old, the tolerance of patients over 70 years old is poor, is 0.5, indicating that the to-be-tested patient is in poor physical condition according to the age assessment, when the age of the i-th to-be-tested patient is less than 70 years old, is 1, indicating that the to-be-tested patient is in normal physical condition according to the age assessment. is the cardiovascular disease identification result of the i-th to-be-tested patient, when the i-th to-be-tested patient has cardiovascular disease, the cardiovascular disease will limit the implementation of treatment, is 0.5, indicating that the to-be-tested patient is in poor physical condition according to the cardiovascular disease condition assessment, when the i-th to-be-tested patient does not have cardiovascular disease, is 1, indicating that the to-be-tested patient is in normal physical condition according to the cardiovascular disease condition assessment.

[0112] According to one embodiment of the present application, toxic T cells are the main force of direct killing of tumors, the higher the density of toxic T cells, the stronger the immune activity, is the ratio of the density of toxic T cells of the i-th to-be-tested patient to the preset toxic T cell density threshold, the larger the ratio, the higher the density of toxic T cells of the i-th to-be-tested patient, and the stronger the immune activity of the i-th to-be-tested patient, can be set to 20 / HPF, indicates the physical condition of the to-be-tested patient according to the density of toxic T cells, the larger, the better the physical condition, is the ratio of CD4+ helper T cells to regulatory T cells of the i-th to-be-tested patient, the helper T cells are anti-tumor type, and the regulatory T cells are pro-tumor type, the larger is a favorable sign for tumor treatment, a ratio of the helper T regulatory cell ratio of the i-th to-be-tested patient to a preset helper T regulatory cell ratio threshold, the larger the ratio, the larger the helper T regulatory cell ratio of the i-th to-be-tested patient, and the more suitable the immune condition of the i-th to-be-tested patient is for tumor treatment, may be set to 2.5, represents the physical condition of the to-be-tested patient determined according to the helper T regulatory cell ratio, the larger, the better the physical condition is, a ratio of the CD68+ macrophage and CD163+ macrophage ratio of the i-th to-be-tested patient to a preset macrophage ratio threshold, the larger the ratio, the larger the macrophage ratio of the i-th to-be-tested patient, and the more suitable the immune condition of the i-th to-be-tested patient is for tumor treatment, the larger, the more favorable sign for tumor treatment, a ratio of the macrophage ratio of the i-th to-be-tested patient to a preset macrophage ratio threshold, the larger the ratio, the larger the macrophage ratio of the i-th to-be-tested patient, and the more suitable the immune condition of the i-th to-be-tested patient is for tumor treatment, may be set to 1.5, represents the physical condition of the to-be-tested patient determined according to the macrophage ratio, the larger, the better the physical condition is.

[0113] According to one embodiment of the present application, represents the health condition of the to-be-tested patient determined according to the metabolic condition, age condition, cardiovascular disease condition and immune condition of the to-be-tested patient, and determines the patient physiological state influence coefficient.

[0114] In this way, the patient physiological state influence coefficient can be determined according to the patient age risk identification result, patient BMI, toxic T cell density, helper T regulatory cell ratio, macrophage ratio, diabetes identification result and cardiovascular disease identification result, and in the calculation process, the health condition of the to-be-tested patient can be evaluated from four aspects of metabolic condition, age condition, cardiovascular disease condition and immune condition, thereby improving the comprehensiveness and accuracy of the patient physiological state influence coefficient.

[0115] According to one embodiment of the present application, in step S8, the medical image data, the laboratory test data and the patient physiological state influence coefficient are processed by the trained tumor deterioration time prediction model to determine the predicted tumor deterioration time.

[0116] Figure 4 An exemplary schematic diagram of determining the predicted tumor deterioration time according to an embodiment of the present application is shown.

[0117] According to one embodiment of the present application, step S8 includes:

[0118] Step S81, determining the muscle layer infiltration degree, ADC value and cervical stroma invasion depth according to the medical image data;

[0119] Step S82, determining the histological type risk level and lymphatic vessel space invasion identification result according to the laboratory test data;

[0120] Step S83, processing the patient physiological state influence coefficient, the muscle layer infiltration degree, the ADC value, the cervical stroma invasion depth, the histological type risk level and the lymphatic vessel space invasion identification result through the trained tumor deterioration time prediction model to obtain the predicted tumor deterioration time.

[0121] For example, according to the pelvic MRI, the T2WI image features are determined, the infiltration degree is determined according to the T2WI image features, such as when the tumor high signal does not penetrate the low signal combination zone, which represents that the infiltration degree is less than 50%, when the tumor high signal penetrates the low signal zone and reaches the outside 1 / 2 of the muscle layer, which represents that the infiltration degree is greater than or equal to 50%, the DWI / ADC value, that is, the ADC value, is determined by using the diffusion weighted imaging (DWI) combined with the apparent diffusion coefficient (ADC) quantitative analysis, the cervical stroma invasion depth is determined by using the MRI sagittal position combined with the axial position joint observation, the histological type is determined by using the histopathology HE staining combined with the immunohistochemistry (IHC) assisted method, the histological type risk level is determined according to the histological type, such as the gland structure is regular, the nuclear atypia is less than 5% for low risk, the histological type risk level is 1, the gland fusion or the cribriform structure accounts for 6% to 50% for medium risk, the histological type risk level is 2, the solid area is greater than 50% or the significant nuclear atypia is high risk, the histological type risk level is 3, the lymphatic vessel space invasion detection is performed according to the HE staining microscopic observation and the IHC double verification, when the detection shows that the LVSI is positive, the lymphatic vessel space invasion identification result is 1.4, otherwise, the lymphatic vessel space invasion identification result is 1, the patient physiological state influence coefficient, the muscle layer infiltration degree, the ADC value, the cervical stroma invasion depth, the histological type risk level and the lymphatic vessel space invasion identification result are processed through the trained tumor deterioration time prediction model to obtain the time when the tumor may occur local recurrence, distant metastasis or pathological grading upgrade, that is, the predicted tumor deterioration time.

[0122] According to one embodiment of the present application, the training step of the tumor deterioration time prediction model comprises:

[0123] Obtaining the historical patient physiological state data, the historical medical image data and the historical laboratory test data of a plurality of historical patients;

[0124] Determining the historical patient physiological state influence coefficient according to the historical patient physiological state data;

[0125] determine a historical muscle layer invasion degree, a historical ADC value, and a historical depth of invasion of cervical stroma according to the historical medical image data;

[0126] determine a historical histological type risk grade and a historical lymphovascular space invasion identification result according to the historical laboratory test data;

[0127] process the historical patient physiological state influence coefficient, the historical muscle layer invasion degree, the historical ADC value, the historical depth of invasion of cervical stroma, the historical histological type risk grade, and the historical lymphovascular space invasion identification result by a tumor progression time prediction model to obtain a historical training progression time;

[0128] obtain historical actual progression times of a plurality of historical patients;

[0129] determine a training loss function of the tumor progression time prediction model according to the historical patient physiological state influence coefficient, the historical training progression time, the historical actual progression time, the historical muscle layer invasion degree, the historical ADC value, the historical depth of invasion of cervical stroma, the historical histological type risk grade, and the historical lymphovascular space invasion identification result;

[0130] train the tumor progression time prediction model according to the training loss function to obtain a trained tumor progression time prediction model.

[0131] For example, the historical patient physiological state data, the historical medical image data and the historical laboratory test data of the historical patient who has completed the treatment in the clinic are acquired; the historical patient physiological state influence coefficient is determined according to the historical patient physiological state data, and the calculation manner of the historical patient physiological state influence coefficient is similar to the calculation manner of the patient physiological state influence coefficient in formula (1), which will not be repeated here; the historical muscle layer infiltration depth, the historical ADC value and the historical cervical stroma invasion depth of the historical patient are determined according to the historical medical image data, and the determination manner of the historical muscle layer infiltration depth, the historical ADC value and the historical cervical stroma invasion depth of the historical patient is the same as the determination manner of the muscle layer infiltration depth, the ADC value and the cervical stroma invasion depth, which will not be repeated here; the historical histological type risk level and the historical lymphatic vessel space invasion identification result of the historical patient are determined according to the historical laboratory test data, and the determination manner of the historical histological type risk level and the historical lymphatic vessel space invasion identification result is the same as the determination manner of the histological type risk level and the lymphatic vessel space invasion identification result, which will not be repeated here; the historical patient physiological state influence coefficient, the historical muscle layer infiltration depth, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type risk level and the historical lymphatic vessel space invasion identification result are processed by the tumor deterioration time prediction model to obtain the time when the tumor of the historical patient is likely to have local recurrence, distant metastasis or pathological grading upgrade, that is, the historical training deterioration time; the historical actual deterioration time of a plurality of historical patients in the actual clinical treatment is acquired; the training loss function of the tumor deterioration time prediction model is determined according to the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical muscle layer infiltration depth, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type risk level and the historical lymphatic vessel space invasion identification result; the tumor deterioration time prediction model is trained according to the training loss function to improve the accuracy of the tumor deterioration time prediction model for tumor deterioration time prediction, and the trained tumor deterioration time prediction model is obtained.

[0132] According to one embodiment of the application, the training loss function of the tumor deterioration time prediction model is determined according to the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical muscle layer infiltration degree, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type risk level and the historical lymphatic vessel space invasion identification result, comprising: determining the training loss function of the tumor deterioration time prediction model according to formula (2) ,

[0133] (2)

[0134] wherein, is the historical training deterioration time of the kth historical patient, a historical actual worsening time of the kth historical patient, a historical ADC value of the kth historical patient, a preset ADC value threshold, a historical muscle layer infiltration degree of the kth historical patient, a preset muscle layer infiltration degree threshold, a historical cervical stroma invasion depth of the kth historical patient, a preset historical cervical stroma invasion depth threshold, a historical histological type risk grade of the kth historical patient, a preset histological type risk grade threshold, a historical lymphatic vessel space invasion identification result of the kth historical patient, , a historical patient physiological state influence coefficient of the kth historical patient, a preset patient physiological state influence coefficient threshold, K is a number of historical patients, k≤K, k and K are positive integers.

[0135] According to one embodiment of the present application, a ratio of the historical muscle layer infiltration degree of the kth historical patient to the preset muscle layer infiltration degree threshold, the larger the ratio, the greater the historical muscle layer infiltration degree of the kth historical patient, and the preset muscle layer infiltration degree threshold can be set to 50%, a ratio of the historical cervical stroma invasion depth of the kth historical patient to the preset historical cervical stroma invasion depth threshold, the larger the ratio, the greater the historical cervical stroma invasion depth of the kth historical patient, which can be set to 3mm, a ratio of the historical histological type risk grade of the kth historical patient to the preset histological type risk grade threshold, the larger the ratio, the higher the historical histological type risk grade of the kth historical patient, and the preset histological type risk grade threshold can be set to 1, a historical lymphatic vessel space invasion identification result of the kth historical patient, when lymphatic vessel space invasion occurs, equals 1.4, indicating that when lymphatic vessel space invasion occurs, the tumor metastasis risk increases by 40%, and when lymphatic vessel space invasion does not occur, equals 1, a ratio of the historical patient physiological state influence coefficient of the kth historical patient to the preset patient physiological state influence coefficient threshold, the larger the ratio, the greater the historical patient physiological state influence coefficient of the kth historical patient, and the preset patient physiological state influence coefficient threshold can be set to 1, a ratio of a historical ADC value of the kth historical patient to a preset ADC value threshold, the larger the ratio, the larger the historical ADC value of the kth historical patient, may be set to , indicates that the historical patient physiological state influence coefficient, the historical ADC value and the historical training deterioration time are positively correlated, for example, the larger the historical patient physiological state influence coefficient, the more conducive the physical health status of the to-be-tested patient to tumor treatment, the larger the value of the historical training deterioration time, the larger the historical ADC value, which generally indicates better treatment effect, indicating that the growth of the tumor is inhibited or delayed, the clinical deterioration speed is slowed down, the larger the value of the historical training deterioration time, the larger the historical muscle layer infiltration degree, the deeper the historical cervical stroma invasion, the higher the historical histological type risk level, the more the historical lymphatic vessel space infiltration recognition result, and the historical training deterioration time are negatively correlated, for example, the larger the historical muscle layer infiltration degree, the more likely it is to break through the uterine serous membrane layer and invade adjacent organs (such as the bladder and rectum), causing the risk of lymph node metastasis to rise sharply, the smaller the value of the historical training deterioration time, the deeper the historical cervical stroma invasion, the faster the tumor spreads to the parauterine tissue, the smaller the value of the historical training deterioration time, the higher the historical histological type risk level, the faster the tumor grows, the lower the sensitivity to radiotherapy, and the smaller the value of the historical training deterioration time, when lymphatic vessel space infiltration occurs, it provides a “high-speed channel” for tumor cells, and the risk of lymph node metastasis rises by 40%, and the smaller the value of the historical training deterioration time. Therefore, the term related to the historical patient physiological state influence coefficient and the historical ADC value is placed in the numerator, and the term related to the historical muscle layer infiltration degree, the historical cervical stroma invasion depth, the historical histological type risk level and the historical lymphatic vessel space infiltration recognition result is placed in the denominator, indicating and the larger the value of the historical training deterioration time, , , and the smaller the value of the historical training deterioration time.

[0136] According to one embodiment of the present application, is the relative error of the kth historical training deterioration time and the historical actual deterioration time, and the relative error of the kth historical training deterioration time and the historical actual deterioration time is obtained by using weighting and averaging the relative error of the historical training deterioration time and the historical actual deterioration time of each historical patient, to obtain a training loss function. In the training process, the training loss function is reduced, thereby reducing the error between the historical training deterioration time and the historical actual deterioration time, thereby improving the prediction accuracy of the tumor deterioration time prediction model for the tumor deterioration time, thereby improving the accuracy of the tumor deterioration time prediction model.

[0137] In this way, the training loss function of the tumor deterioration time prediction model can be determined according to the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical muscle layer infiltration degree, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type danger level and the lymphatic vessel space infiltration recognition result. In the calculation process, the influence of the above data on the error of the historical training deterioration time can be determined according to the influence of the historical patient physiological state influence coefficient, the historical training deterioration time, the historical actual deterioration time, the historical muscle layer infiltration degree, the historical ADC value, the historical cervical stroma invasion depth, the historical histological type danger level and the lymphatic vessel space infiltration recognition result on the tumor deterioration speed, and the training loss function is set based on the influence and the relative error of the historical training deterioration time, so that the tumor deterioration time prediction model reduces the training loss function in the training process, and more accurately improves the precision of the tumor deterioration time prediction model.

[0138] According to an embodiment of the present application, in step S9, the endometrial cancer risk coefficient is determined according to the predicted tumor deterioration time.

[0139] Figure 5 An exemplary schematic diagram of determining the endometrial cancer risk coefficient according to an embodiment of the present application is shown.

[0140] According to an embodiment of the present application, step S9 comprises:

[0141] Step S91, determining a first difference value according to the predicted tumor deterioration time and the set tumor deterioration time threshold value;

[0142] Step S92, determining a first ratio according to the first difference value and the set tumor deterioration time threshold value;

[0143] Step S93, determining the endometrial cancer risk coefficient according to the first ratio.

[0144] For example, the first difference value is determined according to the set tumor deterioration time threshold value minus the predicted tumor deterioration time, and the predicted tumor deterioration time can be set to 6 months; the first ratio is determined according to the ratio of the first difference value to the set tumor deterioration time threshold value; when the first ratio is less than 0, the endometrial cancer risk coefficient is 1, indicating that there is a certain risk; when the first ratio is greater than or equal to 0 and less than or equal to 0.5, the endometrial cancer risk coefficient is 2, indicating that the risk is greater; when the first ratio is greater than 0.5, the endometrial cancer risk coefficient is 3, indicating that the risk is serious.

[0145] According to an embodiment of the present application, in step S10, a detection report is generated according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient.

[0146] For example, according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient, it is determined whether the patient to be tested is diagnosed with endometrial cancer and the risk degree of the diagnosed endometrial cancer.

[0147] The detection method of endometrial cancer based on deep learning according to the embodiment of the present application can perform preliminary investigation on the patient to be tested, determine whether histopathological detection is needed, improve the accuracy of the detection result through histopathological detection, further predict the tumor deterioration time according to the patient physiological state data, medical image data and laboratory test data of the patient to be tested in the case of diagnosing endometrial cancer, evaluate the risk degree of the tumor according to the predicted tumor deterioration time, determine the endometrial cancer risk coefficient and generate a detection report, thereby improving the comprehensiveness and accuracy of the detection result of endometrial cancer. When determining the patient physiological state influence coefficient, the patient physiological state influence coefficient can be determined according to the patient age risk identification result, patient BMI, toxic T cell density, auxiliary regulatory T cell ratio, macrophage ratio, diabetes identification result and cardiovascular disease identification result. In the calculation process, the health status of the patient to be tested can be evaluated from four aspects of metabolic status, age status, cardiovascular disease status and immune status, thereby improving the comprehensiveness and accuracy of the patient physiological state influence coefficient. When determining the training loss function, the training loss function of the tumor deterioration time prediction model can be determined according to the historical patient physiological state influence coefficient, historical training deterioration time, historical actual deterioration time, historical myometrial invasion degree, historical ADC value, historical depth of cervical stroma invasion, historical histological type risk level and lymphatic vessel space invasion identification result. In the calculation process, the influence of the above data on the error of the historical training deterioration time can be determined according to the influence of the historical patient physiological state influence coefficient, historical training deterioration time, historical actual deterioration time, historical myometrial invasion degree, historical ADC value, historical depth of cervical stroma invasion, historical histological type risk level and lymphatic vessel space invasion identification result on the tumor deterioration speed, and the training loss function is set based on the influence and the relative error of the historical training deterioration time, so that the tumor deterioration time prediction model reduces the training loss function in the training process and more accurately improves the precision of the tumor deterioration time prediction model.

[0148] Figure 6 An example block diagram of the detection system of endometrial cancer based on deep learning according to the embodiment of the present application is shown, which comprises:

[0149] A preliminary data module acquires preliminary investigation data of the patient to be tested;

[0150] A preliminary result module determines a preliminary investigation result according to the preliminary investigation data;

[0151] a pathology detection module configured to determine whether a histopathological examination is needed according to the preliminary investigation result;

[0152] a pathology result module configured to acquire hysteroscopy-guided biopsy data of the patient to be examined and determine a histopathological diagnosis result according to the hysteroscopy-guided biopsy data if the histopathological examination is needed;

[0153] a further detection module configured to determine whether further detection is needed according to the histopathological diagnosis result;

[0154] a detailed information module configured to acquire physiological state information, medical image information and laboratory test information of the patient to be examined if the further detection is needed;

[0155] an influence coefficient module configured to determine a physiological state influence coefficient according to the physiological state information of the patient to be examined;

[0156] a prediction time module configured to determine a predicted tumor deterioration time by processing the medical image information, the laboratory test information and the physiological state influence coefficient according to a trained tumor deterioration time prediction model;

[0157] a risk coefficient module configured to determine an endometrial cancer risk coefficient according to the predicted tumor deterioration time;

[0158] a detection report module configured to generate a detection report according to the preliminary investigation result, the histopathological diagnosis result and the endometrial cancer risk coefficient.

[0159] The present application can be a method, apparatus, system, and / or computer program product. Computer program products can include computer readable storage media having computer readable program instructions loaded therewith for performing various aspects of the present application.

[0160] Those skilled in the art will understand that the application described above and illustrated in the accompanying drawings is presented by way of example only and is not limiting. The object of the application has been fully and effectively achieved. The functional and structural principles of the application have been shown and described in the embodiments, and the embodiments of the application can be modified or changed in any way without departing from the principles.

Claims

1.A system for detecting endometrial cancer based on deep learning, characterized by, The method comprises the following steps: A preliminary data module obtains preliminary survey data of a patient to be tested; A preliminary result module determines a preliminary survey result according to the preliminary survey data; A pathological detection module determines whether histopathological detection is needed according to the preliminary survey result; A pathological result module obtains hysteroscopy-guided biopsy data of the patient to be tested if histopathological detection is needed, and determines a histopathological diagnosis result according to the hysteroscopy-guided biopsy data; A deep detection module determines whether further detection is needed according to the histopathological diagnosis result; A detailed information module obtains patient physiological state information, medical image information and laboratory test information of the patient to be tested if further detection is needed; An influence coefficient module determines a patient physiological state influence coefficient according to the patient physiological state information; A prediction time module processes the medical image information, the laboratory test information and the patient physiological state influence coefficient by using a trained tumor deterioration time prediction model to determine a predicted tumor deterioration time; A risk coefficient module determines an endometrial cancer risk coefficient according to the predicted tumor deterioration time; A detection report module generates a detection report according to the preliminary survey result, the histopathological diagnosis result and the endometrial cancer risk coefficient; The training steps of the tumor deterioration time prediction model comprise the following steps: S1, obtaining historical patient physiological state information, historical medical image information and historical laboratory test information of a plurality of historical patients; S2, determining historical patient physiological state influence coefficients according to the historical patient physiological state information; S3, determining historical myometrial invasion degrees, historical ADC values and historical cervical stroma invasion depths according to the historical medical image information; S4, determining historical histological type risk grades and historical lymphatic vessel space invasion recognition results according to the historical laboratory test information; S5, processing the historical patient physiological state influence coefficients, the historical myometrial invasion degrees, the historical ADC values, the historical cervical stroma invasion depths, the historical histological type risk grades and the historical lymphatic vessel space invasion recognition results by using a tumor deterioration time prediction model to obtain historical training deterioration times; S6, obtaining historical actual deterioration times of the plurality of historical patients; S7, determining a training loss function of the tumor deterioration time prediction model according to the historical patient physiological state influence coefficients, the historical training deterioration times, the historical actual deterioration times, the historical myometrial invasion degrees, the historical ADC values, the historical cervical stroma invasion depths, the historical histological type risk grades and the historical lymphatic vessel space invasion recognition results, comprising: according to the formula Training loss function for determining tumor malignancy time prediction model wherein, is a historical training malignancy time of the kth historical patient, is a historical actual malignancy time of the kth historical patient, is a historical ADC value of the kth historical patient, is a preset ADC value threshold, is a historical muscle layer infiltration degree of the kth historical patient, is a preset muscle layer infiltration degree threshold, is a historical cervical stroma invasion depth of the kth historical patient, is a preset historical cervical stroma invasion depth threshold, is a historical histological type risk level of the kth historical patient, is a preset histological type risk level threshold, is a historical lymphatic vessel space infiltration identification result of the kth historical patient, , is a historical patient physiological state influence coefficient of the kth historical patient, is a preset patient physiological state influence coefficient threshold, K is the number of historical patients, k≤K, k and K are positive integers; S8, training the tumor deterioration time prediction model according to the training loss function to obtain a trained tumor deterioration time prediction model. 2.The deep learning-based endometrial cancer detection system of claim 1, wherein The preliminary survey result is determined according to the preliminary survey data, comprising: determining an abnormal bleeding recognition result, an abnormal drainage recognition result and an abnormal pain recognition result according to the preliminary survey data; According to the abnormal bleeding identification result, the abnormal drainage identification result and the abnormal pain identification result, a preliminary investigation result is determined. 3.The deep learning-based endometrial cancer detection system of claim 1, wherein According to the patient physiological state information, a patient physiological state influence coefficient is determined, including: According to the patient physiological state information, a patient age, a patient BMI, a patient disease history, a toxic T cell density, a helper regulatory T cell ratio and a macrophage ratio are determined; According to the patient disease history, a diabetes identification result and a cardiovascular disease identification result are obtained; According to the patient age, a patient age risk identification result is determined; According to the patient age risk identification result, the patient BMI, the toxic T cell density, the helper regulatory T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, a patient physiological state influence coefficient is determined. 4.The deep learning-based endometrial cancer detection system of claim 3, wherein According to the patient age risk identification result, the patient BMI, the toxic T cell density, the helper regulatory T cell ratio, the macrophage ratio, the diabetes identification result and the cardiovascular disease identification result, a patient physiological state influence coefficient is determined, including: According to the formula determining a patient physiological state impact coefficient of the ith to-be-tested patient wherein if is a conditional function, and is a logical operator of "and", is a diabetes identification result of the ith to-be-tested patient, , is a patient BMI of the ith to-be-tested patient, is a patient age risk identification result of the ith to-be-tested patient, , is a cardiovascular disease identification result of the ith to-be-tested patient, , is a toxic T cell density of the ith to-be-tested patient, is a preset toxic T cell density threshold, is a helper regulatory T cell ratio of the ith to-be-tested patient, is a preset helper regulatory T cell ratio threshold, is a macrophage ratio of the ith to-be-tested patient, is a preset macrophage ratio threshold. 5.The deep learning-based endometrial cancer detection system of claim 1, wherein By processing the medical image information, the laboratory test information and the patient physiological state influence coefficient through the trained tumor deterioration time prediction model, a predicted tumor deterioration time is determined, including: According to the medical image information, a muscle layer infiltration degree, an ADC value and a cervical stroma invasion depth are determined; According to the laboratory test information, a histological type risk grade and a lymphatic vessel space invasion identification result are determined; By processing the patient physiological state influence coefficient, the muscle layer infiltration degree, the ADC value, the cervical stroma invasion depth, the histological type risk grade and the lymphatic vessel space invasion identification result through the trained tumor deterioration time prediction model, a predicted tumor deterioration time is obtained. 6.The deep learning-based endometrial cancer detection system of claim 1, wherein According to the predicted tumor deterioration time, an endometrial cancer risk coefficient is determined, including: According to the predicted tumor deterioration time and a set tumor deterioration time threshold, a first difference value is determined; According to the first difference value and the set tumor deterioration time threshold, a first ratio value is determined; According to the first ratio value, an endometrial cancer risk coefficient is determined.

Images