Disease risk assessment method and system based on deep learning

Abstract

The present invention relates to the technical field of disease risk assessment. The present invention discloses a disease risk assessment method and system based on deep learning, including obtaining gastric disease characteristic data and intestinal disease characteristic data of a patient to be detected, calculating a gastric disease influence coefficient based on the gastric disease characteristic data, calculating an intestinal disease influence coefficient based on the intestinal disease characteristic data, and quickly distinguishing the patient's status based on the gastric disease influence coefficient and the intestinal disease influence coefficient, thereby achieving a more comprehensive assessment, inputting a marker interval sequence and a patient's status into a pre-constructed cancer probability prediction model, obtaining a real-time sensitivity probability and a real-time specificity probability, and finally calculating a positive cancer probability based on a priori probability, real-time sensitivity probability and real-time specificity probability, and determining an assessment result according to the positive cancer probability. In the process of evaluating patients, the present invention takes into account the specificity and sensitivity of the marker, so that the accuracy of the assessment is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of disease risk assessment, and more specifically, to a disease risk assessment method and system based on deep learning. Background Art

[0002] Disease risk prediction refers to the use of statistics, machine learning or other predictive models to assess the likelihood of an individual developing a certain disease. The prediction is usually based on analysis and calculations based on factors such as the individual's health data, lifestyle, and genetic information. Disease risk prediction is often used to predict whether a person has cancer. Digestive tract cancer (including gastric cancer, intestinal cancer, etc.) is one of the more common types. Although existing technologies are able to predict and assess the risk of digestive tract cancer through predictive models, this field still faces some challenges and problems.

[0003] For example, Chinese patent publication number CN117169504A provides biomarkers and related prediction systems and applications for detecting gastric cancer-related parameters. Different biomarkers are used to construct a diagnostic model for gastric cancer, which can be used to efficiently predict whether an individual has gastric cancer. Chinese patent publication number CN116519954A provides a method, system and biomarker for constructing a colorectal cancer detection model. By analyzing proteins with significant differences in the blood of colorectal cancer patients and normal people, proteins that can be used alone or in combination as biomarkers for early prediction of the risk of colorectal cancer are screened out, as well as proteins containing the above biomarkers for predicting whether an individual has colorectal cancer.

[0004] Although the above patents all disclose the content of predicting cancer risk through markers, most of the above patents make a single prediction and lack a more comprehensive evaluation. In addition, in the medical field, the specificity and sensitivity of markers are two very important characteristics in disease prediction. However, existing technologies rarely consider these two factors at the same time, resulting in the inability to effectively improve the accuracy of the prediction.

[0005] In view of this, the present invention proposes a disease risk assessment method and system based on deep learning to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a disease risk assessment method and system based on deep learning.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] Disease risk assessment methods based on deep learning include:

[0009] S10: Obtaining gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculating a gastric disease influence coefficient based on the gastric disease characteristic data, calculating an intestinal disease influence coefficient based on the intestinal disease characteristic data, and determining a patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient, wherein the patient status includes suffering from a gastric disease, suffering from an intestinal disease, and not suffering from the disease;

[0010] S20: When the patient is suffering from a stomach disease or an intestinal disease, obtain the medical record data of the patient to be tested, and determine the prior probability according to the medical record data;

[0011] S30: acquiring marker characteristic data of the patient to be detected based on the patient status, determining a marker interval sequence based on the marker characteristic data, inputting the marker interval sequence and the patient status into a pre-built cancer probability prediction model, and obtaining a real-time sensitivity probability and a real-time specificity probability;

[0012] S40: Calculate the positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability, and determine the evaluation result according to the positive cancer probability.

[0013] Furthermore, the gastric disease characteristic data includes a pepsinogen I / II ratio, a gastrin G-17 value, and a cancer antigen CA72-4 concentration value. The method for calculating the gastric disease influence coefficient based on the gastric disease characteristic data includes:

[0014] GDF= ;

[0015] In the formula, GDF is the gastric disease influence coefficient, is the gastrin G-17 value, is the concentration value of cancer antigen CA72-4, is the pepsinogen I / II ratio, is the hyperbolic cosine function, is a logarithmic function with base 2, is the inverse tangent function, is a constant.

[0016] Furthermore, the intestinal disease characteristic data include peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value. The method for calculating the intestinal disease impact coefficient based on the intestinal disease characteristic data includes:

[0017] IDF= ;

[0018] Where, IDF is the intestinal disease impact coefficient, is a logarithmic function with base 10, is the carcinoembryonic antigen CEA concentration value, is the hyperbolic sine function, is the ratio of peripheral blood neutrophils to white blood cells, is the ratio of lymphocyte to leukocyte concentration, is the concentration value of cancer antigen CA19-9, is a natural constant, , are all weight factors.

[0019] Furthermore, the method for determining the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient includes:

[0020] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is not ill;

[0021] If the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is suffering from gastric disease;

[0022] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold, the patient status is suffering from intestinal disease.

[0023] Furthermore, the method of determining the prior probability based on the medical record data includes:

[0024] Traverse the medical record data, determine the first probability corresponding to the patient's age, the second probability corresponding to the number of patients, and the third probability corresponding to the number of bad habits according to the mapping relationship, and take the sum of the first probability, the second probability and the third probability as the prior probability.

[0025] Furthermore, the method for determining the marker interval sequence based on the marker characteristic data includes:

[0026] The marker characteristic data are traversed to determine the marker concentration interval corresponding to each element in the marker characteristic data, and the obtained marker concentration intervals are aggregated to generate a marker interval sequence.

[0027] Furthermore, the method for constructing a cancer probability prediction model includes:

[0028] A sample data set is obtained, wherein the sample data set includes a historical marker interval sequence, a historical patient status, a historical sensitivity probability, and a historical specificity probability. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical marker interval sequence and the historical patient status in the sample training set are used as input data of the regression network, and the historical sensitivity probability and the historical specificity probability in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the real-time sensitivity probability and the real-time specificity probability. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as a cancer probability prediction model.

[0029] Further, the method for calculating the positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability includes:

[0030] PHR= ;

[0031] Where PHR is the probability of positive cancer, is the real-time sensitivity probability, is the prior probability, is the real-time specificity probability.

[0032] A disease risk assessment system based on deep learning, which is used to implement the above-mentioned disease risk assessment method based on deep learning, includes:

[0033] Differentiation module: used to obtain gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculate the gastric disease influence coefficient based on the gastric disease characteristic data, calculate the intestinal disease influence coefficient based on the intestinal disease characteristic data, and determine the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient. The patient status includes suffering from gastric disease, suffering from intestinal disease and not suffering from the disease;

[0034] Prior module: used to obtain the medical record data of the patient to be tested when the patient's condition is gastric disease or intestinal disease, and determine the prior probability based on the medical record data;

[0035] Probability generation module: used to obtain marker feature data of the patient to be tested based on the patient status, determine the marker interval sequence based on the marker feature data, input the marker interval sequence and the patient status into the pre-built cancer probability prediction model, and obtain the real-time sensitivity probability and real-time specificity probability;

[0036] Evaluation module: used to calculate the positive cancer probability based on the prior probability, real-time sensitivity probability and real-time specificity probability, and determine the evaluation result according to the positive cancer probability.

[0037] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the above-mentioned disease risk assessment method based on deep learning.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention obtains gastric disease characteristic data and intestinal disease characteristic data of a patient to be detected, calculates a gastric disease influence coefficient based on the gastric disease characteristic data, calculates an intestinal disease influence coefficient based on the intestinal disease characteristic data, and quickly distinguishes the patient's status based on the gastric disease influence coefficient and the intestinal disease influence coefficient, thereby achieving a more comprehensive evaluation, and then determines a priori probability based on the medical record data, obtains marker characteristic data of the patient to be detected based on the patient's status, determines a marker interval sequence based on the marker characteristic data, inputs the marker interval sequence and the patient's status into a pre-constructed cancer probability prediction model, obtains a real-time sensitivity probability and a real-time specificity probability, and finally calculates a positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability, and determines an evaluation result based on the positive cancer probability. In the process of evaluating patients, the present invention takes into account the specificity and sensitivity of the marker, thereby effectively improving the accuracy of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of a disease risk assessment method based on deep learning in the present invention;

[0041] Figure 2 A schematic diagram of a computer-readable storage medium in the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example 1

[0044] See also Figure 1 As shown, this embodiment discloses a disease risk assessment method based on deep learning, including:

[0045] S10: Obtain gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculate the gastric disease influence coefficient based on the gastric disease characteristic data, calculate the intestinal disease influence coefficient based on the intestinal disease characteristic data, and determine the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient. The patient status includes suffering from gastric disease, suffering from intestinal disease and not suffering from the disease.

[0046] In this embodiment, the gastric disease characteristic data include but are not limited to the pepsinogen I / II ratio, gastrin G-17 value and cancer antigen CA72-4 concentration value. The pepsinogen I / II ratio refers to the concentration ratio of pepsinogen I (PGI) to pepsinogen II (PGII) in the blood, the gastrin G-17 value refers to the concentration value of gastrin in the blood sample, and cancer antigen CA72-4 is a marker commonly used to characterize gastric cancer. The above-mentioned pepsinogen I / II ratio, gastrin G-17 value and cancer antigen CA72-4 concentration value can all be obtained by testing blood samples.

[0047] Methods for calculating the gastric disease influence coefficient based on gastric disease characteristic data include:

[0048] GDF= ;

[0049] In the formula, GDF is the gastric disease influence coefficient, is the gastrin G-17 value, is the concentration value of cancer antigen CA72-4, is the pepsinogen I / II ratio, is the hyperbolic cosine function, is a logarithmic function with base 2, is the inverse tangent function, is a constant.

[0050] It should be noted that, during the atrophy of the gastric mucosa in gastric cancer, due to the atrophy of the gastric fundus and body, the secretion of pepsinogen I is reduced, resulting in a decrease in the pepsinogen I / II ratio, and the cancer antigen CA72-4 is a tumor marker, which is particularly highly expressed in gastric cancer. Elevated gastrin G-17 values ​​are usually associated with an increased risk of gastric cancer. Therefore, based on the above content, it can be concluded that in this embodiment, the gastric disease impact coefficient is positively correlated with the risk of gastric cancer.

[0051] In this embodiment, the intestinal disease characteristic data include but are not limited to the peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value. The neutrophil to leukocyte concentration ratio and lymphocyte to leukocyte concentration ratio both reflect the body's inflammatory state and immune response to a certain extent. For example, a higher neutrophil to leukocyte concentration ratio usually reflects an increased inflammatory state of the body, and a higher lymphocyte to leukocyte concentration ratio usually reflects a better immune state of the body. Carcinoembryonic antigen is a marker commonly used to characterize intestinal cancer, and cancer antigen CA19-9 is mainly used to monitor the condition of colorectal cancer patients. An increase in the cancer antigen CA19-9 concentration value may be associated with the presence of cancer. The above-mentioned peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value can also be obtained by testing blood samples.

[0052] Methods for calculating the intestinal disease impact coefficient based on intestinal disease characteristic data include:

[0053] IDF= ;

[0054] Where, IDF is the intestinal disease impact coefficient, is a logarithmic function with base 10, is the carcinoembryonic antigen CEA concentration value, is the hyperbolic sine function, is the ratio of peripheral blood neutrophils to white blood cells, is the concentration value of cancer antigen CA19-9, is the ratio of lymphocyte to leukocyte concentration, is a natural constant, , are all weight factors.

[0055] Following the above, a higher neutrophil to leukocyte concentration ratio usually reflects an increased inflammatory state of the body, a higher lymphocyte to leukocyte concentration ratio usually reflects a better immune state of the body, and carcinoembryonic antigen is a tumor marker, which is particularly highly expressed in colorectal cancer. Therefore, in this embodiment, the intestinal disease impact coefficient is positively correlated with the risk of colorectal cancer.

[0056] Methods for determining the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient include:

[0057] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is not ill;

[0058] If the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is suffering from gastric disease;

[0059] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold, the patient status is suffering from intestinal disease.

[0060] It is understandable that the prior art usually determines whether a patient has the corresponding cancer by the concentration value of the marker. However, both intestinal cancer and gastric cancer belong to digestive tract cancers, and many markers are the same, such as cancer antigen CA242 and cancer antigen CA125. Therefore, it is impossible to quickly distinguish them only from the perspective of markers. In this embodiment, the patient's condition is quickly distinguished by the pepsinogen I / II ratio and gastrin G-17 value corresponding to the stomach, and the peripheral blood neutrophil to white blood cell concentration ratio and lymphocyte to white blood cell concentration ratio corresponding to the intestine.

[0061] It should be noted that, in the present embodiment, there may be a situation where the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold. At this time, it indicates that the patient's physical condition is poor and the probability of illness is very high. No subsequent analysis is required, and colonoscopy, gastroscopy or CT scan should be performed, and corresponding treatment should be carried out in time.

[0062] S20: When the patient is suffering from a stomach disease or an intestinal disease, obtain the medical record data of the patient to be tested, and determine the prior probability based on the medical record data.

[0063] In this embodiment, the medical record data includes but is not limited to the patient's age, number of patients and number of bad habits. It can be understood that the number of patients refers to the number of relatives of the patient suffering from the corresponding cancer, and the number of bad habits refers to the number of hobbies that cause the corresponding cancer. For example, alcoholism will increase the risk of colon cancer, and long-term intake of high-salt foods will increase the risk of gastric cancer. By pre-establishing a corresponding inquiry form to inquire about the patient, after the patient fills in the inquiry form, the system can directly obtain the number of patients and the number of bad habits from the inquiry form.

[0064] Methods for determining prior probabilities based on medical record data include:

[0065] Traverse the medical record data, determine the first probability corresponding to the patient's age, the second probability corresponding to the number of patients, and the third probability corresponding to the number of bad habits according to the mapping relationship, and take the sum of the first probability, the second probability and the third probability as the prior probability.

[0066] In this embodiment, the corresponding age range can be determined according to the patient's age, and then the first probability can be determined according to the corresponding mapping relationship. The first probability, the second probability and the third probability mentioned above are all predetermined by experts in this field. The patient's age is positively correlated with the first probability. Similarly, the number of patients is positively correlated with the second probability, and the number of bad habits is positively correlated with the third probability. The above method is only an exemplary description, and the present invention is not limited to this.

[0067] S30: Obtain marker characteristic data of the patient to be tested based on the patient status, determine the marker interval sequence based on the marker characteristic data, input the marker interval sequence and the patient status into a pre-constructed cancer probability prediction model to obtain a real-time sensitivity probability and a real-time specificity probability.

[0068] It should be noted that the marker characteristic data in this embodiment changes dynamically according to the patient's condition. For example, when the patient is suffering from a gastric disease, the marker characteristic data includes a cancer antigen CA72-4 concentration value, a cancer antigen CA242 concentration value, and a cancer antigen CA125 concentration value. Of course, the marker characteristic data for gastric diseases may also include SDC2 gene methylation, TERT gene methylation, RNF180 gene methylation, Septin9 gene methylation, etc., which need to be quantified by percentage. This embodiment does not go into details. When the patient is suffering from an intestinal disease, the marker characteristic data includes a carcinoembryonic antigen CEA concentration value, a cancer antigen CA242 concentration value, and a cancer antigen CA19-9 concentration value. Of course, the marker characteristic data for intestinal diseases may also include SDC2 gene methylation, TERT gene methylation, RNF180 gene methylation, Septin9 gene methylation, etc., which need to be quantified by percentage. This embodiment does not go into details. The disease marker characteristic data may also include KRAS gene mutation, SDC2 gene methylation, SFRP2 gene methylation, Septin9 gene methylation, TFPI2 gene methylation, BCAT1 gene methylation, NPY gene methylation, FGF5 gene methylation, PDX1 gene methylation, ADHFE1 gene methylation, PPP2R5C gene methylation, IKZF1 gene methylation, BCAN gene methylation, VAV3 gene methylation, etc., which also need to be quantified by percentage. The acquisition method may be to obtain DNA from a blood sample and then sequence the DNA to obtain data such as SDC2 gene methylation and SFRP2 gene methylation. This embodiment is not limited to this.

[0069] The method for determining a marker interval sequence based on marker characteristic data includes:

[0070] The marker characteristic data are traversed to determine the marker concentration interval corresponding to each element in the marker characteristic data, and the obtained marker concentration intervals are aggregated to generate a marker interval sequence.

[0071] Taking cancer antigen CA242 as an example, its concentration value under normal conditions is usually between 0 and 37 U / mL, while in cancer patients, it can rise to more than 100 U / mL. Therefore, 40 U / mL can be used as the length of each concentration interval to generate multiple marker concentration intervals corresponding to cancer antigen CA242. It should be understood that this division method is only an exemplary description, and the intervals can be divided into equal distances or unequal distances. This embodiment does not impose too many restrictions on this.

[0072] The method for constructing a cancer probability prediction model includes:

[0073] A sample data set is obtained, wherein the sample data set includes a historical marker interval sequence, a historical patient status, a historical sensitivity probability, and a historical specificity probability. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical marker interval sequence and the historical patient status in the sample training set are used as input data of the regression network, and the historical sensitivity probability and the historical specificity probability in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting real-time sensitivity probability and real-time specificity probability. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as a cancer probability prediction model. The regression network is preferably a deep neural network model.

[0074] It should be noted that specificity probability and sensitivity probability are two key indicators in medical evaluation tests. Specificity probability refers to the probability that the test can correctly identify healthy individuals, and sensitivity probability refers to the probability that the test can correctly identify diseased individuals. In cancer screening, tests with high sensitivity can detect more early cases, and high specificity means that the test can correctly exclude healthy individuals and reduce the misdiagnosis rate.

[0075] Methods for obtaining historical sensitivity probability include:

[0076] ;

[0077] In the formula, is the historical sensitivity probability, is the number of true positive patients, is the number of false negative patients.

[0078] Methods for obtaining historical specific probabilities include:

[0079] ;

[0080] In the formula, is the history-specific probability, is the number of true negative patients, is the number of false positive patients.

[0081] It can be understood that a true positive refers to a person who actually has the disease but is correctly identified as positive by the test results, and a false positive refers to a person who actually does not have the disease but is incorrectly identified as positive by the test results. A true negative refers to a person who actually does not have the disease but is correctly identified as negative by the test results, and a false negative refers to a person who actually has the disease but is incorrectly identified as negative by the test results.

[0082] The above-mentioned historical sensitivity probabilities and historical specificity probabilities are obtained by personnel in this field through analysis of a large amount of medical data. The medical data include the patient's corresponding test data and the patient's test results. The patient's test results refer to whether the patient is identified as a true positive, a false negative, a true negative or a false positive. The patient's corresponding test data include a marker interval sequence and a patient status. It can be understood that since the number of patients is large enough, the same marker interval sequence and patient status can have true positive patients, false negative patients, true negative patients and false positive patients, so that the corresponding sensitivity probability and specificity probability can be calculated, and the corresponding relationship can be established, thereby forming a sample data set for subsequent model training.

[0083] S40: Calculate the positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability, and determine the evaluation result according to the positive cancer probability.

[0084] Methods for calculating the probability of positive cancer based on prior probability, real-time sensitivity probability, and real-time specificity probability include:

[0085] PHR= ;

[0086] Where PHR is the probability of positive cancer, is the real-time sensitivity probability, is the prior probability, is the real-time specificity probability.

[0087] The assessment results include mild risk, moderate risk and severe risk. The methods for determining the assessment results based on the probability of positive cancer include:

[0088] When the positive cancer probability is less than or equal to the preset first cancer probability threshold, the assessment result is a slight risk;

[0089] When the positive cancer probability is greater than the preset first cancer probability threshold and less than the preset second cancer probability threshold, the assessment result is moderate risk;

[0090] When the positive cancer probability is greater than or equal to the preset second cancer probability threshold, the assessment result is severe risk.

[0091] It can be understood that, in this embodiment, a mild risk represents a low probability that the patient has cancer, a moderate risk represents a high probability that the patient has cancer, and a severe risk represents an extremely high probability that the patient has cancer.

[0092] In this embodiment, gastric disease characteristic data and intestinal disease characteristic data of the patient to be detected are obtained, the gastric disease influence coefficient is calculated based on the gastric disease characteristic data, the intestinal disease influence coefficient is calculated based on the intestinal disease characteristic data, and the patient's status is quickly distinguished based on the gastric disease influence coefficient and the intestinal disease influence coefficient, thereby achieving a more comprehensive evaluation. Then, the prior probability is determined based on the medical record data, and the marker characteristic data of the patient to be detected is obtained based on the patient's status. The marker interval sequence is determined based on the marker characteristic data, and the marker interval sequence and the patient's status are input into a pre-constructed cancer probability prediction model to obtain a real-time sensitivity probability and a real-time specificity probability. Finally, the positive cancer probability is calculated based on the prior probability, the real-time sensitivity probability and the real-time specificity probability, and the evaluation result is determined based on the positive cancer probability. In the process of evaluating the patient, this embodiment takes into account the specificity and sensitivity of the marker, so that the accuracy of the evaluation is effectively improved.

[0093] Example 2

[0094] This embodiment provides a disease risk assessment system based on deep learning on the basis of embodiment 1, including:

[0095] Differentiation module: used to obtain gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculate the gastric disease influence coefficient based on the gastric disease characteristic data, calculate the intestinal disease influence coefficient based on the intestinal disease characteristic data, and determine the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient. The patient status includes suffering from gastric disease, suffering from intestinal disease and not suffering from the disease.

[0096] In this embodiment, the gastric disease characteristic data include but are not limited to the pepsinogen I / II ratio, gastrin G-17 value and cancer antigen CA72-4 concentration value. The pepsinogen I / II ratio refers to the concentration ratio of pepsinogen I (PGI) to pepsinogen II (PGII) in the blood, the gastrin G-17 value refers to the concentration value of gastrin in the blood sample, and cancer antigen CA72-4 is a marker commonly used to characterize gastric cancer. The above-mentioned pepsinogen I / II ratio, gastrin G-17 value and cancer antigen CA72-4 concentration value can all be obtained by testing blood samples.

[0097] The method for calculating the gastric disease influence coefficient based on the gastric disease characteristic data includes:

[0098] GDF= ;

[0099] In the formula, GDF is the gastric disease influence coefficient, is the gastrin G-17 value, is the concentration value of cancer antigen CA72-4, is the pepsinogen I / II ratio, is the hyperbolic cosine function, is a logarithmic function with base 2, is the inverse tangent function, is a constant.

[0100] It should be noted that, during the atrophy of the gastric mucosa in gastric cancer, due to the atrophy of the gastric fundus and body, the secretion of pepsinogen I is reduced, resulting in a decrease in the pepsinogen I / II ratio, and the cancer antigen CA72-4 is a tumor marker, which is particularly highly expressed in gastric cancer. Elevated gastrin G-17 values ​​are usually associated with an increased risk of gastric cancer. Therefore, based on the above content, it can be concluded that in this embodiment, the gastric disease impact coefficient is positively correlated with the risk of gastric cancer.

[0101] In this embodiment, the intestinal disease characteristic data include but are not limited to the peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value. The neutrophil to leukocyte concentration ratio and lymphocyte to leukocyte concentration ratio both reflect the body's inflammatory state and immune response to a certain extent. For example, a higher neutrophil to leukocyte concentration ratio usually reflects an increased inflammatory state of the body, and a higher lymphocyte to leukocyte concentration ratio usually reflects a better immune state of the body. Carcinoembryonic antigen is a marker commonly used to characterize intestinal cancer, and cancer antigen CA19-9 is mainly used to monitor the condition of colorectal cancer patients. An increase in the cancer antigen CA19-9 concentration value may be associated with the presence of cancer. The above-mentioned peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value can also be obtained by testing blood samples.

[0102] Methods for calculating the intestinal disease impact coefficient based on intestinal disease characteristic data include:

[0103] IDF= ;

[0104] Where, IDF is the intestinal disease impact coefficient, is a logarithmic function with base 10, is the carcinoembryonic antigen CEA concentration value, is the hyperbolic sine function, is the ratio of peripheral blood neutrophils to white blood cells, is the concentration value of cancer antigen CA19-9, is the ratio of lymphocyte to leukocyte concentration, is a natural constant, , are all weight factors.

[0105] Following the above, a higher neutrophil to leukocyte concentration ratio usually reflects an increased inflammatory state of the body, a higher lymphocyte to leukocyte concentration ratio usually reflects a better immune state of the body, and carcinoembryonic antigen is a tumor marker, which is particularly highly expressed in colorectal cancer. Therefore, in this embodiment, the intestinal disease impact coefficient is positively correlated with the risk of colorectal cancer.

[0106] Methods for determining the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient include:

[0107] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is not ill;

[0108] If the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is suffering from gastric disease;

[0109] If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold, the patient status is suffering from intestinal disease.

[0110] It should be noted that, in the present embodiment, there may be a situation where the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold. At this time, it indicates that the patient's physical condition is poor and the probability of illness is very high. No subsequent analysis is required, and colonoscopy, gastroscopy or CT scan should be performed, and corresponding treatment should be carried out in time.

[0111] Prior module: used to obtain the medical record data of the patient to be tested when the patient's condition is gastric disease or intestinal disease, and determine the prior probability based on the medical record data.

[0112] In this embodiment, the medical record data includes but is not limited to the patient's age, number of patients and number of bad habits. It can be understood that the number of patients refers to the number of relatives of the patient suffering from the corresponding cancer, and the number of bad habits refers to the number of hobbies that cause the corresponding cancer. For example, alcoholism will increase the risk of colon cancer, and long-term intake of high-salt foods will increase the risk of gastric cancer. By pre-establishing a corresponding inquiry form to inquire about the patient, after the patient fills in the inquiry form, the system can directly obtain the number of patients and the number of bad habits from the inquiry form.

[0113] Methods for determining prior probabilities based on medical record data include:

[0114] Traverse the medical record data, determine the first probability corresponding to the patient's age, the second probability corresponding to the number of patients, and the third probability corresponding to the number of bad habits according to the mapping relationship, and take the sum of the first probability, the second probability and the third probability as the prior probability.

[0115] In this embodiment, the corresponding age range can be determined according to the patient's age, and then the first probability can be determined according to the corresponding mapping relationship. The first probability, the second probability and the third probability mentioned above are all predetermined by experts in this field. The patient's age is positively correlated with the first probability. Similarly, the number of patients is positively correlated with the second probability, and the number of bad habits is positively correlated with the third probability. The above method is only an exemplary description, and the present invention is not limited to this.

[0116] Probability generation module: used to obtain marker characteristic data of the patient to be tested based on the patient status, determine the marker interval sequence based on the marker characteristic data, input the marker interval sequence and the patient status into the pre-built cancer probability prediction model, and obtain real-time sensitivity probability and real-time specificity probability.

[0117] The method for determining a marker interval sequence based on marker characteristic data includes:

[0118] The marker characteristic data are traversed to determine the marker concentration interval corresponding to each element in the marker characteristic data, and the obtained marker concentration intervals are aggregated to generate a marker interval sequence.

[0119] The method for constructing a cancer probability prediction model includes:

[0120] A sample data set is obtained, wherein the sample data set includes a historical marker interval sequence, a historical patient status, a historical sensitivity probability, and a historical specificity probability. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical marker interval sequence and the historical patient status in the sample training set are used as input data of the regression network, and the historical sensitivity probability and the historical specificity probability in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting real-time sensitivity probability and real-time specificity probability. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as a cancer probability prediction model. The regression network is preferably a deep neural network model.

[0121] It should be noted that specificity probability and sensitivity probability are two key indicators in medical evaluation tests. Specificity probability refers to the probability that the test can correctly identify healthy individuals, and sensitivity probability refers to the probability that the test can correctly identify diseased individuals. In cancer screening, tests with high sensitivity can detect more early cases, and high specificity means that the test can correctly exclude healthy individuals and reduce the misdiagnosis rate.

[0122] Methods for obtaining historical sensitivity probability include:

[0123] ;

[0124] In the formula, is the historical sensitivity probability, is the number of true positive patients, is the number of false negative patients.

[0125] Methods for obtaining historical specific probabilities include:

[0126] ;

[0127] In the formula, is the history-specific probability, is the number of true negative patients, is the number of false positive patients.

[0128] It can be understood that a true positive refers to a person who actually has the disease but is correctly identified as positive by the test results, and a false positive refers to a person who actually does not have the disease but is incorrectly identified as positive by the test results. A true negative refers to a person who actually does not have the disease but is correctly identified as negative by the test results, and a false negative refers to a person who actually has the disease but is incorrectly identified as negative by the test results.

[0129] The above-mentioned historical sensitivity probabilities and historical specificity probabilities are obtained by personnel in this field through analysis of a large amount of medical data. The medical data include the patient's corresponding test data and the patient's test results. The patient's test results refer to whether the patient is identified as a true positive, a false negative, a true negative or a false positive. The patient's corresponding test data include a marker interval sequence and a patient status. It can be understood that since the number of patients is large enough, the same marker interval sequence and patient status can have true positive patients, false negative patients, true negative patients and false positive patients, so that the corresponding sensitivity probability and specificity probability can be calculated, and the corresponding relationship can be established, thereby forming a sample data set for subsequent model training.

[0130] Evaluation module: used to calculate the positive cancer probability based on the prior probability, real-time sensitivity probability and real-time specificity probability, and determine the evaluation result according to the positive cancer probability.

[0131] Methods for calculating the probability of positive cancer based on prior probability, real-time sensitivity probability, and real-time specificity probability include:

[0132] PHR= ;

[0133] Where PHR is the probability of positive cancer, is the real-time sensitivity probability, is the prior probability, is the real-time specificity probability.

[0134] The assessment results include mild risk, moderate risk and severe risk. The methods for determining the assessment results based on the probability of positive cancer include:

[0135] When the positive cancer probability is less than or equal to the preset first cancer probability threshold, the assessment result is a slight risk;

[0136] When the positive cancer probability is greater than the preset first cancer probability threshold and less than the preset second cancer probability threshold, the assessment result is moderate risk;

[0137] When the positive cancer probability is greater than or equal to the preset second cancer probability threshold, the assessment result is severe risk.

[0138] It can be understood that, in this embodiment, a mild risk represents a low probability that the patient has cancer, a moderate risk represents a high probability that the patient has cancer, and a severe risk represents an extremely high probability that the patient has cancer.

[0139] Example 3

[0140] This embodiment discloses an electronic device, including a power supply, an interface, a keyboard, a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the disease risk assessment method based on deep learning provided by the above methods is implemented. The interface includes a network interface and a data interface. The network interface includes a wired or wireless interface, and the data interface includes an input or output interface.

[0141] Since the electronic device introduced in this embodiment is an electronic device used to implement the disease risk assessment method based on deep learning in the embodiment of the present application, based on the disease risk assessment method based on deep learning introduced in the embodiment of the present application, the technical personnel of the field can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of the field implement the electronic device used in the disease risk assessment method based on deep learning in the embodiment of the present application, it belongs to the scope of protection of this application.

[0142] Example 4

[0143] like Figure 2 As shown, this embodiment discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the above-mentioned disease risk assessment method based on deep learning is implemented.

[0144] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0145] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0146] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0147] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0150] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0151] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0152] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A disease risk assessment method based on deep learning, characterized in that: include: S10: Obtaining gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculating a gastric disease influence coefficient based on the gastric disease characteristic data, calculating an intestinal disease influence coefficient based on the intestinal disease characteristic data, and determining a patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient, wherein the patient status includes suffering from a gastric disease, suffering from an intestinal disease, and not suffering from the disease; S20: When the patient is suffering from a stomach disease or an intestinal disease, obtain the medical record data of the patient to be tested, and determine the prior probability according to the medical record data; S30: acquiring marker characteristic data of the patient to be detected based on the patient status, determining a marker interval sequence based on the marker characteristic data, inputting the marker interval sequence and the patient status into a pre-built cancer probability prediction model, and obtaining a real-time sensitivity probability and a real-time specificity probability; S40: calculating the positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability, and determining the evaluation result according to the positive cancer probability; The gastric disease characteristic data include a pepsinogen I / II ratio, a gastrin G-17 value, and a cancer antigen CA72-4 concentration value. The method for calculating the gastric disease influence coefficient based on the gastric disease characteristic data includes: In the formula, GDF is the gastric disease influence coefficient, Ghl is the gastrin G-17 value, and Can 72-4 is the concentration of cancer antigen CA72-4, Psr is the ratio of pepsinogen I / II, cosh(·) is the hyperbolic cosine function, log2(·) is the logarithmic function with base 2, and tan -1 (·) is the inverse tangent function, k is a constant; The intestinal disease characteristic data include peripheral blood neutrophil to leukocyte concentration ratio, lymphocyte to leukocyte concentration ratio, cancer antigen CA19-9 concentration value and carcinoembryonic antigen CEA concentration value. The method for calculating the intestinal disease influence coefficient based on the intestinal disease characteristic data includes: Where, IDF is the intestinal disease impact coefficient, log 10 (·) is a logarithmic function with base 10, PEC is the carcinoembryonic antigen CEA concentration, sinh(·) is a hyperbolic sine function, NWR is the ratio of neutrophils to leukocytes in peripheral blood, LWR is the ratio of lymphocytes to leukocytes, Can 19-9 is the concentration value of cancer antigen CA19-9, e is a natural constant, and f1, f2, and f3 are all weight factors; The method for determining the prior probability based on the medical record data comprises: Traverse the medical record data, determine the first probability corresponding to the patient's age, the second probability corresponding to the number of patients, and the third probability corresponding to the number of bad habits according to the mapping relationship, and take the sum of the first probability, the second probability and the third probability as the prior probability; The method for calculating the positive cancer probability based on the prior probability, the real-time sensitivity probability and the real-time specificity probability comprises: Where PHR is the probability of positive cancer, SEY RE is the real-time sensitivity probability, PRY is the prior probability, SPY is RE is the real-time specificity probability.

2. The disease risk assessment method based on deep learning according to claim 1, characterized in that: The method for determining the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient comprises: If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is not ill; If the gastric disease impact coefficient is greater than or equal to the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is less than the preset intestinal disease impact coefficient threshold, the patient status is suffering from gastric disease; If the gastric disease impact coefficient is less than the preset gastric disease impact coefficient threshold, and the intestinal disease impact coefficient is greater than or equal to the preset intestinal disease impact coefficient threshold, the patient status is suffering from intestinal disease.

3. The disease risk assessment method based on deep learning according to claim 1, characterized in that: The method for determining a marker interval sequence based on marker feature data comprises: The marker characteristic data are traversed to determine the marker concentration interval corresponding to each element in the marker characteristic data, and the obtained marker concentration intervals are aggregated to generate a marker interval sequence.

4. The disease risk assessment method based on deep learning according to claim 1, characterized in that: The method for constructing the cancer probability prediction model includes: A sample data set is obtained, wherein the sample data set includes a historical marker interval sequence, a historical patient status, a historical sensitivity probability, and a historical specificity probability. The sample data set is divided into a sample training set and a sample test set, and a regression network is constructed. The historical marker interval sequence and the historical patient status in the sample training set are used as input data of the regression network, and the historical sensitivity probability and the historical specificity probability in the sample training set are used as output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the real-time sensitivity probability and the real-time specificity probability. The initial regression network is tested using the sample test set, and the initial regression network that satisfies a preset error value is output as a cancer probability prediction model.

5. A disease risk assessment system based on deep learning, which is used to implement the disease risk assessment method based on deep learning according to any one of claims 1 to 4, characterized in that: include: Differentiation module: used to obtain gastric disease characteristic data and intestinal disease characteristic data of the patient to be tested, calculate the gastric disease influence coefficient based on the gastric disease characteristic data, calculate the intestinal disease influence coefficient based on the intestinal disease characteristic data, and determine the patient status based on the gastric disease influence coefficient and the intestinal disease influence coefficient. The patient status includes suffering from gastric disease, suffering from intestinal disease and not suffering from the disease; Priori module: used to obtain the medical record data of the patient to be tested when the patient's condition is gastric disease or intestinal disease, and determine the prior probability based on the medical record data; Probability generation module: used to obtain marker feature data of the patient to be tested based on the patient status, determine the marker interval sequence based on the marker feature data, input the marker interval sequence and the patient status into the pre-built cancer probability prediction model, and obtain the real-time sensitivity probability and real-time specificity probability; Evaluation module: used to calculate the positive cancer probability based on the prior probability, real-time sensitivity probability and real-time specificity probability, and determine the evaluation result according to the positive cancer probability.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the deep learning-based disease risk assessment method according to any one of claims 1 to 4 is implemented.

Images