Establishment method and system of leukemia artificial intelligence preliminary screening model based on blood routine data

Abstract

The invention provides a method and a system for establishing a leukemia artificial intelligence preliminary screening model based on blood routine data. The leukemia artificial intelligence preliminary screening model establishing method comprises the steps that S1, electronic medical record data of leukemia patients and non-leukemia patients are collected and processed, and a plurality of training databases and a verification database are obtained; s2, an analysis module is adopted to train the training databases respectively, a plurality of corresponding rule sets are obtained in sequence, and the rule sets with the false negative rate smaller than 5% are summarized to obtain a final rule set, namely a leukemia risk screening model; s3, verifying a verification database by adopting the leukemia risk screening model; and S4, iterating the leukemia risk screening model for multiple times to obtain an optimal model of the leukemia risk screening model. The leukemia artificial intelligence primary screening model established by the method can be used for rapidly and accurately screening the leukemia.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to the field of leukemia screening technology based on AI technology. Background Art

[0002] Leukemia is a type of malignant proliferative disease of hematopoietic stem cells, characterized by the uncontrolled malignant proliferation of immature blood cells in the bone marrow. It is also one of the top ten malignant tumors in China. The occurrence of the disease is related to factors such as genetics, environmental pollution, and differences in living habits of people in different regions. According to statistics in 2019, the male mortality rate was 4.5 / 100,000, higher than 3.8 / 100,000 female patients. Children and adolescents are prone to occur in the age group of under 5 years old and 5-14 years old, with acute lymphocytic leukemia being the most serious, accounting for the first place in the incidence and mortality of childhood tumors. It can be seen that leukemia has caused serious damage to people's health, and also brought serious burdens to society and families.

[0003] Acute leukemia patients often do not know that they have leukemia. They usually go to the hospital for treatment based on the early symptoms, such as colds and fevers, bleeding gums, subcutaneous bleeding, anemia, etc. The initial diagnosis of leukemia currently relies mainly on the examination of white blood cell morphology. Due to the high intensity of work and fatigue caused by continuous work, the efficiency of manual inspection is low. Therefore, rapid and accurate screening for leukemia has great clinical significance for saving patients' lives. Summary of the invention

[0004] In order to overcome the above technical defects, the first aspect of the present invention provides a method for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data, which comprises:

[0005] Step S1: using the analysis module to train several training databases respectively, to obtain several corresponding rule sets in turn, and to combine and aggregate several rule sets with false negative rates less than 5% to obtain the final rule set, i.e., the leukemia risk screening model;

[0006] Step S2: The leukemia risk screening model performs statistics on the verification database to obtain the false negative rate. The rule set with a false negative rate of <5% is judged as the rule set that meets the verification standard. If the false negative rate of the rule set is ≥5%, return to step S2, adjust the settings before starting the rule training, start the rule training again, and obtain a new rule set until the false negative rate is <5%;

[0007] Step S3: Iterate the leukemia risk screening model multiple times, and find the optimal model of the leukemia risk screening model under the constraints of a specific number of iterations and calculation time through multiple rounds of model calculations.

[0008] Furthermore, before step S1, the method further comprises:

[0009] Step S11: Prepare medical record data; collect patient electronic medical records from the hospital electronic medical record platform, and collect electronic medical records of leukemia patients and non-leukemia patients with diagnosis results;

[0010] Step S12: Medical feature extraction: extracting leukemia medical features from the qualified electronic medical record data obtained in step S11, extracting medical features and medical feature values; the leukemia medical features include basic information, blood routine items, blood routine research parameters, and XN instrument Q-Flag information;

[0011] Step S13: feature data standardization and data cleaning: perform feature data standardization on the big data of leukemia clinical and laboratory tests obtained in step S12, remove data with missing values, and obtain a standard data sample, which includes a standard medical feature data set and a standard diagnosis result set, and the data in the standard medical feature data set and the data in the standard diagnosis result set are in a one-to-one correspondence;

[0012] Step S14: Feature screening: The standard data is statistically analyzed and rule-mined using Laboman AI software to screen out characteristic parameters related to leukemia, wherein the characteristic parameters related to leukemia include: total number of specimens, number of true positive specimens, number of true negative specimens, number of predicted positive specimens, number of predicted negative specimens, false negative rate, false positive rate, true negative rate, true positive rate, release rate, total effective rate, positive predictive value, negative predictive value, specificity, sensitivity, Youden index, false negative rate-confidence interval, T test, key case coverage, and 19 statistical indicators;

[0013] Step S15: Data set splitting: pre-segment the data and divide the data equally into several training databases and a verification database.

[0014] Furthermore, the blood routine test items include 35 specific characteristic fields, as shown in the following table:

[0015]

[0016]

[0017] The blood routine research parameters include 52 specific feature fields, as shown in the following table:

[0018]

[0019]

[0020]

[0021] The XN instrument Q-Flag information includes 14 specific feature fields, as shown in the following table:

[0022] Project Name explain Blasts / Abn Lympho? Possibility of blasts / lymphocyte abnormalities Blasts?* There is a possibility of explosion Abn Lympho?* Possibility of lymphocyte abnormality Left Shift? Possibility of shifting left A typical Lympho? Possibility of atypical lymphocytes NRBC Present High nucleated red blood cell count IG Present Immature granulocytes RBC Agglutination? Possibility of red blood cell agglutination Turbidity / HGB Interf? Possibility of Celiac Disease Influencing HGB Iron Deficiency? Possibility of iron deficiency HGB Defect? Possibility of HGB abnormality Fragments? Possibility of red blood cell fragmentation Anemia anemia PLT Clumps? The possibility of PLT agglomeration .

[0023] Further, step S13 includes:

[0024] S131 Feature data standardization: Establish a standard library and a leukemia professional database, and manually construct a standard library of blood routine items, research parameters, Q-Flag information and other medical entity nouns based on the leukemia professional database. The standard library contains the standard name of each medical noun and similar names that have appeared, and is encoded for unique identification to form a feature standard library. For the medical features and medical feature values ​​extracted in step S12, replace different expressions of the same feature with the feature standard library, unify the feature description, and obtain standardized medical feature data. Specifically, the replacement of medical feature values ​​mainly involves the replacement of symbols, letters, words, units, and medical codes, and unifies symbols, letters, words, units, and medical codes;

[0025] S132 Data cleaning: Perform the following preprocessing on the data: (1) Mark feature data and label information; (2) If a large amount of key feature data information of the specimen is missing, discard the specimen; (3) If the label information of the specimen is missing, discard the specimen; (4) Fill the empty value field with the average value of the field; For fields with a very large missing ratio, such as >=80%, discard the field.

[0026] Furthermore, the several training databases include training database A, training database B, and training database C, and the step S1 includes: (1) before starting rule training, pre-setting in the analysis module: model training intensity, combination condition number, maximum training round, false negative rate threshold, operation time limit, automatic merging of redundant rules on or off, key must be on or off, 7 settings; importing training database A into the analysis module; then starting rule training, and obtaining a preliminary rule set A after the operation is completed; for the obtained rule set A, using single rule overfitting, single rule false negative rate, single rule false positive rate, 3 indicators to optimize the rule set Then; (2) according to the mined rule set, check the statistical results, if the false negative rate is less than 5%, judge it as the rule set that meets the training standard; if the false negative rate is not less than 5%, return to step (1), adjust the settings before starting rule training, modify the number of combined conditions, false negative rate threshold, and running time limit, and start rule training again until the false negative rate is less than 5%, and obtain the final rule set A; (3) Repeat step S1, use database B and database C to obtain rule set B and rule set C respectively; (4) Finally, rule set A, rule set B, and rule set C are summarized to obtain the final rule set, which is called leukemia risk screening model.

[0027] The second aspect of the present application provides a system for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data, which includes:

[0028] An analysis module, wherein the analysis module is used to train the plurality of training databases, obtain a plurality of corresponding rule sets in turn, and aggregate a plurality of rule sets with false negative rates of less than 5% to obtain a final rule set, i.e., a leukemia risk screening model; and is also used to use the leukemia risk screening model to perform statistics on the verification database to obtain a false negative rate, and a rule set with a false negative rate of less than 5% is judged as a rule set that meets the verification standard. If the false negative rate of the rule set is ≥5%, the settings before starting the rule training are adjusted, and the rule training is started again to obtain a new rule set until the false negative rate is less than 5%;

[0029] The iteration module is used to perform multiple iterations on the leukemia risk screening model, and find the optimal model of the leukemia risk screening model under the constraints of a specific number of iterations and calculation time through multiple rounds of model calculations.

[0030] Furthermore, the analysis module includes n groups of sub-models, n groups of statistical modules and a group of selection modules; each of the sub-models includes a decision tree algorithm module, the sub-model divides the feature data, continuously constructs suitable rules according to the decision tree algorithm module, and repeats the steps of the decision tree algorithm module several times to select the qualified model of the current round, and finally counts and selects a group of optimal models through the collaborative calculation of n groups of sub-models; the statistical module is used to use a group of rules obtained by model mining as input, and calculate various statistical indicators according to the actual annotated label data in the sample set; the selection module is used to use a series of statistical indicators of statistical modules as input, and select a group of sub-models with the best false negative rate and false positive rate indicators from the sub-models corresponding to a series of statistical data.

[0031] Furthermore, the system for establishing the leukemia artificial intelligence preliminary screening model based on routine blood data further includes a data acquisition and processing module, and the data acquisition and processing module includes:

[0032] A medical record data collection module, which is used to collect electronic medical records of patients from the hospital's electronic medical record platform, and collect electronic medical records of leukemia patients and non-leukemia patients with diagnosis results;

[0033] The medical feature extraction module is used to extract leukemia medical features from the qualified electronic medical record data obtained by the medical record data collection module, and extract medical features and medical feature values; the leukemia medical features include basic information, blood routine items, blood routine research parameters, and XN instrument Q-Flag information; the blood routine research parameters include 52 specific feature fields, as shown in the following table:

[0034]

[0035]

[0036]

[0037] The XN instrument Q-Flag information includes 14 specific feature fields, as shown in the following table:

[0038] Project Name explain Blasts / Abn Lympho? Possibility of blasts / lymphocyte abnormalities Blasts?* There is a possibility of explosion Abn Lympho?* Possibility of lymphocyte abnormality Left Shift? Possibility of shifting left A typical Lympho? Possibility of atypical lymphocytes NRBC Present High nucleated red blood cell count IG Present Immature granulocytes RBC Agglutination? Possibility of red blood cell agglutination Turbidity / HGB Interf? Possibility of Celiac Disease Influencing HGB Iron Deficiency? Possibility of iron deficiency HGB Defect? Possibility of HGB abnormality Fragments? Possibility of red blood cell fragmentation Anemia anemia PLT Clumps? The possibility of PLT agglomeration

[0039] A feature data standardization and data cleaning module, which is used to standardize the feature data of the big data of leukemia clinical and testing, remove data with missing values, and obtain a standard data sample. The standard data sample includes a standard medical feature data set and a standard diagnosis result set. The data in the standard medical feature data set and the data in the standard diagnosis result set are in a one-to-one correspondence;

[0040] A feature screening module, which is used to perform statistics and rule mining on standard data using Laboman AI software to screen out characteristic parameters related to leukemia, wherein the characteristic parameters related to leukemia include: total number of specimens, number of true positive specimens, number of true negative specimens, number of predicted positive specimens, number of predicted negative specimens, false negative rate, false positive rate, true negative rate, true positive rate, release rate, total effective rate, positive predictive value, negative predictive value, specificity, sensitivity, Youden index, false negative rate-confidence interval, T test, key case coverage, and 19 statistical indicators;

[0041] The data set splitting module is used to pre-split the data and divide the data into several training databases and a verification database.

[0042] Furthermore, feature data standardization and data cleaning include:

[0043] A feature data standardization module, wherein the feature data standardization module is used to establish a standard library and a leukemia professional database, and artificially construct a standard library of blood routine items, research parameters, Q-Flag information and other medical entity nouns based on the leukemia professional database. The standard library contains the standard name of each medical noun and similar names that have appeared, and is encoded for unique identification to form a feature standard library. For the extracted medical features and medical feature values, different expressions of the same feature are replaced with the feature standard library to unify the feature description and obtain standardized medical feature data. Specifically, the replacement of medical feature values ​​mainly involves the replacement of symbols, letters, words, units, and medical codes, and the symbols, letters, words, units, and medical codes are unified.

[0044] The data cleaning module is used to perform the following preprocessing on the data: (1) marking feature data and label information; (2) if a large amount of key feature data information of the specimen is missing, the specimen is discarded; (3) if the label information of the specimen is missing, the specimen is discarded; (4) the empty value field is filled with the average value of the field; for fields with a very large missing ratio, such as ≥80%, the field is discarded.

[0045] Furthermore, the plurality of training databases include training database A, training database B, and training database C, and the analysis module is used to: (1) before starting rule training, pre-set in the analysis module: model training intensity, number of combined conditions, maximum training rounds, false negative rate threshold, running time limit, automatic merging of redundant rules on or off, and key hit on or off, seven settings; import the training database A into the analysis module; then start rule training, and after the running is completed, obtain a preliminary rule set A; for the obtained rule set A, use three indicators, single rule overfitting, single rule false negative rate, and single rule false positive rate, to optimize ization rules; (2) according to the mined rule set, check the statistical results, if the false negative rate is less than 5%, judge the rule set as the rule set that meets the training standard; if the false negative rate is ≥5%, return to step (1), adjust the settings before starting the rule training, modify the number of combined conditions, false negative rate threshold, and running time limit, and start the rule training again until the false negative rate is less than 5%, and obtain the final rule set A; (3) Repeat (1) and (2), use database B and database C to obtain rule set B and rule set C respectively; (4) Finally, rule set A, rule set B, and rule set C are summarized to obtain the final rule set, which is called the leukemia risk screening model.

[0046] Compared with the prior art, the above technical solution has the following beneficial effects:

[0047] This application uses the learning ability of the decision tree algorithm to divide the data into several training data groups and a verification data group. Several training data groups are used to establish rules, so that multiple sets of rules can be obtained. The final merging of rules can achieve a reduction in the false negative rate, reaching a false negative rate of less than 1.5%, which is far superior to the statistical requirement of a false negative rate of less than 5%. The leukemia artificial intelligence initial screening model established by the present invention can quickly and accurately screen out leukemia, which has great clinical significance for saving patients' lives. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the artificial intelligence initial screening method for leukemia based on routine blood data in this application;

[0049] Figure 2 Flowchart for establishing an artificial intelligence screening model for leukemia;

[0050] Figure 3This is a structural diagram of the system for establishing an artificial intelligence initial screening model for leukemia based on routine blood data. DETAILED DESCRIPTION

[0051] The advantages of the present invention are further described below in conjunction with the accompanying drawings and specific embodiments. Those skilled in the art should understand that the following specific description is illustrative rather than restrictive, and should not be used to limit the scope of protection of the present invention.

[0052] like Figure 1-Figure 2 As shown, this embodiment provides an artificial intelligence initial screening method for leukemia based on routine blood data, which includes steps 1 to 3.

[0053] Step 1: Establish a leukemia risk screening model, including steps 11 to 14.

[0054] like Figure 3 As shown, this embodiment establishes a leukemia risk screening model by establishing a system for an artificial intelligence initial screening model for leukemia based on routine blood data. The system includes: a data acquisition and processing module, an analysis module, and an iteration module.

[0055] The method for establishing a leukemia risk screening model using the above-mentioned system includes steps 11 to 14:

[0056] Step 11: Data collection and processing module: Collect electronic medical record data of leukemia patients and non-leukemia patients, extract medical features and medical feature values, standardize and clean the feature data, screen out feature parameters related to leukemia, and finally divide the data into several training databases and a verification database according to time periods, so as to serve as samples to be analyzed.

[0057] Step 11 includes steps 111 to 115:

[0058] Step 111: Prepare medical record data; collect patient electronic medical records from the hospital's electronic medical record platform, and collect electronic medical records of leukemia patients and non-leukemia patients with diagnosis results.

[0059] The diagnosis result is a method for collecting electronic medical records of leukemia patients: comparing the electronic medical record diagnosis result with the disease name in the leukemia disease name database to obtain the electronic medical record of the leukemia patient;

[0060] The electronic medical record collection method of the non-leukemia patients is to collect the data of the patients and the physical examination personnel in the same period, excluding the electronic medical records of patients with unclear medical history and incomplete test data; the medical record data includes medical records, test results, doctor's orders, operation records, nursing records, and real diagnosis results, and the test results include medical characteristics and thresholds;

[0061] Obtain qualified electronic medical record data;

[0062] The leukemia disease name database contains various medical disease names that can be judged as leukemia.

[0063] Step 112: Medical feature extraction: extract leukemia medical features from the qualified electronic medical record data obtained in step 111, extract medical features and medical feature values; the leukemia medical features include basic information, blood routine test items, blood routine research parameters, and XN instrument Q-Flag information.

[0064] The basic information table includes: gender, age, specimen number, date, medical record number, 5 specific feature fields;

[0065] The blood routine test items have a total of 35 specific characteristic fields, see Table 1 for details; the blood routine research parameters have a total of 52 specific characteristic fields, see Table 2 for details; the XN instrument Q-Flag information has a total of 14 specific characteristic fields, see Table 3 for details.

[0066] Table 1 Detailed list of blood routine items

[0067]

[0068]

[0069] Table 2 Detailed list of blood routine study parameters

[0070]

[0071]

[0072]

[0073] Table 3 XN instrument Q-Flag information details

[0074]

[0075]

[0076] The medical characteristic values ​​are specific values ​​of each medical characteristic in basic information, blood routine items, research parameters, and Q-Flag information characteristics.

[0077] Obtain big data on leukemia clinical and laboratory data.

[0078] Step 113: feature data standardization and data cleaning: perform feature data standardization on the big data of leukemia clinical and laboratory tests obtained in step 112, remove data with missing values, and obtain a standard data sample, which includes a standard medical feature data set and a standard diagnosis result set. The data in the standard medical feature data set and the data in the standard diagnosis result set are in a one-to-one correspondence;

[0079] Step 113 includes 1131-1132:

[0080] Step 1131: Standardization of feature data: using conventional technical methods in the art to establish a standard library and a leukemia professional database, and artificially constructing a standard library of blood routine items, research parameters, Q-Flag information and other medical entity nouns based on the leukemia professional database. The standard library contains the standard name of each medical noun and similar names that have appeared, and is encoded for unique identification to form a feature standard library. For the medical features and medical feature values ​​extracted in step 112, different expressions of the same feature are replaced with the feature standard library, and the feature description is unified to obtain standardized medical feature data. Specifically, the replacement of medical feature values ​​mainly involves the replacement of symbols, letters, words, units, and medical codes, and the symbols, letters, words, units, and medical codes are unified.

[0081] Step 1132 Data cleaning: Perform the following preprocessing on the data: (1) Mark feature data and label information; (2) If a large amount of key feature data information of the specimen is missing, discard the specimen; (3) If the label information of the specimen is missing, discard the specimen; (4) Fill the empty value field with the average value of the field; For fields with a very large missing ratio, such as ≥80%, discard the field.

[0082] Step 114: Feature screening: The standard data is statistically analyzed and rule-mined using Laboman AI software to screen out feature parameters related to leukemia, wherein the feature parameters related to leukemia include: total number of specimens, number of true positive specimens, number of true negative specimens, number of predicted positive specimens, number of predicted negative specimens, false negative rate, false positive rate, true negative rate, true positive rate, release rate, total effective rate, positive predictive value, negative predictive value, specificity, sensitivity, Youden index, false negative rate-confidence interval, T test, key case coverage, and 19 statistical indicators;

[0083] The data were summarized using Excel, and the label set included: specimen number, date, medical record number, and positive label. Leukemia-positive cases were marked as 1, and non-leukemia-negative cases were marked as 0.

[0084] Laboman AI software was used to perform statistics and rule mining on standard data, screen out parameters related to leukemia, and mine a set of judgment rules.

[0085] The Laboman AI software is the Laboman AI rule system (blood version) software, which is a medical inspection rule mining and analysis software suitable for use in hospitals. It can perform statistical analysis based on specimen result data and microscopic data collected by medical inspection equipment, and mine effective judgment rules. Its main functions include: creating new experiments, importing experiments, rule training, label judgment rule settings, project team settings, etc.

[0086] The standard data were statistically analyzed and rule-mined using Laboman AI software. The rules mined included 19 statistical indicators, including the total number of specimens, the number of positive specimens (real), the number of negative specimens (real), the number of positive specimens (predicted), the number of negative specimens (predicted), the false negative rate, the false positive rate, the true negative rate, the true positive rate, the release rate, the total effective rate, the positive predictive value, the negative predictive value, the specificity, the sensitivity, the Youden index, the false negative rate-confidence interval, the T test, and the coverage rate of key cases. These statistical items are commonly used in statistics, and Laboman AI software belongs to the prior art.

[0087] The present invention uses the above-mentioned statistical method and Laboman AI software to perform statistics and rule mining to obtain the false negative rate. We set the false negative rate to be less than 0.05%. It can be considered that the obtained rules have an extremely significant correlation with the diagnosis of leukemia, and the model established by this set of rules is reasonable.

[0088] The false negative rate is the ratio of the number of leukemias missed by the leukemia risk screening model in the test data to the total number of test data;

[0089] The false positive rate is the ratio of the number of leukemias misdiagnosed by the leukemia risk screening model in the test data to the total number of test data;

[0090] The true negative rate is the ratio of the number of true non-leukemias confirmed by the leukemia risk screening model in the test data to the total number of test data;

[0091] The true positive rate is the ratio of the number of true leukemias confirmed by the leukemia risk screening model in the test data to the total number of test data;

[0092] The release rate is the ratio of the sum of the number of true non-leukemias and the number of missed leukemias determined by the leukemia risk screening model in the test data to the total number of test data;

[0093] The total effective rate is the ratio of the sum of the number of true non-leukemias and the number of true leukemias determined by the leukemia risk screening model in the test data to the total number of test data;

[0094] The total effective rate is the ratio of the sum of the number of true non-leukemias and the number of true leukemias determined by the leukemia risk screening model in the test data to the total number of test data;

[0095] The positive predictive value = number of true positive cases / (number of true positive cases + number of false positive cases)*100%;

[0096] The negative predictive value = number of true negative cases / (number of true negative cases + number of false negative cases)*100%;

[0097] The sensitivity = number of true positives / (number of true positives + number of false negatives)*100%;

[0098] The specificity = number of true negatives / (number of true negatives + number of false positives))*100%;

[0099] The Youden index = (sensitivity + specificity) - 1;

[0100] The false negative rate - confidence interval = the statistical result of 1000 50% random samplings;

[0101] The T test, in is the sample mean, μ is the population mean, σ X is the sample standard deviation, n is the sample size;

[0102] The key case coverage rate is the number of marked cases determined by the rules accounting for the total number of marked cases in the database.

[0103] Step 115: Dataset splitting: pre-segment the data and divide the data equally into several training databases and a validation database.

[0104] The data is pre-split and divided into four databases A, B, C, and D according to time periods. The first three databases A, B, and C are used to establish rules, and the last database D is used to verify the rules.

[0105] Step 12: Analysis module: The analysis module is used to train the several training databases respectively, to obtain several corresponding rule sets in turn, and several rule sets with false negative rates less than 5% are combined and summarized to obtain the final rule set, i.e., the leukemia risk screening model.

[0106] The analysis module consists of n groups of sub-models, statistical modules and a group of selection modules. The sub-model divides the feature data, continuously builds suitable rules according to the decision tree algorithm module, and repeats the algorithm module steps n times to select the qualified model for the current round. Through the collaborative calculation of n groups of sub-models, a group of optimal models are finally counted and selected. The statistical module takes a group of rules mined from a set of models as input, and calculates various statistical indicators such as false negative rate, false positive rate, specificity, sensitivity, etc. according to the actual labeled data in the sample set. The selection module takes a series of statistical indicators of statistical modules as input, and selects a group of sub-models with the best false negative rate, false positive rate and other indicators in the sub-models corresponding to a series of statistical data.

[0107] The several training databases include training database A, training database B, and training database C. Step 12 includes: (121) Excel data can be imported into Laboman AI software. In the importing step, key cases can be marked, and the coverage rate of key cases can be calculated during statistics; the training database A is imported into Laboman AI, which is called training data A. This part of the data is used for rule mining and establishment; before starting rule training, the following seven settings are pre-set in the analysis module: model training intensity, number of combined conditions, maximum training rounds, false negative rate threshold, running time limit, automatic merging of redundant rules on or off, and key hit on or off; the training database A is imported into the analysis module; then rule training is started, and a preliminary rule set A is obtained after the operation is completed; for the obtained rule set A, three indicators, single rule overfitting, single rule false negative rate, and single rule false positive rate, are used to optimize the rules; (122) According to the mined rule set, the statistical results are checked and re- Check whether the false negative rate result is less than 5%. If the rule set with a false negative rate of less than 5% is judged as a rule set that meets the training standard; if the false negative rate is not less than 5%, return to step (121), adjust the settings before starting the rule training, focus on modifying the number of combined conditions, the false negative rate threshold, and the running time limit, and start the rule training again until the false negative rate is less than 5%, and obtain the final rule set A; (123) Repeat step 12, and use database B and database C to obtain rule set B and rule set C respectively; (124) Finally, rule set A, rule set B, and rule set C are summarized to obtain the final rule set, which is called the leukemia risk screening model. Among them, rule set A is specifically shown in Table 4, rule set B is specifically shown in Table 5, rule set C is specifically shown in Table 6, and the final rule set is specifically shown in Table 7. The number of combined conditions is the maximum number of parameters that each mined rule is combined with. The false negative rate threshold is the ratio of the number of white blood cells correctly predicted by the leukemia risk screening model in the test data to the total number of leukemias predicted by the leukemia risk screening model.

[0108] Table 4 Rule Set A

[0109] Serial number Rules Details 1 [PLT-I]<=130.0AND[RBC]<=3.015AND[Age]<=49.5 2 [NEUT#&]<=2.165AND[RBC]<=3.015AND[Age]>49.5 3 [MO-WY]<=53.5AND[RBC]>3.015AND[Age]<=56.0 4 [LYMP#&]<=0.105AND[MO-WY]>53.5AND[RBC]>3.015 5 [EO%]<=30.85AND[NE-WY]<=1047.5AND[NE-WZ]>1045.5 6 [EO%]>30.85AND[NE-WY]<=1047.5 7 [LY-WY]<=1176.5AND[MONO%]>9.05AND[NE-WY]>1047.5 8 [MCV]<=99.4AND[NE-WY]>382.0AND[RDW-SD]<=29.5 9 [MCV]<=99.4AND[NE-WZ]<=53.0AND[RDW-SD]>29.5 10 [EO#]<  0.005AND[MCV]>99.4AND[NRBC%]<0.6 11 [MCV]>99.4AND[NRBC%]>0.6AND[Age]<=67.0

[0110] Table 5 Rule Set B

[0111] Serial number Rules Details 1 [WBC]<=1.935AND[age]>22.0AND[age]<=52.5 2 [WBC] <= 1.935 AND [age] > 52.5 3 [LY-Y]>87.35AND[NE-WX]<=346.5AND[WBC]>1.935 4 [LY-X]>84.05AND[NE-WX]>346.5AND[WBC]>1.935 5 [NE-WY]<=664.0AND[NE-WZ]>845.5AND[age]<=50.5 6 [MacroR]>4.0AND[NE-WY]>664.0AND[Age]<=50.5 7 [MO-Z]<=42.5AND[NEUT#&]>1.84AND[AGE]>50.5 8 [IG%]>16.635AND[MO-Z]>42.5AND[Age]>50.5 9 [EO%]<=32.05AND[LY-Z]>85.8AND[LYMPH#]<=2.945 10 [EO%]<=1.2AND[LYMPH#]>2.945 11 [EO%]>32.05AND[NE-SFL]>140.5 12 [LY-WX]>496.5AND[MO-WY]<=157.0AND[age]<=62.5 13 [MO-WY]>157.0AND[MO-Y]<=142.5AND[RDW-SD]<=6.0 14 [MO-WY]>157.0AND[MO-Y]>142.5

[0112] Table 6 Rule Set C

[0113]

[0114]

[0115] Table 7 Final rule set

[0116] Serial number rule 1 [PLT-I]<=130.0AND[RBC]<=3.015AND[Age]<=49.5 2 [NEUT#&]<=2.165AND[RBC]<=3.015AND[Age]>49.5 3 [MO-WY]<=53.5AND[RBC]>3.015AND[Age]<=56.0 4 [LYMP#&]<=0.105AND[MO-WY]>53.5AND[RBC]>3.015 5 [EO%]<=30.85AND[NE-WY]<=1047.5AND[NE-WZ]>1045.5 6 [EO%]>30.85AND[NE-WY]<=1047.5 7 [LY-WY]<=1176.5AND[MONO%]>9.05AND[NE-WY]>1047.5 8 [MCV]<=99.4AND[NE-WY]>382.0AND[RDW-SD]<=29.5 9 [MCV]<=99.4AND[NE-WZ]<=53.0AND[RDW-SD]>29.5 10 [EO#]<  0.005AND[MCV]>99.4AND[NRBC%]<0.6 11 [MCV]>99.4AND[NRBC%]>0.6AND[Age]<=67.0 12 [WBC]<=1.935AND[age]>22.0AND[age]<=52.5 13 [WBC] <= 1.935 AND [age] > 52.5 14 [LY-Y]>87.35AND[NE-WX]<=346.5AND[WBC]>1.935 15 [LY-X]>84.05AND[NE-WX]>346.5AND[WBC]>1.935 16 [NE-WY]<=664.0AND[NE-WZ]>845.5AND[age]<=50.5 17 [MacroR]>4.0AND[NE-WY]>664.0AND[Age]<=50.5 18 [MO-Z]<=42.5AND[NEUT#&]>1.84AND[AGE]>50.5 19 [IG%]>16.635AND[MO-Z]>42.5AND[Age]>50.5 20 [EO%]<=32.05AND[LY-Z]>85.8AND[LYMPH#]<=2.945 21 [EO%]<=1.2AND[LYMPH#]>2.945 22 [EO%]>32.05AND[NE-SFL]>140.5 23 [LY-WX]>496.5AND[MO-WY]<=157.0AND[age]<=62.5 24 [MO-WY]>157.0AND[MO-Y]<=142.5AND[RDW-SD]<=6.0 25 [MO-WY]>157.0AND[MO-Y]>142.5 26 [NE-SSC]<=165.7AND[NEUT#&]<=1.845AND[RBC]<=3.69 27 [NEUT#&]<=1.845AND[RBC]>3.69AND[RDW-CV]<=11.8 28 [NEUT#&]>1.845AND[RBC]<=3.575AND[Age]<=49.5 29 [MO-WX]<=107.0AND[NEUT#&]>1.845AND[AGE]>49.5 30 [MONO%]<=5.4AND[NE-WX]<=382.5AND[NEUT#&]<=2.98 31 [BA-D%]>1.285AND[NE-WX]<=382.5AND[NEUT#&]>2.98 32 [LY-X]>87.05AND[NE-WX]>382.5AND[age]>18.5

[0117] The method of using the final rule set in Table 7 (i.e., leukemia risk screening model) to judge leukemia is as follows: 1. Each rule has a combination of 2-3 items, and they are in an AND (and) relationship, that is, they need to meet the judgment requirements at the same time to be a positive patient, which refers to leukemia patients here; 2. The rule set is a combination of several rules, and each rule in the rule set is in an OR (or) relationship, that is, if any rule is met, it is a positive patient, which refers to leukemia patients here. For example, the first rule in the final rule set is: [PLT-I] <= 130.0 AND [RBC] <= 3.015 AND [age] <= 49.5. When the patient's test results meet ([PLT-I] <= 130.0 and [RBC] <= 3.015 and [age] <= 49.5) at the same time, it indicates that the patient may be a leukemia patient, otherwise the second rule is verified. When none of the 32 rules determines that the patient has leukemia, the patient is prompted to be a non-leukemia patient.

[0118] Step 13: The leukemia risk screening model performs statistics on the verification database to obtain the false negative rate. The rule set with a false negative rate of less than 5% is judged as the rule set that meets the verification standard. If the false negative rate of the rule set is ≥5%, return to step 12, adjust the settings before starting rule training, start rule training again, and obtain a new rule set until the false negative rate is less than 5%.

[0119] The leukemia risk screening model performs statistics on the training database D obtained in step 115. Like the training data, the total number of specimens, the number of positive specimens (real), the number of negative specimens (real), the number of positive specimens (predicted), the number of negative specimens (predicted), the false negative rate, the false positive rate, the true negative rate, the true positive rate, the release rate, the total effective rate, the positive predictive value, the negative predictive value, the specificity, the sensitivity, the Youden index, the false negative rate-confidence interval, the T test, the key case coverage rate, and 19 statistical indicators can be obtained. Focus on checking whether the false negative rate result is less than 5%. If the rule set with a false negative rate of less than 5% is judged as the rule set that has been verified to meet the standard. If the false negative rate is not less than 5%, return to step 12, adjust the settings before starting the rule training, focus on modifying the number of combined conditions, the false negative rate threshold, and the running time limit, and start the rule training again to obtain a new rule set until the false negative rate is less than 5%.

[0120] The final rule set was validated using the validation database D, and statistics showed that the false negative rate was 1.30%.

[0121] Step 14: Iteration module: Iterate the leukemia risk screening model multiple times, and find the optimal model of the leukemia risk screening model under the specific number of iterations and calculation time limits through multiple rounds of model calculations. The iteration module is a multiple iteration of the analysis module, and aims to find the optimal model under the specific number of iterations and calculation time limits through multiple rounds of model calculations.

[0122] Step 2: Organize the data of patients to be screened.

[0123] The patient data to be screened are sorted according to the method of step 112 to step 115. The user data includes medical characteristics and medical characteristic values.

[0124] The leukemia medical characteristics include basic information, blood routine test items, blood routine research parameters, and XN instrument Q-Flag information. The blood routine test items have a total of 35 specific feature fields, see Table 1 for details; the blood routine research parameters have a total of 52 specific feature fields, see Table 2 for details; the XN instrument Q-Flag information has a total of 14 specific feature fields, see Table 3 for details.

[0125] The medical characteristic values ​​are specific values ​​of each medical characteristic in basic information, blood routine items, blood routine research parameters, and XN instrument Q-Flag information characteristics.

[0126] Obtain big data on leukemia clinical and laboratory data.

[0127] Step 3: Substitute the data of the patient to be screened into the leukemia risk screening model for calculation, and finally obtain the leukemia risk prediction result.

[0128] For example, some test results of the patient to be screened are male, 41 years old, PLT-I=120.0, WBC=19.73, RBC=2.97, HGB=94, HCT=0.296, MCV=29.6, MCHC=31.6, etc. According to the final rule set in Table 7, the patient is judged to be positive. Later, the patient's clinical diagnosis was confirmed to be positive.

[0129] The laboratory department can quickly and accurately screen out leukemia patients based on routine blood test results and use the leukemia risk screening model.

[0130] It should be noted that the embodiments of the present invention have better practicability and do not impose any form of limitation on the present invention. Any technician familiar with the field may use the technical content disclosed above to change or modify it into an equivalent effective embodiment. However, any modification or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for establishing an artificial intelligence initial screening model for leukemia based on routine blood data, characterized in that: include: Step S1: using the analysis module to train several training databases respectively, to obtain several corresponding rule sets in turn, and to combine and aggregate several rule sets with false negative rates less than 5% to obtain the final rule set, i.e., the leukemia risk screening model; Step S2: The leukemia risk screening model performs statistics on the verification database to obtain the false negative rate. The rule set with a false negative rate of <5% is judged as the rule set that meets the verification standard. If the false negative rate of the rule set is ≥5%, return to step S2, adjust the settings before starting the rule training, start the rule training again, and obtain a new rule set until the false negative rate is <5%; Step S3: Iterate the leukemia risk screening model multiple times, and find the optimal model of the leukemia risk screening model under the constraints of a specific number of iterations and calculation time through multiple rounds of model calculations.

2. The method for establishing an artificial intelligence primary screening model for leukemia based on routine blood data according to claim 1, characterized in that: Before step S1, the method further comprises: Step S11: Prepare medical record data; collect patient electronic medical records from the hospital electronic medical record platform, and collect electronic medical records of leukemia patients and non-leukemia patients with diagnosis results; Step S12: Medical feature extraction: extracting leukemia medical features from the qualified electronic medical record data obtained in step S11, extracting medical features and medical feature values; the leukemia medical features include basic information, blood routine items, blood routine research parameters, and XN instrument Q-Flag information; Step S13: feature data standardization and data cleaning: perform feature data standardization on the big data of leukemia clinical and laboratory tests obtained in step S12, remove data with missing values, and obtain a standard data sample, which includes a standard medical feature data set and a standard diagnosis result set, and the data in the standard medical feature data set and the data in the standard diagnosis result set are in a one-to-one correspondence; Step S14: Feature screening: The standard data is statistically analyzed and rule-mined using Laboman AI software to screen out characteristic parameters related to leukemia, wherein the characteristic parameters related to leukemia include: total number of specimens, number of true positive specimens, number of true negative specimens, number of predicted positive specimens, number of predicted negative specimens, false negative rate, false positive rate, true negative rate, true positive rate, release rate, total effective rate, positive predictive value, negative predictive value, specificity, sensitivity, Youden index, false negative rate-confidence interval, T test, key case coverage, and 19 statistical indicators; Step S15: Data set splitting: pre-segment the data and divide the data equally into several training databases and a verification database.

3. The method for establishing an artificial intelligence primary screening model for leukemia based on routine blood data according to claim 2, characterized in that: The blood routine research parameters include 52 specific feature fields, as shown in the following table: The XN instrument Q-Flag information includes 14 specific feature fields, as shown in the following table: 。 4. The method for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data according to claim 1, characterized in that: Step S13 includes: S131 Feature data standardization: Establish a standard library and a leukemia professional database, and manually construct a standard library of blood routine items, blood routine research parameters, XN instrument Q-Flag information and other medical entity nouns based on the leukemia professional database. The standard library contains the standard name of each medical noun and similar names that have appeared, and encodes them for unique identification to form a feature standard library. For the medical features and medical feature values ​​extracted in step S12, replace different expressions of the same feature with the feature standard library, unify the feature description, and obtain standardized medical feature data. Specifically, the replacement of medical feature values ​​mainly involves the replacement of symbols, letters, words, units, and medical codes, and unifies symbols, letters, words, units, and medical codes; S132 Data cleaning: Perform the following preprocessing on the data: (1) Mark feature data and label information; (2) If a large amount of key feature data information is missing, discard the sample; (3) If the label information of the sample is missing, discard the sample; (4) Fill the empty value field with the average value of the field; For fields with a very large missing ratio, such as ≥80%, discard the field.

5. The method for establishing an artificial intelligence primary screening model for leukemia based on routine blood data according to claim 1, characterized in that: The plurality of training databases include training database A, training database B, and training database C. Step S1 includes: (1) before starting rule training, presetting in the analysis module: model training intensity, number of combined conditions, maximum training rounds, false negative rate threshold, running time limit, automatic merging of redundant rules on or off, and key hit on or off, seven settings; importing training database A into the analysis module; then starting rule training, and obtaining a preliminary rule set A after the running is completed; for the obtained rule set A, using three indicators, single rule overfitting, single rule false negative rate, and single rule false positive rate, to optimize the rules; (2) According to the mined rule set, check the statistical results. If the rule set with a false negative rate lower than 5% is judged as a rule set that meets the training standard; if the false negative rate is not lower than 5%, return to step (1), adjust the settings before starting the rule training, modify the number of combined conditions, false negative rate threshold, and running time limit, and start the rule training again until the false negative rate is lower than 5%, and obtain the final rule set A; (3) Repeat step S1, and use database B and database C to obtain rule set B and rule set C respectively; (4) Finally, rule set A, rule set B, and rule set C are aggregated to obtain the final rule set, which is called the leukemia risk screening model.

6. A system for establishing an artificial intelligence initial screening model for leukemia based on routine blood data, characterized in that: include: An analysis module, wherein the analysis module is used to train the plurality of training databases, obtain a plurality of corresponding rule sets in turn, and aggregate a plurality of rule sets with false negative rates of less than 5% to obtain a final rule set, i.e., a leukemia risk screening model; and is also used to use the leukemia risk screening model to perform statistics on the verification database to obtain a false negative rate, and a rule set with a false negative rate of less than 5% is judged as a rule set that meets the verification standard. If the false negative rate of the rule set is ≥5%, the settings before starting the rule training are adjusted, and the rule training is started again to obtain a new rule set until the false negative rate is less than 5%; The iteration module is used to perform multiple iterations on the leukemia risk screening model, and find the optimal model of the leukemia risk screening model under the constraints of a specific number of iterations and calculation time through multiple rounds of model calculations.

7. The system for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data according to claim 6, characterized in that: The analysis module includes n groups of sub-models, n groups of statistical modules and a group of selection modules; each of the sub-models includes a decision tree algorithm module, the sub-model divides the feature data, continuously constructs suitable rules according to the decision tree algorithm module, and repeats the steps of the decision tree algorithm module several times to select the qualified model of the current round, and finally counts and selects a group of optimal models through the collaborative calculation of n groups of sub-models; the statistical module is used to use a group of rules obtained by model mining as input, and calculate various statistical indicators according to the actual annotated label data in the sample set; the selection module is used to use a series of statistical indicators of statistical modules as input, and select a group of sub-models with the best false negative rate and false positive rate indicators in the sub-models corresponding to a series of statistical data.

8. The system for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data according to claim 6, characterized in that: Further comprising a data acquisition and processing module, the data acquisition and processing module comprising: A medical record data collection module, which is used to collect electronic medical records of patients from the hospital's electronic medical record platform, and collect electronic medical records of leukemia patients and non-leukemia patients with diagnosis results; A medical feature extraction module, which is used to extract leukemia medical features from the qualified electronic medical record data obtained by the medical record data collection module, and extract medical features and medical feature values; the leukemia medical features include basic information, blood routine items, blood routine research parameters, and XN instrument Q-Flag information; The blood routine research parameters include 52 specific feature fields, as shown in the following table: The XN instrument Q-Flag information includes 14 specific feature fields, as shown in the following table: A feature data standardization and data cleaning module, which is used to standardize the feature data of the big data of leukemia clinical and testing, remove data with missing values, and obtain a standard data sample. The standard data sample includes a standard medical feature data set and a standard diagnosis result set. The data in the standard medical feature data set and the data in the standard diagnosis result set are in a one-to-one correspondence; A feature screening module, which is used to perform statistics and rule mining on standard data using Laboman AI software to screen out characteristic parameters related to leukemia, wherein the characteristic parameters related to leukemia include: total number of specimens, number of true positive specimens, number of true negative specimens, number of predicted positive specimens, number of predicted negative specimens, false negative rate, false positive rate, true negative rate, true positive rate, release rate, total effective rate, positive predictive value, negative predictive value, specificity, sensitivity, Youden index, false negative rate-confidence interval, T test, key case coverage, and 19 statistical indicators; The data set splitting module is used to pre-split the data and divide the data into several training databases and a verification database.

9. The system for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data according to claim 6, characterized in that: Feature data standardization and data cleaning include: A feature data standardization module, which is used to establish a standard library and a leukemia professional database, and artificially construct a standard library of blood routine items, blood routine research parameters, XN instrument Q-Flag information and other medical entity nouns based on the leukemia professional database. The standard library contains the standard name of each medical noun and similar names that have appeared, and is encoded for unique identification to form a feature standard library. For the extracted medical features and medical feature values, different expressions of the same feature are replaced with the feature standard library to unify the feature description and obtain standardized medical feature data. Specifically, the replacement of medical feature values ​​is mainly the replacement of symbols, letters, words, units, and medical codes, and the symbols, letters, words, units, and medical codes are unified; The data cleaning module is used to perform the following preprocessing on the data: (1) marking feature data and label information; (2) if a large amount of key feature data information of the specimen is missing, the specimen is discarded; (3) if the label information of the specimen is missing, the specimen is discarded; (4) the empty value field is filled with the average value of the field; for fields with a very large missing ratio, such as ≥80%, the field is discarded.

10. The system for establishing an artificial intelligence preliminary screening model for leukemia based on routine blood data according to claim 6, characterized in that: The plurality of training databases include training database A, training database B, and training database C. The analysis module is used to: (1) before starting rule training, pre-set the following seven settings in the analysis module: model training intensity, number of combined conditions, maximum training rounds, false negative rate threshold, running time limit, automatic merging of redundant rules on or off, and key hits on or off; import the training database A into the analysis module; then start rule training, and after the running is completed, obtain a preliminary rule set A; for the obtained rule set A, use three indicators, single rule overfitting, single rule false negative rate, and single rule false positive rate, to optimize the rule set. Then; (2) according to the mined rule set, check the statistical results. If the rule set with a false negative rate of less than 5% is judged as a rule set that meets the training standard; if the false negative rate is ≥5%, return to step (1), adjust the settings before starting rule training, modify the number of combined conditions, false negative rate threshold, and running time limit, and start rule training again until the false negative rate is less than 5%, and obtain the final rule set A; (3) Repeat (1) and (2), and use database B and database C to obtain rule set B and rule set C respectively; (4) Finally, rule set A, rule set B, and rule set C are summarized to obtain the final rule set, which is called the leukemia risk screening model.

Images