A malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data

Abstract

The application discloses a kind of based on blood data malignant tumor patient personalized nutrition assessment dynamic analysis method, it is related to personalized nutrition assessment technical field, including: with multilayer perception machine neural network as basic model, the data set S2 of blood routine and biochemical data is trained, obtain model_1;The historical blood routine and biochemical data of the person to be tested are added in data set S2 and constitute data set S3, and the weight of data set S3 is re-assigned;The model structure and parameter of model_1 are reloaded, the first layer parameter S3 of model_1 is fixed, the second layer and third layer parameter of model_1 are activated to continue training, and the training epoch is 200, obtain model_2.The application provides customized nutrition status assessment according to the data of each person's unique physiological change and treatment process, compared with general model, more can fit individual difference, and simultaneously, the ability of model is retrained by continuously incorporating new data, with the accumulation of patient historical data, the prediction accuracy and adaptability of model will be continuously improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical examination and disease identification, and particularly relates to a malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data. BACKGROUND

[0002] With the gradual increase of the incidence of malignant tumors, the incidence of complications such as malnutrition also shows an upward trend, but the clinical nutrition treatment rate of malignant tumor patients is low, and the malnutrition caused by malignant tumors urgently needs the attention of patients and medical staff.

[0003] At present, although there are nutrition assessment scales such as NRS2002, the method has the following shortcomings when applied to the medical industry:

[0004] (1) The scale evaluation method is highly subjective and has a long evaluation period, and cannot timely find the nutritional status of malignant tumor patients;

[0005] (2) The scale evaluation score is fixed, and cannot realize precise medical treatment for individual cases, and the evaluation method for a person can only use the same scale for evaluation, and the evaluation is highly subjective.

[0006] At present, although artificial intelligence methods are used for nutrition assessment, the model test results are poor when the data distribution of the test person, the detection instrument or the model generalization performance is insufficient. SUMMARY

[0007] To solve the above problems, the present application provides a malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data, which uses artificial intelligence methods to accurately evaluate the nutritional status of tumor patients.

[0008] To achieve the above purpose, the present application provides a malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data, comprising the following steps:

[0009] (1) Obtain a data set S2 based on blood routine and biochemical data through data screening and screening, and the training set: test set in the data set S2 = 4:1;

[0010] (2) Take a multi-layer perception machine MLP as a basic model, train the data set S2, and obtain model_1;

[0011] (3) Put the historical blood routine and biochemical data set D1 of the person to be tested, add the standardized data set D1 to the data set S2 to form a data set S3, and reassign weights to the data set S3;

[0012] (4) Load the model structure and parameters of model_1, fix the first layer parameters of model_1, activate the second and third layer parameters of model_1 to continue training the data set S3, and obtain model_2;

[0013] (5) Test the blood routine and biochemical data of the person to be tested by using model_2 to obtain a personalized classification result.

[0014] Preferably, in step (2), the input dimension when training the data set S2 includes:

[0015] 'sex', 'age', 'RBC', 'MCV', 'PDW', 'WBC', 'NEUT%', 'LYMPH%', 'EO%', 'BASO%', 'NEUT#', 'LYMPH#', 'BASO#', 'HGB', 'HCT', 'MCH', 'MCHC', 'R-CV', 'PLT', 'MPV', 'PCT', 'MONO#', 'MONO%', 'EO#', 'GLU', 'CREA', 'UREA', 'ALB', 'ALT', 'AST' and the label of whether malnutrition exists.

[0016] Preferably, in step (2), the training epoch is 200.

[0017] Preferably, step (3) comprises: adding the historical blood routine and biochemical data D1 of the person to be tested to the data set S2 after standardization to form a data set S3, D1 is other data sources not belonging to S2, and the test set T1 of the S2 data set is unchanged; reassigning weights to the data set S3, D1:S2=1.05:1.

[0018] Preferably, in step (4), when the second and third layer parameters of model_1 are activated to continue training, the epoch is 200.

[0019] The malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data has the following beneficial effects:

[0020] (1) According to the data of the unique physiological changes and treatment process of each person, a customized nutrition status assessment is provided, which is more suitable for individual differences than a general model, and at the same time, the model has the ability to continuously incorporate new data for retraining, and as the accumulation of patient historical data, the prediction accuracy and adaptability of the model will continuously improve;

[0021] (2) In the incremental learning process of model_1, the first layer parameters of model_1 are fixed, and the second and third layer parameters of model_1 are activated for further training, and the test accuracy of the obtained model_2 on the test data of the test patient is obviously improved.

[0022] The technical solutions of the present application will be further described in detail below with the aid of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A schematic diagram of the overall process of the malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data is shown in the figure.

[0024] Figure 2 A flowchart for data selection. DETAILED DESCRIPTION

[0025] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be further described in detail below with the aid of the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present application and do not limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0026] Embodiment one

[0027] A malignant tumor patient personalized nutrition assessment dynamic analysis method based on blood data, the process is shown in the figure, and specifically includes the following steps: Figure 1

[0028] S1, data selection and processing, as shown in the figure. Figure 2

[0029] ​​The blood-based routine blood and biochemical dataset S1 was selected, with dimensions of sex, age, red blood cell count (RBC), mean corpuscular volume (MCV), platelet distribution width (PDW), white blood cell count (WBC), neutrophil percentage (NEUT%), lymphocyte percentage (LYMPH%), eosinophil percentage (EO%), basophil percentage (BASO%), neutrophil absolute value (NEUT#), lymphocyte absolute value (LYMPH#), basophil absolute value (BASO#), hemoglobin (HGB), hematocrit (HCT), mean corpuscular hemoglobin content (MCH), mean corpuscular hemoglobin concentration (MCHC), red blood cell volume distribution width (R-CV), platelet count (PLT), mean platelet volume (MPV), platelet hematocrit (PCT), monocyte absolute value (MONO#), monocyte percentage (MONO%), eosinophil absolute value (EO#), glucose (GLU), creatinine (CREA), urea (UREA), albumin (ALB), alanine aminotransferase (ALT), and aspartate aminotransferase (AST), a total of 3404 cases, and the dataset was from the People's Hospital of Wuhai City.

[0030] After the data set S1 was subjected to the data set S2, a total of 2969 cases of malignant tumor cases were obtained, including 669 cases of malnutrition cases (positive cases) and 2300 cases of normal nutrition cases (negative cases).

[0031] Inclusion criteria: ① Patients diagnosed with tumor; ② Age 18-90 years old; ③ Conscious, patients and their families willing to cooperate with this survey.

[0032] Exclusion criteria: ① Installation of cardiac pacemaker or arterial stent surgery; ② Patients with severe complications such as severe bleeding, severe infection, etc.; ③ Unable to cooperate for other reasons.

[0033] After the data set S2 was preprocessed and feature selected, the training set T0 was divided into 2375 cases, and the test set T1 was divided into 594 cases.

[0034] S2, train the data set S2 with the multi-layer perception (MLP) as the basic model. The input dimension is selected as'sex', 'age', 'RBC', 'MCV', 'PDW', 'WBC', 'NEUT%', 'LYMPH%', 'EO%', 'BASO%', 'NEUT#', 'LYMPH#', 'BASO#', 'HGB', 'HCT', 'MCH', 'MCHC', 'R-CV', 'PLT', 'MPV', 'PCT', 'MONO#', 'MONO%', 'EO#', 'GLU', 'CREA', 'UREA', 'ALB', 'ALT', 'AST', and the label of whether malnutrition exists. After 200 epochs, model_1 is obtained.

[0035] S3, the blood routine and biochemical data history data of the blood of the person to be tested are proposed, and the model_1 is used for incremental learning, the data of the person is added to the data set S2, and the data is re-weighed.

[0036] Specifically, it includes:

[0037] S31, the blood routine and biochemical data history data set D1 of the blood of the person to be tested is added to the data set S2 after standardization to form a data set S3, D1 is other data source not belonging to S2, and the test set T1 of the data set S2 is not changed.

[0038] S32, re-assign weights to the data set S3, D1:S2=1.05:1.

[0039] The amount of historical data of the person to be tested is not large, so by using different matching weights, the data of the person to be tested can be given a higher sampling ratio in the model training link, and the importance of the data of the person to be tested can be strengthened.

[0040] S4, reload the saved model structure and parameters of model_1, and retrain the data set S3. In order to be able to preserve the generalization ability of the model, the first layer network (input layer) with the largest parameter amount is fixed, and the second layer (hidden layer) and the third layer (output layer) network are released for training, and model_2 is obtained.

[0041] The first layer model has the largest parameter amount, so preserving the first layer network can make the fitting performance of the model for the data set S2 not weaken, and at the same time, the parameter amount of the second layer and the third layer network is small, which can be well fitted to the data of the person to be tested. In this way, the generalization performance of the model can be guaranteed in the process of fitting new data.

[0042] S5, using the trained model_2 to test the data of the to-be-tested person not in S2, and obtaining personalized accurate results. The evaluation indexes used in the test results are: TPR, TNR, and ACC.

[0043] The calculation formulas are respectively:

[0044] TPR = TP / (TP+FP);

[0045] TNR = TN / (TN+FP);

[0046] ACC = (TP+TN) / (TP+FP+FN+TN);

[0047] Where TP represents the number of correct identification as positive; TN represents the number of correct identification as negative; FP represents the number of false identification as positive; FN represents the number of false identification as negative; TPR represents the ratio of correct identification as positive to all that should be positive, i.e. sensitivity; TNR represents the ratio of correct identification as negative to all that should be negative, i.e. specificity; and ACC represents the accuracy of the model prediction classification result.

[0048] Test Example One

[0049] Using model_1 and model_2 to test the test set T1, the test results of the two models are shown in Table 1.

[0050] Table 1 Comparison of Algorithm Results

[0051]

[0052] Test Example Two

[0053] From the newly added data of model_1 and model_2 that have not been trained and meet the inclusion and exclusion criteria, a test subject with more test times is extracted, and his blood routine and biochemical data are used to test using model_1 and model_2. The test results are shown in Table 2.

[0054] Table 2

[0055]

[0056]

[0057] From the analysis in Table 2, it can be seen that the test results of model_1 on the data of the test person are very poor due to data distribution or instrument differences, with a correct rate of only 1 / 13, while the test results of the trained model_2 on the test person are improved to 8 / 13. According to Table 1, the accuracy of model_1 and model_2 is comparable for the test set T1 constructed for the S2 data set, indicating that the method does not overfit or deviate the model due to the addition of D1 data, but ensures the generalization performance of the model, and has better test results for data with poor test results due to data distribution or instrument differences.

[0058] Table 3

[0059]

[0060]

[0061] In Table 3, a number of test subjects with more experimental times are re-drawn, and their blood routine and biochemical data are tested using model_2, and the test results are shown in Table 3 above. It can be seen that due to patient non-compliance, sample problems or other factors, there are missing scores. In the traditional sense, when facing real score gaps, doctors often have difficulty making accurate judgments, not only affecting the diagnosis efficiency, but also possibly leading to ineffective or worsening of the patient's treatment plan. However, with the help of advanced machine learning algorithms, even in the absence of part of the real score, reliable dynamic grading evaluation probability values can be obtained to help doctors understand the possibility distribution of the patient's state and obtain more comprehensive information to support their decision-making process. This shows that the method can perform personalized dynamic nutritional assessment on patients with malignant tumors and solve the problem of differences between patients.

[0062] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application rather than limit them, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the present application, and these modifications or equivalent replacements also cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A dynamic analysis method for personalized nutritional assessment of a malignant tumor patient based on blood data, characterized in that, Comprising the following steps: (1) Obtain a data set S2 based on blood routine and biochemical data through data collection and screening, and the training set: test set = 4:1 in the data set S2; (2) Take the multilayer perception MLP as the basic model, train the data set S2, and obtain model_1; (3) Put forward the historical blood routine and biochemical data set D1 of the person to be tested, add the standardized data set D1 to the data set S2 to form a data set S3, and reassign weights to the data set S3; (4) Load the model structure and parameters of model_1, fix the first layer parameters of model_1, activate the second layer and third layer parameters of model_1, continue to train the data set S3, and obtain model_2; (5) Test the blood routine and biochemical data of the person to be tested by using model_2, and obtain the individualized classification result; In step (2), the input dimension when training the data set S2 includes: 'sex', 'age', 'RBC', 'MCV', 'PDW', 'WBC', 'NEUT%', 'LYMPH%', 'EO%', 'BASO%', 'NEUT#', 'LYMPH#', 'BASO#', 'HGB', 'HCT', 'MCH', 'MCHC', 'R-CV', 'PLT', 'MPV', 'PCT', 'MONO#', 'MONO%', 'EO#', 'GLU', 'CREA', 'UREA', 'ALB', 'ALT', 'AST' and the label of whether there is malnutrition; In step (3), the data set D1 is other data sources not belonging to S2, and the test set T1 of the data set S2 does not change; reassign weights to the data set S3, D1:S2 = 1.05:

1.

2. The dynamic analysis method for personalized nutritional assessment of a malignant tumor patient based on blood data according to claim 1, characterized in that, In step (2), the training epoch is 200. 3.The method of claim 1, wherein, In step (4), when the second layer and third layer parameters of model_1 are activated to continue training, the epoch is 200.

Images