Intelligent prediction method for risk of lack of nutrient elements in human body based on conventional inspection data

Abstract

The invention belongs to the technical field of medical data information processing, and discloses a human nutrient element deficiency risk intelligent prediction method based on conventional test data, comprising the following steps: acquiring basic personal information and conventional test data of a subject; performing data cleaning on the acquired data information; the individual nutrition status is comprehensively evaluated, and risk labels of nutrient element deficiency are marked in a standardized mode; training a human nutrient element deficiency risk prediction model; determining an optimal threshold value of the human nutrient element deficiency risk prediction model; and evaluating the feasibility of human nutrient element deficiency risk prediction based on conventional test data. By adopting the human nutrient element deficiency risk intelligent prediction method based on the conventional inspection data, the capability of identifying the human nutrient element deficiency risk is comprehensively improved, the detection cost is reduced, the detection time is shortened, dynamic monitoring and immediate feedback of the nutrition status are realized, and the human nutrient element deficiency risk intelligent prediction method is suitable for popularization and application. The method has the characteristics of comprehensiveness, high efficiency, accuracy and real-time dynamic monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data information processing, and in particular to an intelligent prediction method for the risk of human nutritional element deficiency based on conventional test data. Background Art

[0002] Human health depends on the balanced intake of multiple nutrients, including vitamins, minerals, proteins, fats and carbohydrates, which each have different physiological functions in the human body and work together to maintain the normal functioning of life. The diversity of nutritional needs requires individuals to consume a wide range of foods in their daily diet to ensure comprehensive and adequate nutritional support. However, with the rapid changes in modern lifestyles, people are facing unprecedented nutritional challenges. The fast-paced pace of life has led to irregular diets and the prevalence of fast food culture. In addition, environmental pollution, nutritional losses during food processing, and personal bad eating habits have made the lack of human nutrients a common phenomenon. These challenges not only affect the normal absorption and utilization of nutrients, but also increase the risk of malnutrition and health problems.

[0003] The lack of nutrients in the human body can directly lead to a series of health problems, including growth retardation, decreased immune function, anemia, bone abnormalities, and neurological symptoms. Long-term deficiency may also increase the risk of chronic diseases. Therefore, early detection and dynamic monitoring of nutrient deficiencies play an important role in ensuring that individuals receive comprehensive and balanced nutritional support. This not only helps to detect and correct nutritional problems in a timely manner, but also ensures that everyone receives appropriate nutritional support according to their health needs, thereby promoting the improvement of overall health levels.

[0004] However, today's human nutritional element determination technology often focuses on the detection of a single nutrient, ignoring the integrity of the nutritional system and the interaction between nutrients. This approach cannot fully assess the nutritional status of an individual and may lead to deviations in the assessment results. For example, if the role of vitamin C in promoting iron absorption is ignored, it will directly affect the accurate assessment of anemia risk. Based on this, in the face of the complex situation of multiple nutritional element deficiencies, multi-label recognition technology integrates multiple detection principles and signal processing technologies to achieve simultaneous detection of multiple nutrients, greatly improving detection efficiency. However, the high cost, cumbersome operating procedures and dependence on professionals have made biochemical detection technology encounter obstacles in its popularization. Simplifying the detection process, reducing costs, and improving the convenience and popularity of detection technology are also issues that need to be urgently addressed.

[0005] In addition, the nutritional status of the human body is affected by many factors and presents dynamic changes. Therefore, building an effective dynamic monitoring mechanism to track and evaluate the nutritional status of individuals in real time is of great significance for nutritional intervention and health management. Therefore, the present invention proposes an intelligent prediction method for the risk of human nutritional element deficiency based on conventional test data. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent prediction method for the risk of human nutrient deficiency based on routine test data, aiming to predict the risk results of human nutrient deficiency and provide strong support for health management.

[0007] To achieve the above object, the present invention provides an intelligent prediction method for the risk of human nutrient deficiency based on routine test data, which is characterized by comprising the following steps:

[0008] S1. Obtain the basic personal information and routine test data of the subjects;

[0009] S2, performing data cleaning on the data information obtained in step S1;

[0010] S3. Comprehensively assess individual nutritional status and standardize risk labels for nutritional deficiencies;

[0011] S4. Training of risk prediction model for human nutritional deficiency;

[0012] S5. Determine the optimal threshold of the risk prediction model for human nutrient deficiency;

[0013] S6. Assess the feasibility of predicting the risk of human nutritional deficiency based on routine test data.

[0014] Preferably, in step S1, the basic information of the subject includes age, gender, and medical history; the routine test data of the subject includes blood routine, urine routine, biochemical test, and fecal routine.

[0015] Preferably, in step S2, the data information obtained in step S1 is cleaned, and the specific process is as follows:

[0016] S21, clear the abnormal number;

[0017] Based on statistical methods, identify samples that are too large or too small from the normal data range; for outliers, choose to directly eliminate them or use appropriate statistical methods to correct them according to the data distribution;

[0018] S22, fill in missing data;

[0019] Analyze the type and distribution of missing data and determine the type of missing data; select different filling methods to fill the missing information in the data set according to different data characteristics and business backgrounds to ensure the integrity of data analysis;

[0020] S23. Use the Z-score standardization formula to adjust the distribution of the data as follows:

[0021]

[0022] Where X is the original data, μ is the mean, and σ is the standard deviation. Each data point is converted to a new standardized value.

[0023] Preferably, in step S3, the nutritional status of the individual is comprehensively assessed, and the risk labels of nutritional deficiency are standardized and marked. The specific process is as follows:

[0024] S31. Collect biochemical analysis results of different biological samples from subjects and analyze the concentration levels of specific nutrients therein; biological samples include blood, urine, hair, and saliva; specific nutrients include vitamins, minerals, fatty acids, and amino acids.

[0025] S32. Integrate nutrient concentration information from different biological samples to comprehensively assess individual nutritional status;

[0026] S33. Systematize the routine test data of the subjects, i.e. label them with standardized nutrient deficiency risk labels.

[0027] Preferably, in step S4, a human nutrient deficiency risk prediction model is trained, and the specific process is as follows:

[0028] S41, performing multi-label classification training on the conventional inspection data according to the labels annotated in step S33, and constructing a human nutrient deficiency risk prediction model based on a fully connected neural network algorithm;

[0029] S42. Introduce the sigmoid activation function to predict the independent probability of each label in the multi-label classification task; the output range of the sigmoid function is between 0 and 1; use the binary cross entropy loss function to perform independent binary classification for each label.

[0030] Preferably, in step S5, the optimal threshold of the human nutrient element deficiency risk prediction model is determined by maximizing the Youden index method, as shown below:

[0031] Youden index = sensitivity + specificity - 1 (2);

[0032]

[0033] Preferably, in step S6, the feasibility assessment includes an overall assessment of the human nutrient element deficiency risk prediction model and an assessment of the prediction performance of each element; wherein the overall assessment of the human nutrient element deficiency risk prediction model adopts the top five accuracy and accuracy evaluation indicators; the prediction performance evaluation indicators of each element are accuracy, sensitivity and specificity;

[0034]

[0035] Therefore, the present invention adopts the above-mentioned intelligent prediction method for human nutrient deficiency risk based on routine test data, and adopts multi-label classification method and threshold optimization technology to comprehensively improve the ability to identify the risk of multiple nutrient deficiency in the human body. Relying on routine test data, without the need for complex biochemical equipment, it provides a comprehensive assessment of nutritional status; through threshold optimization, the recognition accuracy is improved, the misjudgment is reduced, the detection cost and time are reduced, the dynamic monitoring and instant feedback of nutritional status are realized, and the precise formulation of personalized nutritional intervention plans is assisted, which has the characteristics of comprehensiveness, efficiency, accuracy and real-time dynamic monitoring.

[0036] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart for realizing the intelligent prediction method of human nutrient deficiency risk based on conventional test data of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0039] like Figure 1 As shown, the intelligent prediction method for human nutrient element deficiency risk based on conventional test data of the present invention comprises the following steps:

[0040] S1. Obtain basic personal information and routine test data of the subjects.

[0041] Basic information of the subjects, including age, gender, medical history, etc.

[0042] Routine test data of the subjects, including blood routine, urine routine, biochemical test, stool routine, etc.

[0043] S2. Clean the data information obtained in step S1.

[0044] S21, clear the abnormal number;

[0045] Based on statistical methods, such as box plots and Z-score tests, samples that are too large or too small from the normal data range are identified; outliers caused by measurement errors, equipment failures, human interference, etc. are directly eliminated or corrected using appropriate statistical methods based on data distribution. Outliers in the data set are removed to ensure the accuracy and reliability of the analysis results.

[0046] S22, fill in missing data;

[0047] Analyze the type and distribution of missing data to determine whether it is randomly missing, completely randomly missing, or systematically missing; consider different filling methods for different data characteristics and business backgrounds, such as mean filling, median filling, mode filling, model-based filling, etc. Fill in the missing information in the data set to ensure the integrity of data analysis.

[0048] S23. Use the Z-score standardization formula to adjust the distribution of data so that each variable is at the same numerical level, which is convenient for subsequent data analysis and model training, as shown below:

[0049]

[0050] Where X is the original data, μ is the mean, and σ is the standard deviation. Each data point is converted to a new standardized value.

[0051] S3. Comprehensively assess individual nutritional status and standardize risk labels for nutritional deficiencies.

[0052] S31. Collect biochemical analysis results of different biological samples from subjects, such as blood, urine, hair, saliva, etc., and analyze the concentration levels of specific nutrients such as vitamins, minerals, fatty acids, amino acids, etc.

[0053] S32. Integrate nutrient concentration information from different biological samples to comprehensively assess individual nutritional status.

[0054] S33. Systematize the routine test data of the subjects, i.e. label them with standardized nutrient deficiency risk labels.

[0055] S4. Train the risk prediction model for human nutritional deficiency.

[0056] S41. Perform multi-label classification training on the routine inspection data according to the labels marked in step S33, and build a human nutrient deficiency risk prediction model based on a fully connected neural network algorithm.

[0057] S42, introduce the sigmoid activation function to predict the independent probability of each label in the multi-label classification task; the output range of the sigmoid function is between 0 and 1. Then, use the binary cross-entropy loss function to effectively handle the independent binary classification problem of each label.

[0058] S5. Determine the optimal threshold of the human nutrient deficiency risk prediction model.

[0059] The optimal threshold of the human nutrient deficiency risk prediction model is determined by maximizing the Youden index method to ensure that the selected threshold achieves the best prediction performance in this category. The Youden index calculation formula is as follows:

[0060] Youden Index = Sensitivity + Specificity - 1 (2)

[0061]

[0062]

[0063] S6. Assess the feasibility of predicting the risk of human nutritional deficiency based on routine test data.

[0064] The feasibility assessment includes two main aspects: the overall assessment of the human nutrient element deficiency risk prediction model and the assessment of the prediction performance of each element. Among them, the overall assessment of the human nutrient element deficiency risk prediction model uses evaluation indicators such as the top five accuracy and accuracy; the prediction performance evaluation indicators of each element are accuracy, sensitivity and specificity.

[0065]

[0066] Example

[0067] This example predicts the risk of vitamin and mineral deficiency based on routine blood test data. The specific process is as follows:

[0068] S1. Obtain the basic personal information of the subjects, including age, gender, medical history, and routine blood data.

[0069] S2. Clean the data information obtained in step S1, including removing abnormal numbers, filling missing data, and adjusting the distribution of data using the Z-score standardization formula. The specific process is as follows:

[0070] S21. Use the statistical method of Z-score test to identify samples that are too large or too small from the normal data range; for outliers confirmed to be caused by measurement errors, equipment failures, human interference, etc., choose to directly eliminate them or use appropriate statistical methods to correct them according to the data distribution, so as to eliminate outliers in the data set and ensure the accuracy and reliability of the analysis results.

[0071] S22. Analyze the type and distribution of missing data to determine whether it is randomly missing, completely randomly missing, or systematically missing. For different data characteristics and business backgrounds, different filling methods need to be considered (such as mean filling, median filling, mode filling, model-based filling, etc.) to reduce missing information in the data set and ensure the integrity of data analysis.

[0072] S23. Use the Z-score standardization formula to adjust the data distribution so that each variable is at the same numerical level, which is convenient for subsequent data analysis and model training. The Z-score standardization calculation formula is as follows:

[0073]

[0074] Where X is the original data, μ is the mean, and σ is the standard deviation. Each data point is converted to a new standardized value.

[0075] S3. Comprehensively assess individual nutritional status and standardize risk labels for nutritional deficiencies.

[0076] S31. Collect biochemical analysis results of different biological samples from subjects, such as blood, urine, hair, saliva, etc., and analyze their vitamin and mineral concentration levels.

[0077] S32. Integrate nutrient concentration information from different biological samples to comprehensively assess individual nutritional status.

[0078] S33. Systematically process the blood test data of the subjects, i.e. mark them with standardized labels of nutritional deficiency risks.

[0079] S4. Train vitamin and mineral deficiency risk prediction model.

[0080] S41. Perform multi-label classification training on routine blood test data according to the labels.

[0081] S42. Use a fully connected neural network to construct a vitamin and mineral deficiency risk prediction model, in which routine blood test data is used as input and the vitamin and mineral deficiency risk is used as output to train the vitamin and mineral deficiency risk prediction model.

[0082] The model consists of a 24-dimensional input layer, 3 fully connected layers, a Dropout layer, and a 17-dimensional output layer. The number of neurons in the fully connected layers are 416, 352, and 160. The Adam optimizer and BinaryCrossentropy loss function are used to compile the neural network.

[0083] S5. Determine the optimal threshold of the human nutrient deficiency risk prediction model.

[0084] The optimal threshold of the human nutrient deficiency risk prediction model was determined by maximizing the Youden index method to ensure that the selected threshold achieved the best prediction performance in this category. The specific results are shown in Table 2.

[0085] S6. Assess the feasibility of predicting the risk of human nutritional deficiency based on routine test data.

[0086] As shown in Table 1 , the top five accuracy rates and accuracy evaluation indicators were used to evaluate the vitamin and mineral deficiency risk prediction model as a whole; as shown in Table 2 , the prediction performance of each element was evaluated by accuracy, sensitivity, and specificity.

[0087] Table 1 Model performance list

[0088] Model Name Top five accuracy Accuracy Risk prediction model for human nutrient deficiency 83.62％ 56.26％

[0089] Table 2 Test set index list

[0090]

[0091]

[0092] The experimental results show that the top five accuracy rates reached 83.62%, which means that for most test samples, the present invention can accurately include the real risks of nutritional deficiency in the top five categories with the highest prediction probabilities. At the same time, the accuracy rate also reached 56.26%. Although it is lower than the top five accuracy rates, it still shows that the present invention has a certain ability to directly predict the correct category. The experimental results prove the feasibility of the intelligent prediction method for human nutritional deficiency risks based on routine test data of the present invention. By deeply mining and analyzing the rich information in routine test data, it can effectively identify potential nutritional deficiency risks, providing strong support for clinical nutrition assessment and the formulation of personalized intervention plans.

[0093] Therefore, the present invention adopts the above-mentioned intelligent prediction method for human nutrient deficiency risk based on routine test data, and adopts multi-label classification method and threshold optimization technology to comprehensively improve the ability to identify the risk of multiple nutrient deficiency in the human body. Relying on routine test data, without the need for complex biochemical equipment, it provides a comprehensive assessment of nutritional status; through threshold optimization, the recognition accuracy is improved, the misjudgment is reduced, the detection cost and time are reduced, the dynamic monitoring and instant feedback of nutritional status are realized, and the precise formulation of personalized nutritional intervention plans is assisted, which has the characteristics of comprehensiveness, efficiency, accuracy and real-time dynamic monitoring.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. An intelligent prediction method for the risk of human nutritional deficiency based on routine test data, characterized in that: The following steps are involved: S1. Obtain the basic personal information and routine test data of the subjects; S2, performing data cleaning on the data information obtained in step S1; S3. Comprehensively assess individual nutritional status and standardize risk labels for nutritional deficiencies; S4. Training a model for predicting the risk of human nutritional deficiency; S5. Determine the optimal threshold of the risk prediction model for human nutrient deficiency; S6. Assess the feasibility of predicting the risk of human nutritional deficiency based on routine test data.

2. The method for intelligently predicting the risk of human nutrient deficiency based on routine test data according to claim 1 is characterized in that: In step S1, the basic information of the subject includes age, gender, and medical history; the routine test data of the subject includes blood routine, urine routine, biochemical test, and fecal routine.

3. The intelligent prediction method for human nutrient deficiency risk based on routine test data according to claim 1 is characterized in that: In step S2, the data information obtained in step S1 is cleaned. The specific process is as follows: S21, clear the abnormal number; Based on statistical methods, identify samples that are too large or too small from the normal data range; for outliers, choose to directly eliminate them or use appropriate statistical methods to correct them according to the data distribution; S22, fill in missing data; Analyze the type and distribution of missing data and determine the type of missing data; select different filling methods to fill the missing information in the data set according to different data characteristics and business backgrounds to ensure the integrity of data analysis; S23. Use the Z-score standardization formula to adjust the distribution of the data as follows: Among them, X is the original data, μ is the mean, σ is the standard deviation, and each data point is converted to a new standardized value.

4. The intelligent prediction method for human nutrient deficiency risk based on routine test data according to claim 1 is characterized in that: In step S3, the nutritional status of individuals is comprehensively assessed and risk labels of nutritional deficiency are standardized. The specific process is as follows: S31. Collect biochemical analysis results of different biological samples from subjects and analyze the concentration levels of specific nutrients therein; biological samples include blood, urine, hair, and saliva; specific nutrients include vitamins, minerals, fatty acids, and amino acids. S32. Integrate nutrient concentration information from different biological samples to comprehensively assess individual nutritional status; S33. Systematize the routine test data of the subjects, i.e. label them with standardized nutrient deficiency risk labels.

5. The intelligent prediction method for human nutrient deficiency risk based on routine test data according to claim 4 is characterized in that: In step S4, a human nutrient deficiency risk prediction model is trained, and the specific process is as follows: S41, performing multi-label classification training on the conventional inspection data according to the labels marked in step S33, and constructing a human nutrient deficiency risk prediction model based on a fully connected neural network algorithm; S42. Introduce the sigmoid activation function to predict the independent probability of each label in the multi-label classification task; the output range of the sigmoid function is between 0 and 1; use the binary cross entropy loss function to perform independent binary classification for each label.

6. The intelligent prediction method for human nutrient deficiency risk based on routine test data according to claim 4 is characterized in that: In step S5, the optimal threshold of the human nutrient deficiency risk prediction model is determined by maximizing the Youden index method, as shown below: Youden index = sensitivity + specificity - 1 (2); 7. The intelligent prediction method for human nutrient deficiency risk based on routine test data according to claim 1, characterized in that: In step S6, the feasibility assessment includes an overall assessment of the human nutrient element deficiency risk prediction model and an assessment of the prediction performance of each element; wherein the overall assessment of the human nutrient element deficiency risk prediction model uses the top five accuracy and accuracy evaluation indicators; the prediction performance evaluation indicators of each element are accuracy, sensitivity and specificity;

Images