Weight optimization integration method and device for biological reference interval establishment

Abstract

The invention provides a weight optimization integration method and device for biological reference interval establishment, and the method comprises the steps: generating simulated healthy subject data according to the data distribution characteristics of a target test item of a healthy subject; generating simulated patient data according to the data distribution of the target inspection items of the patient; inserting the simulated patient data into the simulated healthy subject data to form a comprehensive test result; respectively calculating the health probability of each data point in the comprehensive test result; and determining an optimal reference interval of the test item according to the health probability of each data point. According to the method, data of healthy subjects and patients are simulated by applying an integrated algorithm of a machine learning technology, physiological states of different crowds can be reflected more accurately, more accurate reference interval establishment is realized, and scientificity and effectiveness of clinical decisions are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of reference interval construction, and in particular to a weight optimization integration method and device for establishing biological reference intervals. Background Art

[0002] In medicine, biology and other scientific fields, it is often necessary to test a specific indicator to evaluate the physiological condition or other related factors of biological individuals or samples. In order to conduct effective testing, it is necessary to establish a reference interval, which includes the indicator range of normal individuals or samples, which can help doctors or researchers judge the abnormality of biological individuals or samples. The determination of the reference interval is an important step. In the prior art, the methods for establishing the reference interval are mainly divided into direct sampling technology and indirect sampling technology. Direct sampling technology establishes the reference interval by collecting data from healthy people. This method is time-consuming, costly and cumbersome. Indirect sampling technology establishes the reference interval by analyzing the big data in clinical tests and separating the data of healthy people from the mixed data. Although the cost is low, there are differences in the processing results of different algorithms for the data, resulting in bias in the reference interval and affecting clinical decision-making. In addition, there is currently no corresponding guide to explain which method should be used for different projects, which increases the uncertainty of establishing the reference interval. Summary of the invention

[0003] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a weight optimization integration method and device for establishing biological reference intervals based on machine learning.

[0004] To achieve the above objectives, the present invention provides the following solutions:

[0005] A weight optimization integration method for establishing biological reference intervals, comprising:

[0006] Step 1: Generate simulated healthy subject data based on the data distribution characteristics of the target test items of healthy subjects and calculate the actual value of the reference interval;

[0007] Step 2: Generate simulated patient data according to the data distribution of the patient's target test items;

[0008] Step 3: insert the simulated patient data into the simulated healthy subject data to form a comprehensive test result;

[0009] Step 4: Calculate the health probability of each data point in the comprehensive test results respectively;

[0010] Step 5: Determine the healthy subject population based on the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject population.

[0011] Preferably, the step 1: generating simulated healthy subject data according to the data distribution characteristics of the target test items of healthy subjects, comprises:

[0012] When the data distribution of the target test item of healthy subjects is a normal distribution, the simulated healthy subject data of normal distribution is generated by the rnorm function based on the mean and standard deviation of the target test item data;

[0013] When the data distribution of the target test item of healthy subjects is right-skewed, the simulated healthy subject data of right-skewed distribution is generated by the rlnorm function based on the mean and standard deviation of the target test item data;

[0014] When the data distribution of the target test items of healthy subjects is left-skewed, right-skewed distribution data is generated through the rlnorm function based on the mean and standard deviation of the target test item data, and the right-skewed distribution data is inverted to form simulated healthy subject data with left-skewed distribution.

[0015] Preferably, the step 2: generating simulated patient data according to the data distribution of the target test items of the patient, comprises:

[0016] Use the rnorm function to generate simulated patient data based on the mean and standard deviation of the patient's target test data.

[0017] Preferably, the step 3: inserting the simulated patient data into the simulated healthy subject data to form a comprehensive test result comprises:

[0018] The insertion ratio of the simulated patient data is randomly determined within the range of 0 to 20%, and the simulated patient data is inserted into the high-value area and the low-value area of ​​the simulated healthy subject data so that the high-value area and the low-value area contain the simulated patient data.

[0019] Preferably, the step 5: determining the optimal reference interval of the test item according to the health probability of each data point includes:

[0020] Use refineR and kosmic algorithms to calculate the health probability of each data point in the comprehensive test results;

[0021] When both refineR and kosmic algorithms judge a data point to be healthy, a weight w is assigned to the corresponding data point;

[0022] When the refineR algorithm determines that a data point is healthy and the kosmic algorithm determines that it is unhealthy, or when the refineR algorithm determines that a data point is unhealthy and the kosmic algorithm determines that it is healthy, a weight of 1-w is assigned to the corresponding data point;

[0023] Multiply the data point with the corresponding weight to obtain the weighted value, and calculate the 2.5% and 97.5% quantiles of the weighted value as the upper and lower limits of the reference interval;

[0024] The mean square error is used to calculate the difference between the upper and lower limits of the reference interval and the actual lower limit;

[0025] Traverse all weight values ​​and select the reference interval corresponding to the weight with the smallest mean square error as the best reference interval.

[0026] The present invention also provides a weight optimization integrated device for establishing a biological reference interval, comprising:

[0027] A healthy subject data simulation module is used to generate simulated healthy subject data according to the data distribution characteristics of the healthy subject target test items and calculate the actual value of the reference interval;

[0028] A patient data simulation module, used for generating simulated patient data according to the data distribution of the patient's target test items;

[0029] A data insertion module, used to insert simulated patient data into simulated healthy subject data to form a comprehensive test result;

[0030] A health probability calculation module is used to calculate the health probability of each data point in the comprehensive test results;

[0031] The optimal reference interval determination module is used to determine the healthy subject group according to the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject group.

[0032] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0033] The present invention provides a weight optimization integration method for establishing biological reference intervals. Compared with the prior art, on the one hand, the present invention proposes an integration technology for establishing reference intervals. On the other hand, the present invention is based on simulated data of healthy subjects and patients, applies machine learning technology, and trains the weight w of the reference interval integration algorithm for test items with different distribution characteristics, thereby achieving more accurate reference interval establishment and ensuring the scientificity and effectiveness of clinical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0035] Figure 1 A flow chart of a weight optimization integrated method for establishing biological reference intervals provided by the present invention;

[0036] Figure 2 This is a schematic diagram of the health probability calculation provided by the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0039] The terms "first", "second", "third" and "fourth" in the specification and claims of the present application and the drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a series of steps, processes, methods, etc. are not limited to the listed steps, but may optionally include steps that are not listed, or may optionally include other step elements inherent to these processes, methods, products or devices.

[0040] In order to make the above-mentioned objectives, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] See also Figure 1-2 , a weight optimization ensemble method for establishing biological reference intervals, comprising:

[0042] Step 1: Generate simulated healthy subject data according to the data distribution characteristics of the target test items of healthy subjects, and calculate the actual value of the reference interval;

[0043] In the step 1, when the data distribution of the target test item of healthy subjects is a normal distribution, the simulated healthy subject data of the normal distribution is generated by the rnorm function based on the mean and standard deviation of the target test item data;

[0044] When the data distribution of the target test item of healthy subjects is right-skewed, the simulated healthy subject data of right-skewed distribution is generated by the rlnorm function based on the mean and standard deviation of the target test item data;

[0045] When the data distribution of the target test items of healthy subjects is left-skewed, right-skewed distribution data is generated through the rlnorm function based on the mean and standard deviation of the target test item data, and the right-skewed distribution data is inverted to form simulated healthy subject data with left-skewed distribution.

[0046] Step 2: Generate simulated patient data according to the data distribution of the patient's target test items;

[0047] In step 2, the present invention can generate simulated patient data using the rnorm function according to the mean and standard deviation of the data of the target test items of the patient.

[0048] Step 3: insert the simulated patient data into the simulated healthy subject data to form a comprehensive test result;

[0049] Furthermore, in step 3, the present invention can randomly determine the insertion ratio of the simulated patient data within the range of 0 to 20%, and insert the simulated patient data into the high value area and the low value area of ​​the simulated healthy subject data, so that the high value area and the low value area contain the simulated patient data.

[0050] Step 4: Calculate the health probability of each data point in the comprehensive test results respectively;

[0051] Step 5: Determine the healthy subject population based on the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject population.

[0052] Further, the step 5 comprises:

[0053] Use refineR and kosmic algorithms to calculate the health probability of each data point in the comprehensive test results;

[0054] When both refineR and kosmic algorithms judge a data point to be healthy, a weight w is assigned to the corresponding data point;

[0055] When the refineR algorithm determines that a data point is healthy and the kosmic algorithm determines that it is unhealthy, or when the refineR algorithm determines that a data point is unhealthy and the kosmic algorithm determines that it is healthy, a weight of 1-w is assigned to the corresponding data point;

[0056] Multiply the data point with the corresponding weight to obtain the weighted value, and calculate the 2.5% and 97.5% quantiles of the weighted value as the upper and lower limits of the reference interval;

[0057] The mean square error is used to calculate the difference between the upper and lower limits of the reference interval and the actual lower limit;

[0058] Traverse all weight values ​​and select the reference interval corresponding to the weight with the smallest mean square error as the best reference interval.

[0059] The following is a further description of a weight optimization integration method for establishing a biological reference interval according to the present invention in conjunction with a specific embodiment:

[0060] Data generation: Healthy subject data generation: Generate simulated healthy subject data based on different distribution characteristics such as normal, right-skewed, and left-skewed, and set different means and standard deviations. Patient data insertion: Randomly insert simulated patient data into the high-value area and low-value area of ​​healthy subjects, and the insertion ratio is randomly determined in the range of 0 to 20%.

[0061] The detailed steps are as follows: Set parameters: Set the mean and standard deviation for each distribution type (normal distribution, right-skewed distribution, and left-skewed distribution) to generate corresponding data.

[0062] 1.1 Generate healthy subject data: Generate normal distribution data through rnorm function. Generate right-skewed distribution (lognormal distribution) data through rlnorm function. Generate left-skewed distribution data by reversing the right-skewed distribution data.

[0063] 1.2 Generate patient data: Set the mean and standard deviation of the patient data, and use the rnorm function to generate simulated patient data. Determine the insertion ratio: Randomly determine the insertion ratio in the range of 0 to 20%, and calculate the number of patient data that need to be inserted.

[0064] 1.3 Insert patient data: Insert the patient data into the high-value area and the low-value area of ​​the healthy subject data so that the high-value area and the low-value area contain the patient data.

[0065] Initial weight generation: Generate weights between 0.55 and 0.95 in steps of 0.05.

[0066] Fitness evaluation: The fitness probability of each data point is calculated using refineR and kosmic algorithms respectively.

[0067] Kosmic algorithm and refineR algorithm determine healthy subjects: As shown in the following formula, after obtaining the distribution of healthy subjects, kosmic algorithm and refineR algorithm calculate the ratio of the algorithm predicted value frequency (Fap) and the actual total frequency (Ftotal) in different value segments to determine the probability (P) that the subject in the data segment is a healthy subject. When P ≥ 0.5, it indicates that the data is the data of healthy subjects. When P < 0.5, it indicates that the data is the data of patients (see Figure 2 ).

[0068] For each data point, if both algorithms judge it as healthy, the data point is given a weight of w, and if only one algorithm judges it as healthy, the data point is given a weight of 1-w. Then the 2.5% and 97.5% quantiles of the weighted data are calculated as the upper and lower limits of the reference interval. Finally, the mean square error (MSE) under this weight configuration is calculated, that is, the difference between the estimated value and the actual value, to evaluate its fitness.

[0069] Weight optimization: Traverse all generated weights, calculate the upper and lower limits of the reference interval and the mean square error (MSE) under each weight configuration. Select the weight with the smallest mean square error as the optimal weight for the corresponding data type test item.

[0070] Modular design: Each indirect sampling technique (such as refineR and kosmic algorithms) is implemented as an independent module with standardized interfaces, including data processing and health probability calculation. New indirect sampling techniques can be seamlessly integrated into existing systems, enhancing the scalability of the algorithm.

[0071] In practical applications, after the user uploads the data, the system calculates the data distribution parameters, determines the data distribution type, selects the corresponding weight value w according to the data distribution type, and uses refineR and kosmic algorithms to calculate the health probability respectively. For data points that are judged to be healthy by both algorithms at the same time, a weight w is assigned; for data points that are judged to be healthy by only one algorithm, a weight 1-w is assigned. Weighted quantiles are calculated according to different weights as the biological reference intervals of the corresponding test items. In addition, the present invention can also combine multiple algorithms to complete the construction process of the reference interval. In addition to refineR and kosmic algorithms, a third algorithm (such as the expectation maximization algorithm) is introduced. Based on the two algorithms selected by the user, weight training and optimization are performed for data with different distributions, and then weights are selected according to the characteristics of the patient data. For data points that are judged to be healthy by both algorithms at the same time, a weight w is assigned; for data points that are judged to be healthy by only one algorithm, a weight 1-w is assigned. Weighted quantiles are calculated according to different weights as the biological reference intervals of the corresponding test items.

[0072] The present invention draws on the theory of machine learning, dynamically adjusts weights, combines refineR and kosmic algorithms, and uses weight optimization technology to achieve accurate calculation of reference intervals. Compared with traditional methods, the present invention improves the accuracy and reliability of reference intervals, reduces bias, and ensures the scientificity and effectiveness of clinical decision-making. Through modular design, the scalability of the algorithm is enhanced, which facilitates the integration of new indirect sampling technologies into the system and improves the adaptability and flexibility of the algorithm.

[0073] The present invention also provides a weight optimization integrated device for establishing a biological reference interval, comprising:

[0074] A healthy subject data simulation module is used to generate simulated healthy subject data according to the data distribution characteristics of the healthy subject target test items and calculate the actual value of the reference interval;

[0075] A patient data simulation module, used for generating simulated patient data according to the data distribution of the patient's target test items;

[0076] A data insertion module, used to insert simulated patient data into simulated healthy subject data to form a comprehensive test result;

[0077] A health probability calculation module is used to calculate the health probability of each data point in the comprehensive test results;

[0078] The optimal reference interval determination module is used to determine the healthy subject group according to the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject group.

[0079] Compared with the prior art, the beneficial effects of a weight optimization integration device for establishing biological reference intervals provided by the present invention are the same as the beneficial effects of a weight optimization integration method for establishing biological reference intervals described in the above technical solution, which will not be repeated here.

[0080] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the method disclosed in the embodiment, since it corresponds to the device disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the device part description.

[0081] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A weight optimization integration method for establishing biological reference intervals, characterized in that: include: Step 1: Generate simulated healthy subject data according to the data distribution characteristics of the target test items of healthy subjects, and calculate the actual value of the reference interval; Step 2: Generate simulated patient data according to the data distribution of the patient's target test items; Step 3: insert the simulated patient data into the simulated healthy subject data to form a comprehensive test result; Step 4: Calculate the health probability of each data point in the comprehensive test results respectively; Step 5: Determine the healthy subject population based on the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject population.

2. A weight optimization integration method for establishing biological reference intervals according to claim 1, characterized in that: The step 1: generating simulated healthy subject data according to the data distribution characteristics of the target test items of healthy subjects, comprises: When the data distribution of the target test item of healthy subjects is a normal distribution, the simulated healthy subject data of normal distribution is generated by the rnorm function based on the mean and standard deviation of the target test item data; When the data distribution of the target test item of healthy subjects is right-skewed, the simulated healthy subject data of right-skewed distribution is generated by the rlnorm function based on the mean and standard deviation of the target test item data; When the data distribution of the target test items of healthy subjects is left-skewed, right-skewed distribution data is generated through the rlnorm function based on the mean and standard deviation of the target test item data, and the right-skewed distribution data is inverted to form simulated healthy subject data with left-skewed distribution.

3. A weight optimization integration method for establishing biological reference intervals as claimed in claim 2, characterized in that: The step 2: generating simulated patient data according to the data distribution of the target test items of the patient, comprises: Use the rnorm function to generate simulated patient data based on the mean and standard deviation of the patient's target test data.

4. A weight optimization integration method for establishing biological reference intervals as claimed in claim 3, characterized in that: The step 3: inserting the simulated patient data into the simulated healthy subject data to form a comprehensive test result, including: The insertion ratio of the simulated patient data is randomly determined within the range of 0 to 20%, and the simulated patient data is inserted into the high-value area and the low-value area of ​​the simulated healthy subject data so that the high-value area and the low-value area contain the simulated patient data.

5. A weight optimization integration method for establishing biological reference intervals as claimed in claim 4, characterized in that: The step 5: determining a healthy subject group according to the health probability of each data point, and determining the optimal reference interval of the test item based on the healthy subject group, includes: Use refineR and kosmic algorithms to calculate the health probability of each data point in the comprehensive test results; When both refineR and kosmic algorithms judge a data point to be healthy, a weight w is assigned to the corresponding data point; When the refineR algorithm determines that a data point is healthy and the kosmic algorithm determines that it is unhealthy, or when the refineR algorithm determines that a data point is unhealthy and the kosmic algorithm determines that it is healthy, a weight of 1-w is assigned to the corresponding data point; Multiply the data point with the corresponding weight to obtain the weighted value, and calculate the 2.5% and 97.5% quantiles of the weighted value as the upper and lower limits of the reference interval; The mean square error is used to calculate the difference between the upper and lower limits of the reference interval and the actual lower limit; Traverse all weight values ​​and select the reference interval corresponding to the weight with the smallest mean square error as the best reference interval.

6. A weight optimization integrated device for establishing biological reference intervals, characterized in that: include: A healthy subject data simulation module is used to generate simulated healthy subject data according to the data distribution characteristics of the healthy subject target test items and calculate the actual value of the reference interval; A patient data simulation module, used for generating simulated patient data according to the data distribution of the patient's target test items; A data insertion module, used to insert simulated patient data into simulated healthy subject data to form a comprehensive test result; A health probability calculation module is used to calculate the health probability of each data point in the comprehensive test results; The optimal reference interval determination module is used to determine the healthy subject group according to the health probability of each data point, and determine the optimal reference interval of the test item based on the healthy subject group.

Images