Optimization method of chemiluminescence calibration curve based on artificial intelligence algorithm

Abstract

The present invention relates to a chemiluminescence calibration curve optimization method based on an artificial intelligence algorithm, characterized in that: by collecting and preprocessing inter-laboratory quality assessment measurement values, target values ​​and calibration curve related data determined by the chemiluminescence method, selecting and applying a suitable AI algorithm, such as a random forest, a support vector machine, and a neural network method, a least squares method is used to establish and optimize a regression model, and the calibration curve is adjusted to reduce bias and improve detection accuracy and correct indoor quality control target values, and the model performance is evaluated using mean square error and determination coefficient. This method not only improves the accuracy and consistency of detection, but also provides strong technical support for clinical laboratories and promotes the development of precision medicine. For manufacturers of chemiluminescence instruments, this technological advancement and effective improvement can improve brand effects and enhance market competitiveness.

Description

Technical Field

[0001] The technical field of this study involves medical testing technology, particularly the use of artificial intelligence (AI) algorithms to accurately optimize the calibration curve of chemiluminescence. Background Art

[0002] As a highly sensitive medical detection technology, chemiluminescence plays an important role in clinical diagnosis. However, chemiluminescence detection systems from different manufacturers may have significant biases when measuring the same marker, affecting the accuracy and consistency of the test results. In order to correct these biases and improve the accuracy of the measurement, clinical laboratories usually use external quality assessment (EQA) as a key external quality control activity, which not only helps to identify and correct systematic errors within the laboratory, but also plays a decisive role in improving the accuracy and reliability of the test results and internal quality control (IQC) to jointly monitor the quality of laboratory testing. A variety of quality control schemes are usually used in the laboratory, such as Westgard Sigma rules, to select the optimal quality control scheme, thereby improving quality control efficiency and continuous improvement. In addition, the optimization of the calibration curve of the chemiluminescence method is a key step to ensure the accuracy of the test results. The methods for optimizing the calibration curve include the use of multi-point calibration and two-point correction techniques to ensure the stability and accuracy of the calibration curve. In clinical laboratories, these optimization measures help to improve the consistency and accuracy of the test results, and for manufacturers, optimizing the calibration curve can increase the brand effect of the instrument and user praise. Through these methods, the application of chemiluminescence in clinical laboratories has been significantly improved, ensuring the accuracy and consistency of test results.

[0003] The main technical challenges faced in the field of the present invention include ensuring the quality and consistency of medical data, improving the interpretability and transparency of AI models, solving data bias problems to enhance model generalization capabilities, responding to regulatory requirements for clinical deployment of AI systems, enhancing clinical professionals' trust and acceptance of AI technology, establishing corresponding regulations and regulatory mechanisms, and ensuring close integration of technological development with clinical needs. The present invention aims to improve detection accuracy, enhance laboratory quality management, and promote the progress of precision medicine by optimizing the chemiluminescence calibration curve.

[0004] The application of AI technology in the field of clinical testing also faces challenges, including the abstract nature of algorithms, which makes it difficult to define the technical scope and protection boundaries in patent applications, the rapid development of existing technologies, which increases the requirements for algorithm novelty and creativity, and the uncertainty of patent review results. In addition, factors such as the inconsistency of standards and specifications among different medical institutions, data isolation and information islands, data security and privacy issues, and differences in technology and equipment also restrict the smooth implementation of mutual recognition of test results among medical institutions.

[0005] The innovation of this invention is to optimize the chemiluminescence calibration curve by using AI algorithms (such as random forest, support vector machine (SVM), neural network), and to adjust the calibration curve by establishing an accurate regression equation through the least squares method, so as to reduce bias, improve the accuracy of the test results and correct the IQC target value. The model performance is judged by the mean square error (MSE) and the coefficient of determination (R 2 ) to ensure the prediction accuracy and fitting effect of the model. This method not only improves the scientificity and practicality, but also significantly improves the accuracy and consistency of detection, provides strong technical support for clinical laboratories, and promotes the development of precision medicine. Summary of the invention

[0006] The present invention collects and preprocesses relevant data of chemiluminescence determination, selects and applies appropriate AI algorithms (such as random forest, support vector machine, neural network) to establish and optimize regression models. The calibration curve is adjusted to reduce the bias of test results and IQC and improve the detection accuracy. 2 This method not only improves the accuracy and consistency of detection, but also provides strong technical support for clinical laboratories, promotes the development of precision medicine, and provides chemiluminescence instrument manufacturers with opportunities for technological advancement and brand effect enhancement.

[0007] The beneficial effects of the present invention include significantly improving detection accuracy, optimizing the chemiluminescence calibration curve through an AI algorithm, and reducing the bias of test results and IQC. At the same time, it enhances data utilization efficiency, mines data patterns and trends, and provides more accurate information. In addition, it improves laboratory quality management, monitors and warns EQA and IQC data in real time, and enhances the reliability and consistency of test results. Optimizing the calibration curve effectively reduces system errors and corrects errors caused by equipment calibration, personnel operation, etc. The present invention also improves detection efficiency, dynamically adapts to experimental conditions and sample characteristics, reduces repeated testing, and improves the reliability of clinical decision-making, providing doctors with more accurate data support and promoting technological progress.

[0008] The present invention has a wide range of applications, and is particularly suitable for medical testing situations that require high precision and consistency, such as clinical diagnosis and laboratory quality control. Its application scenarios include stability monitoring of chemiluminescence detection systems, quality risk assessment, and optimization of calibration curves. Through real-time monitoring and intelligent early warning, the present invention can sensitively identify and dynamically monitor potential quality risks in the analysis process, and assist laboratory personnel in taking preventive and corrective measures in a timely manner.

[0009] In terms of social and economic benefits, the present invention significantly enhances the accuracy of diagnosis by improving the accuracy and consistency of detection, providing patients with more accurate diagnostic results and improving laboratory quality control management. In addition, by reducing misdiagnosis and missed diagnosis, reducing medical costs, promoting the application of AI technology in the field of medical testing, and promoting technological progress and innovation. For chemiluminescence instrument manufacturers, this technological progress and effective improvement can enhance brand effect and market competitiveness.

[0010] The beneficial effects of the present invention are as follows: 1. Significantly improve detection accuracy, the adjusted calibration curve has higher sensitivity and specificity, and the bias is significantly reduced compared with traditional methods; 2. Enhance data utilization efficiency, the AI ​​algorithm can process a large amount of complex data, mine data patterns and trends, and provide more accurate information; 3. Improve laboratory quality management, real-time monitoring and early warning of EQA and IQC data, and enhance the consistency and reliability of test results; 4. Effectively reduce system errors, and optimize the calibration curve to correct errors caused by equipment calibration, personnel operation, etc.; 5. Improve detection efficiency, dynamically adapt to experimental conditions and sample characteristics, and reduce repeated detection; 6. Improve the reliability of clinical decision-making, provide doctors with more accurate data support, provide guidance for research and practice, verify the effectiveness of AI algorithms in detection, and promote technological progress. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 The present invention takes the determination of carbohydrate antigen 50 (CA50) as an example to illustrate the positive effects of the calibration curves before and after correction. DETAILED DESCRIPTION

[0012] Example 1

[0013] 1. Data Collection Phase

[0014] 1. Collect data: Taking carbohydrate antigen 50 (CA50) as an example, collect EQA data, IQC data and calibration curve data of CA50 from the laboratory. The present invention takes the first tumor marker B project of the Clinical Laboratory Center of the National Health Commission in 2024 as an example. The batch numbers are 202411, 202412, 202413, 202414 and 202415, respectively. A total of 363 laboratories across the country participated.

[0015] In June 2024, the Laboratory Department of Baodi Hospital Affiliated to Tianjin Medical University measured the EQA results of the CA50 project of the National Health Commission's Clinical Laboratory Center, including laboratory measurement values, reported target values ​​and biases of samples of different batches, to evaluate the accuracy and consistency of laboratory tests, as shown in Table 1.

[0016] Table 1 Laboratory measurement values ​​and target values ​​of the EQA CA50 project of the National Health Commission Clinical Laboratory Center on June 5, 2024

[0017]

[0018] 2. CA50 calibration curve data collection The CA50 calibration data and curve of the Laboratory Department of Baodi Hospital Affiliated to Tianjin Medical University in June 2024 are collected. The specific data are shown in Table 2. The calibration curve is as follows Figure 1 .

[0019] Table 2 Calibration data of CA50 project of Baodi Hospital Affiliated to Tianjin Medical University

[0020]

[0021] 3. IQC results of CA50 project of Baodi Hospital Affiliated to Tianjin Medical University in June 2024

[0022] The CA50 IQC data of the Laboratory Department of Baodi Hospital Affiliated to Tianjin Medical University in June 2024 were collected, see Table 3.

[0023] Table 3 IQC results of CA50 project in June

[0024]

[0025] 3. Instruments and Reagents

[0026] The instrument used in this study is the Antu AutoLumo A2000 Plus chemiluminescence analyzer, produced by Zhengzhou Antu Bioengineering Co., Ltd. The CA50 kit and IQC are supporting products, and the reagent batch number is 20231008. The IQC batch number is: 2023720.

[0027] 2. Data Preprocessing Stage

[0028] Data cleaning: Check and remove outliers and erroneous data in the data set, such as measurement values ​​that deviate from the normal range (such as IQC exceeding ±3SD value, EQA exceeding the allowable range, and the value of the calibration curve drift point). Standardization: Standardize the data to eliminate the impact of different experimental conditions and sample characteristics on the data, so that it meets the input requirements of the AI ​​algorithm.

[0029] The Z-score standardization method is used, and its calculation formula is:

[0030]

[0031] Where X is the original data point; μ is the mean of the data set; σ is the standard deviation of the data set; and Z is the standardized data point.

[0032] Algorithm screening: According to data characteristics and research objectives, select suitable AI algorithms from random forest, support vector machine (SVM), neural network and other algorithms, and select neural network algorithm for nonlinear relationships.

[0033] 4. Model Construction

[0034] The preprocessed dataset is divided into a training set and a test set in a ratio of 70%:30%. The training set data is used to train the selected AI algorithm, and the model parameters are adjusted to optimize the model performance.

[0035] 1. Calibration equation and model establishment The CA50 calibration curve data is input into the DeepSeekAI tool, and converted into the original calibration curve through Python code, and the binary multivariate equation is derived using the AI ​​algorithm. The CA50 calibration curve data, EQA measurement value and target value are further input into the DeepSeekAI tool, and a regression model is established through Python code, which will provide the intercept and slope of the regression equation. These regression equation parameters are substituted into the original calibration curve equation to generate a new calibration equation and a new calibration curve.

[0036] Original calibration curve equation: C = 1.2 × 10 -12 RLU 2 +3.5×10 -6 .RLU+0.001 Where: C is the concentration value, RLU is the measured luminescence value.

[0037] The regression model established based on EQA measured values ​​and target values ​​yielded a slope of 0.95204 and an intercept of -0.2305.

[0038] C adjusted =0.95204×C-0.2305, that is, C=(C adjusted +0.2305) / 0.95 is substituted into the original calibration curve equation:

[0039] C = 1.2 × 10 -12 RLU 2 +3.5×10 -6 .RLU+0.001

[0040] C adjusted =0.95204(1.2×10 -12 RLU 2 +3.5×10 -6 .RLU+0.001)-0.2305

[0041] Result: C adjusted =1.14×10 -12 RLU 2 +3.325×10-6 .RLU-0.22955

[0042] 2. Curve generation Using the curve python code generated by DeepSeekAI tool, input the Bohrium tool to output the CA50 original calibration curve and the calibrated calibration curve after adjusting the EQA measured value and target value, such as Figure 1 Curve 1 is the original calibration curve: C = 1.2 × 10 -12 RLU 2 +3.5×10 -6 .RLU+0.001; Curve 2 is the calibration curve after calibration: C adjusted =1.14×10 -12 RLU 2 +3.325×10 -6 .RLU-0.22955.

[0043] 5. Result Verification Phase

[0044] Use Python code to implement model performance evaluation:

[0045] 1. Calculate MSE and R based on the regression model 2 MSE = 1.23 and R 2 =0.9876. The smaller the MSE value, the higher the prediction accuracy of the model; R 2 : Indicates the degree to which the model explains the variability of the data. The closer the value is to 1, the better the model fit is.

[0046] 2. The EQA quality control products were retested using the adjusted curve, and the bias was calculated and the changes before and after correction were compared. The batch numbers of the EQA quality control products are: 202411, 202412, 202413, 202414 and 202415.

[0047] Comparison of the calibration curve equation measured values ​​and bias before and after adjustment. A bias percentage within ±5% is generally considered to be low bias, indicating that the predicted value is very close to the true value and the prediction accuracy of the model is high. After correction, the CA50 data showed low bias. See Table 4.

[0048] Table 4 Laboratory measurement values ​​and target values ​​of the EQA CA50 project of the National Health Commission Clinical Laboratory Center on June 5, 2024

[0049]

[0050] 3. Changes in IQC calculated target values ​​before and after adjustment of the calibration curve In June 2024, the average target value of CA50 IQC calculation is: 29.98U / ml. The theoretically calculated value based on the adjusted calibration curve is: 28.49U / ml, which reduces the bias percentage, corrects the systematic deviation, improves accuracy and consistency, and can continue to verify the positive effects of the calibrated CA50 calibration curve through IQC and EQA practices.

[0051] VI. Clinical Application Stage

[0052] exist Figure 1 The optimized calibration curve was applied to actual CA50 testing. After EQA and IQC testing practices, the accuracy and consistency were significantly improved, providing support for clinical laboratory quality management and promoting the development of precision medicine.

[0053] VII. Scope of protection of the present invention

[0054] It covers the methods of optimizing chemiluminescence calibration curves using AI algorithms (such as random forests, support vector machines, and neural networks), including key steps such as data preprocessing, model building, performance evaluation, calibration curve adjustment, EQA and IQC data analysis, result verification, and clinical application, as well as calibration curve adjustment by chemiluminescence instrument manufacturers. The final scope of protection will be determined based on the patent application documents and the claims authorized by the Patent Office.

Claims

1. A chemiluminescence calibration curve optimization method based on artificial intelligence algorithm, characterized in that: a. Optimize the calibration curve of chemiluminescence method using artificial intelligence algorithm; b. Collect and preprocess relevant data of chemiluminescence determination; c. Select and apply appropriate AI algorithms to establish and optimize regression models; d. Reduce bias and improve detection accuracy by adjusting the calibration curve; e. Use mean square error and determination coefficient to evaluate model performance.

2. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: The data preprocessing includes data cleaning and standardization, checking and removing outliers and erroneous data in the data set; standardizing the data to eliminate the impact of different experimental conditions and sample characteristics on the data, so that it meets the input requirements of the AI ​​algorithm.

3. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: The selection of the AI ​​algorithm is based on data characteristics and research objectives, and is screened from algorithms such as random forest, support vector machine (SVM), neural network, etc.

4. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: The calibration curve optimization step includes establishing a regression model according to the EQA measurement value, the target value and the original calibration curve, and optimizing the parameters of the calibration curve by using a neural network algorithm and a square method.

5. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: The model performance was evaluated by mean square error and determination coefficient to ensure the prediction accuracy and fitting effect of the model.

6. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: Positive effects are monitored and evaluated by determining IQC and EQA through adjusted calibration curves.

7. The method for optimizing the chemiluminescence calibration curve based on an artificial intelligence algorithm according to claim 1, characterized in that: The technology is applicable to different types of clinical test data, including but not limited to test results in the fields of hematology, biochemistry, immunology, etc.

Images