High-sensitivity ATP fluorescence detection method and system

Abstract

The invention provides a high-sensitivity ATP fluorescence detection method and system. Wherein the optimal ratio of luciferase to fluorescein is determined, a reaction mixture containing luciferase and fluorescein in the optimal ratio is prepared, and it is ensured that a remarkably enhanced fluorescence signal is generated under the low-concentration ATP condition; based on the reaction mixture, controlling a reaction environment by applying a micro-fluidic chip technology, so that the reaction mixture and the to-be-detected sample are fully mixed in a dark environment, and meanwhile, adjusting the reaction temperature by applying a dynamic temperature control technology to obtain a mixed reaction system; capturing and separating a fluorescence signal generated by the mixed reaction system within a set time by using a high-sensitivity photomultiplier, generating a pure target fluorescence signal, and measuring the intensity of the pure target fluorescence signal at the same time; the intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be detected is calculated and generated. According to the technical scheme provided by the invention, the ATP detection sensitivity and accuracy are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of biosensing and detection technology, and in particular to a high-sensitivity ATP fluorescence detection method and system. Background Art

[0002] In the fields of biomedicine, food safety, environmental monitoring, etc., there is a growing demand for rapid and accurate detection of the concentration of adenosine triphosphate (ATP) in samples. ATP is a key molecule for intracellular energy transfer, and its concentration can reflect important information such as microbial activity or contamination level.

[0003] There are many ATP fluorescence detection methods on the market, among which the detection method based on luciferase-luciferin system is more common. This system indirectly determines the ATP content in the sample by luciferase catalyzing the reaction of luciferin and ATP to produce a fluorescent signal. In order to improve the detection efficiency and accuracy, some studies have also combined microfluidics and advanced photoelectric detection devices to optimize the entire process.

[0004] However, when dealing with low-concentration ATP, traditional methods have a high detection limit because the fluorescence signal is weak and difficult to be effectively captured. It is difficult to accurately control factors such as reaction temperature under ordinary laboratory conditions, which affects the consistency and repeatability of the fluorescence signal. Existing methods usually require manual correction of background noise and lack effective algorithm support, making the conversion process from raw fluorescence signal to final result complicated and error-prone.

[0005] In summary, although there are currently a variety of ATP fluorescence detection schemes, they still show limitations when facing specific challenges. This proposal aims to solve the above problems by optimizing the ratio of luciferase to luciferin, using microfluidic chip technology to improve reaction conditions, and combining advanced signal processing technology with machine learning models, in order to achieve a more efficient and accurate measurement of ATP concentration. Summary of the invention

[0006] The embodiments of the present application provide a highly sensitive ATP fluorescence detection method and system to solve the problems of low sensitivity and accuracy of ATP detection in the prior art.

[0007] In a first aspect, the present invention provides a highly sensitive ATP fluorescence detection method, comprising:

[0008] Determine the optimal ratio of luciferase to luciferin, and prepare a reaction mixture containing the optimal ratio of luciferase to luciferin to ensure a significantly enhanced fluorescence signal under low ATP concentration conditions;

[0009] Based on the reaction mixture, the reaction environment is controlled by using microfluidic chip technology, so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal, thereby obtaining a mixed reaction system;

[0010] Using a high-sensitivity photomultiplier tube, the fluorescence signal generated by the mixed reaction system is captured within a set time, and the fluorescence signal is processed using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to separate and generate a pure target fluorescence signal, and the intensity of the pure target fluorescence signal is measured at the same time;

[0011] Based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model trained with a large number of ATP standard solution data sets of different concentrations is used, and the model parameters are optimized using a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

[0012] Optionally, the method uses a high-sensitivity photomultiplier tube to capture the fluorescence signal generated by the mixed reaction system within a set time, and uses a digital phase-locked amplification technology and an adaptive noise elimination algorithm to process the fluorescence signal, separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal, including:

[0013] Using a high-sensitivity photomultiplier tube, based on a dynamic integration time window of the fluorescence signal intensity change, combined with an adaptive threshold algorithm, the fluorescence signal generated by the mixed reaction system is efficiently captured to obtain a preliminary captured signal;

[0014] According to the preliminary captured signal, a digital phase-locked amplification technology based on frequency tracking is used to analyze the signal spectrum through a fast Fourier transform algorithm, and a frequency closest to the target fluorescence signal is selected as a synchronization reference signal, and the preliminary captured signal is phase-locked to obtain an amplified target fluorescence signal;

[0015] Based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to perform noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal;

[0016] The pure target fluorescence signal is digitized by using a built-in high-precision analog-to-digital converter, and a self-correction algorithm is used to correct the nonlinear error in the analog-to-digital conversion process, so as to accurately measure the intensity of the pure target fluorescence signal and generate final fluorescence signal intensity data.

[0017] Optionally, the use of a high-sensitivity photomultiplier tube, based on a dynamic integration time window of the change in fluorescence signal intensity, combined with an adaptive threshold algorithm, to efficiently capture the fluorescence signal generated by the mixed reaction system to obtain a preliminary captured signal includes:

[0018] Using a high-sensitivity photomultiplier tube, based on the average fluorescence signal intensity determined in a preliminary experiment, an initial integration time window is set to start preliminarily capturing the fluorescence signal generated by the mixed reaction system to obtain a preliminarily captured fluorescence signal intensity;

[0019] According to the intensity of the fluorescent signal initially captured, an adaptive threshold algorithm is used to evaluate the signal intensity in real time. When the signal intensity is detected to exceed the preset threshold range, the length of the integration time window is dynamically adjusted to obtain an optimized integration time window.

[0020] Based on the optimized integration time window, the fluorescence signal generated by the mixed reaction system is continuously captured using a high-sensitivity photomultiplier tube, and the integration time window is updated immediately after each capture to ensure that the best signal capture performance is maintained throughout the detection process to obtain a continuously captured fluorescence signal;

[0021] All of the continuously captured fluorescence signals are aggregated to form a preliminary captured signal, providing high-quality raw data for subsequent processing using a digital phase-locked amplification technology based on frequency tracking.

[0022] Optionally, the method of using a digital phase-locked amplification technology based on frequency tracking according to the preliminary captured signal, analyzing the signal spectrum through a fast Fourier transform algorithm, selecting a frequency closest to the target fluorescence signal as a synchronization reference signal, performing phase locking processing on the preliminary captured signal, and obtaining an amplified target fluorescence signal includes:

[0023] Using a fast Fourier transform algorithm, performing spectrum analysis on the preliminary captured signal, identifying multiple frequency components in the signal, and obtaining a preliminary spectrum analysis result;

[0024] According to a preset frequency range and signal intensity threshold, a frequency closest to the target fluorescence signal is selected from the preliminary spectrum analysis result as a synchronization reference signal to generate a selected synchronization reference signal;

[0025] Based on the selected synchronization reference signal, a digital phase-locked amplification technology based on frequency tracking is adopted to adjust the phase-locked loop parameters so that the phase-locked loop output signal maintains phase synchronization with the selected synchronization reference signal, and the preliminary captured signal is phase-locked to generate a phase-locked signal;

[0026] The phase-locked signal is amplified using the adjusted amplifier gain to ensure that the intensity of the target fluorescence signal is significantly improved while reducing the impact of background noise, thereby generating an amplified target fluorescence signal.

[0027] Optionally, based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to perform noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal, including:

[0028] Based on the amplified target fluorescence signal, a convolutional neural network is used to construct an adaptive noise elimination model, wherein the adaptive noise elimination model is used to monitor and analyze signal characteristics in real time;

[0029] Using the adaptive noise elimination model, the amplified target fluorescence signal is analyzed in real time to identify the target component and the noise component in the signal to obtain the analysis result;

[0030] According to the analysis result, automatically adjusting the parameters of the adaptive noise elimination model, optimizing the noise elimination effect, ensuring that the noise component in the amplified target fluorescence signal is effectively removed, and generating an optimized noise elimination model;

[0031] The optimized noise elimination model is used to perform noise removal processing on the amplified target fluorescence signal to generate a pure target fluorescence signal, thereby providing a clean signal source for subsequent signal intensity measurement.

[0032] Optionally, based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to perform noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal, including:

[0033] Based on the amplified target fluorescence signal S amp , using convolutional neural network to build an adaptive noise removal model M CNN , the adaptive noise cancellation model M CNN Used to monitor and analyze signal characteristics in real time;

[0034] Using the adaptive noise cancellation model M CNN , the amplified target fluorescence signal S amp Perform real-time analysis to identify target components in the signal.

[0035] According to the analysis result R, the adaptive noise elimination model M is automatically adjusted CNNparameters to optimize the noise elimination effect and ensure that the amplified target fluorescence signal S amp The noise component N in the image is effectively removed to generate an optimized noise elimination model M' CNN ;

[0036] The optimized noise elimination model M' is calculated by the following formula: CNN :

[0037] M′ CNN =M CNN +W·ΔP

[0038] Among them, ΔP is the parameter increment automatically adjusted according to the analysis result R, W is the weight coefficient, and the weight coefficient W is expressed as:

[0039]

[0040] Among them, α is the sensitivity coefficient, S threshold is the preset target component intensity threshold. This formula ensures that when the target component intensity S target Approaching or exceeding the threshold S threshold When , the weight coefficient W is close to 1, thus increasing the strength of parameter adjustment;

[0041] Using the optimized noise elimination model M' CNN , the amplified target fluorescence signal S amp Perform noise removal to generate a pure target fluorescence signal S clean , providing a clean signal source for subsequent signal strength measurement;

[0042] The noise removal processing formula is calculated as follows:

[0043] S clean =S amp -β·N

[0044] Among them, β is the noise removal coefficient, and the noise removal coefficient β is expressed as:

[0045]

[0046] Among them, γ is an exponential factor. This formula ensures that when the target component intensity S target When the noise component N is relatively strong, the noise removal coefficient β is close to 1, thereby removing the noise more effectively.

[0047] Optionally, the method uses a built-in high-precision analog-to-digital converter to digitize the pure target fluorescence signal, uses a self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, accurately measures the intensity of the pure target fluorescence signal, and generates final fluorescence signal intensity data, including:

[0048] The pure target fluorescence signal is digitized by using a built-in high-precision analog-to-digital converter, and the analog signal is converted into a digital signal to obtain a digitized pure target fluorescence signal;

[0049] Based on the digitized pure target fluorescence signal, a self-correction algorithm is used to analyze nonlinear errors that may exist in the analog-to-digital conversion process, including quantization errors and gain errors, to obtain error analysis results;

[0050] According to the error analysis results, adjusting the parameters of the self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, optimize the quality of the digital signal, and generate a corrected digital signal;

[0051] The corrected digital signal is used to accurately measure the intensity of the pure target fluorescence signal to generate final fluorescence signal intensity data, thereby ensuring the accuracy and reliability of the measurement result.

[0052] Optionally, based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized by a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test result, including:

[0053] Using a large number of data sets of ATP standard solutions with different concentrations, the large number of data sets of ATP standard solutions with different concentrations include ATP standard solutions with known concentrations and their corresponding fluorescence signal intensities, preprocessing the data sets to obtain a preprocessed data set;

[0054] Based on the preprocessed data set, a machine learning nonlinear regression model is constructed, and the model parameters are optimized using a gradient descent algorithm to ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity and the ATP concentration, thereby generating a trained nonlinear regression model;

[0055] Analyzing the intensity of the pure target fluorescence signal using the trained nonlinear regression model, taking the intensity of the pure target fluorescence signal as input, and calculating the concentration value of ATP in the sample to be tested through model prediction to obtain a preliminary concentration value;

[0056] Through multiple iterations of optimizing model parameters, the accuracy and reliability of model predictions are verified, and the concentration value of ATP in the sample to be tested is generated to ensure the accuracy and reliability of the test results.

[0057] Optionally, based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized by a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test result, including:

[0058] A large number of data sets of ATP standard solutions with different concentrations are used, wherein the data set contains ATP standard solutions C with known concentrations. i and its corresponding fluorescence signal intensity I i , preprocess the data set to obtain a preprocessed data set (C i , I i );

[0059] Based on the preprocessed data set (C i , I i ), construct a machine learning nonlinear regression model f(I; θ), use the gradient descent algorithm to optimize the model parameters, ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity I and the ATP concentration C, and generate a trained nonlinear regression model f trained (I;θ);

[0060] The parameter update formula of the gradient descent algorithm is as follows:

[0061]

[0062] Among them, θ j is the jth parameter in the model parameter vector θ, θ=(θ 0 ,θ 1 ,…,θ n ) represents the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, and I ij is the jth eigenvalue of the i-th sample;

[0063] Using the trained nonlinear regression model f trained (I; θ), the intensity of the pure target fluorescence signal I clean Analyze the intensity of the pure target fluorescence signal I clean As input, the concentration value C of ATP in the sample to be tested is calculated by model prediction pred , get the initial concentration value C pred ;

[0064] The prediction formula is as follows: C pred =f trained (I clean ;θ)

[0065] The model parameters are optimized through multiple iterations to verify the accuracy and reliability of the model prediction and generate the ATP concentration value C in the sample to be tested. final , ensuring the accuracy and reliability of the test results;

[0066] The optimization formula is as follows:

[0067]

[0068] C final =f trained (I clean θ opt )

[0069] Among them, θ opt is the optimized model parameter vector; C i is a standard solution of ATP with known concentration, I i is the corresponding fluorescence signal intensity, (C i , I i ) is the preprocessed data set, θ=(θ 0 ,θ 1 ,…,θ n ) is the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, I ij is the jth eigenvalue of the ith sample, f(I;θ) is the nonlinear regression model, f trained (I; θ) is the trained nonlinear regression model, I clean is the intensity of the pure target fluorescence signal, C pred is the initial concentration value, C final is the final concentration of ATP in the sample to be tested, θ opt is the optimized model parameter vector.

[0070] In a second aspect, the present application provides a highly sensitive ATP fluorescence detection system, comprising:

[0071] A configuration module is used to determine the optimal ratio of luciferase to luciferin, configure a reaction mixture containing the optimal ratio of luciferase to luciferin, and ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions;

[0072] A mixing module is used to control the reaction environment based on the reaction mixture by using microfluidic chip technology, so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal, so as to obtain a mixed reaction system;

[0073] A separation and measurement module is used to capture the fluorescence signal generated by the mixed reaction system within a set time using a high-sensitivity photomultiplier tube, and process the fluorescence signal using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal;

[0074] The analysis and calculation module is used to analyze the intensity of the pure target fluorescence signal based on the intensity of the pure target fluorescence signal, use a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, and use a gradient descent algorithm to optimize model parameters. The concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

[0075] In the embodiment of the present application, the optimal ratio of luciferase to luciferin is determined, and a reaction mixture containing luciferase and luciferin in the optimal ratio is configured to ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions; based on the reaction mixture, the reaction environment is controlled by using microfluidic chip technology so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal to obtain a mixed reaction system; a high-sensitivity photomultiplier tube is used to capture the fluorescence signal generated by the mixed reaction system within a set time, and a digital phase-locked amplification technology and an adaptive noise elimination algorithm are used to process the fluorescence signal, separate and generate a pure target fluorescence signal, and measure the intensity of the pure target fluorescence signal; based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized by using a gradient descent algorithm, the intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated by model prediction to ensure the accuracy and reliability of the detection results.

[0076] The technical solution of this application has the following beneficial effects:

[0077] The present application determines the optimal ratio of luciferase to luciferin and configures a reaction mixture containing the optimal ratio. The present invention can produce a significantly enhanced fluorescence signal under low ATP concentration conditions, thereby improving the detection sensitivity. The microfluidic chip technology and dynamic temperature control technology are applied to ensure the stability of the reaction environment and the reaction speed, thereby improving the stability of the fluorescence signal. The high-sensitivity photomultiplier tube combined with the digital phase-locked amplification technology and the adaptive noise elimination algorithm can efficiently capture and process the fluorescence signal, separate and generate a pure target fluorescence signal, and accurately measure its intensity. The machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations uses a gradient descent algorithm to optimize the model parameters, thereby further improving the accuracy of the analysis of the intensity of the pure target fluorescence signal, thereby ensuring the accuracy and reliability of the ATP concentration value in the sample to be tested.

[0078] Furthermore, the present application uses a high-sensitivity photomultiplier tube and a dynamic integration time window based on the change of fluorescence signal intensity combined with an adaptive threshold algorithm. The present invention can efficiently capture the fluorescence signal generated by the mixed reaction system and obtain a preliminary captured signal; a digital phase-locked amplification technology based on frequency tracking is used to analyze the signal spectrum through a fast Fourier transform algorithm, and the frequency closest to the target fluorescence signal is selected as the synchronization reference signal for phase locking processing to effectively amplify the target fluorescence signal; an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, automatically adjust the algorithm parameters, remove the noise components, and obtain a pure target fluorescence signal; finally, the signal is digitized using a built-in high-precision analog-to-digital converter, and a self-correction algorithm is used to correct nonlinear errors, thereby ensuring the accuracy of the fluorescence signal intensity measurement, thereby providing a high-quality data source for subsequent analysis.

[0079] Furthermore, the present application uses a large number of data sets of ATP standard solutions of different concentrations and preprocesses them. The present invention constructs a machine learning nonlinear regression model, which can accurately reflect the nonlinear relationship between the fluorescence signal intensity and the ATP concentration; the gradient descent algorithm is used to optimize the model parameters to ensure the accuracy and generalization ability of the model; the trained nonlinear regression model is used to analyze the intensity of the pure target fluorescence signal, and the concentration value of ATP in the sample to be tested is calculated by model prediction to obtain a preliminary concentration value; the model parameters are optimized through multiple iterations to verify the accuracy and reliability of the model prediction, and finally the concentration value of ATP in the sample to be tested is generated, ensuring the high accuracy and reliability of the test results. This method not only improves the detection efficiency, but also enhances the credibility of the test results, and is suitable for various application scenarios that require high-precision ATP concentration detection.

[0080] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0082] Figure 1 A flow chart of a highly sensitive ATP fluorescence detection method provided in an embodiment of the present application;

[0083] Figure 2 A schematic diagram of the structure of a highly sensitive ATP fluorescence detection system provided in an embodiment of the present application;

[0084] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0085] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0086] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0087] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0088] Figure 1 A flow chart of a highly sensitive ATP fluorescence detection method is provided for the present application embodiment, as shown in Figure 1 As shown, the method includes:

[0089] 101. Determine the optimal ratio of luciferase to luciferin, prepare a reaction mixture containing the optimal ratio of luciferase to luciferin, and ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions;

[0090] In this step, luciferase is an enzyme that can catalyze the oxidation of luciferin to produce light. Luciferin is a compound that can emit light under the action of luciferase. The optimal ratio is the ratio of luciferase to luciferin determined experimentally to ensure that a significantly enhanced fluorescence signal is produced under low ATP concentration conditions.

[0091] First, through a series of preliminary experiments, the optimal ratio of luciferase to luciferin was determined so that the fluorescence signal intensity reached the maximum value in the presence of low concentration of ATP.

[0092] Secondly, according to the determined optimal ratio, a reaction mixture containing luciferase and luciferin is prepared to provide necessary chemical reaction conditions for subsequent detection.

[0093] In the example of the present application, it is assumed that in water quality testing, a standard ATP solution of known concentration is added to a mixture of luciferase and luciferin in different proportions, and the intensity of the fluorescence signal is measured and recorded.

[0094] The ratio with the strongest fluorescence signal was selected as the optimal ratio, and the reaction mixture was prepared using this ratio for ATP detection in actual water samples.

[0095] 102. Based on the reaction mixture, the reaction environment is controlled by using microfluidic chip technology so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal to obtain a mixed reaction system;

[0096] In this step, microfluidic chip technology is a technology that uses a network of tiny channels to process or analyze trace amounts of liquid. Dynamic temperature control technology is a technology that optimizes reaction conditions by adjusting the reaction temperature in real time.

[0097] First, the prepared reaction mixture and the sample to be tested are introduced into the microfluidic chip and fully mixed in a dark environment.

[0098] Secondly, dynamic temperature control technology is used to adjust the reaction temperature as needed to speed up the reaction and improve the stability of the fluorescence signal.

[0099] In this application example, it is assumed that in water quality testing, water samples and reaction mixtures are injected into different channels of a microfluidic chip, and their flow and mixing are controlled by micropumps and valves. Then, the built-in heating element and temperature sensor are used to monitor and adjust the reaction temperature in real time to ensure that the reaction is carried out at the optimal temperature, thereby improving the detection sensitivity and accuracy.

[0100] 103. Using a high-sensitivity photomultiplier tube to capture the fluorescence signal generated by the mixed reaction system within a set time, and using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to process the fluorescence signal, separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal;

[0101] In this step, the high-sensitivity photomultiplier tube is a highly sensitive photoelectric conversion device that can convert weak light signals into electrical signals. Digital phase-locked amplification technology is a signal processing technology used to extract useful signals from noise. The adaptive noise removal algorithm is an algorithm that automatically adjusts parameters to remove noise.

[0102] First, a highly sensitive photomultiplier tube is used to capture the fluorescence signal generated by the mixed reaction system.

[0103] Secondly, digital phase-locked amplification technology and adaptive noise elimination algorithm are used to process the captured signals and separate the pure target fluorescence signals.

[0104] Finally, the intensity of the pure target fluorescence signal is measured to provide a data basis for subsequent analysis.

[0105] In this application example, it is assumed that in water quality testing, a photomultiplier tube is placed in the fluorescence detection area of ​​the microfluidic chip to capture the fluorescence signal after the water sample reacts. The frequency of the fluorescence signal is locked by digital phase-locked amplification technology to remove background noise; the signal is further purified by combining the adaptive noise elimination algorithm. Finally, pure fluorescence signal intensity data is obtained for subsequent ATP concentration calculation.

[0106] Optionally, the step 103 uses a high-sensitivity photomultiplier tube to capture the fluorescence signal generated by the mixed reaction system within a set time, and uses a digital phase-locked amplification technology and an adaptive noise elimination algorithm to process the fluorescence signal, separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal, including: using a high-sensitivity photomultiplier tube, based on a dynamic integration time window of the fluorescence signal intensity change, combined with an adaptive threshold algorithm, to efficiently capture the fluorescence signal generated by the mixed reaction system to obtain a preliminary captured signal; according to the preliminary captured signal, using a digital phase-locked amplification technology based on frequency tracking, and analyzing the signal spectrum through a fast Fourier transform algorithm. , select the frequency closest to the target fluorescence signal as the synchronization reference signal, perform phase locking processing on the preliminary captured signal, and obtain the amplified target fluorescence signal; based on the amplified target fluorescence signal, use an adaptive noise elimination algorithm based on deep learning, monitor and analyze the signal in real time through a convolutional neural network, automatically adjust the algorithm parameters, and perform noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal; use a built-in high-precision analog-to-digital converter to digitize the pure target fluorescence signal, use a self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, accurately measure the intensity of the pure target fluorescence signal, and generate final fluorescence signal intensity data.

[0107] Optionally, the use of a high-sensitivity photomultiplier tube in step 103, based on a dynamic integration time window of the change in fluorescence signal intensity, combined with an adaptive threshold algorithm, to efficiently capture the fluorescence signal generated by the mixed reaction system to obtain a preliminary captured signal, includes: using a high-sensitivity photomultiplier tube, based on an average fluorescence signal intensity determined in a preliminary experiment, setting an initial integration time window, starting to preliminarily capture the fluorescence signal generated by the mixed reaction system, and obtaining a preliminary captured fluorescence signal intensity; using an adaptive threshold algorithm to evaluate the signal intensity in real time based on the preliminary captured fluorescence signal intensity, and when it is detected that the signal intensity exceeds a preset threshold range, dynamically adjusting the length of the integration time window to obtain an optimized integration time window; based on the optimized integration time window, using a high-sensitivity photomultiplier tube to continuously capture the fluorescence signal generated by the mixed reaction system, and updating the integration time window immediately after each capture to ensure that the best signal capture performance is maintained throughout the detection process, and to obtain a continuously captured fluorescence signal; all the continuously captured fluorescence signals are aggregated to form a preliminary captured signal to provide high-quality raw data for subsequent processing using a digital phase-locked amplification technology based on frequency tracking.

[0108] Optionally, the method in step 103 adopts a digital phase-locked amplification technology based on frequency tracking according to the preliminary capture signal, analyzes the signal spectrum through a fast Fourier transform algorithm, selects a frequency closest to the target fluorescence signal as a synchronization reference signal, performs phase locking processing on the preliminary capture signal, and obtains an amplified target fluorescence signal, including: performing spectrum analysis on the preliminary capture signal using a fast Fourier transform algorithm, identifying multiple frequency components in the signal, and obtaining a preliminary spectrum analysis result; selecting a frequency closest to the target fluorescence signal from the preliminary spectrum analysis result as a synchronization reference signal according to a preset frequency range and signal strength threshold, and generating a selected synchronization reference signal; based on the selected synchronization reference signal, adopting a digital phase-locked amplification technology based on frequency tracking, adjusting the phase-locked loop parameters so that the phase-locked loop output signal maintains phase synchronization with the selected synchronization reference signal, performing phase locking processing on the preliminary capture signal, and generating a phase locking signal; and amplifying the phase locking signal using the adjusted amplifier gain to ensure that the intensity of the target fluorescence signal is significantly improved while reducing the influence of background noise, and generating an amplified target fluorescence signal.

[0109] Optionally, the method in step 103 uses an adaptive noise elimination algorithm based on deep learning based on the amplified target fluorescence signal, monitors and analyzes the signal in real time through a convolutional neural network, automatically adjusts the algorithm parameters, performs noise removal processing on the amplified target fluorescence signal, and obtains a pure target fluorescence signal, including: constructing an adaptive noise elimination model based on the amplified target fluorescence signal using a convolutional neural network, and the adaptive noise elimination model is used to monitor and analyze signal characteristics in real time; using the adaptive noise elimination model, performing real-time analysis on the amplified target fluorescence signal, identifying target components and noise components in the signal, and obtaining an analysis result; according to the analysis result, automatically adjusting the parameters of the adaptive noise elimination model, optimizing the noise elimination effect, ensuring that the noise components in the amplified target fluorescence signal are effectively removed, and generating an optimized noise elimination model; using the optimized noise elimination model, performing noise removal processing on the amplified target fluorescence signal, generating a pure target fluorescence signal, and providing a clean signal source for subsequent signal intensity measurement.

[0110] Optionally, the step 103 of using a built-in high-precision analog-to-digital converter to digitize the pure target fluorescence signal, using a self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, accurately measuring the intensity of the pure target fluorescence signal, and generating final fluorescence signal intensity data includes: using a built-in high-precision analog-to-digital converter to digitize the pure target fluorescence signal, converting the analog signal into a digital signal, and obtaining a digitized pure target fluorescence signal; based on the digitized pure target fluorescence signal, using a self-correction algorithm to analyze nonlinear errors that may exist in the analog-to-digital conversion process, including quantization errors and gain errors, to obtain error analysis results; according to the error analysis results, adjusting the parameters of the self-correction algorithm to correct the nonlinear errors in the analog-to-digital conversion process, optimize the quality of the digital signal, and generate a corrected digital signal; using the corrected digital signal to accurately measure the intensity of the pure target fluorescence signal, generate final fluorescence signal intensity data, and ensure the accuracy and reliability of the measurement results.

[0111] In this step, the dynamic integration time window is an integration time setting that is automatically adjusted according to the change in fluorescence signal intensity to optimize signal capture efficiency. The adaptive threshold algorithm is an algorithm that can adjust the threshold in real time according to the signal intensity, and is used to determine whether the signal intensity requires adjustment of the integration time window. The fast Fourier transform algorithm is an algorithm that efficiently calculates discrete Fourier transforms and is used to analyze signal spectra. Digital phase-locked amplification technology: a technology that extracts useful signals by locking the signal frequency, often used to separate target signals from noise. Convolutional neural network is a deep learning model that is particularly suitable for image and signal processing tasks and can automatically identify and classify features in signals. A high-precision analog-to-digital converter is a device that converts analog signals into digital signals with high precision. The self-correction algorithm is an algorithm used to correct nonlinear errors in the analog-to-digital conversion process to ensure the accuracy of digital signals.

[0112] First, a high-sensitivity photomultiplier tube is used to capture the fluorescence signal. The initial integration time window is set according to the average fluorescence signal intensity determined in the preliminary experiment. The adaptive threshold algorithm is used to evaluate the signal intensity in real time. When the signal intensity exceeds the preset threshold range, the length of the integration time window is dynamically adjusted. Based on the optimized integration time window, the fluorescence signal is continuously captured and the integration time window is updated in real time to ensure the best signal capture performance. All continuously captured fluorescence signals are aggregated to form a preliminary capture signal.

[0113] Secondly, the fast Fourier transform algorithm is used to perform spectrum analysis on the preliminary capture signal to identify multiple frequency components. According to the preset frequency range and signal strength threshold, the frequency closest to the target fluorescence signal is selected as the synchronization reference signal. The digital phase-locked amplification technology is used to adjust the phase-locked loop parameters to keep the output signal in phase with the synchronization reference signal. The preliminary capture signal is phase-locked and the amplified target fluorescence signal is generated using the adjusted amplifier gain.

[0114] Furthermore, an adaptive noise elimination model based on a convolutional neural network is constructed for real-time monitoring and analysis of signal characteristics. The model is used to perform real-time analysis on the amplified target fluorescence signal to identify the target component and the noise component. According to the analysis results, the model parameters are automatically adjusted to optimize the noise elimination effect. The optimized noise elimination model is generated, and the amplified target fluorescence signal is subjected to noise removal processing to generate a pure target fluorescence signal.

[0115] Finally, a high-precision analog-to-digital converter is used to convert the pure target fluorescence signal into a digital signal. A self-correction algorithm is used to analyze the nonlinear errors in the analog-to-digital conversion process, including quantization error and gain error. According to the error analysis results, the self-correction algorithm parameters are adjusted to optimize the digital signal quality. The corrected digital signal is used to accurately measure the intensity of the pure target fluorescence signal to generate the final fluorescence signal intensity data.

[0116] In the embodiment of the present application, it is assumed that in water quality testing, in order to improve the sensitivity and accuracy of ATP fluorescence detection, the above scheme can be adopted.

[0117] First, in water sample detection, a high-sensitivity photomultiplier tube is used to capture the fluorescence signal. The initial integration time window is set based on the average fluorescence signal intensity determined by the preliminary experiment. During the detection process, the adaptive threshold algorithm is used to evaluate the signal intensity in real time and dynamically adjust the integration time window to ensure the best performance of signal capture. All captured fluorescence signals are aggregated to form a preliminary capture signal.

[0118] Secondly, the fast Fourier transform algorithm is used to perform spectrum analysis on the preliminary captured signal to identify multiple frequency components. According to the preset frequency range and signal intensity threshold, the frequency closest to the target fluorescence signal is selected as the synchronization reference signal. The digital phase-locked amplification technology is used to adjust the phase-locked loop parameters so that the output signal is phase-synchronized with the synchronization reference signal, thereby amplifying the target fluorescence signal.

[0119] Furthermore, an adaptive noise elimination model based on convolutional neural network is constructed to monitor and analyze the amplified target fluorescence signal in real time. The model can identify the target component and noise component in the signal and automatically adjust the parameters to optimize the noise elimination effect. Finally, a pure target fluorescence signal is generated.

[0120] Finally, a high-precision analog-to-digital converter is used to convert the pure target fluorescence signal into a digital signal. A self-correction algorithm is used to analyze and correct the nonlinear error in the analog-to-digital conversion process to ensure the accuracy of the digital signal. The corrected digital signal is used to accurately measure the fluorescence signal intensity and generate the final fluorescence signal intensity data for subsequent ATP concentration calculations.

[0121] Through these steps, high-sensitivity and high-accuracy ATP fluorescence detection can be achieved in water quality testing, ensuring the reliability and repeatability of the test results.

[0122] This application considers that in the high-sensitivity ATP fluorescence detection method, in order to improve the purity of the signal and the accuracy of the detection, it is necessary to perform noise removal on the amplified target fluorescence signal. Here, an adaptive noise removal algorithm based on deep learning is adopted, which monitors and analyzes the signal characteristics in real time through a convolutional neural network, and automatically adjusts the algorithm parameters, thereby effectively removing the noise components.

[0123] Optionally, the step 103 uses an adaptive noise elimination algorithm based on deep learning based on the amplified target fluorescence signal, monitors and analyzes the signal in real time through a convolutional neural network, automatically adjusts algorithm parameters, and performs noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal, including:

[0124] Based on the amplified target fluorescence signal S amp , using convolutional neural network to build an adaptive noise removal model M CNN , the adaptive noise cancellation model N CNN Used to monitor and analyze signal characteristics in real time;

[0125] Using the adaptive noise cancellation model N CNN , the amplified target fluorescence signal S amp Perform real-time analysis to identify target components in the signal.

[0126] According to the analysis result R, the adaptive noise elimination model M is automatically adjusted CNN parameters to optimize the noise elimination effect and ensure that the amplified target fluorescence signal S amp The noise component N in the image is effectively removed to generate an optimized noise elimination model M' CNN ;

[0127] The optimized noise elimination model M' is calculated by the following formula: CNN :

[0128] M′ CNN =M CNN+W·ΔP

[0129] Among them, ΔP is the parameter increment automatically adjusted according to the analysis result R, W is the weight coefficient, and the weight coefficient W is expressed as:

[0130]

[0131] Among them, α is the sensitivity coefficient, S threshold is the preset target component intensity threshold. This formula ensures that when the target component intensity S target Approaching or exceeding the threshold S threshold When , the weight coefficient W is close to 1, thus increasing the strength of parameter adjustment;

[0132] Using the optimized noise elimination model M' CNN , the amplified target fluorescence signal S amp Perform noise removal to generate a pure target fluorescence signal S clean , providing a clean signal source for subsequent signal strength measurement;

[0133] The noise removal processing formula is calculated as follows:

[0134] S clean =S amp -β·N

[0135] Among them, β is the noise removal coefficient, and the noise removal coefficient β is expressed as:

[0136]

[0137] Among them, γ is an exponential factor. This formula ensures that when the target component intensity S target When the noise component N is relatively strong, the noise removal coefficient β is close to 1, thereby removing the noise more effectively.

[0138] Dynamically adjust the adaptive noise cancellation model M through the weight coefficient W and the parameter increment ΔP CNN The parameters of the model enable the model to self-optimize according to the current signal characteristics. The design of the weight coefficient W ensures that when the target component intensity S target Approaching or exceeding the threshold S threshold When the target component intensity S is increased, the parameter adjustment strength is increased to better adapt to signal changes. The design of the noise removal coefficient β ensures that when the target component intensity S target When the noise component N is relatively strong, the noise is removed more effectively while retaining the target signal.

[0139] The following is a brief introduction to how to obtain the parameters of the formula:

[0140] Among them, the sensitivity coefficient α can be determined by experiment, and the appropriate α value is selected so that W is target Close to S threshold There is an obvious response. The target component intensity threshold S threshold Through statistical analysis of experimental data, a reasonable threshold is determined, which should be able to distinguish normal signals from noise signals. The exponential factor γ is determined through experiments, and the appropriate γ value is selected so that β is target It is close to 1 when it is stronger than N, and decreases when N is larger. The parameter increment ΔP is calculated by gradient descent or other optimization algorithms during model training, which represents the change in the parameter in each iteration.

[0141] Assume that the following parameters have been obtained in water quality testing:

[0142] Sensitivity coefficient α = 10; preset target component intensity threshold S threshold =500; exponential factor γ = 2;

[0143] And the following data were obtained from actual testing:

[0144] The amplified target fluorescence signal S amp =800; the preliminary analysis result R shows the target component intensity S target =700; noise component N = 300;

[0145] Calculate the weight coefficient W:

[0146]

[0147] W≈1

[0148] Due to e -2000 is very close to 0, so W is close to 1.

[0149] Calculate the parameter increment ΔP:

[0150] Assume that the parameter increment ΔP obtained through the model training process is 0.01;

[0151] Generate optimized noise elimination model M' CNN :

[0152] M′ CNN =M CNN +W·ΔP

[0153] M′ CNN =M CNN +1 0.01

[0154] M′ CNN =M CNN +0.01

[0155] Calculate the noise removal coefficient β:

[0156]

[0157] β=0.49

[0158] Generate pure target fluorescence signal S clean :

[0159] S clean =S amp -β·N

[0160] S clean =800-0.49·300

[0161] S clean =653

[0162] Through the above calculation, the pure target fluorescence signal S is obtained. clean =653. This shows that after the adaptive noise elimination algorithm is processed, the noise component in the original signal is effectively removed and the target signal is enhanced. The weight coefficient W≈1, indicating that the current target component strength S target =700 is much higher than the threshold S threshold =500, so the model parameter adjustment is stronger, which helps to better adapt to signal changes. Noise removal coefficient β = 0.49, indicating that when the target component intensity S target =700 and the noise component N = 300, the noise removal effect is stronger, but the noise will not be completely removed to avoid distortion of the target signal. Pure target fluorescence signal S clean =653, compared with the original signal S amp =800, some noise components are removed, a purer signal is obtained, and the accuracy and reliability of detection are improved.

[0163] This treatment method can significantly improve the quality of ATP fluorescence signals in water quality testing, ensuring the accuracy and reliability of test results.

[0164] 104. Based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized using a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

[0165] In this step, the machine learning nonlinear regression model is a model based on data training that can fit nonlinear relationships. The gradient descent algorithm is an optimization algorithm used to minimize the loss function and thus optimize the model parameters.

[0166] First, a machine learning nonlinear regression model was trained using a large dataset of ATP standard solutions with different concentrations.

[0167] Secondly, the gradient descent algorithm was used to optimize the model parameters to ensure that the model could accurately reflect the nonlinear relationship between fluorescence signal intensity and ATP concentration.

[0168] Finally, the trained model is used to analyze the intensity of the pure target fluorescence signal to predict and generate the concentration value of ATP in the sample to be tested.

[0169] In the example of this application, it is assumed that in water quality testing, a series of ATP standard solutions with known concentrations are used to collect corresponding fluorescence signal intensity data and construct a training data set. Secondly, a nonlinear regression model is trained, and the model parameters are optimized using a gradient descent algorithm. Finally, the pure target fluorescence signal intensity is input into the trained model to predict and output the ATP concentration value in the water sample to be tested, ensuring the accuracy and reliability of the test results.

[0170] Optionally, the intensity of the pure target fluorescence signal in step 104 is analyzed by using a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, and a gradient descent algorithm is used to optimize model parameters. The concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test result, including: using a large number of data sets of ATP standard solutions with different concentrations, wherein the large number of data sets of ATP standard solutions with different concentrations contain ATP standard solutions with known concentrations and their corresponding fluorescence signal intensities, and preprocessing the data sets to obtain a preprocessed data set ; Based on the preprocessed data set, a machine learning nonlinear regression model is constructed, and the model parameters are optimized by a gradient descent algorithm to ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity and the ATP concentration, and a trained nonlinear regression model is generated; using the trained nonlinear regression model, the intensity of the pure target fluorescence signal is analyzed, and the intensity of the pure target fluorescence signal is used as input, and the concentration value of ATP in the sample to be tested is calculated by model prediction to obtain a preliminary concentration value; optimizing the model parameters through multiple iterations, verifying the accuracy and reliability of the model prediction, generating the concentration value of ATP in the sample to be tested, and ensuring the accuracy and reliability of the test results.

[0171] In this step, the machine learning nonlinear regression model is a model based on data training, which can fit the nonlinear relationship between fluorescence signal intensity and ATP concentration. The gradient descent algorithm is an optimization algorithm that optimizes model performance by iteratively updating model parameters to minimize the error between the predicted value and the actual value. Preprocessing is to clean and standardize the raw data to improve the effect of model training. The data set is a data set containing ATP standard solutions of known concentrations and their corresponding fluorescence signal intensities, which is used to train and verify the model.

[0172] First, collect a large number of ATP standard solutions of different concentrations and record their corresponding fluorescence signal intensities. Preprocess the collected data, including removing outliers, normalizing or standardizing, to ensure data quality. Generate a preprocessed data set to prepare for subsequent model training.

[0173] Secondly, select a suitable machine learning algorithm to build a nonlinear regression model. Use the gradient descent algorithm to optimize the model parameters and reduce the error between the predicted value and the actual value through multiple iterations. Evaluate the model performance through cross-validation and other methods to ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity and ATP concentration. Generate a trained nonlinear regression model for subsequent concentration prediction.

[0174] Furthermore, the intensity of the pure target fluorescence signal is used as input to the trained nonlinear regression model. The model predicts and outputs the concentration value of ATP in the sample to be tested based on the input fluorescence signal intensity to obtain a preliminary concentration value.

[0175] Furthermore, the model parameters are further optimized through multiple iterations to ensure that the model can maintain good prediction performance on new data. The prediction accuracy of the model is verified through the validation set or test set to ensure the reliability of the model in practical applications.

[0176] Finally, the concentration value of ATP in the sample to be tested is generated to ensure the accuracy and reliability of the test results.

[0177] In the embodiments of the present application, it is assumed that in water quality testing, in order to accurately determine the ATP concentration in a water sample, the above scheme can be adopted.

[0178] First, collect a series of ATP standard solutions with known concentrations and measure their corresponding fluorescence signal intensities. These data may come from laboratory experiments or historical data records. Preprocess the data, including steps such as removing outliers, normalization or standardization, to ensure data consistency and high quality.

[0179] Secondly, a nonlinear regression model based on a neural network was constructed using the preprocessed data set. The model parameters were optimized by the gradient descent algorithm to ensure that the model could accurately reflect the nonlinear relationship between the fluorescence signal intensity and the ATP concentration. The model performance was evaluated by cross-validation to ensure the generalization ability of the model on new data.

[0180] Furthermore, in actual water quality testing, the intensity of the pure target fluorescence signal is input into the trained nonlinear regression model. The model predicts and outputs the concentration value of ATP in the water sample to be tested based on the input fluorescence signal intensity to obtain a preliminary concentration value.

[0181] Finally, the model parameters are further optimized through multiple iterations to ensure that the model can maintain good prediction performance on new water quality samples. The prediction accuracy of the model is verified using a validation set or a test set to ensure the reliability of the model in practical applications. Finally, the concentration value of ATP in the water sample to be tested is generated to ensure the accuracy and reliability of the test results.

[0182] Through these steps, high-precision and high-reliability ATP concentration determination can be achieved in water quality testing, ensuring the effectiveness and accuracy of water quality monitoring.

[0183] This application considers that in a highly sensitive ATP fluorescence detection method, in order to accurately predict the ATP concentration in the sample to be tested, a machine learning nonlinear regression model f(I; θ) is used to establish the relationship between the fluorescence signal intensity I and the ATP concentration C. The model is trained by a large number of data sets of ATP standard solutions with different concentrations, and the model parameters are optimized using a gradient descent algorithm to ensure that the model can accurately reflect this nonlinear relationship.

[0184] Optionally, the intensity of the pure target fluorescence signal in step 104 is analyzed by using a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, and a gradient descent algorithm is used to optimize model parameters. The concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test result, including:

[0185] A large number of data sets of ATP standard solutions with different concentrations are used, wherein the data set contains ATP standard solutions C with known concentrations. i and its corresponding fluorescence signal intensity I i , preprocess the data set to obtain a preprocessed data set (C i , I i );

[0186] Based on the preprocessed data set (C i , I i), construct a machine learning nonlinear regression model f(I; θ), use the gradient descent algorithm to optimize the model parameters, ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity I and the ATP concentration C, and generate a trained nonlinear regression model f trained (I;θ);

[0187] The parameter update formula of the gradient descent algorithm is as follows:

[0188]

[0189] Among them, θ j is the jth parameter in the model parameter vector θ, θ=(θ 0 ,θ 1 ,…,θ n ) represents the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, and I ij is the jth eigenvalue of the i-th sample;

[0190] Using the trained nonlinear regression model f trained (I; θ), the intensity of the pure target fluorescence signal I clean Analyze the intensity of the pure target fluorescence signal I clean As input, the concentration value C of ATP in the sample to be tested is calculated by model prediction pred , get the initial concentration value C pred ;

[0191] The prediction formula is as follows: C pred =f trained (I clean ;θ)

[0192] The model parameters are optimized through multiple iterations to verify the accuracy and reliability of the model prediction and generate the ATP concentration value C in the sample to be tested. final , ensuring the accuracy and reliability of the test results;

[0193] The optimization formula is as follows:

[0194]

[0195] C final =f trained (I clean θ opt )

[0196] Among them, θ opt is the optimized model parameter vector; C i is a standard solution of ATP with known concentration, I i is the corresponding fluorescence signal intensity, (C i , Ii ) is the preprocessed data set, θ=(θ 0 ,θ 1 ,…,θ n ) is the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, I ij is the jth eigenvalue of the ith sample, f(I;θ) is the nonlinear regression model, f trained (I; θ) is the trained nonlinear regression model, I clean is the intensity of the pure target fluorescence signal, C pred is the initial concentration value, C final is the final concentration of ATP in the sample to be tested, θ opt is the optimized model parameter vector.

[0197] The raw data is cleaned and standardized to improve the effect of model training. i , I i ) contains a standard solution C of known concentration of ATP i and its corresponding fluorescence signal intensity I i . Construct a nonlinear regression model f(I; θ), where θ is the model parameter vector. Use the gradient descent algorithm to optimize the model parameter θ so that the model can accurately fit the nonlinear relationship between the light signal intensity I and the ATP concentration C. Use the trained model f trained (I;θ) intensity of pure target fluorescence signal I clean Analyze and predict the ATP concentration value C in the sample to be tested pred The model parameters are optimized through multiple iterations to generate the final ATP concentration value C final。

[0198] The following is a brief introduction to the design reasons of each sub-item of the formula:

[0199] This sub-item determines the adjustment direction and size of the model parameters by calculating the average of the product of the prediction error of each sample and the corresponding eigenvalue, thereby gradually reducing the overall prediction error.

[0200] The following is a brief introduction to how to obtain the parameters of the formula:

[0201] Dataset (C i , I i ) is obtained through experimental measurement, including a series of ATP standard solutions with known concentrations and their corresponding fluorescence signal intensities. The learning rate α is determined through experiments, and a suitable α is selected to balance the acquisition speed and stability. The number of samples m is determined according to the actual data set size. The eigenvalue I ijIt is directly obtained from experimental data. The model parameter vector θ is automatically learned through the training process.

[0202] Assume that the following data has been obtained:

[0203] Dataset (C i , I i ) contains 100 samples, of which C i and I i are the known concentrations of ATP standard solutions and the corresponding fluorescence signal intensities; learning rate α = 0.01; model parameter vector θ = (θ 0 ,θ 1 ,…,θ n )The initial value is a random value; the intensity of the pure target fluorescence signal I clean =653;

[0204] Data preprocessing: Assume that the dataset (C i , I i ) has been preprocessed, such as normalization or standardization.

[0205] Construct a nonlinear regression model:

[0206] Assume that the model form is a polynomial regression model f(I;θ)=θ 0 +θ 1 I+θ 2 I 2 +…+θ n I n ;

[0207] Gradient descent algorithm optimizes model parameters:

[0208] Initial parameter vector θ = (0.1, 0.2, 0.3, ..., 0.1);

[0209] Iteratively update the parameter θ through the gradient descent algorithm until the model converges;

[0210] Trained nonlinear regression model:

[0211] After multiple iterations, the trained model parameter vector θ is obtained trained =(0.1, 0.2, 0.3,…, 0.1).

[0212] Predict ATP concentration;

[0213] Preliminary forecast:

[0214] C pred =f trained (I clean θ trained )

[0215] Cpred =0.1+0.2·653+0.3·653 2 +…+0.1 653 n

[0216] Assume that after calculation, we get C pred =10.5μM.

[0217] Optimize model parameters:

[0218] By optimizing the model parameters through multiple iterations, the optimized parameter vector θ is obtained. opt =(0.1, 0.2, 0.3,…, 0.1).

[0219] Final prediction:

[0220] C final =f trained (I clean θ opt )

[0221] C final =0.1+0.2·653+0.3·653 2 +…+0.1 653 n

[0222] Assume that after calculation, we get C final =10.7μM.

[0223] Through the above calculation, the ATP concentration value C in the sample to be tested is obtained. final =10.7μM. This shows that the trained and optimized nonlinear regression model can accurately predict the ATP concentration in water quality samples. pred = 10.5 μM is the intensity of the pure target light signal I obtained by the trained model clean =653 for preliminary prediction and obtain preliminary concentration value. final =10.7 μM is the final ATP concentration value obtained by optimizing the model parameters through multiple iterations to further improve the prediction accuracy.

[0224] This treatment method can significantly improve the accuracy and reliability of ATP concentration detection in water quality testing, ensuring the accuracy and consistency of the test results.

[0225] Figure 2 A structural schematic diagram of a highly sensitive ATP fluorescence detection system is provided for the present application embodiment, as shown in Figure 2 As shown, the system includes:

[0226] Configuration module 21, used to determine the optimal ratio of luciferase to luciferin, configure a reaction mixture containing the optimal ratio of luciferase to luciferin, and ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions;

[0227] A mixing module 22 is used to control the reaction environment based on the reaction mixture by using microfluidic chip technology, so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal, so as to obtain a mixed reaction system;

[0228] The separation and measurement module 23 is used to capture the fluorescence signal generated by the mixed reaction system within a set time using a high-sensitivity photomultiplier tube, and process the fluorescence signal using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to separate and generate a pure target fluorescence signal, and at the same time measure the intensity of the pure target fluorescence signal;

[0229] The analysis and calculation module 24 is used to analyze the intensity of the pure target fluorescence signal based on the intensity of the pure target fluorescence signal, use a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, and use a gradient descent algorithm to optimize model parameters. The concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

[0230] Figure 2 The highly sensitive ATP fluorescence detection system can be performed Figure 1 The implementation principle and technical effect of the high-sensitivity ATP fluorescence detection method described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the high-sensitivity ATP fluorescence detection system in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0231] In one possible design, Figure 2 The highly sensitive ATP fluorescence detection system of the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0232] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0233] The processing component 32 is used to: determine the optimal ratio of luciferase to luciferin, configure a reaction mixture containing the optimal ratio of luciferase to luciferin, and ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions; based on the reaction mixture, use microfluidic chip technology to control the reaction environment so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and use dynamic temperature control technology to adjust the reaction temperature, improve the reaction speed and the stability of the fluorescence signal, and obtain a mixed reaction system; use a high-sensitivity photomultiplier tube to capture the fluorescence signal generated by the mixed reaction system within a set time, and use digital phase-locked amplification technology and adaptive noise elimination algorithm to process the fluorescence signal, separate and generate a pure target fluorescence signal, and measure the intensity of the pure target fluorescence signal; based on the intensity of the pure target fluorescence signal, use a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, use a gradient descent algorithm to optimize model parameters, analyze the intensity of the pure target fluorescence signal, and calculate and generate the ATP concentration value in the sample to be tested through model prediction to ensure the accuracy and reliability of the detection result.

[0234] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0235] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0236] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0237] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0238] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0239] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0240] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a highly sensitive ATP fluorescence detection method.

[0241] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0242] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0243] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0244] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A highly sensitive ATP fluorescence detection method, characterized in that: include: Determine the optimal ratio of luciferase to luciferin, and prepare a reaction mixture containing the optimal ratio of luciferase to luciferin to ensure a significantly enhanced fluorescence signal under low ATP concentration conditions; Based on the reaction mixture, the reaction environment is controlled by using microfluidic chip technology, so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal, thereby obtaining a mixed reaction system; Using a high-sensitivity photomultiplier tube, the fluorescence signal generated by the mixed reaction system is captured within a set time, and the fluorescence signal is processed using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to separate and generate a pure target fluorescence signal, and the intensity of the pure target fluorescence signal is measured at the same time; Based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model trained with a large number of ATP standard solution data sets of different concentrations is used, and the model parameters are optimized using a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

2. The highly sensitive ATP fluorescence detection method according to claim 1, characterized in that: The method uses a high-sensitivity photomultiplier tube to capture the fluorescence signal generated by the mixed reaction system within a set time, and uses a digital phase-locked amplification technology and an adaptive noise elimination algorithm to process the fluorescence signal, separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal, including: Using a high-sensitivity photomultiplier tube, based on a dynamic integration time window of the fluorescence signal intensity change, combined with an adaptive threshold algorithm, the fluorescence signal generated by the mixed reaction system is efficiently captured to obtain a preliminary captured signal; According to the preliminary captured signal, a digital phase-locked amplification technology based on frequency tracking is used to analyze the signal spectrum through a fast Fourier transform algorithm, and a frequency closest to the target fluorescence signal is selected as a synchronization reference signal, and the preliminary captured signal is phase-locked to obtain an amplified target fluorescence signal; Based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to perform noise removal processing on the amplified target fluorescence signal to obtain a pure target fluorescence signal; The pure target fluorescence signal is digitized by using a built-in high-precision analog-to-digital converter, and a self-correction algorithm is used to correct the nonlinear error in the analog-to-digital conversion process, so as to accurately measure the intensity of the pure target fluorescence signal and generate final fluorescence signal intensity data.

3. The highly sensitive ATP fluorescence detection method according to claim 2, characterized in that: The method uses a high-sensitivity photomultiplier tube, based on a dynamic integration time window of the fluorescence signal intensity change, combined with an adaptive threshold algorithm, to efficiently capture the fluorescence signal generated by the mixed reaction system to obtain a preliminary captured signal, including: Using a high-sensitivity photomultiplier tube, based on the average fluorescence signal intensity determined in a preliminary experiment, an initial integration time window is set to start preliminarily capturing the fluorescence signal generated by the mixed reaction system to obtain a preliminarily captured fluorescence signal intensity; According to the intensity of the fluorescent signal initially captured, an adaptive threshold algorithm is used to evaluate the signal intensity in real time. When the signal intensity is detected to exceed the preset threshold range, the length of the integration time window is dynamically adjusted to obtain an optimized integration time window. Based on the optimized integration time window, the fluorescence signal generated by the mixed reaction system is continuously captured using a high-sensitivity photomultiplier tube, and the integration time window is updated immediately after each capture to ensure that the best signal capture performance is maintained throughout the detection process to obtain a continuously captured fluorescence signal; All of the continuously captured fluorescence signals are aggregated to form a preliminary captured signal, providing high-quality raw data for subsequent processing using a digital phase-locked amplification technology based on frequency tracking.

4. The highly sensitive ATP fluorescence detection method according to claim 2, characterized in that: The method comprises: according to the preliminary captured signal, using a digital phase-locked amplification technology based on frequency tracking, analyzing the signal spectrum through a fast Fourier transform algorithm, selecting a frequency closest to the target fluorescence signal as a synchronization reference signal, performing phase locking processing on the preliminary captured signal, and obtaining an amplified target fluorescence signal, including: Using a fast Fourier transform algorithm, performing spectrum analysis on the preliminary captured signal, identifying multiple frequency components in the signal, and obtaining a preliminary spectrum analysis result; According to a preset frequency range and signal intensity threshold, a frequency closest to the target fluorescence signal is selected from the preliminary spectrum analysis result as a synchronization reference signal to generate a selected synchronization reference signal; Based on the selected synchronization reference signal, a digital phase-locked amplification technology based on frequency tracking is adopted to adjust the phase-locked loop parameters so that the phase-locked loop output signal maintains phase synchronization with the selected synchronization reference signal, and the preliminary captured signal is phase-locked to generate a phase-locked signal; The phase-locked signal is amplified using the adjusted amplifier gain to ensure that the intensity of the target fluorescence signal is significantly improved while reducing the impact of background noise, thereby generating an amplified target fluorescence signal.

5. The highly sensitive ATP fluorescence detection method according to claim 2, characterized in that: Based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to remove noise from the amplified target fluorescence signal to obtain a pure target fluorescence signal, including: Based on the amplified target fluorescence signal, a convolutional neural network is used to construct an adaptive noise elimination model, wherein the adaptive noise elimination model is used to monitor and analyze signal characteristics in real time; Using the adaptive noise elimination model, the amplified target fluorescence signal is analyzed in real time to identify the target component and the noise component in the signal to obtain the analysis result; According to the analysis result, automatically adjusting the parameters of the adaptive noise elimination model, optimizing the noise elimination effect, ensuring that the noise component in the amplified target fluorescence signal is effectively removed, and generating an optimized noise elimination model; The optimized noise elimination model is used to perform noise removal processing on the amplified target fluorescence signal to generate a pure target fluorescence signal, thereby providing a clean signal source for subsequent signal intensity measurement.

6. The highly sensitive ATP fluorescence detection method according to claim 5, characterized in that: Based on the amplified target fluorescence signal, an adaptive noise elimination algorithm based on deep learning is used to monitor and analyze the signal in real time through a convolutional neural network, and the algorithm parameters are automatically adjusted to remove noise from the amplified target fluorescence signal to obtain a pure target fluorescence signal, including: Based on the amplified target fluorescence signal S amp , using convolutional neural network to build an adaptive noise removal model M CNN , the adaptive noise cancellation model M CNN Used to monitor and analyze signal characteristics in real time; Using the adaptive noise cancellation model M CNN , the amplified target fluorescence signal S amp Perform real-time analysis to identify target components in the signal. According to the analysis result R, the adaptive noise elimination model M is automatically adjusted CNN parameters to optimize the noise elimination effect and ensure that the amplified target fluorescence signal S amp The noise component N in the image is effectively removed to generate an optimized noise elimination model M' CNN ; The optimized noise elimination model M' is calculated by the following formula: CNN : M′ CNN =M CNN +W·ΔP Among them, ΔP is the parameter increment automatically adjusted according to the analysis result R, W is the weight coefficient, and the weight coefficient W is expressed as: Among them, α is the sensitivity coefficient, S threshold is the preset target component intensity threshold. This formula ensures that when the target component intensity S target Approaching or exceeding the threshold S threshold When , the weight coefficient W is close to 1, thus increasing the strength of parameter adjustment; Using the optimized noise elimination model M' CNN , the amplified target fluorescence signal S amp Perform noise removal to generate a pure target fluorescence signal S clean , providing a clean signal source for subsequent signal strength measurement; The noise removal processing formula is calculated as follows: S clean =S amp -β·N Among them, β is the noise removal coefficient, and the noise removal coefficient β is expressed as: Among them, γ is an exponential factor. This formula ensures that when the target component intensity S target When the noise component N is relatively strong, the noise removal coefficient β is close to 1, thereby removing the noise more effectively.

7. The highly sensitive ATP fluorescence detection method according to claim 2, characterized in that: The method utilizes a built-in high-precision analog-to-digital converter to digitize the pure target fluorescence signal, uses a self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, accurately measures the intensity of the pure target fluorescence signal, and generates final fluorescence signal intensity data, including: The pure target fluorescence signal is digitized by using a built-in high-precision analog-to-digital converter, and the analog signal is converted into a digital signal to obtain a digitized pure target fluorescence signal; Based on the digitized pure target fluorescence signal, a self-correction algorithm is used to analyze nonlinear errors that may exist in the analog-to-digital conversion process, including quantization errors and gain errors, to obtain error analysis results; According to the error analysis results, adjusting the parameters of the self-correction algorithm to correct the nonlinear error in the analog-to-digital conversion process, optimize the quality of the digital signal, and generate a corrected digital signal; The corrected digital signal is used to accurately measure the intensity of the pure target fluorescence signal to generate final fluorescence signal intensity data, thereby ensuring the accuracy and reliability of the measurement result.

8. The highly sensitive ATP fluorescence detection method according to claim 1, characterized in that: Based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized by a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results, including: Using a large number of data sets of ATP standard solutions with different concentrations, the large number of data sets of ATP standard solutions with different concentrations include ATP standard solutions with known concentrations and their corresponding fluorescence signal intensities, preprocessing the data sets to obtain a preprocessed data set; Based on the preprocessed data set, a machine learning nonlinear regression model is constructed, and the model parameters are optimized using a gradient descent algorithm to ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity and the ATP concentration, thereby generating a trained nonlinear regression model; Analyzing the intensity of the pure target fluorescence signal using the trained nonlinear regression model, taking the intensity of the pure target fluorescence signal as input, and calculating the concentration value of ATP in the sample to be tested through model prediction to obtain a preliminary concentration value; Through multiple iterations of optimizing model parameters, the accuracy and reliability of model predictions are verified, and the concentration value of ATP in the sample to be tested is generated to ensure the accuracy and reliability of the test results.

9. The highly sensitive ATP fluorescence detection method according to claim 8, characterized in that: Based on the intensity of the pure target fluorescence signal, a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations is used, and the model parameters are optimized by a gradient descent algorithm. The intensity of the pure target fluorescence signal is analyzed, and the concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results, including: A large number of data sets of ATP standard solutions with different concentrations are used, wherein the data set contains ATP standard solutions C with known concentrations. i and its corresponding fluorescence signal intensity I i , preprocess the data set to obtain a preprocessed data set (C i , I i ); Based on the preprocessed data set (C i , I i ), construct a machine learning nonlinear regression model f(I; θ), use the gradient descent algorithm to optimize the model parameters, ensure that the model can accurately reflect the nonlinear relationship between the fluorescence signal intensity I and the ATP concentration C, and generate a trained nonlinear regression model f trained (I;θ); The parameter update formula of the gradient descent algorithm is as follows: Among them, θ j is the jth parameter in the model parameter vector θ, θ=(θ0,θ1,…,θ n ) represents the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, and I ij is the jth eigenvalue of the i-th sample; Using the trained nonlinear regression model f trained (I; θ), the intensity of the pure target fluorescence signal I clean Analyze the intensity of the pure target fluorescence signal I clean As input, the concentration value C of ATP in the sample to be tested is calculated by model prediction pred , get the initial concentration value C pred ; The prediction formula is as follows: C pred =f trained (I clean ;θ) The model parameters are optimized through multiple iterations to verify the accuracy and reliability of the model prediction and generate the ATP concentration value C in the sample to be tested. final , ensuring the accuracy and reliability of the test results; The optimization formula is as follows: Among them, θ opt is the optimized model parameter vector; C i is a standard solution of ATP with known concentration, I i is the corresponding fluorescence signal intensity, (C i , I i ) is the preprocessed data set, θ=(θ0,θ1,…,θ n ) is the parameter vector of the model, α is the learning rate, m is the number of samples in the dataset, I ij is the jth eigenvalue of the ith sample, f(I;θ) is the nonlinear regression model, f trained (I; θ) is the trained nonlinear regression model, I clean is the intensity of the pure target fluorescence signal, C pred is the initial concentration value, C final is the final concentration of ATP in the sample to be tested, θ opt is the optimized model parameter vector.

10. A highly sensitive ATP fluorescence detection system, characterized in that: include: A configuration module is used to determine the optimal ratio of luciferase to luciferin, configure a reaction mixture containing the optimal ratio of luciferase to luciferin, and ensure that a significantly enhanced fluorescence signal is generated under low ATP concentration conditions; A mixing module is used to control the reaction environment based on the reaction mixture by using microfluidic chip technology, so that the reaction mixture and the sample to be tested are fully mixed in a dark environment, and the reaction temperature is adjusted by using dynamic temperature control technology to improve the reaction speed and the stability of the fluorescence signal, so as to obtain a mixed reaction system; A separation and measurement module is used to capture the fluorescence signal generated by the mixed reaction system within a set time using a high-sensitivity photomultiplier tube, and process the fluorescence signal using a digital phase-locked amplification technology and an adaptive noise elimination algorithm to separate and generate a pure target fluorescence signal, and simultaneously measure the intensity of the pure target fluorescence signal; The analysis and calculation module is used to analyze the intensity of the pure target fluorescence signal based on the intensity of the pure target fluorescence signal, use a machine learning nonlinear regression model obtained by training a large number of ATP standard solution data sets with different concentrations, and use a gradient descent algorithm to optimize model parameters. The concentration value of ATP in the sample to be tested is calculated and generated through model prediction to ensure the accuracy and reliability of the test results.

Images