PCR detection fluorescence intensity data processing system and method

Abstract

The invention relates to the technical field of biotechnology detection data processing, discloses a PCR detection fluorescence intensity data processing system and method, and aims to solve the problems of low signal-to-noise ratio of fluorescence signals, inaccurate baseline deduction and poor Ct value judgment consistency in the prior art. The method comprises the following steps: processing an original fluorescence signal through adaptive wavelet threshold noise reduction, dynamically determining and correcting a baseline interval, extracting amplification curve characteristic parameters, dynamically adjusting a fluorescence threshold line according to amplification efficiency, identifying an abnormal amplification mode by using a pre-trained deep learning model, and finally integrating results to generate a comprehensive detection report. By adopting the technical scheme, the signal-to-noise ratio of the fluorescence signal and the baseline deduction precision can be remarkably improved, the consistency of the Ct value is improved, multiple abnormal amplification modes are automatically detected, and a full-automatic processing flow is realized.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology detection data processing technology, specifically relating to a PCR detection fluorescence intensity data processing system and method. Background Technology

[0002] Polymerase chain reaction (PCR) technology, as a core detection method in the field of molecular biology, has been widely used in pathogen detection, gene expression analysis, and genetic disease screening. The accuracy of its detection results is highly dependent on the quality of fluorescence signal acquisition and processing. In real-time quantitative PCR, the system determines the initial amount of target nucleic acid by monitoring the change in fluorescence intensity in each amplification cycle. Therefore, the stability, signal-to-noise ratio, and baseline correction accuracy of the fluorescence data directly determine the reliability of the quantitative results.

[0003] However, existing PCR fluorescence intensity data processing technologies still face multiple challenges: First, the raw fluorescence signal is easily affected by fluctuations in the instrument's optical system, temperature differences between reaction wells, and uneven reagent batches, resulting in high data noise levels; second, traditional baseline correction methods often use fixed cyclic intervals or simple linear fitting, which are difficult to adapt to the dynamic characteristics of amplification curves of different samples and are prone to introducing systematic biases; third, the setting of the threshold generally relies on manual experience or globally unified standards, and cannot be adaptively adjusted for the amplification efficiency of individual reaction wells, affecting the accuracy of Ct value determination.

[0004] Finally, existing systems lack intelligent identification and correction mechanisms for abnormal amplification curves (such as non-specific amplification, primer dimers, or amplification inhibition), which can easily lead to false positive or false negative results in complex clinical sample testing. These problems severely restrict the reliability and automation level of PCR testing in high-throughput, high-precision applications, necessitating an integrated processing system and method capable of intelligent noise reduction of fluorescence data, dynamic baseline correction, adaptive threshold determination, and abnormal curve identification. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a PCR detection fluorescence intensity data processing system and method, which can effectively solve the problems in the background technology. To achieve the above objective, this invention provides the following technical solution: On the one hand, a PCR detection fluorescence intensity data processing system includes: The raw fluorescence signal acquisition module is used to receive multi-channel raw fluorescence intensity time-series data from the photoelectric detection unit of the PCR instrument, and to perform preliminary format conversion and timestamp alignment of the fluorescence signal of each reaction well. The signal preprocessing and noise filtering module is connected to the output of the original fluorescence signal acquisition module. It uses an adaptive wavelet threshold noise reduction algorithm to perform multi-scale decomposition and reconstruction of the original fluorescence signal to suppress high-frequency random noise and low-frequency baseline drift. The dynamic baseline correction module communicates with the signal preprocessing and noise filtering module. Based on the local weighted scatter smoothing algorithm, it dynamically determines the baseline interval in the early stage of the exponential growth of the amplification curve, and uses the nonlinear least squares method to fit the baseline trend to perform fluorescence background subtraction. The amplification curve feature extraction module receives baseline-corrected fluorescence data, calculates the first and second derivatives of fluorescence intensity for each cycle, and extracts the inflection point position, plateau fluorescence value, and slope characteristics of the exponential growth interval of the curve. An adaptive threshold determination module, coupled to the amplification curve feature extraction module, dynamically adjusts the fluorescence threshold line based on amplification efficiency and signal-to-noise ratio. The threshold level is associated with the specific amplification stage of each reaction well. The abnormal curve identification and classification module integrates a pre-trained convolutional neural network model to perform multi-dimensional analysis of the morphological features of the amplification curve and identify abnormal patterns including non-specific amplification, premature signal saturation, or amplification inhibition. The results integration and output module aggregates the processing results from each module, calculates the Ct value, and generates a comprehensive test report including confidence intervals and quality assessment indicators.

[0006] Preferably, the adaptive wavelet threshold denoising algorithm executed by the signal preprocessing and noise filtering module is as follows: The original fluorescence sequence was subjected to 5-level wavelet packet decomposition to obtain wavelet coefficients in different frequency bands; an adaptive threshold was calculated based on the statistical distribution characteristics of the coefficients at each decomposition level, and the calculation formula is as follows: in, The threshold for the j-th layer, Let N be the standard deviation estimate of the wavelet coefficients of the j-th layer, and N be the signal length. Wavelet coefficients above the threshold are shrunk using a soft threshold function, and coefficients below the threshold are set to zero. Finally, the denoised fluorescence signal is obtained through wavelet reconstruction.

[0007] Furthermore, the specific process by which the dynamic baseline correction module determines the baseline interval includes: Within the amplification cycle range of 1 to 15, the coefficient of variation of local fluorescence values ​​was calculated using a 3-cycle window. The interval where the coefficient of variation remained below a preset threshold of 0.05 for five consecutive cycles was defined as the stable baseline interval. Within the determined baseline interval, a locally weighted regression algorithm was used to smoothly fit the fluorescence values, with the following weighting function: Where wi is the weight of the i-th cycle, xi is the cycle number, xc is the current fitting point, and h is the bandwidth parameter with a value of 3; the fitted curve is used as the baseline and subtracted from the original fluorescence signal.

[0008] Furthermore, the dynamic adjustment mechanism of the adaptive threshold determination module includes: The amplification efficiency E is calculated based on four consecutive cycles of the exponential growth phase of the amplification curve, using the following formula: Where slope is the linear regression slope of the exponential growth phase; when the amplification efficiency E is in the range of 1.8 to 2.2, the threshold is set to 12 times the baseline fluorescence standard deviation; when the amplification efficiency E exceeds this range, the threshold coefficient is adjusted proportionally to ensure that the threshold line is always in the middle region of the exponential growth phase.

[0009] Preferably, the convolutional neural network model used in the abnormal curve identification and classification module has a multi-layer convolutional structure, and the input is a standardized amplification curve sequence; the first convolutional layer uses 32 filters of size 3×1 to extract local morphological features; the second convolutional layer uses 64 filters of size 3×1 to perform deep feature abstraction; the pooling layer uses max pooling to reduce the feature dimension; the fully connected layer maps the features to a 128-dimensional latent space; the output layer uses the Softmax function to calculate the probability distribution of the curve belonging to four categories: normal, non-specific amplification, premature signal saturation, or amplification inhibition.

[0010] Furthermore, the amplification curve feature extraction module also calculates multiple kinetic parameters of the curve, including the duration of the exponential growth phase, the maximum fluorescence intensity growth rate, and the area under the curve; these parameters, together with the Ct value, constitute a multidimensional feature vector, which is used for subsequent quality assessment and result verification.

[0011] On the other hand, a method for processing PCR detection fluorescence intensity data includes the following specific steps: Step S110: Collect raw fluorescence intensity time-series data from multiple reaction wells using the photoelectric detection system of the PCR instrument, and perform preliminary time alignment and format standardization on the data. Step S120: Adaptive wavelet threshold denoising is performed on the original fluorescence signal. Multi-scale decomposition and coefficient shrinkage effectively suppress instrument noise and environmental interference. Step S130: Based on local coefficient of variation analysis, the baseline interval of the amplification curve is dynamically determined, and the local weighted regression algorithm is used to fit the baseline trend for fluorescence background correction. Step S140: Extract multiple characteristic parameters from the corrected amplification curve, including fluorescence derivative characteristics, inflection point position, and kinetic characteristics of the exponential growth phase. Step S150: Dynamically adjust the fluorescence threshold line according to the specific amplification efficiency of each reaction well to ensure the accuracy and consistency of Ct value determination; Step S160: Use a pre-trained deep learning model to perform multi-class classification of the amplification curve morphology, and automatically identify and label abnormal amplification patterns. Step S170: Integrate all processing results, calculate the final Ct value, and generate a test report including quality indicators and confidence intervals.

[0012] In summary, this application includes at least one of the following beneficial technical effects: 1. By combining adaptive wavelet threshold denoising technology with dynamic baseline correction technology, high-frequency random noise and low-frequency baseline drift are effectively suppressed, significantly improving the signal-to-noise ratio of fluorescence signals. At the same time, the baseline interval is dynamically determined based on local coefficient of variation analysis, and the baseline is fitted using a local weighted regression algorithm, which greatly improves the accuracy of baseline subtraction, thereby reducing the systematic bias introduced by background fluorescence fluctuations.

[0013] 2. By establishing an adaptive threshold determination mechanism based on amplification efficiency, the limitations of the traditional fixed threshold method are overcome. The fluorescence threshold line is dynamically adjusted according to the efficiency of the specific amplification stage of each reaction well, ensuring that the threshold line is always in the middle region of the exponential growth stage. This significantly improves the consistency and accuracy of Ct value determination among samples with different amplification efficiencies.

[0014] 3. By integrating a pre-trained convolutional neural network model, an intelligent recognition system capable of multi-dimensional analysis of amplification curve morphology was constructed. It can automatically and accurately detect various abnormal amplification modes such as non-specific amplification, premature signal saturation, and amplification inhibition, significantly reducing the risk of false positives and false negatives in complex sample detection and enhancing the reliability of the results.

[0015] 4. By integrating multiple processes such as signal acquisition, noise reduction, baseline correction, feature extraction, threshold determination, anomaly identification, and result integration into a single processing flow, the system achieves fully automated processing from raw data to the final report, reducing the need for manual intervention. It is particularly suitable for high-throughput PCR testing scenarios in clinical diagnosis and scientific research, effectively improving testing efficiency and standardization. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall technical solution architecture proposed in this invention.

[0017] Figure 2 This is a schematic diagram of the core principle framework of adaptive wavelet threshold denoising and dynamic baseline correction in this invention.

[0018] Figure 3This is a flowchart illustrating the logical flow of amplification curve feature extraction and adaptive threshold determination in this invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific embodiments according to the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments.

[0020] Example 1 In the context of high-throughput PCR detection in clinical molecular diagnostic laboratories, this embodiment provides a PCR detection fluorescence intensity data processing system, see [link to documentation]. Figure 1 and Figure 2 The system includes a raw fluorescence signal acquisition module, a signal preprocessing and noise filtering module, a dynamic baseline correction module, an amplification curve feature extraction module, an adaptive threshold determination module, an abnormal curve identification and classification module, and a result integration and output module. Specifically: The A100 raw fluorescence signal acquisition module is used to receive multi-channel raw fluorescence intensity time-series data from the photoelectric detection unit of the PCR instrument, and to perform preliminary format conversion and timestamp alignment of the fluorescence signal of each reaction well.

[0021] The A110 raw fluorescence signal acquisition module establishes a physical connection with the photoelectric detection unit of the real-time fluorescence quantitative PCR instrument through an optical fiber interface, and continuously receives the raw fluorescence intensity time series data of each detection channel of the 96-well reaction plate at a sampling frequency of 0.1 seconds per cycle.

[0022] The A120 original fluorescence signal acquisition module has a built-in multi-channel signal demodulation circuit that converts the analog voltage signal output by the photomultiplier tube into a 16-bit digital signal to form a standardized data format unit. At the same time, it uses a hardware clock synchronization mechanism to align the fluorescence acquisition time point of each reaction well with millisecond-level precision, ensuring that the cyclic data between different channels are strictly comparable.

[0023] The A130 data format standardization unit converts raw binary stream data into a structured data table. The structured data table includes five core fields: well number, detection channel identifier, cycle number, fluorescence intensity value, and timestamp. The integrity of data transmission is verified by CRC checksum.

[0024] The A200 signal preprocessing and noise filtering module is connected to the output of the original fluorescence signal acquisition module. It uses an adaptive wavelet threshold denoising algorithm to perform multi-scale decomposition and reconstruction of the original fluorescence signal, suppressing high-frequency random noise and low-frequency baseline drift.

[0025] The A210 signal preprocessing and noise filtering module receives standardized fluorescence sequences from the original fluorescence signal acquisition module via a high-speed data bus. The core processing unit of the signal preprocessing and noise filtering module adopts an adaptive wavelet threshold noise reduction algorithm. First, it performs 5-level wavelet packet decomposition on the original fluorescence sequence of each reaction well. Then, it decomposes the signal into different frequency bands through the Daubechies wavelet basis function to obtain a wavelet coefficient matrix including approximation coefficients and detail coefficients.

[0026] During the coefficient processing stage, the system calculates an adaptive threshold based on the statistical distribution characteristics of the coefficients in each decomposition layer. The specific calculation formula is as follows: in, The threshold for the j-th layer, is the standard deviation estimate of the wavelet coefficients of the j-th layer, and N is the signal length.

[0027] For wavelet coefficients above a threshold, a soft thresholding function is used for nonlinear shrinkage, with the shrinkage magnitude proportional to the absolute value of the coefficient. For coefficients below the threshold, they are directly set to zero to eliminate random noise. Finally, the signal is reconstructed through inverse wavelet transform. Boundary extension techniques are used during the reconstruction process to prevent edge effects, and the output fluorescence sequence with improved signal-to-noise ratio is sent to the downstream processing unit.

[0028] The A300, a dynamic baseline correction module, communicates with the signal preprocessing and noise filtering module. Based on a local weighted scatter smoothing algorithm, it dynamically determines the baseline interval in the early stage of the exponential growth of the amplification curve and uses nonlinear least squares method to fit the baseline trend for fluorescence background subtraction.

[0029] The A310 dynamic baseline correction module establishes a data exchange channel with the signal preprocessing and noise filtering module through shared memory. The baseline interval identification unit of the dynamic baseline correction module calculates the coefficient of variation of local fluorescence values ​​in the range of amplification cycles 1 to 15 with a sliding window of 3 cycles and a window sliding step size of 1 cycle.

[0030] A320, the coefficient of variation is calculated as the ratio of the standard deviation to the mean. When the coefficient of variation values ​​for five consecutive cycles are all below the preset threshold of 0.05, the system determines that the interval is a stable baseline interval.

[0031] In the A330, within a defined baseline interval, the dynamic baseline correction module includes a baseline fitting unit. This unit uses a locally weighted regression algorithm to smoothly fit the fluorescence values, with the following weighting function: in Let be the weight of the i-th loop. For the cycle number, Here is the current fitting point, and h is the bandwidth parameter with a value of 3.

[0032] During the fitting process, the local polynomial coefficients are solved by iterative reweighted least squares method. The obtained fitted curve is used as the baseline signal and subtracted from the original fluorescence signal to output the baseline-corrected relative fluorescence intensity data.

[0033] The A400 amplification curve feature extraction module receives baseline-corrected fluorescence data, calculates the first and second derivatives of fluorescence intensity for each cycle, and extracts the inflection point position, plateau fluorescence value, and slope characteristics of the exponential growth interval of the curve.

[0034] The amplification curve feature extraction module obtains the baseline-corrected fluorescence data matrix through direct memory access. The amplification curve feature extraction module includes a differential calculation unit, an inflection point detection unit, an exponential growth interval identification unit, and a plateau period determination unit.

[0035] The A410 differential calculation unit uses the central difference method to calculate the first derivative of fluorescence intensity for each cycle. The difference step size is set to 2 cycles to balance calculation accuracy and noise sensitivity. The second derivative calculation is based on the first derivative result and the central difference method is applied again.

[0036] The A420 inflection point detection unit determines the inflection point of the amplification curve by analyzing the zero-crossing point of the second derivative, and records the fluorescence intensity value at the inflection point as a characteristic parameter.

[0037] The A430 exponential growth interval identification unit determines the linear interval after logarithmic transformation of fluorescence intensity through sliding window regression analysis, and calculates the regression slope of the exponential growth interval as the exponential growth rate feature.

[0038] A430, the plateau phase determination unit monitors the interval where the fluorescence intensity change rate is continuously lower than the threshold of 0.01, and takes the average fluorescence intensity of the last third of this interval as the plateau phase fluorescence value.

[0039] In addition, the A440 amplification curve feature extraction module also calculates multiple kinetic parameters of the curve, including the duration of the exponential growth phase, the maximum fluorescence intensity growth rate, and the area under the curve. These parameters, together with the Ct value, constitute a 128-dimensional feature vector for subsequent quality assessment and result verification.

[0040] The A500 adaptive threshold determination module, coupled to the amplification curve feature extraction module, dynamically adjusts the fluorescence threshold line based on amplification efficiency and signal-to-noise ratio. Its threshold level is associated with the specific amplification stage of each reaction well.

[0041] The adaptive threshold determination module establishes a loosely coupled connection with the amplification curve feature extraction module through a message queue. It includes an amplification efficiency calculation unit, a threshold dynamic adjustment unit, and a threshold line drawing unit.

[0042] The A510 amplification efficiency calculation unit calculates amplification efficiency based on four consecutive cycles of data from the exponential growth phase of the amplification curve. The specific formula is as follows: Where slope is the linear regression slope during the exponential growth phase.

[0043] The A520's threshold dynamic adjustment unit monitors the amplification efficiency E in real time. When E is between 1.8 and 2.2, the system sets the fluorescence threshold to 12 times the baseline fluorescence standard deviation. When E is below 1.8, the threshold coefficient is increased by 0.5 for every 0.1 unit decrease. When E is above 2.2, the threshold coefficient is decreased by 0.3 for every 0.1 unit increase.

[0044] The A530 threshold line plotting unit overlays the adjusted threshold level onto the amplification curve as a horizontal line, ensuring that the threshold line is always in the middle region of the exponential growth phase, providing a stable reference benchmark for Ct value calculation.

[0045] The A600's abnormal curve recognition and classification module integrates a pre-trained convolutional neural network model to perform multi-dimensional analysis of the morphological characteristics of amplification curves, identifying abnormal patterns including non-specific amplification, premature signal saturation, or amplification inhibition. The abnormal curve recognition and classification module integrates a convolutional neural network model pre-trained with 100,000 labeled samples. This model includes an input layer, a first convolutional layer, a second convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0046] A610, the input layer receives the normalized amplification curve sequence. The normalization process includes normalizing the fluorescence intensity to the 0-1 range and aligning the cycle number to 50 fixed points.

[0047] In the A620, the first convolutional layer uses 32 filters of size 3×1 to extract the local morphological features of the amplification curve through a sliding window with a stride of 1. The activation function is a ReLU nonlinear transformation.

[0048] In the A630, the second convolutional layer uses 64 filters of size 3×1 for deep feature abstraction, and maintains the feature map size unchanged through padding operations.

[0049] The A640 pooling layer uses max pooling with a kernel size of 2×1, reducing the feature dimension to 50% of the original size.

[0050] The A650 fully connected layer maps the pooled features to a 128-dimensional latent space and uses the Dropout mechanism to prevent overfitting.

[0051] The A660 output layer uses the Softmax function to calculate the probability distribution of curves belonging to four categories: normal, non-specific amplification, premature signal saturation, or amplification inhibition. The classification threshold is set to 0.85, and samples below this threshold are marked as pending verification.

[0052] The A700 results integration and output module aggregates the processing results from various modules, calculates the Ct value, and generates a comprehensive test report including confidence intervals and quality assessment indicators.

[0053] The results integration and output module aggregates intermediate results and final data from various processing modules through a database connection pool. This module includes a Ct value calculation unit, a quality assessment unit, and a report generation engine.

[0054] The A710 Ct value calculation unit uses the second derivative maximum value method to determine the amplification starting point, and combines it with the adaptive threshold line intersection method for cross-validation to output Ct value results with confidence intervals.

[0055] The A720 quality assessment unit integrates 12 indicators, including amplification efficiency, curve morphology score, and signal-to-noise ratio, and generates a quality score of 0-100 through a weighted scoring model.

[0056] The A730 report generation engine integrates raw data, processing procedures, result parameters, and quality assessments into a structured test report. The report format conforms to the requirements of the Clinical Laboratory Standards Association and supports PDF, XML, and JSON output formats. Digital signatures ensure the integrity and immutability of the report.

[0057] In its specific implementation, this embodiment also provides a method for processing PCR detection fluorescence intensity data. The system executes this processing method, see [link to relevant documentation]. Figure 3 As shown, the steps include: Step S110: Collect raw fluorescence intensity time-series data from multiple reaction wells using the photoelectric detection system of the PCR instrument. During the acquisition process, time alignment and format standardization are performed simultaneously to ensure data consistency and comparability.

[0058] Step S120: Adaptive wavelet threshold denoising is performed on the original fluorescence signal. Through multi-scale decomposition and coefficient shrinkage, instrument noise and environmental interference are effectively suppressed, and signal quality is improved.

[0059] Step S130: Based on local coefficient of variation analysis, the baseline interval of the amplification curve is dynamically determined, and a local weighted regression algorithm is used to fit the baseline trend to correct the fluorescence background and eliminate systematic bias.

[0060] Step S140: Extract multiple feature parameters from the corrected amplification curve, including fluorescence derivative characteristics, inflection point position, and kinetic characteristics of the exponential growth phase, to construct a complete curve feature description system.

[0061] In step S150, the fluorescence threshold line is dynamically adjusted according to the specific amplification efficiency of each reaction well, and the accuracy and consistency of Ct value determination are ensured through a real-time feedback mechanism.

[0062] Step S160: Use a pre-trained deep learning model to perform multi-class classification of the amplification curve morphology, automatically identify and label abnormal amplification patterns, and reduce the risk of misjudgment.

[0063] Step S170: Integrate all processing results, calculate the final Ct value, and generate a test report including quality indicators and confidence intervals, thus completing the fully automated data processing workflow.

[0064] Example 2 In the application scenario of mobile platforms for rapid pathogen detection, the system is optimized and configured to meet the specific needs of portable real-time quantitative PCR instruments.

[0065] The original fluorescence signal acquisition module is compatible with the communication protocols of various brands of portable devices. It receives fluorescence data through USB or Bluetooth interfaces. A resampling algorithm has been developed to address the differences in sampling frequencies between devices, ensuring data standardization.

[0066] The signal preprocessing and noise filtering module adjusts the wavelet decomposition level to 4 to address the high noise levels in portable devices. The standard deviation estimation in the threshold calculation formula adopts a robust statistical method to avoid outlier interference.

[0067] The dynamic baseline correction module shortens the baseline recognition interval to cycles 1 to 10 for rapid amplification reagents, and adjusts the coefficient of variation threshold to 0.08 to accommodate more severe initial fluctuations.

[0068] The amplification curve feature extraction module, designed for the characteristics of rapid amplification curve morphology, shortens the differential calculation step to one loop, thereby improving the temporal resolution of feature extraction.

[0069] The adaptive threshold determination module, designed for the generally high amplification efficiency of rapid amplification reagents, extends the upper limit of the normal efficiency range to 2.4, and accordingly optimizes the threshold adjustment coefficient to adjust by 0.2 coefficients for every 0.1 unit change.

[0070] The abnormal curve identification and classification module loads a lightweight convolutional neural network model, halving the number of convolutional kernels to meet the computing resource limitations of mobile devices, while maintaining a classification accuracy of no less than 95%.

[0071] The results integration and output module adds an instant messaging interface, which can upload test results to the cloud database in real time via 4G / 5G network, and supports pushing brief test reports through mobile applications.

[0072] The entire system runs on an embedded hardware platform with a multi-threaded architecture. The data acquisition and preprocessing threads have the highest priority to ensure real-time performance. The feature extraction and classification threads use task queue management to balance the computational load. The result output thread uses an asynchronous communication mechanism to avoid blocking the main processing flow. The system performs self-test calibration upon startup, including fluorescence intensity benchmark calibration, temperature sensor calibration, and communication link testing, to ensure operational reliability in mobile environments.

[0073] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.

[0074] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment includes only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A PCR detection fluorescence intensity data processing system, characterized in that, include: The raw fluorescence signal acquisition module is used to receive multi-channel raw fluorescence intensity time-series data from the photoelectric detection unit of the PCR instrument, and to perform format conversion and timestamp alignment of the fluorescence signal of each reaction well. The signal preprocessing and noise filtering module is connected to the original fluorescence signal acquisition module and uses an adaptive wavelet threshold noise reduction algorithm to denoise the original fluorescence signal. The dynamic baseline correction module is connected to the signal preprocessing and noise filtering module. It dynamically determines the baseline interval and fits the baseline trend to perform fluorescence background subtraction based on the local weighted scatter smoothing algorithm. The amplification curve feature extraction module receives baseline-corrected fluorescence data, calculates the fluorescence intensity derivative, and extracts the inflection point position, plateau fluorescence value, and slope characteristics of the exponential growth interval of the curve. An adaptive threshold determination module and a coupled amplification curve feature extraction module dynamically adjust the fluorescence threshold line based on amplification efficiency and signal-to-noise ratio. The abnormal curve identification and classification module integrates a pre-trained convolutional neural network model to analyze the shape of the amplification curve in order to identify abnormal amplification patterns. The results integration and output module aggregates the processing results from each module, calculates the Ct value, and generates a comprehensive test report including confidence intervals and quality assessment indicators.

2. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The adaptive wavelet threshold denoising algorithm executed by the signal preprocessing and noise filtering module is as follows: The original fluorescence sequence was decomposed into five layers of wavelet packets to obtain wavelet coefficients in different frequency bands. An adaptive threshold was calculated based on the statistical distribution characteristics of the coefficients in each decomposition layer. Wavelet coefficients above the threshold were shrunk using a soft threshold function, while coefficients below the threshold were set to zero. Finally, the denoised fluorescence signal was obtained through wavelet reconstruction.

3. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The specific process by which the dynamic baseline correction module determines the baseline interval includes: Within the amplification cycle range of 1 to 15, the coefficient of variation of local fluorescence values ​​is calculated using 3 cycles as a window; the interval where the coefficient of variation is below the preset threshold of 0.05 for 5 consecutive cycles is determined as the stable baseline interval; within the determined baseline interval, the fluorescence values ​​are smoothly fitted using a local weighted regression algorithm; the fitted curve is used as the baseline and subtracted from the original fluorescence signal.

4. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The dynamic adjustment mechanism of the adaptive threshold determination module includes: The amplification efficiency E is calculated based on four consecutive cycles of the exponential growth phase of the amplification curve. When the amplification efficiency E is in the range of 1.8 to 2.2, the threshold is set to 12 times the baseline fluorescence standard deviation. When the amplification efficiency E exceeds this range, the threshold coefficient is adjusted proportionally to ensure that the threshold line is always in the middle region of the exponential growth phase.

5. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The abnormal curve identification and classification module uses a convolutional neural network model with a multi-layer convolutional structure. The input is a standardized amplification curve sequence. The first convolutional layer uses 32 filters of size 3×1 to extract local morphological features. The second convolutional layer uses 64 filters of size 3×1 to perform deep feature abstraction. The pooling layer uses max pooling to reduce the feature dimension. The fully connected layer maps the features to a 128-dimensional latent space. The output layer uses the Softmax function to calculate the probability distribution of the curve belonging to four categories: normal, non-specific amplification, premature signal saturation, or amplification inhibition.

6. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The amplification curve feature extraction module also calculates multiple kinetic parameters of the curve, including the duration of the exponential growth phase, the maximum fluorescence intensity growth rate, and the area under the curve; these parameters, together with the Ct value, constitute a multidimensional feature vector, which is used for subsequent quality assessment and result verification.

7. The PCR detection fluorescence intensity data processing system according to claim 1, characterized in that, The result integration and output module collects intermediate results and final data from each processing module through a database connection pool; the Ct value calculation unit uses the second derivative maximum value method to determine the amplification starting point, and combines it with the adaptive threshold line intersection method for cross-validation to output Ct value results with confidence intervals. The quality assessment unit integrates amplification efficiency, curve morphology score, and signal-to-noise ratio index to generate a quality score through a weighted scoring model; the report generation engine integrates raw data, processing procedures, result parameters, and quality assessment into a structured test report.

8. The PCR detection fluorescence intensity data processing system according to claim 2, characterized in that, When performing wavelet packet decomposition, the signal preprocessing and noise filtering module uses the Daubechies wavelet basis function to decompose the signal into different frequency bands, and employs boundary extension techniques during wavelet reconstruction to prevent edge effects.

9. The PCR detection fluorescence intensity data processing system according to claim 3, characterized in that, When the dynamic baseline correction module uses the local weighted regression algorithm for smooth fitting, it uses the iterative reweighted least squares method to solve the local polynomial coefficients, and the sliding window has a sliding step size of 1 cycle.

10. A method for processing PCR detection fluorescence intensity data, characterized in that, include: Step S110: Collect raw fluorescence intensity time-series data from multiple reaction wells using the photoelectric detection system of the PCR instrument, and perform preliminary time alignment and format standardization on the data. Step S120: Adaptive wavelet threshold denoising is performed on the original fluorescence signal. Multi-scale decomposition and coefficient shrinkage effectively suppress instrument noise and environmental interference. Step S130: Based on local coefficient of variation analysis, the baseline interval of the amplification curve is dynamically determined, and the local weighted regression algorithm is used to fit the baseline trend for fluorescence background correction. Step S140: Extract multiple characteristic parameters from the corrected amplification curve, including fluorescence derivative characteristics, inflection point position, and kinetic characteristics of the exponential growth phase. Step S150: Dynamically adjust the fluorescence threshold line according to the specific amplification efficiency of each reaction well to ensure the accuracy and consistency of Ct value determination; Step S160: Use a pre-trained deep learning model to perform multi-class classification of the amplification curve morphology, and automatically identify and label abnormal amplification patterns. Step S170: Integrate all processing results, calculate the final Ct value, and generate a test report including quality indicators and confidence intervals.

Images