A method and system for processing methylmalonic acid mass spectrometry signals and test results

Abstract

The application provides a methylmalonic acid mass spectrum signal and detection result processing method and system. The method comprises the following steps: obtaining original mass spectrum data and preprocessing, using wavelet transform to perform multi-scale decomposition on the preprocessed mass spectrum signal, obtaining wavelet coefficients of different scales, inputting a preset deep residual network for denoising processing, obtaining enhanced wavelet coefficients, performing signal reconstruction through wavelet inverse transform, obtaining a denoised mass spectrum signal, performing peak detection and integration, obtaining a methylmalonic acid quantitative result, inputting a preset anomaly detection model to identify whether there is a potential measurement deviation, and if there is, performing adaptive correction according to historical quality control data; the application fuses wavelet transform and deep residual network, combines anomaly detection and adaptive correction mechanism, removes background noise, enhances target peak signal, improves methylmalonic acid quantitative precision, reduces false positive and false negative risk, solidifies the data set, feature extraction rule, isolated forest parameter, score calculation logic and anomaly judgment rule, and realizes the processing of methylmalonic acid mass spectrum signal and detection result.

Description

Technical Field

[0001] This application relates to the field of mass spectrometry detection signal processing technology, and more specifically, to a method and system for processing methylmalonic acid mass spectrometry signals and detection results. Background Technology

[0002] Methylmalonic acid (MMA) is an important intermediate product in the propionyl-CoA metabolic pathway in the body, and its concentration level in blood or urine is a key biomarker for diagnosing inherited metabolic diseases such as methylmalonic acidemia. Liquid chromatography-tandem mass spectrometry (LC-MS / MS) has become the mainstream technique for the quantitative detection of MMA due to its high sensitivity, high specificity, and multiple reaction monitoring capabilities. However, in actual detection, LC-MS / MS technology faces the following prominent problems: First, the biological sample matrix is ​​complex, containing matrix interferences such as phospholipids and salts. These interferences generate background noise and matrix effects during mass spectrometry detection, causing the target peak signal to be masked or distorted. Second, in the low concentration region, the target ion signal of MMA is weak, resulting in a low signal-to-noise ratio, making it difficult for conventional signal processing methods to effectively distinguish the true signal from random noise. Third, factors such as instrument response drift between different batches and column aging introduce systematic errors, making it difficult to guarantee the consistency of quantitative results across batches. Existing mass spectrometry signal processing methods mainly employ traditional digital filtering techniques such as moving average filtering and Savitzky-Golay smoothing filtering. While these methods are effective in handling high-frequency random noise, they have the following limitations: First, traditional filtering methods are usually based on fixed parameters and cannot adaptively adjust according to signal characteristics, which can easily cause distortion or broadening of the target peak signal while removing noise. Second, they lack the ability to specifically enhance weak signals at low concentrations, and the risk of false positives and false negatives in quantitative results increases significantly when the MMA concentration is close to the detection limit. Third, they lack the ability to detect anomalies across batches and adaptively correct errors, making it difficult to cope with systematic errors caused by instrument drift and matrix variations.

[0003] Therefore, there is an urgent need for a signal processing method that can integrate the multi-scale signal analysis capabilities of wavelet transform with the nonlinear feature extraction capabilities of deep learning, in order to automatically remove background noise, enhance target peak signals, identify potential measurement biases and perform adaptive correction, solidify datasets, feature extraction rules, isolated forest parameters, score calculation logic and anomaly judgment rules, and improve the accuracy and reliability of quantitative MMA detection.

[0004] Effective technical solutions are urgently needed to address the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for processing methylmalonic acid mass spectrometry signals and detection results. By fusing wavelet transform and deep residual network, combined with anomaly detection and adaptive correction mechanisms, background noise can be removed, target peak signals can be enhanced, the quantitative accuracy of methylmalonic acid can be improved, and the risk of false positives and false negatives can be reduced. The data set, feature extraction rules, isolated forest parameters, score calculation logic and anomaly judgment rules are solidified to achieve the processing of methylmalonic acid mass spectrometry signals and detection results.

[0006] Firstly, this application provides a method for processing methylmalonic acid mass spectrometry signals and detection results, including the following steps:

[0007] The raw mass spectrometry data and historical quality control data of liquid chromatography-tandem mass spectrometry were acquired. The raw mass spectrometry data included chromatographic retention time, mass-to-charge ratio and corresponding ion intensity signals. Baseline correction and normalization preprocessing were performed to obtain preprocessed mass spectrometry signals.

[0008] Based on the historical quality control data, wavelet transform is used to perform multi-scale decomposition on the preprocessed mass spectrometry signal to obtain wavelet coefficients at different scales.

[0009] The wavelet coefficients are input into a preset deep residual network for denoising to obtain enhanced wavelet coefficients;

[0010] The enhanced wavelet coefficients are reconstructed using inverse wavelet transform to obtain a denoised mass spectrum signal.

[0011] Peak detection and integration were performed on the denoised mass spectrometry signal to obtain the quantitative results of methylmalonic acid;

[0012] The quantitative result of methylmalonic acid is input into a preset anomaly detection model to obtain the potential measurement deviation identification result. If there is a potential measurement deviation, the quantitative result of methylmalonic acid is adaptively corrected according to historical quality control data to obtain the quantitative correction result of methylmalonic acid. If there is no potential measurement deviation, the quantitative result of methylmalonic acid is output.

[0013] Optionally, in the method for processing methylmalonic acid mass spectrometry signals and detection results described in this application, the step of performing multi-scale decomposition of the preprocessed mass spectrometry signal using wavelet transform based on the historical quality control data to obtain wavelet coefficients at different scales includes:

[0014] The signal-to-noise ratio (SNR) of historical mass spectrometry signals in the historical quality control data is calculated and analyzed, and the number of decomposition layers that maximizes the SNR is taken as the number of decomposition layers for the current batch.

[0015] Based on the number of decomposition levels, the preprocessed mass spectrum signal is subjected to multi-level discrete wavelet decomposition using Daubechies wavelet basis functions to obtain approximation coefficients and detail coefficients. Wavelet coefficients of different scales are then constructed based on the approximation coefficients and detail coefficients.

[0016] Optionally, in the method for processing methylmalonic acid mass spectrometry signals and detection results described in this application, the training process of the preset deep residual network includes:

[0017] Obtain the standard mass spectrometry signal corresponding to the methylmalonic acid standard sample with a known concentration from the historical quality control data;

[0018] Extract the standard wavelet coefficients of the standard mass spectrum signal after wavelet transform multi-scale decomposition;

[0019] The target wavelet coefficients corresponding to the labeled denoised target signal are used as labels, wherein the denoised target signal is obtained by superimposing and averaging the mass spectrometry signals of repeated measurements in the historical quality control data;

[0020] Using the standard wavelet coefficients as training input features and the target wavelet coefficients as output, the network parameters of the preset deep residual network are trained through the backpropagation algorithm and the mean square error loss function to obtain the preset deep residual network.

[0021] Optionally, in the method for processing methylmalonic acid mass spectrometry signals and detection results described in this application, the step of performing peak detection and integration on the denoised mass spectrometry signal to obtain methylmalonic acid quantitative results includes:

[0022] The target peak region corresponding to methylmalonic acid is obtained by locating the target peak of the denoised mass spectrometry signal using a preset continuous wavelet transform peak detection algorithm.

[0023] The ion intensity signal of the chromatographic peak region is integrated to obtain the peak area of ​​methylmalonic acid.

[0024] The quantitative results of methylmalonic acid were obtained by analyzing the peak area of ​​methylmalonic acid and the preset standard curve.

[0025] Optionally, in the method for processing methylmalonic acid mass spectrometry signals and detection results described in this application, the construction process of the preset anomaly detection model includes:

[0026] The historical methylmalonic acid quantitative results of multiple batches in the historical quality control data are obtained to form a model training sample set;

[0027] Extract the statistical distribution characteristics of each historical methylmalonic acid quantification result in the model training sample set;

[0028] Based on the statistical distribution characteristics, an isolated forest is constructed. For each isolated tree, a feature dimension and a random split value on that dimension are recursively selected to divide the sample space into two subspaces until only one sample remains in the subspace or the maximum depth is reached. The path length of each sample is recorded.

[0029] The mean of the path lengths is calculated to obtain the average path length of the sample. Anomaly scores are obtained by calculating and analyzing the average path length. If the anomaly score is greater than the preset anomaly score threshold, it is identified as a potential measurement bias. The dataset, feature extraction rules, isolated forest parameters, score calculation logic and anomaly judgment rules are solidified to construct a preset anomaly detection model.

[0030] Optionally, in the method for processing methylmalonic acid mass spectrometry signals and detection results described in this application, the step of adaptively correcting the methylmalonic acid quantification results based on historical quality control data to obtain methylmalonic acid quantification correction results includes:

[0031] Obtain the measured values ​​of quality control samples and the corresponding standard values ​​of quality control samples from historical quality control data. The measured values ​​of quality control samples include the measured values ​​of low-concentration quality control samples and the measured values ​​of high-concentration quality control samples.

[0032] By comparing and analyzing the measured values ​​of the low-concentration quality control samples and the high-concentration quality control samples with the standard values ​​of the quality control samples, the deviation rates of the low-concentration quality control samples and the high-concentration quality control samples are obtained.

[0033] The batch correction factor is obtained by weighting and averaging the deviation rates of the low-concentration quality control samples and the high-concentration quality control samples.

[0034] The quantitative result of methylmalonic acid is corrected according to the batch correction factor to obtain the corrected quantitative result of methylmalonic acid.

[0035] Secondly, this application provides a system for processing methylmalonic acid mass spectrometry signals and detection results, the system comprising:

[0036] The data acquisition module is used to acquire the raw mass spectrometry data of the liquid chromatography-tandem mass spectrometry and output the raw mass spectrometry data to the preprocessing module;

[0037] The preprocessing module is used to perform baseline correction and normalization preprocessing on the raw mass spectrometry data to obtain a preprocessed mass spectrometry signal;

[0038] The wavelet transform module is used to perform multi-scale decomposition of the preprocessed mass spectrum signal using wavelet transform to obtain wavelet coefficients at different scales.

[0039] A deep residual network module is used to input the wavelet coefficients into a preset deep residual network for denoising processing to obtain enhanced wavelet coefficients;

[0040] The signal reconstruction module is used to reconstruct the signal from the enhanced wavelet coefficients by inverse wavelet transform to obtain a denoised mass spectrum signal.

[0041] The quantitative calculation module is used to perform peak detection and integration on the denoised mass spectrometry signal to obtain the quantitative result of methylmalonic acid;

[0042] An anomaly detection module is used to input the quantitative results of methylmalonic acid into a preset anomaly detection model to identify whether there is a potential measurement bias in the quantitative results of methylmalonic acid.

[0043] An adaptive correction module is used to adaptively correct the quantitative result of methylmalonic acid based on historical quality control data when potential measurement deviations are identified, so as to obtain the corrected quantitative result of methylmalonic acid.

[0044] Optionally, the methylmalonic acid mass spectrometry signal and detection result processing system described in this application further includes: a memory and a processor, wherein the memory includes a program for processing methylmalonic acid mass spectrometry signals and detection results, and when the program for processing methylmalonic acid mass spectrometry signals and detection results is executed by the processor, it implements the following steps:

[0045] The raw mass spectrometry data and historical quality control data of liquid chromatography-tandem mass spectrometry were acquired. The raw mass spectrometry data included chromatographic retention time, mass-to-charge ratio and corresponding ion intensity signals. Baseline correction and normalization preprocessing were performed to obtain preprocessed mass spectrometry signals.

[0046] Based on the historical quality control data, wavelet transform is used to perform multi-scale decomposition on the preprocessed mass spectrometry signal to obtain wavelet coefficients at different scales.

[0047] The wavelet coefficients are input into a preset deep residual network for denoising to obtain enhanced wavelet coefficients;

[0048] The enhanced wavelet coefficients are reconstructed using inverse wavelet transform to obtain a denoised mass spectrum signal.

[0049] Peak detection and integration were performed on the denoised mass spectrometry signal to obtain the quantitative results of methylmalonic acid;

[0050] The quantitative result of methylmalonic acid is input into a preset anomaly detection model to obtain the potential measurement deviation identification result. If there is a potential measurement deviation, the quantitative result of methylmalonic acid is adaptively corrected according to historical quality control data to obtain the quantitative correction result of methylmalonic acid. If there is no potential measurement deviation, the quantitative result of methylmalonic acid is output.

[0051] Optionally, in the methylmalonic acid mass spectrometry signal and detection result processing system described in this application, the step of performing multi-scale decomposition of the preprocessed mass spectrometry signal using wavelet transform based on the historical quality control data to obtain wavelet coefficients at different scales includes:

[0052] The signal-to-noise ratio (SNR) of historical mass spectrometry signals in the historical quality control data is calculated and analyzed, and the number of decomposition layers that maximizes the SNR is taken as the number of decomposition layers for the current batch.

[0053] Based on the number of decomposition levels, the preprocessed mass spectrum signal is subjected to multi-level discrete wavelet decomposition using Daubechies wavelet basis functions to obtain approximation coefficients and detail coefficients. Wavelet coefficients of different scales are then constructed based on the approximation coefficients and detail coefficients.

[0054] Optionally, in the methylmalonic acid mass spectrometry signal and detection result processing system described in this application, the training process of the preset deep residual network includes:

[0055] Obtain the standard mass spectrometry signal corresponding to the methylmalonic acid standard sample with a known concentration from the historical quality control data;

[0056] Extract the standard wavelet coefficients of the standard mass spectrum signal after wavelet transform multi-scale decomposition;

[0057] The target wavelet coefficients corresponding to the labeled denoised target signal are used as labels, wherein the denoised target signal is obtained by superimposing and averaging the mass spectrometry signals of repeated measurements in the historical quality control data;

[0058] Using the standard wavelet coefficients as training input features and the target wavelet coefficients as output, the network parameters of the preset deep residual network are trained through the backpropagation algorithm and the mean square error loss function to obtain the preset deep residual network.

[0059] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0060] (1) This invention achieves adaptive removal of background noise and effective enhancement of target peak signals in mass spectrometry signals by integrating the multi-scale signal decomposition capability of wavelet transform with the nonlinear feature extraction capability of deep residual networks. Wavelet transform decomposes the signal into different frequency scales, enabling the separation of noise components and signal components in a multi-scale space; deep residual networks learn the complex nonlinear mapping relationship between noisy wavelet coefficients and denoised wavelet coefficients through residual learning structures. Compared with traditional threshold denoising methods, it can more effectively preserve the morphological features of the target peak signal and significantly improve the signal-to-noise ratio in low-concentration regions.

[0061] (2) The present invention uses a preset anomaly detection model to detect anomalies in the quantitative results of methylmalonic acid and uses the isolated forest algorithm to identify potential measurement deviations from the normal distribution. This can promptly detect abnormalities such as instrument drift and matrix interference, thereby reducing the risk of false positives and false negatives in the quantitative results.

[0062] (3) This invention achieves adaptive correction of methylmalonic acid quantification results by calculating batch correction factors based on historical quality control data, effectively eliminating cross-batch systematic errors and improving the consistency and traceability of quantification results. At the same time, by monitoring cross-batch correction factors, it can provide early warning of instrument performance degradation or reagent batch differences, providing data support for laboratory quality management.

[0063] (4) The preset deep residual network of the present invention is trained using historical quality control data covering different concentration levels, different matrix types and different instrument conditions, so that the trained network has the ability to generalize across batches and sample types, without having to retrain the model for each new batch, thus reducing the threshold for practical application. Attached Figure Description

[0064] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 A flowchart illustrating a method for processing methylmalonic acid mass spectrometry signals and detection results, provided in an embodiment of this application;

[0066] Figure 2 A flowchart illustrating the cross-batch correction factor monitoring mechanism of a method for processing methylmalonic acid mass spectrometry signals and detection results provided in this application embodiment;

[0067] Figure 3 This is a system block diagram of a methylmalonic acid mass spectrometry signal and detection result processing system provided in an embodiment of this application. Detailed Implementation

[0068] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0069] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0070] Example 1

[0071] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for processing methylmalonic acid mass spectrometry signals and detection results according to some embodiments of this application. The method for processing methylmalonic acid mass spectrometry signals and detection results includes the following steps:

[0072] S11. Obtain raw mass spectrometry data and historical quality control data of liquid chromatography-tandem mass spectrometry. The raw mass spectrometry data includes chromatographic retention time, mass-to-charge ratio and corresponding ion intensity signal. Baseline correction and normalization preprocessing are performed to obtain preprocessed mass spectrometry signal.

[0073] Specifically, samples containing methylmalonic acid were detected using liquid chromatography-tandem mass spectrometry (LC-MS / MS), and raw mass spectrometry data were acquired. The raw mass spectrometry data were represented as a two-dimensional matrix, with one dimension being the chromatographic retention time (in minutes) and the other dimension being the mass-to-charge ratio (m / z). The matrix element values ​​were the ion intensity signals at the corresponding positions (in counts per second). In this embodiment, a reversed-phase C18 column was used for liquid chromatography, with water and methanol (containing 0.1% formic acid) as the mobile phase. The gradient elution program was: 0-2 minutes, 5% methanol; 2-8 minutes, 5%-95% methanol linear gradient; 8-10 minutes, 95% methanol; 10-10.5 minutes, 95%-5% methanol; 10.5-12 minutes, 5% methanol equilibration. Tandem mass spectrometry was performed using an electrospray ionization source in negative ion mode and multiple reaction monitoring (MRM) mode. The precursor ion m / z of methylmalonic acid was 117.1, and the daughter ions m / z was 73.1 (quantitative ion) and 55.1 (qualitative ion). The sampling frequency of the raw mass spectrometry data was 10 Hz, i.e., one data point was acquired every 0.1 seconds.

[0074] Baseline correction employs an asymmetric least squares algorithm. Its core idea is to fit the signal baseline using asymmetric weighted least squares, ensuring the baseline follows the lower envelope of the signal while avoiding fitting signal peaks. The smoothing parameter λ of the asymmetric least squares algorithm is set to 10. 5 Up to 10 7 The asymmetric weight parameter p is set to 0.001 to 0.01, and the number of iterations is set to 10 to 20. After baseline correction, the fitted baseline is subtracted to obtain the baseline-corrected signal. The baseline-corrected signal is then normalized to scale the signal intensity to the [0,1] interval. The normalization formula is:

[0075] ;

[0076] in, This represents the intensity value of the baseline-corrected signal at time t. and These are the minimum and maximum values ​​of the signal after baseline correction, respectively. This is the preprocessed mass spectrometry signal after normalization. Normalization eliminates differences in signal intensity scale between different batches and sample types, providing a unified input scale for subsequent wavelet transform and deep residual network processing.

[0077] S12. Based on the historical quality control data, wavelet transform is used to perform multi-scale decomposition on the preprocessed mass spectrometry signal to obtain wavelet coefficients at different scales.

[0078] Specifically, the Daubechies wavelet basis functions (db4 to db8) are used to perform multi-level discrete wavelet decomposition on the preprocessed mass spectrum signal. The discrete wavelet decomposition process is as follows: In the j-th level decomposition, the low-pass decomposition filter h(k) and the high-pass decomposition filter g(k) are convolved with the approximation coefficients of the previous level, and downsampling is performed to obtain the approximation coefficients A_j and detail coefficients D_j of the j-th level. The decomposition formula is as follows:

[0079] ;

[0080] ;

[0081] in, Let be the approximation coefficients for the j-th layer. Let j be the detail coefficients of the j-th layer. These are the coefficients of the low-pass decomposition filter. These are the coefficients of the high-pass decomposition filter. For the (j-1)th level approximation coefficients, For sampling point index, This is the filter coefficient index.

[0082] The number of decomposition levels in the multi-level discrete wavelet decomposition is dynamically determined based on the signal-to-noise ratio (SNR) analysis of historical quality control data. Specifically, the SNR of the mass spectrometry signal in the historical quality control data is calculated at different decomposition levels (3 to 6 levels). The SNR calculation formula is as follows: ,in, For signal-to-noise ratio, For signal power, The noise power is used. The number of decomposition layers that maximizes the signal-to-noise ratio is selected as the number of decomposition layers for the current batch. In this embodiment, for methylmalonic acid mass spectrometry signals, the optimal number of decomposition layers is typically 4 to 5. Approximate coefficients are obtained after decomposition. (J represents the total number of decomposition levels) and the detail coefficients for each level, D1, D2, ..., D... J These coefficients constitute a multi-scale wavelet coefficient set.

[0083] S13. Input the wavelet coefficients into a preset deep residual network for denoising to obtain enhanced wavelet coefficients;

[0084] Specifically, the pre-defined deep residual network includes an input layer, multiple residual blocks, and an output layer. The input layer receives multi-scale wavelet coefficients as input features, and the input feature vector corresponding to each data point is a vector composed of the approximation coefficients and detail coefficients of that point at different scales. The number of residual blocks is 4 to 8, and each residual block includes a first convolutional layer, a first batch of normalization layers, a ReLU activation function layer, a second convolutional layer, a second batch of normalization layers, and a skip connection layer connected in sequence.

[0085] In each residual block, the first convolutional layer uses 1D convolution with a kernel size of 3 to 7 and a kernel count of 32 to 128, using the same padding method. The output of the first convolutional layer is fed into the second convolutional layer after passing through a first batch of normalization layers and a ReLU activation function. The second convolutional layer uses the same parameter settings as the first convolutional layer, and its output is added to the input of the residual block through a skip connection after passing through a second batch of normalization layers, and then output after passing through a ReLU activation function. The formula for the residual connection is:

[0086] ;

[0087] in, For the input of the residual block, For residual mapping function, This is the set of weight parameters for the convolutional layer. This is the output of the residual block. Through residual connections, the network directly learns the residual mapping instead of the complete mapping, effectively alleviating the gradient vanishing problem in deep networks and enabling the network to learn a more refined mapping relationship between noise features and signal features.

[0088] The output layer uses a 1D convolutional layer with one convolutional kernel and a linear activation function. The output dimension is the same as the input dimension, and the output is enhanced multi-scale wavelet coefficients.

[0089] The training process of the pre-defined deep residual network includes the following steps: (1) Obtain the mass spectrometry signal corresponding to the standard sample with known concentration in the historical quality control data, wherein the historical quality control data includes measurement records of quality control samples of multiple batches and multiple sample types, and each quality control sample has a known standard concentration value and the corresponding original mass spectrometry signal; (2) Perform the same preprocessing and wavelet transform on the original mass spectrometry signal as in steps S11-S12 to obtain the multi-scale wavelet coefficients for training; (3) Use the wavelet coefficients corresponding to the pure mass spectrometry signal under high signal-to-noise ratio conditions as the annotation target. (The mass spectrometry signals measured repeatedly in the historical quality control data are superimposed and averaged.) The wavelet coefficients corresponding to the noisy mass spectrometry signals under low signal-to-noise ratio conditions are used as input to construct training sample pairs. (4) The mean square error loss function is used to perform backpropagation training through the Adam optimizer. The learning rate is set to 0.001 to 0.0001, the batch size is set to 32 to 128, the training rounds are set to 100 to 300 rounds, and an early stopping strategy is used to prevent overfitting. (5) After training, the network parameters are saved to obtain the preset depth residual network.

[0090] The multi-scale wavelet coefficients obtained in step S12 are input into a pre-trained preset deep residual network, and the network outputs enhanced wavelet coefficients. In the enhanced wavelet coefficients, the high-frequency detail coefficients corresponding to noise are suppressed, while the signal features corresponding to the methylmalonic acid target peak are enhanced and preserved.

[0091] S14. The enhanced wavelet coefficients are reconstructed by inverse wavelet transform to obtain a denoised mass spectrum signal;

[0092] Specifically, the inverse wavelet transform is the inverse process of discrete wavelet decomposition. Using the enhanced approximation coefficients and enhanced detail coefficients at each level, it performs upsampling and convolution operations through a low-pass reconstruction filter and a high-speed reconstruction filter to reconstruct the original signal layer by layer. The reconstruction formula is as follows:

[0093] ;

[0094] in, These are the coefficients of the low-pass reconstruction filter. For the coefficients of the high-pass reconstruction filter, when If the value is an integer, the corresponding coefficient value is used; otherwise, zero is used (i.e., upsampling and zero-interpolation operation). Reconstruction is performed layer by layer from layer J to layer 0, and the reconstruction result of layer 0 is the denoised mass spectrometry signal. The denoised mass spectrometry signal retains the morphological information of the methylmalonic acid target peak (including peak position, peak height, and peak area), while significantly reducing background noise.

[0095] S15. Peak detection and integration are performed on the denoised mass spectrometry signal to obtain the quantitative result of methylmalonic acid.

[0096] Specifically, a pre-defined continuous wavelet transform peak detection algorithm is used to locate the target peak in the denoised mass spectrometry signal. The core idea of ​​continuous wavelet transform peak detection is to convolve the signal with wavelet functions at different scales to generate local maxima at the peak center. Peak localization is achieved by detecting the positions of local maxima at different scales.

[0097] The continuous wavelet transform, employing the Mexican Hat wavelet function with a scale parameter ranging from 1 to 20, was applied to the denoised mass spectrometry signal to obtain a scale-position two-dimensional coefficient matrix. Local maxima were detected along the position dimension within this matrix, and maxima at different scales were connected to form ridges. The positions between the start and end scales of these ridges represent the chromatographic peak regions. Based on the known retention time range of methylmalonic acid (4.5-5.5 minutes in this embodiment), the corresponding chromatographic peak regions were selected from these peak regions.

[0098] The ion intensity signal of the chromatographic peak region is then integrated. The integration method uses the trapezoidal integration method, accumulating the signal within the chromatographic peak region over a time step using a trapezoidal surface. The calculation formula is as follows:

[0099] ;

[0100] in, To denoise the mass spectrometry signal at time The intensity value, This represents the number of data points within the chromatographic peak region. represents the peak area of ​​methylmalonic acid.

[0101] The quantitative results of methylmalonic acid were calculated based on the peak area and a pre-defined standard curve. The pre-defined standard curve was constructed as follows: a series of methylmalonic acid standard solutions of known concentrations (concentration gradients of 0.05, 0.1, 0.5, 1.0, 5.0, 10.0, 50.0, and 100.0 μmol / L) were prepared and detected under the same liquid chromatography-tandem mass spectrometry conditions. The peak areas corresponding to each concentration were obtained, and a linear fit was performed with concentration as the x-axis and peak area as the y-axis to obtain the standard curve equation.

[0102] ;

[0103] in, This represents the concentration of methylmalonic acid (μmol / L). The slope The intercept is used. Substituting the peak area of ​​methylmalonic acid in the sample to be tested into the standard curve equation, the quantitative result (concentration value) of methylmalonic acid is calculated.

[0104] S16. Input the quantitative result of methylmalonic acid into a preset anomaly detection model to obtain the potential measurement deviation identification result. If it is a potential measurement deviation, then adaptively correct the quantitative result of methylmalonic acid according to historical quality control data to obtain the quantitative correction result of methylmalonic acid. If there is no potential measurement deviation, then output the quantitative result of methylmalonic acid.

[0105] Specifically, the preset anomaly detection model is an anomaly detection model based on the Isolation Forest algorithm. The basic principle of the Isolation Forest algorithm is: to recursively and randomly select features and segmentation values ​​to divide the sample space. Since the number of anomalies is small and they are very different from normal points, they can be isolated with a small number of divisions on average, so the path length of anomalies is short.

[0106] The construction process of the pre-defined anomaly detection model includes the following steps: (1) Obtain the quantitative results of methylmalonic acid from multiple batches in the historical quality control data, and construct a historical quantitative result dataset. Each data sample includes the measured concentration value, measurement time, sample type, and batch number characteristics of the quality control sample; (2) Extract the statistical distribution characteristics of the quantitative results of methylmalonic acid, including: within-batch mean, within-batch standard deviation, within-batch coefficient of variation, between-batch mean offset, trend slope of continuous measurement values, and percentage deviation from the historical mean; (3) Construct an isolated forest based on the statistical distribution characteristics. The specific parameters are set as: number of isolated trees. Set to 100 to 500, subsample size Set to 256, maximum tree depth set to ceil( (4) For each isolated tree, recursively select a feature dimension and a random split value on that dimension to divide the sample space into two subspaces until only one sample remains in the subspace or the maximum depth is reached, and record the path length of each sample; (5) Calculate the anomaly score, the formula for which is: ,in, For the sample Abnormal scores, The number of training samples, For the sample Average path length across all isolated trees This is a normalization factor for the average path length of the binary search tree. The anomaly score ranges from (0,1), with values ​​closer to 1 indicating a more likely anomaly and closer to 0.5 indicating a more likely normal sample. An anomaly score threshold of 0.6 to 0.7 is set; when the anomaly score of the methylmalonic acid quantification result exceeds this threshold, it is identified as a potential measurement bias.

[0107] Adaptive calibration includes the following steps: (1) Obtaining the measured values ​​of the quality control samples in the current batch and the corresponding standard values ​​of the quality control samples. The quality control samples are known concentration samples that are detected synchronously in the same batch as the samples to be tested, including low concentration quality control samples (e.g., 0.5 μmol / L) and high concentration quality control samples (e.g., 50.0 μmol / L), and their standard values ​​are preset known concentration values; (2) Calculating the deviation rate between the measured values ​​of the quality control samples and the standard values ​​of the quality control samples, the deviation rate is calculated using the following formula: ,in, The deviation rate, These are the measured values ​​for the quality control samples. (3) Calculate the batch correction factor based on the deviation rate. Calculate the deviation rates for low-concentration and high-concentration quality control samples respectively. and The overall batch correction factor is calculated using a weighted average method. ,in, Batch correction factor and The weights for low-concentration and high-concentration quality control samples are respectively set to 0.5 in this embodiment; (4) The batch correction factor is used to correct the quantitative results of methylmalonic acid, and the correction formula is: ,in, The results of methylmalonic acid quantification before correction. The corrected quantitative correction results for methylmalonic acid are shown below. This is the batch correction factor. Through the above adaptive correction, the influence of systematic deviations such as instrument drift and column aging on the quantitative results can be compensated, achieving adaptive correction across batches and sample types. If no potential measurement deviation is identified in step S16, the methylmalonic acid quantitative result obtained in step S15 is directly output as the final quantitative result, without the need for adaptive correction.

[0108] Example 2: Wavelet Transform Multi-Scale Decomposition Parameter Optimization

[0109] This embodiment further optimizes the parameters of wavelet transform multi-scale decomposition based on embodiment 1.

[0110] In step S12, the selection of wavelet basis functions has a significant impact on the denoising effect. This embodiment compares the denoising effects of Daubechies series (db2, db4, db6, db8), Symlets series (sym4, sym6, sym8), and Coiflets series (coif3, coif5) wavelet basis functions on methylmalonic acid mass spectrometry signals. Evaluation metrics include signal-to-noise ratio improvement (ΔSNR), mean square error (MSE), and peak shape fidelity (measured by correlation coefficient R²).

[0111] Table 1 shows a comparison of the denoising effects of different wavelet basis functions:

[0112] Table 1. Comparison of denoising effects of different wavelet basis functions

[0113]

[0114] As shown in Table 1, the Daubechies db4 wavelet basis function, with a decomposition level of 4 layers, achieves a signal-to-noise ratio improvement (ΔSNR) of 12.15 dB, a mean square error (MSE) of 0.0021, and a peak shape fidelity (R²) of 0.978, demonstrating the best overall denoising effect among the tested wavelet basis functions. Therefore, in this embodiment, the db4 wavelet basis function and 4-layer decomposition are preferred as the wavelet transform parameters for the methylmalonic acid mass spectrometry signal.

[0115] Example 3: Deep Residual Network Structure and Training Optimization

[0116] This embodiment further optimizes the structure and training strategy of the preset deep residual network based on embodiment 1.

[0117] In step S13, the structural parameters of the deep residual network have a significant impact on the denoising performance. This embodiment compares the effects of different numbers of residual blocks (4, 6, 8), different kernel sizes (3, 5, 7), and different numbers of kernels (32, 64, 128) on the denoising effect. Table 2 below shows the comparison of the denoising effects of different network structures:

[0118] Table 2 Comparison of denoising effects of residual network structures with different depths

[0119]

[0120] As shown in Table 2, when the number of residual blocks is 6, the kernel size is 5, and the number of convolutional kernels is 64, the signal-to-noise ratio (SNR) improvement (ΔSNR) reaches 12.87 dB, and the inference time is only 2.1 ms, achieving a good balance between denoising performance and computational efficiency. When the number of residual blocks increases to 8 or the number of convolutional kernels increases to 128, although the ΔSNR improves slightly, the number of parameters and inference time increase significantly, resulting in diminishing marginal returns. Therefore, this embodiment preferably uses a network structure with 6 residual blocks, a kernel size of 5, and 64 convolutional kernels.

[0121] In terms of training strategy, the following optimization measures are adopted in this embodiment: (1) Data augmentation: random time shift (±0.2 minutes), random amplitude scaling (0.8-1.2 times) and random addition of Gaussian white noise (signal-to-noise ratio 10-30 dB) are applied to the mass spectrometry signals in the training set to expand the diversity of training samples; (2) Learning rate decay: a cosine annealing learning rate scheduling strategy is adopted, with an initial learning rate of 0.001, a minimum learning rate of 0.00001, and a cycle of 50 rounds; (3) Regularization: a Dropout layer is added after each convolutional layer, with a dropout rate of 0.1 to 0.3 to prevent overfitting; (4) Early stopping strategy: the loss value is monitored on the validation set, and training is stopped when the validation loss no longer decreases for 20 consecutive rounds, and the model parameters with the minimum validation loss are saved. Through the above optimization measures, the preset deep residual network achieves a signal-to-noise ratio improvement of 12.87±0.35 dB on the test set, with a mean square error of 0.0019±0.0003 and a peak fidelity R² of 0.982±0.005, which is significantly better than the traditional wavelet thresholding denoising method (ΔSNR of 7.52 dB, R² of 0.943) and the Savitzky-Golay filtering method (ΔSNR of 5.38 dB, R² of 0.921).

[0122] Furthermore, this embodiment also implements a dynamic determination mechanism for the number of decomposition layers. For each batch of newly acquired mass spectrometry data, the system automatically calculates the signal-to-noise ratio (SNR) for each number of decomposition layers within the range of 3 to 6 layers, and selects the decomposition layer corresponding to the maximum SNR as the optimal number of decomposition layers for the current batch. This dynamic determination mechanism can adapt to the differences in signal characteristics between different batches and different sample types, ensuring the stability of the denoising effect.

[0123] Please refer to Figure 2 , Figure 2 This is a flowchart of a cross-batch correction factor monitoring mechanism for a method of processing methylmalonic acid mass spectrometry signals and detection results in some embodiments of this application.

[0124] It is worth mentioning that, according to an example of the present invention, it also includes:

[0125] S21. Obtain the batch correction factors of multiple historical batches to form a historical sequence of correction factors;

[0126] S22. Calculate the historical mean of the correction factor based on the historical sequence of the correction factor;

[0127] S23. Compare the batch correction factor of the current batch with the historical average of the correction factor to obtain the correction factor deviation;

[0128] S24. Compare the correction factor deviation with the preset standard deviation of the historical sequence of the correction factor;

[0129] S251. If the deviation of the correction factor is greater than the preset standard deviation, a batch anomaly warning is generated.

[0130] S252. If the deviation of the correction factor is less than or equal to the preset standard deviation, no warning will be output.

[0131] Specifically, batch correction factors from multiple historical batches are obtained, and a time series of batch correction factors is constructed. The deviation between the current batch's batch correction factor and the mean of historical batch correction factors is calculated. If this deviation exceeds twice the standard deviation of historical batch correction factors, a batch anomaly warning is generated, prompting operators to check the instrument status or reagent batch.

[0132] Example 4: Processing System for Methylmalonic Acid Mass Spectrometry Signals and Detection Results

[0133] Please refer to Figure 3 , Figure 3 This is a system block diagram of a methylmalonic acid mass spectrometry signal and detection result processing system according to some embodiments of this application.

[0134] This embodiment provides a system for processing methylmalonic acid mass spectrometry signals and detection results, including a memory and a processor. The memory includes a method program for processing methylmalonic acid mass spectrometry signals and detection results. When the processor executes the method program for processing methylmalonic acid mass spectrometry signals and detection results, it implements the steps of the method described in any one of Embodiments 1 to 3.

[0135] Specifically, the system also includes a data acquisition module, a preprocessing module, a wavelet transform module, a deep residual network module, a signal reconstruction module, a quantitative calculation module, an anomaly detection module, and an adaptive correction module.

[0136] The data acquisition module is used to acquire raw mass spectrometry data from liquid chromatography-tandem mass spectrometry (LC-MS / MS), including chromatographic retention time and ion intensity signals corresponding to mass-to-charge ratio. The data acquisition module communicates with the LC-MS / MS instrument, receiving raw data acquired by the instrument in real time via a data interface, or importing pre-stored raw data files into the system.

[0137] The preprocessing module performs baseline correction and normalization preprocessing on the raw mass spectrometry data to obtain a preprocessed mass spectrometry signal. The preprocessing module integrates an asymmetric least squares baseline correction algorithm and a normalization algorithm, enabling automatic execution of baseline correction and normalization operations.

[0138] The wavelet transform module is used to perform multi-scale decomposition of preprocessed mass spectrometry signals using wavelet transform to obtain wavelet coefficients at different scales. The wavelet transform module supports various wavelet basis functions (including the Daubechies, Symlets, and Coiflets series) and dynamic decomposition level determination functionality.

[0139] The deep residual network module is used to input wavelet coefficients into a preset deep residual network for denoising, obtaining enhanced wavelet coefficients. Internally, the deep residual network module contains a pre-trained deep residual network model, supporting the loading, updating, and inference of model parameters. The deep residual network module utilizes the signal characteristics of historical quality control data for model training, and this historical quality control data is stored in the system's historical database.

[0140] The signal reconstruction module is used to reconstruct the signal from the enhanced wavelet coefficients through inverse wavelet transform to obtain a denoised mass spectrum signal.

[0141] The quantitative calculation module is used to perform peak detection and integration on the denoised mass spectrometry signal to obtain the quantitative result of methylmalonic acid. The quantitative calculation module integrates continuous wavelet transform peak detection algorithm, trapezoidal integration algorithm, and standard curve calculation functions.

[0142] The anomaly detection module is used to input the quantitative results of methylmalonic acid into a preset anomaly detection model to identify whether there are potential measurement biases in the quantitative results of methylmalonic acid. The anomaly detection module internally deploys an anomaly detection model based on the isolated forest algorithm.

[0143] The adaptive correction module is used to adaptively correct the methylmalonic acid quantification results based on historical quality control data when potential measurement deviations are identified, obtaining a corrected methylmalonic acid quantification result. The adaptive correction module also monitors the changing trend of the correction factor across batches, generating a batch anomaly warning when the correction factor fluctuates abnormally.

[0144] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0145] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0146] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0148] The above description is merely a preferred embodiment of the present invention and does not limit the scope of patent protection of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for processing methylmalonic acid mass spectrometry signals and detection results, characterized in that, Includes the following steps: The raw mass spectrometry data and historical quality control data of liquid chromatography-tandem mass spectrometry were acquired. The raw mass spectrometry data included chromatographic retention time, mass-to-charge ratio and corresponding ion intensity signals. Baseline correction and normalization preprocessing were performed to obtain preprocessed mass spectrometry signals. Based on the historical quality control data, wavelet transform is used to perform multi-scale decomposition on the preprocessed mass spectrometry signal to obtain wavelet coefficients at different scales. The wavelet coefficients are input into a preset deep residual network for denoising to obtain enhanced wavelet coefficients; The enhanced wavelet coefficients are reconstructed using inverse wavelet transform to obtain a denoised mass spectrum signal. Peak detection and integration were performed on the denoised mass spectrometry signal to obtain the quantitative results of methylmalonic acid; The quantitative result of methylmalonic acid is input into a preset anomaly detection model to obtain the potential measurement deviation identification result. If there is a potential measurement deviation, the quantitative result of methylmalonic acid is adaptively corrected according to historical quality control data to obtain the quantitative correction result of methylmalonic acid. If there is no potential measurement deviation, the quantitative result of methylmalonic acid is output.

2. The method for processing methylmalonic acid mass spectrometry signals and detection results according to claim 1, characterized in that, The step of performing multi-scale decomposition of the preprocessed mass spectrometry signal using wavelet transform based on the historical quality control data to obtain wavelet coefficients at different scales includes: The signal-to-noise ratio (SNR) of historical mass spectrometry signals in the historical quality control data is calculated and analyzed, and the number of decomposition layers that maximizes the SNR is taken as the number of decomposition layers for the current batch. Based on the number of decomposition levels, the preprocessed mass spectrum signal is subjected to multi-level discrete wavelet decomposition using Daubechies wavelet basis functions to obtain approximation coefficients and detail coefficients. Wavelet coefficients of different scales are then constructed based on the approximation coefficients and detail coefficients.

3. The method for processing methylmalonic acid mass spectrometry signals and detection results according to claim 1, characterized in that, The training process of the preset depth residual network includes: Obtain the standard mass spectrometry signal corresponding to the methylmalonic acid standard sample with a known concentration from the historical quality control data; Extract the standard wavelet coefficients of the standard mass spectrum signal after wavelet transform multi-scale decomposition; The target wavelet coefficients corresponding to the labeled denoised target signal are used as labels, wherein the denoised target signal is obtained by superimposing and averaging the mass spectrometry signals of repeated measurements in the historical quality control data; Using the standard wavelet coefficients as training input features and the target wavelet coefficients as output, the network parameters of the preset deep residual network are trained through the backpropagation algorithm and the mean square error loss function to obtain the preset deep residual network.

4. The method for processing methylmalonic acid mass spectrometry signals and detection results according to claim 1, characterized in that, The step of peak detection and integration of the denoised mass spectrometry signal to obtain the quantitative result of methylmalonic acid includes: The target peak region corresponding to methylmalonic acid is obtained by locating the target peak of the denoised mass spectrometry signal using a preset continuous wavelet transform peak detection algorithm. The ion intensity signal of the chromatographic peak region is integrated to obtain the peak area of ​​methylmalonic acid. The quantitative results of methylmalonic acid were obtained by analyzing the peak area of ​​methylmalonic acid and the preset standard curve.

5. The method for processing methylmalonic acid mass spectrometry signals and detection results according to claim 1, characterized in that, The construction process of the preset anomaly detection model includes: The historical methylmalonic acid quantitative results of multiple batches in the historical quality control data are obtained to form a model training sample set; Extract the statistical distribution characteristics of each historical methylmalonic acid quantification result in the model training sample set; Based on the statistical distribution characteristics, an isolated forest is constructed. For each isolated tree, a feature dimension and a random split value on that dimension are recursively selected to divide the sample space into two subspaces until only one sample remains in the subspace or the maximum depth is reached. The path length of each sample is recorded. The mean of the path lengths is calculated to obtain the average path length of the sample. Anomaly scores are obtained by calculating and analyzing the average path length. If the anomaly score is greater than the preset anomaly score threshold, it is identified as a potential measurement bias. The dataset, feature extraction rules, isolated forest parameters, score calculation logic and anomaly judgment rules are solidified to construct a preset anomaly detection model.

6. The method for processing methylmalonic acid mass spectrometry signals and detection results according to claim 1, characterized in that, The adaptive correction of the methylmalonic acid quantification result based on historical quality control data to obtain the methylmalonic acid quantification correction result includes: Obtain the measured values ​​of quality control samples and the corresponding standard values ​​of quality control samples from historical quality control data. The measured values ​​of quality control samples include the measured values ​​of low-concentration quality control samples and the measured values ​​of high-concentration quality control samples. By comparing and analyzing the measured values ​​of the low-concentration quality control samples and the high-concentration quality control samples with the standard values ​​of the quality control samples, the deviation rates of the low-concentration quality control samples and the high-concentration quality control samples are obtained. The batch correction factor is obtained by weighting and averaging the deviation rates of the low-concentration quality control samples and the high-concentration quality control samples. The quantitative result of methylmalonic acid is corrected according to the batch correction factor to obtain the corrected quantitative result of methylmalonic acid.

7. A system for processing methylmalonic acid mass spectrometry signals and detection results, wherein the system is used to implement the method for processing methylmalonic acid mass spectrometry signals and detection results according to any one of claims 1-6, characterized in that, include: The data acquisition module is used to acquire the raw mass spectrometry data of the liquid chromatography-tandem mass spectrometry and output the raw mass spectrometry data to the preprocessing module; The preprocessing module is used to perform baseline correction and normalization preprocessing on the raw mass spectrometry data to obtain a preprocessed mass spectrometry signal; The wavelet transform module is used to perform multi-scale decomposition of the preprocessed mass spectrum signal using wavelet transform to obtain wavelet coefficients at different scales. A deep residual network module is used to input the wavelet coefficients into a preset deep residual network for denoising processing to obtain enhanced wavelet coefficients; The signal reconstruction module is used to reconstruct the signal from the enhanced wavelet coefficients by inverse wavelet transform to obtain a denoised mass spectrum signal. The quantitative calculation module is used to perform peak detection and integration on the denoised mass spectrometry signal to obtain the quantitative result of methylmalonic acid; An anomaly detection module is used to input the quantitative results of methylmalonic acid into a preset anomaly detection model to identify whether there is a potential measurement bias in the quantitative results of methylmalonic acid. An adaptive correction module is used to adaptively correct the quantitative result of methylmalonic acid based on historical quality control data when potential measurement deviations are identified, so as to obtain the corrected quantitative result of methylmalonic acid.

8. The processing system for methylmalonic acid mass spectrometry signals and detection results according to claim 7, characterized in that, It also includes a memory and a processor, wherein the memory contains a program for processing methylmalonic acid mass spectrometry signals and detection results, and when the program for processing methylmalonic acid mass spectrometry signals and detection results is executed by the processor, it performs the following steps: The raw mass spectrometry data and historical quality control data of liquid chromatography-tandem mass spectrometry were acquired. The raw mass spectrometry data included chromatographic retention time, mass-to-charge ratio and corresponding ion intensity signals. Baseline correction and normalization preprocessing were performed to obtain preprocessed mass spectrometry signals. Based on the historical quality control data, wavelet transform is used to perform multi-scale decomposition on the preprocessed mass spectrometry signal to obtain wavelet coefficients at different scales. The wavelet coefficients are input into a preset deep residual network for denoising to obtain enhanced wavelet coefficients; The enhanced wavelet coefficients are reconstructed using inverse wavelet transform to obtain a denoised mass spectrum signal. Peak detection and integration were performed on the denoised mass spectrometry signal to obtain the quantitative results of methylmalonic acid; The quantitative result of methylmalonic acid is input into a preset anomaly detection model to obtain the potential measurement deviation identification result. If there is a potential measurement deviation, the quantitative result of methylmalonic acid is adaptively corrected according to historical quality control data to obtain the quantitative correction result of methylmalonic acid. If there is no potential measurement deviation, the quantitative result of methylmalonic acid is output.

9. The processing system for methylmalonic acid mass spectrometry signals and detection results according to claim 8, characterized in that, The step of performing multi-scale decomposition of the preprocessed mass spectrometry signal using wavelet transform based on the historical quality control data to obtain wavelet coefficients at different scales includes: The signal-to-noise ratio (SNR) of historical mass spectrometry signals in the historical quality control data is calculated and analyzed, and the number of decomposition layers that maximizes the SNR is taken as the number of decomposition layers for the current batch. Based on the number of decomposition levels, the preprocessed mass spectrum signal is subjected to multi-level discrete wavelet decomposition using Daubechies wavelet basis functions to obtain approximation coefficients and detail coefficients. Wavelet coefficients of different scales are then constructed based on the approximation coefficients and detail coefficients.

10. The processing system for methylmalonic acid mass spectrometry signals and detection results according to claim 8, characterized in that, The training process of the preset depth residual network includes: Obtain the standard mass spectrometry signal corresponding to the methylmalonic acid standard sample with a known concentration from the historical quality control data; Extract the standard wavelet coefficients of the standard mass spectrum signal after wavelet transform multi-scale decomposition; The target wavelet coefficients corresponding to the labeled denoised target signal are used as labels, wherein the denoised target signal is obtained by superimposing and averaging the mass spectrometry signals of repeated measurements in the historical quality control data; Using the standard wavelet coefficients as training input features and the target wavelet coefficients as output, the network parameters of the preset deep residual network are trained through the backpropagation algorithm and the mean square error loss function to obtain the preset deep residual network.

Images