An automated chromatographic analysis system and method for PET pharmaceutical quality control

Abstract

The application discloses an automatic chromatographic analysis system and method for PET drug quality control, and relates to the field of radiopharmaceutical analysis.The analysis method comprises the following steps: continuously acquiring an original response signal in a chromatographic separation process; extracting a change rate and a change rate trend of a leading edge and a trailing edge of a candidate chromatographic peak as morphological characteristic parameters in real time; dynamically generating a starting identification time window and an ending identification time window according to the morphological characteristic parameters; establishing a tracking association between a peak boundary and baseline noise fluctuation; correcting a starting and ending boundary position in real time; and finally completing integration and calculating radiochemical purity in the dynamic window.The application does not need to rely on a fixed response threshold, can adapt to signal amplitude attenuation and noise fluctuation, ensures consistency of radiochemical purity calculation results under different analysis time nodes, and thus improves reliability of quality control decision.

Description

Technical Field

[0001] This invention relates to the field of radiopharmaceutical analysis technology, and in particular to an automated chromatographic analysis system and method for PET drug quality control. Background Technology

[0002] Positron emission tomography (PET) drugs are a special class of drugs that undergo continuous physical decay within a very short time. The half-life of their active ingredient, the radionuclide, is typically very short, meaning that the radioactivity of the sample continues to decrease during quality control analysis after the drug's production is completed. The radiochemical purity of PET drugs is a key indicator determining their safe use in clinical diagnosis, and high-performance liquid chromatography coupled with a radioactive detector is currently the mainstream technique for determining radiochemical purity.

[0003] In traditional chromatographic analysis, peak identification and integration typically rely on pre-set fixed response thresholds or fixed peak width rules. When the detector output signal amplitude exceeds a certain preset value, the system determines the peak has begun; otherwise, it determines the peak has ended. This static method works well for routine drug analysis where signal amplitudes are relatively stable. However, when applied to the quality control of PET drugs, due to the continuous presence of radioactive decay, the same chromatographic peak in the same sample can exhibit drastically different absolute amplitudes in the early and later stages of analysis. As decay intensifies, the peak amplitude gradually approaches the fluctuation level of background noise. The fixed threshold either gets set too high, completely missing decayed peaks, or it gets set too low, incorrectly identifying noise spikes as chromatographic peaks. Furthermore, fixed peak width rules struggle to adapt to subtle changes in peak shape during decay, leading to improper truncation of the peak front or rear, and significant deviations in the integrated area. Consequently, when the same batch of samples is analyzed at different time points, the calculated radiochemical purity results are inconsistent, severely interfering with quality control personnel's accurate assessment of drug quality. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an automated chromatographic analysis system and method for PET drug quality control.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Firstly, the present invention provides an automated chromatographic analysis method for PET drug quality control, comprising the following steps:

[0006] S1. During the chromatographic separation process, the raw response signal output by the detector is continuously acquired to form a response data stream that changes over time;

[0007] S2. During the transmission of the response data stream, the morphological characteristic parameters of the leading and trailing edges of the candidate chromatographic peaks are extracted in real time. The morphological characteristic parameters include at least the rate of change of the response amplitude over time and the trend of the rate of change.

[0008] S3. Based on the morphological characteristics of the candidate chromatographic peak front, dynamically generate the initial identification time window of the chromatographic peak, wherein the initial boundary of the initial identification time window corresponds to the position where the response signal first systematically deviates from the baseline noise fluctuation pattern.

[0009] S4. Based on the morphological characteristics of the trailing edge of the candidate chromatographic peak, dynamically generate the termination identification time window for the chromatographic peak, wherein the termination boundary of the termination identification time window corresponds to the position where the response signal decays to the point where its rate of change is indistinguishable from the baseline noise fluctuation pattern.

[0010] S5. Within the dynamic time window determined by the start and end identification time windows, establish a tracking correlation between the chromatographic peak start and end boundaries and baseline noise fluctuations, and correct the position of the start and end boundaries in real time according to the instantaneous fluctuations of the baseline noise.

[0011] S6. Based on the corrected start and end boundaries, integrate the original response signal within the dynamic time window to calculate the chromatographic peak area of ​​each component.

[0012] S7. Calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

[0013] In a preferred embodiment of the present invention, the morphological characteristic parameters further include: the duration and amplitude increment of the peak leading edge from the starting point to the peak apex, and the duration and amplitude decrease of the peak trailing edge from the peak apex to the ending point.

[0014] In a preferred embodiment of the present invention, in step S3, the specific method for dynamically generating the initial identification time window is as follows: based on the change trajectory of the rising slope of the peak front, a change curve reflecting the acceleration of the response signal from the baseline to the peak is fitted, and the position where the response signal first systematically deviates from the baseline noise fluctuation law is deduced based on the fitted curve as the theoretical starting point, and the theoretical starting point is determined as the starting boundary.

[0015] In a preferred embodiment of the present invention, in step S4, the specific method for dynamically generating the termination identification time window is as follows: a curve reflecting the gradual decay of the response signal from the peak to the baseline is fitted based on the trajectory of the change in the slope of the peak's fall. Based on the fitted curve, the theoretical termination point is deduced as the position where the rate of change of the response signal decays to the point where it is indistinguishable from the baseline noise fluctuation. This theoretical termination point is then determined as the termination boundary.

[0016] In a preferred embodiment of the present invention, in step S5, the specific method for establishing tracking association and real-time correction of start and end boundaries is as follows: real-time acquisition of the instantaneous fluctuation amplitude of baseline noise within the current time window, dynamic comparison of the instantaneous fluctuation amplitude with the baseline noise level at the peak front starting point, and temporary correction of the start and end boundary positions within the current time window based on the comparison results when the noise fluctuation amplitude undergoes a sudden change exceeding the expected range.

[0017] In a preferred embodiment of the present invention, step S6 further includes linear interpolation of the endpoints outside the boundary before integration, so that they fall on a common baseline level to eliminate the influence of baseline drift on the integration area.

[0018] In a preferred embodiment of the present invention, in step S7, the radiochemical purity value is calculated by: performing a ratio calculation between the peak area of ​​the target component and the sum of the peak areas of all components to generate a radiochemical purity value, and comparing the radiochemical purity value with the product release standard to output a quality control judgment result.

[0019] In a preferred embodiment of the present invention, the determination method for the baseline noise fluctuation pattern is as follows: by comparing the difference between the rate of change of the response amplitude within the current time window and the average rate of change of the previous non-peak interval, it is determined whether the trend of the rate of change has changed abruptly.

[0020] In a preferred embodiment of the present invention, the method further includes: after dynamically generating the start identification time window and the end identification time window, performing logical verification on the start boundary and the end boundary, and triggering a signal quality alarm when the start boundary is later than the end boundary.

[0021] In a second aspect, the present invention provides an automated chromatographic analysis system for PET drug quality control, for implementing the analytical method as described in any one of the above-mentioned methods, the analysis system comprising:

[0022] The signal acquisition module is used to continuously acquire the raw response signal output by the detector during the chromatographic separation process, forming a response data stream that changes over time;

[0023] The morphological feature extraction module is used to extract the morphological feature parameters of the leading and trailing edges of candidate chromatographic peaks in real time during the transmission of the response data stream.

[0024] The window generation module is used to dynamically generate the start and end recognition time windows of a chromatographic peak based on the morphological characteristics of the leading and trailing edges of the candidate chromatographic peak.

[0025] The boundary correction module is used to establish a tracking correlation between the start and end boundaries of the chromatographic peak and the baseline noise fluctuation within the dynamic time window determined by the start and end identification time window, and to correct the position of the start and end boundaries in real time according to the instantaneous fluctuation of the baseline noise.

[0026] The integration calculation module is used to integrate the original response signal within the dynamic time window based on the corrected start and end boundaries, and calculate the chromatographic peak area of ​​each component.

[0027] The purity calculation and judgment module is used to calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

[0028] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0029] (1) This invention provides an automated chromatographic analysis system and method for PET drug quality control. By extracting the morphological characteristics of the leading and trailing edges of chromatographic peaks and dynamically generating start and end recognition time windows, the determination of peak boundaries no longer depends on a fixed response threshold. This ensures that the trajectory of the rising and falling slopes of the chromatographic peaks remains relatively stable during radioactive decay, because decay changes the absolute amplitude of the signal without altering the basic shape of the chromatographic peak. Consequently, peak recognition is insensitive to signal amplitude attenuation. Compared to traditional methods that suffer from missed peak recognition or boundary shifts after signal decay, this invention can accurately pinpoint the theoretical start and end positions of each chromatographic peak even under conditions of continuously decreasing radioactivity, thereby ensuring the consistency of radiochemical purity calculation results for the same batch of samples at different analysis time points.

[0030] (2) This invention establishes a tracking correlation mechanism between the chromatographic peak start and end boundaries and baseline noise fluctuations, and corrects the position of the start and end boundaries in real time according to the instantaneous fluctuations of the baseline noise. The system continuously monitors the real-time noise fluctuation amplitude within a dynamic time window and dynamically compares it with the baseline noise level at the peak front starting point. When abnormal fluctuations such as noise spikes occur, the boundary position is temporarily adjusted according to the comparison results to exclude noise interference from the integration region, thereby improving the robustness of peak boundary determination. Unlike the fixed noise tolerance threshold in existing schemes, this invention enables the peak integration region to adaptively follow changes in noise level, avoiding misjudging noise spikes as chromatographic peaks and preventing the incorrect removal of true low-profile chromatographic peaks, thereby improving the accuracy of radiochemical purity calculation.

[0031] (3) The present invention replaces the peak identification logic that relies on static amplitude comparison in traditional chromatographic analysis with window adaptive logic that relies on dynamic morphological evolution trend. Since the scheme is designed for the physical process of continuous signal decay caused by radioactive decay, its judgment basis is the kinematic characteristics of the chromatographic peak itself rather than the absolute amplitude. Therefore, it can overcome the failure problem of the fixed threshold method in the signal decay scenario. In practical applications, the operator does not need to repeatedly adjust the integral parameters according to the change of sample activity. The system can automatically adapt to the complete analysis process from high activity to low activity, making the PET drug quality control operation more standardized, thereby reducing the uncertainty caused by human intervention and improving the reliability and repeatability of quality control results. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic flowchart of an automated chromatographic analysis method for PET drug quality control according to the present invention;

[0034] Figure 2 This is a schematic diagram of the dynamic identification of chromatographic peaks, boundary correction, and integral calculation of the present invention;

[0035] Figure 3 This is a structural block diagram of an automated chromatographic analysis system for PET drug quality control according to the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] 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. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0038] The automated chromatographic analysis system and method for PET drug quality control provided in this invention are applied to the quality control stage after the production of radiopharmaceuticals for positron emission tomography (PET). A typical application scenario is in the quality control department of a medical cyclotron accelerator laboratory or a radiopharmaceutical manufacturing company. Operators will prepare the samples to be tested... 18 F-FDG or FLT PET drug samples are injected into a radioactive detector coupled with a high-performance liquid chromatograph. Then, the automated chromatographic analysis system described in this invention is started, which can complete the identification, integration and radiochemical purity calculation of chromatographic peaks without human intervention, and output a quality control report that meets the requirements of Good Manufacturing Practice (GMP) for pharmaceuticals.

[0039] Example 1:

[0040] like Figure 1 As shown, an automated chromatographic analysis method for PET drug quality control includes the following steps:

[0041] S1. During the chromatographic separation process, the raw response signal output by the detector is continuously acquired to form a response data stream that changes over time;

[0042] S2. During the transmission of the response data stream, extract the morphological characteristic parameters of the leading and trailing edges of the candidate chromatographic peaks in real time. The morphological characteristic parameters include at least the rate of change of the response amplitude over time and the trend of the rate of change.

[0043] S3. Based on the morphological characteristics of the candidate chromatographic peak front, dynamically generate the initial identification time window of the chromatographic peak, wherein the initial boundary of the initial identification time window corresponds to the position where the response signal first systematically deviates from the baseline noise fluctuation pattern.

[0044] S4. Based on the morphological characteristics of the trailing edge of the candidate chromatographic peak, dynamically generate the termination identification time window for the chromatographic peak, wherein the termination boundary of the termination identification time window corresponds to the position where the response signal decays to the point where its rate of change is indistinguishable from the baseline noise fluctuation pattern.

[0045] S5. Within the dynamic time window determined by the start and end identification time windows, establish a tracking correlation between the chromatographic peak start and end boundaries and baseline noise fluctuations, and correct the position of the start and end boundaries in real time according to the instantaneous fluctuations of the baseline noise.

[0046] S6. Based on the corrected start and end boundaries, integrate the original response signal within the dynamic time window to calculate the chromatographic peak area of ​​each component.

[0047] S7. Calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

[0048] The concept of this embodiment is to abandon the traditional chromatographic analysis approach of relying on fixed response thresholds or preset peak width rules for peak identification. Instead, it utilizes the morphological evolution characteristics of the chromatographic peak itself, namely the slope and changing trend of the peak front and rear edges, to dynamically infer the theoretical starting and ending points of each chromatographic peak. This method also establishes a dynamic tracking correlation between the peak boundary and real-time baseline noise, enabling the peak integration region to adaptively respond to noise fluctuations. This solves the problem of inaccurate peak identification and inconsistent radiochemical purity calculation results caused by the continuous decay of PET drugs and the signal amplitude attenuation. This improves the reliability and consistency of quality control decisions.

[0049] The core challenge in implementing this embodiment lies in the fact that the half-life of radionuclides is typically very short; for example, the half-life of fluorine-18 is approximately 110 minutes. This means that the sample activity may decrease by more than 30% during a single analysis, rendering the absolute signal amplitude, which is relied upon by traditional methods, unsuitable as a stable benchmark under such dynamic changes. Furthermore, chromatographic baseline noise is not constant; mobile phase pulses, temperature fluctuations, or thermal noise from electronic components can all generate interference signals with randomly varying amplitudes. The specific and challenging problem this embodiment aims to solve is accurately distinguishing between a true chromatographic peak with an amplitude close to background and a pure noise spike, under conditions of continuously decaying signal amplitude accompanied by non-stationary noise.

[0050] Example 2:

[0051] This embodiment is a further refinement based on the above embodiment 1.

[0052] In this embodiment, in the specific engineering implementation of step S1, the automated chromatography analysis system is connected to the analog signal output terminal of the chromatography detector via a data acquisition card. Once a sample analysis is initiated, the system reads the voltage signal output by the detector at fixed time intervals. This voltage signal is proportional to the radioactivity measured by the detector. The system groups the voltage value at each reading moment and its corresponding timestamp into an ordered pair and stores them in a memory buffer in chronological order, thus forming a response data stream that extends continuously over time. This data stream includes both the baseline noise signal output by the detector when the mobile phase is a pure solvent, with amplitude randomly fluctuating around zero, and the chromatographic peak signals generated when each chemical component in the sample passes through the detector sequentially, with amplitudes first increasing and then decreasing.

[0053] It should be noted that the raw response signal refers to the electrical signal directly output by the chromatographic detector without any smoothing or filtering, with voltage or current as its physical dimensions. The response data stream, on the other hand, refers to a sequence of time-amplitude data points collected at equal time intervals along a continuous time axis, with a sampling frequency between 5 Hz and 20 Hz to ensure sufficient resolution for the morphology of chromatographic peaks.

[0054] Understandably, radioactive decay follows an exponential law, but the events of a single photon arriving at the detector are discrete and random. Directly counting these discrete events would introduce Poisson noise. This step, by continuously and equally spaced readings of the integrated voltage at the detector's analog output, essentially performs a time-based accumulation and averaging of the randomly occurring decay events within each sampling period. This converts the discrete and noisy count rate signal into a continuous and more easily processed analog voltage signal, providing smooth raw data for subsequent morphology-based analysis.

[0055] Specifically, it includes the following sub-steps:

[0056] S11. Simultaneously with initiating chromatographic separation and sample analysis, a high-precision hardware timer is started.

[0057] S12. Whenever the timer reaches the preset sampling interval, an analog-to-digital conversion operation is triggered to convert the current analog voltage amplitude output by the detector into a digital value.

[0058] S13. Encapsulate the converted digital value, the absolute timestamp of this sampling, and the cumulative sampling sequence number into a data point.

[0059] S14. Push the data point to a first-in-first-out circular buffer for real-time access in subsequent steps.

[0060] It should be noted that the sampling interval is set based on the expected width of the target chromatographic peak. For columns commonly used in PET drug analysis, the half-peak width (WHM) is typically between 6 s and 30 s. To ensure the accuracy of peak shape reduction, the sampling interval should not exceed 1 / 10 of the minimum expected WHM, with a range of 0.1 s to 0.5 s, corresponding to a sampling frequency of 2 Hz to 10 Hz. The circulating buffer capacity should be able to accommodate at least three times the data points generated by the longest analysis time. For example, if an analysis lasts 30 minutes and the sampling frequency is 10 Hz, the buffer capacity should be at least 3 × 30 × 60 × 10 = 54,000 data points.

[0061] This embodiment converts the continuous analog signal output by the detector into a computer-processable digital signal sequence, forming a high-fidelity data stream that retains all time-series characteristics of the original signal. The challenge of this method lies in how to ensure a high sampling rate to capture peak details without generating excessive data volume that would degrade the system's real-time processing capabilities. Through the aforementioned circular buffer mechanism, a smooth transition between the continuous data stream and discrete processing steps can be achieved.

[0062] In this embodiment, step S2 is further refined as follows: During the continuous push of the response data stream into the buffer, a sliding time window analyzer is activated to monitor newly entering data points in real time. When the response amplitude is detected to be higher than the currently estimated baseline noise level for several consecutive sampling points (e.g., 3-5 consecutive points) and exhibits a monotonically increasing trend, it is determined to be the start of a candidate chromatographic peak. At each subsequent sampling point, the morphological characteristics of the leading and trailing edges of the candidate chromatographic peak are calculated and recorded.

[0063] It should be noted that morphological characteristic parameters are a set of numerical indicators used to quantitatively describe the geometry of chromatographic peaks. The rate of change of the response amplitude over time, i.e., the first derivative, indicates how fast the signal rises or falls, and is referred to as the slope in this field. The trend of the rate of change, i.e., the second derivative, indicates the rate of increase or decrease of the slope, and is used to determine whether the signal is rising at an accelerating, uniform, or decelerating speed, as well as the precise location of the peak apex.

[0064] Understandably, an ideal chromatographic peak theoretically follows a Gaussian distribution or an exponentially modified Gaussian distribution. Its leading edge starts from the baseline, with the slope gradually increasing from zero to its maximum value, and then gradually decreasing to zero to reach the peak apex; its trailing edge exhibits a symmetrical or tailing process. By calculating the first and second derivatives of the original signal, the geometric characteristics of the peak shape can be transformed into precise mathematical indicators. The zero point of the first derivative corresponds to the peak apex, the extreme points of the first derivative correspond to the highest and lowest slope points of the peak, and the zero point of the second derivative corresponds to the inflection point. These rate-of-change parameters are insensitive to the absolute amplitude of the signal but are highly sensitive to the shape of the signal itself, thus overcoming the amplitude attenuation effect caused by radioactive decay.

[0065] Specifically, it includes the following sub-steps:

[0066] S21. Maintain a sliding window of length N. Whenever a new data point x(t) arrives, update the sequence of data points within the window as follows: .

[0067] S22. For each data point within the window, calculate its first derivative d(t) and second derivative d2(t) using the five-point cubic Savitzky-Golay smoothing algorithm. This algorithm can smooth noise while maintaining the width and height of the peak shape relatively well.

[0068] S23. When it is detected that d(t) changes from near zero to a positive value and is greater than 0 for 3 consecutive points, and the amplitude x(t) is higher than the baseline noise level for 3 consecutive points, mark the start of the candidate peak and record this interval as the peak front.

[0069] S24. Within the peak-leading interval, record and store the following parameters: d(t) at each point, d2(t) at each point, and the duration from the starting point to the current point. The magnitude increment from the starting point to the current point .

[0070] S25. When d(t) changes from a positive value to a negative value, and the second derivative d2(t) changes from a negative value to a positive value, mark the point x(t) as the peak.

[0071] S26. After the peak, detect the period where d(t) is negative and the amplitude x(t) continues to decrease, and record this interval as the peak trailing edge. Within the peak trailing edge interval, record and store the following parameters: d(t) at each point, d2(t) at each point, and the duration from the peak to the current point. Amplitude reduction from peak to current point .

[0072] It should be noted that the sliding window length N ranges from 7 to 15 sampling points. For a sampling rate of 10 Hz, N=11 can cover a signal of 1.1 s, which is sufficient to calculate the derivative under low noise. The five-point cubic Savitzky-Golay algorithm has a smoothing window width of 5 points and a fitting polynomial degree of 3. The baseline noise level is estimated by taking the root mean square of the amplitude of all sampling points within the first 5 seconds after the start of the analysis and updating it every 10 seconds.

[0073] In this embodiment, step S3 is to dynamically generate the initial identification time window, and to pass the peak front morphological feature parameters obtained in S2, namely the first derivative sequence d(t) and the second derivative sequence d2(t) during the rising process, as input to a dynamic window generation model.

[0074] It should be noted that the initial identification time window is a dynamically determined time interval, and the left boundary of this interval is the inferred chromatographic peak starting point of the system. Systematic deviation from baseline noise fluctuation refers to the signal change pattern transforming from random, zero-mean white noise characteristics to a directional, non-random upward trend.

[0075] Understandably, the signal amplitude at the peak initiation point is extremely low, completely mixed with noise, and indistinguishable by amplitude alone. However, the kinetic characteristics of the peak initiation point are fundamentally different from those of noise. The first derivative d(t) of noise fluctuates randomly around zero, as does its second derivative d2(t). At the initiation of a chromatographic peak, the second derivative d2(t) initially exhibits a positive value (acceleration), indicating that the signal begins to accelerate. This model uses the point where this acceleration originates to deduce the location where the signal begins to systematically deviate from zero.

[0076] Specifically, it includes the following sub-steps:

[0077] S31. Obtain the first derivative sequence from step S2 that belongs to the candidate peak frontal region. and second derivative sequence , where t i The time is the time of the i-th sampling point within the frontier region.

[0078] S32. Using the least squares method, apply the second derivative sequence... A linear fit is performed on the initial segment of the peak front (the first 3-5 points after the detection of the candidate peak) to obtain a second-order derivative function that varies with time. ,in Represents the rate of change of acceleration. This is the intercept.

[0079] S33, Solving the equation The theoretical acceleration start time is obtained. .

[0080] S34, with Based on this, we trace back a fixed physical time. , obtain the initial boundary . The physical meaning of is the shortest time required for a signal to go from deviating from noise to producing a measurable positive acceleration.

[0081] S35, the left boundary of the dynamic start recognition time window is... The right boundary is temporarily set as the rightmost point of the currently detected peak front.

[0082] It should be noted that the number of data points used for the first derivative fitting is the first 3-7 points of the initial leading edge segment. The specific number is dynamically adjusted based on the signal-to-noise ratio of that segment; more points are used when the signal-to-noise ratio is low to improve fitting stability. (Backtracking time) The value of is in the range of 0.5 s to 2.0 s, and is related to the extracolumn volume and sampling frequency of the chromatographic system. More preferably, it can be set to 1.5 times the time unit corresponding to the theoretical plate height of the chromatographic system. The fitting algorithm used is the standard least squares method, and its objective is to minimize the sum of squared residuals.

[0083] Because traditional fixed threshold methods cannot automatically adjust the judgment starting point when the signal decays, resulting in the peak front being truncated, this embodiment shifts the judgment basis from static amplitude comparison to dynamic inference of morphological evolution trend. By analyzing the acceleration characteristics of the signal emerging from disordered noise, it is equivalent to defining the birth of chromatographic peaks from the kinematic root, which has good adaptability to special objects such as PET drugs with continuously changing signal intensity.

[0084] In this embodiment, step S4 is used to generate a termination identification time window. The peak trailing edge morphological feature parameters obtained in S2, namely the first derivative sequence d(t) and the second derivative sequence d2(t) during the descent process, are input into the same dynamic window generation model.

[0085] It should be noted that the termination boundary of the termination identification time window corresponds to the end point of the chromatographic peak. The indistinguishability of the rate of change with the baseline noise fluctuation pattern means that the signal's rate of change has decayed to the same statistical level as the random fluctuation rate of pure noise. At this point, the signal has been infused with noise and can no longer be effectively predicted based on its trend.

[0086] Understandably, the difficulty in determining peak termination lies in the fact that as the signal slowly decays back to the baseline, the amplitude difference is very small, making it easy to terminate prematurely or delayed. This step also avoids amplitude judgment, instead analyzing the absolute value of the first derivative d(t) (the descent slope). At the trailing edge of the peak, |d(t)| gradually decreases. When the decreasing trend of |d(t)| stops, and its numerical fluctuation enters the range consistent with the fluctuation range of the first derivative of the baseline noise, it can be determined that the signal decay has ended, and what remains is only noise fluctuation. This method essentially defines peak termination as the process of signal kinetic energy depletion.

[0087] Specifically, it includes the following sub-steps:

[0088] S41. Obtain the first derivative sequence from step S2 that belongs to the interval following the candidate peak. and second derivative sequence .

[0089] S42. Calculate the local variance of the absolute value of the descending slope |d(t)| within a short time window in real time. .

[0090] S43. Simultaneously calculate the first derivative of the pure baseline interval (the peakless region where the distance between the peak and the chromatographic peak is sufficiently large). variance .

[0091] S44, Dynamic Calculation Ratio When the R value continues to decrease and falls below a preset threshold for the first time... th At this point, it indicates that the rate of change of the signal is no longer indistinguishable from noise.

[0092] S45. The time point at which this condition is triggered is the theoretical termination point t. end The right boundary of the dynamic termination recognition time window is t. end The left boundary is temporarily set as the peak.

[0093] It should be noted that the short time window length for calculating local variance is set to 2-3 times the sampling period. For example, when the sampling period is 0.1 s, the window length is 0.3 s. Baseline noise variance Recalculate every 5 seconds, using the most recent peakless data point. Ratio threshold R th The value ranges from 1.05 to 1.5. This value determines the sensitivity of the judgment. The smaller the value, the more likely it is to miss the tail of the peak, while the larger the value, the more likely it is to misjudge noise as a peak. The preferred threshold is 1.1.

[0094] In this embodiment, as Figure 2 As shown, the detailed implementation of step S5 is as follows: at the initial boundary t determined in S3 and S4... start and termination boundary t end Within the defined dynamic time window, a start and end boundary correction module is initiated. This module does not directly use t. start and t end Instead of using them as the final integration boundaries, they are fine-tuned based on real-time noise.

[0095] It should be noted that tracking correlation refers to the system continuously monitoring the instantaneous fluctuations of the baseline noise during peak integration and dynamically mathematically correlating the determination results of the peak start and end boundaries with the current noise level. Instantaneous fluctuations refer to the peak-to-peak changes of the noise signal within a very short period of time.

[0096] Understandably, even with the dynamic positioning via S3 and S4, t start and t end This remains a theoretical value based on trend inference; actual baseline noise may contain occasional spikes with large amplitudes. These noise spikes, if within t... start Before or t endSubsequently, a noise suppression threshold may appear, with an amplitude exceeding the theoretical peak initiation amplitude. This step addresses this by establishing a noise suppression threshold, whose value is not fixed but proportional to the real-time noise fluctuation amplitudes before and after the current window. Only when the signal amplitude significantly exceeds this dynamic threshold and persists for a specific duration is it identified as part of a peak.

[0097] Specifically, it includes the following sub-steps:

[0098] S51, in Calculate the real-time peak fluctuation of the baseline noise within the previous 3-second interval. .in, The maximum value of the noise signal. This represents the minimum value of the noise signal.

[0099] S52, in During the subsequent 3-second interval, the real-time peak fluctuation of the baseline noise was calculated. .

[0100] S53, Set dynamic noise tolerance , where α is a safety factor.

[0101] S54, from The system begins scanning forward from the starting point, searching for the first three consecutive points where the amplitude is greater than the current baseline average plus... The position is shifted forward by 0.5 s to serve as the corrected starting boundary. .

[0102] S55, from Starting from the first point, the system scans backwards, finding the last three consecutive points where the amplitude is greater than the current baseline average plus Threshold_noise. This position is then shifted back 0.5 seconds as the corrected termination boundary. .

[0103] S56, if in After or The amplitude exceeded the limit within the previous window. However, peaks with a width narrower than 1 / 3 of the minimum expected half-peak width are marked as noise spikes by the system and removed from the integration interval in the subsequent S6 integration calculation.

[0104] It should be noted that the safety factor α ranges from 1.5 to 3.0. This value ensures that noise spikes must be at least twice as high as the average noise fluctuation to be considered a signal. Three consecutive points are used to confirm that a fluctuation is a genuine signal rise rather than an occasional noise spike; the value of these three points corresponds to 0.3 s at a sampling frequency of 10 Hz. Shifting forward / backward by 0.5 s compensates for potential boundary lag caused by abrupt noise changes; this value can be scaled according to the peak width, typically from 1 / 20 to 1 / 10 of the full width at half maximum (FWHM).

[0105] In this embodiment, as Figure 2 As shown, the detailed implementation of step S6 is as follows: After completing the dynamic boundary correction in S5, the system obtains the final upper and lower limits of integration. and Integration is performed within the interval defined by these two time points.

[0106] It should be noted that in chromatographic analysis, the integral refers to the cumulative area of ​​the response signal curve on the time axis, and this area is proportional to the total radioactivity of the analyte passing through the detector.

[0107] Understandably, the original chromatographic peak signal is continuous time-series data. The area under discrete sampling points is typically calculated using the trapezoidal rule, approximating the total area under the curve as the sum of the areas of right trapezoids formed by multiple consecutive sampling points. To further improve integration accuracy, especially at the beginning and end of the signal rise and fall, special treatment of the boundaries is required.

[0108] Specifically, it includes the following sub-steps:

[0109] S61. Obtain the original response data stream in the interval. All data point sequences within .

[0110] S62, For the two endpoints outside the boundary and The values ​​are linearly interpolated to fall on a common baseline level, thus eliminating the effect of baseline drift on the integral area. The interpolation is based on... The previous 10 points average and The average of the next 10 points.

[0111] S63. Calculate the total area using the compound trapezoidal rule:

[0112] .in, That is, a constant sampling interval Δt.

[0113] S64. Subtract the area formed by the connection from the total area. and The trapezoidal area enclosed by the straight line and the time axis is obtained to get the net peak area after correcting the baseline drift.

[0114] It should be noted that the number of points n pre and n post The typical values are both 10. To adapt to long time series, all calculations are performed using double-precision floating-point numbers to maintain sufficient significant figures.

[0115] In this embodiment, the refinement of step S7 is as follows: According to the peak area of the target component obtained in S6 and the areas of all other impurity and by-product peaks, the radiochemical purity is calculated.

[0116] It should be noted that the radiochemical purity refers to the percentage of the radioactivity in the target radiopharmaceutical in the total radioactivity in the sample. The quality control determination result refers to the qualified or unqualified conclusion obtained by comparing the calculated radiochemical purity with the drug release standard pre-stored in the system.

[0117] It can be understood that the response of the radioactive detector is linear within a certain activity range. Therefore, the integrated area of the chromatographic peak directly represents the radioactivity of each component. By the ratio of the target peak area to the total peak area (the sum of all peak areas), the proportion of the target product in all radioactive substances, that is, the radiochemical purity, can be accurately calculated.

[0118] Specifically, it includes the following sub-steps:

[0119] S71. Summarize the net areas of all the peaks calculated in S6 to obtain the total sum .

[0120] S72. Identify and extract the area of the main peak (according to the preset retention time window) .

[0121] S73. Calculate the radiochemical purity .

[0122] S74. Read the preset radiochemical purity release standard of this product from the database .

[0123] S75. Compare the RCP with . If RCP ≥ , the determination result is qualified; if RCP < , the determination result is unqualified.

[0124] S76. Package the original chromatogram, the marked peaks after integration, the RCP value and the determination result into an electronic data interchange document and output it to a printer, a monitor or a quality control database.

[0125] It should be noted that the preset width of the retention time window for the main peak is ±5% or ±0.2 min of the target retention time, whichever is greater. The final RCP value is retained to two decimal places. The output electronic data exchange document format conforms to the electronic record requirements of the National Medical Products Administration for chromatographic data systems.

[0126] Example 3:

[0127] The technical solution in this embodiment is a further refinement based on the above embodiments, which elaborates on the specific mathematical implementation and internal logic of the dynamic window generation model.

[0128] The structure and operation mechanism of the dynamic window generation model:

[0129] The model is a physical model based on the kinematics principle of chromatography and consists of three functional modules: the leading edge inference module, the trailing edge inference module, and the boundary synchronization module.

[0130] The time t for the frontier inference module to receive sampling points within the frontier interval of the candidate peak i and its second derivative d2(t) i The input is ). The core of this module is a weighted linear regression estimator. Unlike ordinary least squares, this estimator assigns higher weights to the most recent time points to more sensitively reflect the dynamic changes at the onset of the peak. Its weighting function is: , where t n It represents the latest time point, and λ is the decay factor.

[0131] Furthermore, weighted linear regression was used to analyze the data points (t). i , d2(t i By fitting the equation, we obtain the linear equation describing the change of the second derivative with time: The fit is achieved by minimizing the weighted sum of squared residuals. This is achieved. Then, the module solves the equation. The theoretical zero-crossing time t of the second derivative is obtained. z According to kinematic principles, the critical point for peak initiation in a chromatogram actually occurs before acceleration begins. Therefore, the frontier extrapolation module uses an empirical equation to calculate t... z Transformation into the starting boundary of the peak : ,in is the standard error of the fitting parameter k, and γ is a relaxation coefficient. In this formula, This represents the confidence level of the fit; when the signal noise is high, Get bigger It will shift further to the left to ensure that the frontier is not lost. The value of γ ranges from 1.96 to 3.00, corresponding to a statistical confidence interval of 95% to 99.7%.

[0132] In this embodiment, the trailing edge inference module receives the sampling points within the trailing edge interval of the candidate peak at time t. i and its first derivative As input, the core of this module is the exponential decay fitter. An exponential decay model was chosen over a nonlinear model because the tailing behavior of chromatographic peaks theoretically more closely resembles the exponential decay process, especially in PET drug analysis, where tailing is common due to factors such as column adsorption. The fitted function is in the form of:

[0133] , where t p A is the time at the peak, and A is t. p The absolute value of the descent slope at a given point, where τ is a time constant representing the rate of decay.

[0134] The module uses the nonlinear least squares method to analyze the data points (t). i ,d(t i The module performs a fitting operation to solve for parameters A and τ, aiming to minimize the sum of squared residuals. After obtaining the optimal fitting parameters, the module sets an energy depletion threshold E. th Threshold E th It is not fixed, but rather the root mean square of the first derivative of the baseline noise obtained by fitting the peak front. Dynamic association: β is a dimensionless scaling factor. The value of β is based on the fact that the root mean square of the first derivative of the noise can quantify the background jitter intensity of the signal. When the absolute value of the first derivative of the peak signal decays to below this level, it means that the motion signal of the peak has been completely submerged by the background vibration. The value of β ranges from 1.0 to 2.0, preferably 1.5.

[0135] Module solving equations |, i.e., solving The theoretical peak termination time is obtained. .

[0136] The boundary synchronization module receives data from the frontier inference module. and the following inference module The main function of this module is to perform logical checks and, when necessary, corrections. If Later This is a logical error, and the module will trigger a signal quality alarm. Under normal circumstances, < This module is also responsible for handling rare cases where... or When the calculation result exceeds the preset physical limit (for example, Less than the sample injection time, or If the total analysis time exceeds the preset limit, reset it to these physical limits.

[0137] Example 4:

[0138] like Figure 3 As shown, this embodiment provides an automated chromatographic analysis system for PET drug quality control, used to implement any of the automated chromatographic analysis methods in Examples 1-3. The system includes:

[0139] The signal acquisition module is used to continuously acquire the raw response signal output by the detector during the chromatographic separation process, forming a response data stream that changes over time;

[0140] The morphological feature extraction module is used to extract the morphological feature parameters of the leading and trailing edges of candidate chromatographic peaks in real time during the transmission of the response data stream.

[0141] The window generation module is used to dynamically generate the start and end recognition time windows of a chromatographic peak based on the morphological characteristics of the leading and trailing edges of the candidate chromatographic peak.

[0142] The boundary correction module is used to establish a tracking correlation between the start and end boundaries of the chromatographic peak and the baseline noise fluctuation within the dynamic time window determined by the start and end identification time window, and to correct the position of the start and end boundaries in real time according to the instantaneous fluctuation of the baseline noise.

[0143] The integration calculation module is used to integrate the original response signal within the dynamic time window based on the corrected start and end boundaries, and calculate the chromatographic peak area of ​​each component.

[0144] The purity calculation and judgment module is used to calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

[0145] It should be noted that the automated chromatographic analysis system for PET drug quality control can perform the steps in the automated chromatographic analysis method for PET drug quality control as described in the above embodiments, and can achieve the same technical effect. Refer to the description in the above embodiments, which will not be elaborated here.

[0146] The above description is based on the preferred embodiments of the present invention. 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 invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered exemplary and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description, and all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0147] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains 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. An automated chromatographic analysis method for PET drug quality control, characterized in that, Includes the following steps: S1. During the chromatographic separation process, the raw response signal output by the detector is continuously acquired to form a response data stream that changes over time; S2. During the transmission of the response data stream, the morphological characteristic parameters of the leading and trailing edges of the candidate chromatographic peaks are extracted in real time. The morphological characteristic parameters include at least the rate of change of the response amplitude over time and the trend of the rate of change. S3. Based on the morphological characteristics of the candidate chromatographic peak front, dynamically generate the initial identification time window of the chromatographic peak, wherein the initial boundary of the initial identification time window corresponds to the position where the response signal first systematically deviates from the baseline noise fluctuation pattern. S4. Based on the morphological characteristics of the trailing edge of the candidate chromatographic peak, dynamically generate the termination identification time window for the chromatographic peak, wherein the termination boundary of the termination identification time window corresponds to the position where the response signal decays to the point where its rate of change is indistinguishable from the baseline noise fluctuation pattern. S5. Within the dynamic time window determined by the start and end identification time windows, establish a tracking correlation between the chromatographic peak start and end boundaries and baseline noise fluctuations, and correct the position of the start and end boundaries in real time according to the instantaneous fluctuations of the baseline noise. S6. Based on the corrected start and end boundaries, integrate the original response signal within the dynamic time window to calculate the chromatographic peak area of ​​each component. S7. Calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

2. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, The morphological characteristic parameters also include: the duration and amplitude increment of the peak leading edge from the starting point to the peak apex, and the duration and amplitude decrease of the peak trailing edge from the peak apex to the ending point.

3. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, In step S3, the specific method for dynamically generating the initial identification time window is as follows: based on the change trajectory of the rising slope of the peak front, a change curve reflecting the acceleration of the response signal from the baseline to the peak is fitted, and the position where the response signal first systematically deviates from the baseline noise fluctuation law is deduced based on the fitted curve as the theoretical starting point, and the theoretical starting point is determined as the starting boundary.

4. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, In step S4, the specific method for dynamically generating the termination identification time window is as follows: based on the change trajectory of the descent slope of the peak, a change curve reflecting the gradual decay of the response signal from the peak to the baseline is fitted. Based on the fitted curve, the position where the response signal decays to the point where its rate of change is indistinguishable from the baseline noise fluctuation pattern is deduced as the theoretical termination point, and this theoretical termination point is determined as the termination boundary.

5. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, In step S5, the specific method for establishing tracking association and real-time correction of start and end boundaries is as follows: real-time acquisition of the instantaneous fluctuation amplitude of baseline noise within the current time window, dynamic comparison of the instantaneous fluctuation amplitude with the baseline noise level at the peak front starting point, and temporary correction of the start and end boundary positions within the current time window based on the comparison results when the noise fluctuation amplitude shows a sudden change exceeding the expected range.

6. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, In step S6, before integration, linear interpolation is performed on the endpoints outside the boundary to make them fall on a common baseline level, so as to eliminate the effect of baseline drift on the integration area.

7. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, In step S7, the radiochemical purity value is calculated by: taking the ratio of the peak area of ​​the target component to the sum of the peak areas of all components to generate the radiochemical purity value, and comparing the radiochemical purity value with the product release standard to output the quality control judgment result.

8. The automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, The method for determining the baseline noise fluctuation pattern is as follows: by comparing the difference between the rate of change of the response amplitude within the current time window and the average rate of change of the previous non-peak interval, it is determined whether the trend of the rate of change has changed abruptly.

9. An automated chromatographic analysis method for PET drug quality control according to claim 1, characterized in that, Also includes: After dynamically generating the start and end recognition time windows, logical verification is performed on the start and end boundaries. When the start boundary is later than the end boundary, a signal quality alarm is triggered.

10. An automated chromatographic analysis system for PET drug quality control, characterized in that, For implementing the analytical method as described in any one of claims 1-9, the analytical system comprises: The signal acquisition module is used to continuously acquire the raw response signal output by the detector during the chromatographic separation process, forming a response data stream that changes over time; The morphological feature extraction module is used to extract the morphological feature parameters of the leading and trailing edges of candidate chromatographic peaks in real time during the transmission of the response data stream. The window generation module is used to dynamically generate the start and end recognition time windows of a chromatographic peak based on the morphological characteristics of the leading and trailing edges of the candidate chromatographic peak. The boundary correction module is used to establish a tracking correlation between the start and end boundaries of the chromatographic peak and the baseline noise fluctuation within the dynamic time window determined by the start and end identification time window, and to correct the position of the start and end boundaries in real time according to the instantaneous fluctuation of the baseline noise. The integration calculation module is used to integrate the original response signal within the dynamic time window based on the corrected start and end boundaries, and calculate the chromatographic peak area of ​​each component. The purity calculation and judgment module is used to calculate the radiochemical purity value based on the chromatographic peak area of ​​each component and output the quality control judgment result.

Images