A method and system for detecting impurities in an oral liquid composition

Abstract

The present application relates to the field of drug quality detection, and specifically discloses a method and system for detecting impurities in oral liquid components. The method and system are aimed at solving the problems of strong background interference and high subjectivity in the interpretation of oral liquid thin layer chromatography. The test solution, target component control solution and negative control solution lacking the target drug flavor are prepared, and the three solutions are subjected to thin layer chromatography imaging. The lane intensity distribution is extracted from the chromatogram image and the integral along the development direction is calculated to generate the test curve, control curve and negative curve. The first differential fingerprint based on the background subtraction of the negative curve is generated in parallel, and the second differential fingerprint based on the target response suppression of the control curve is generated. The abnormal features are extracted to form an impurity risk judgment and perform re-inspection and conclusion output according to the trigger conditions. Through complementary verification of the two differential fingerprints, interference suppression and abnormal highlighting are achieved, false negatives and missed detections are reduced, and the consistency and reviewability of the results are improved.

Description

Technical Field

[0001] This invention relates to the field of pharmaceutical quality testing, specifically to a method and system for detecting impurities in oral liquid components. Background Technology

[0002] Oral liquid preparations, especially compound traditional Chinese medicine oral liquids, typically contain multiple medicinal extracts and excipients, resulting in complex systems and significant matrix interference. To ensure product quality and medication safety, the production and testing processes require the detection of target components and the identification of key medicinal ingredients. Simultaneously, it is necessary to monitor the risks of impurities and abnormal components arising from production fluctuations, raw material differences, or degradation reactions. In current quality control practices, thin-layer chromatography (TLC) is widely used for the identification of medicinal ingredients and the detection of some components in traditional Chinese medicine preparations due to its low cost, ease of operation, and adaptability to multi-component systems.

[0003] However, existing TLC detection methods still have shortcomings in the identification and judgment of impurities and interferences. On the one hand, TLC results often rely on spot position (such as Rf-related position), color, and intensity as the basis for judgment. Actual interpretation often depends on manual experience or simple comparison, which is greatly affected by factors such as development conditions, color development conditions, sample volume, batch of thin-layer plates, imaging lighting, and imaging parameters. This can easily lead to spot position drift, differences in spot morphology, or changes in background noise, resulting in insufficient consistency in judgment. On the other hand, under the complex matrix conditions of compound oral liquids, excipients, other medicinal components, and their interactions may produce superimposed responses or false positive responses in the vicinity of the target peak. This makes it difficult to clearly distinguish between the target component response, matrix background, and non-target responses (suspected impurities). Relying solely on a single comparison can easily lead to misjudgment, misidentifying background interference as impurities, or masking true anomalies within the background and target responses.

[0004] Furthermore, existing TLC testing procedures have room for improvement in terms of retesting decisions and traceability. In actual testing, there is often a lack of quantifiable and verifiable criteria to support decisions regarding whether retesting is necessary when suspected abnormalities occur, what corrective measures should be used for retesting, and how to attribute the retesting results. This can easily lead to inconsistent retesting strategies, increased duplication of work, and difficulty in establishing a continuous accumulation and management mechanism for batch-to-batch drift and the characteristics of qualified sample groups. Even with the introduction of image acquisition and software assistance, without unified rules for the curve representation of TLC images, differential processing, and abnormal feature extraction, as well as a template or threshold update mechanism based on historical qualified data, it remains difficult to achieve stable, objective, and traceable impurity risk assessment in complex oral liquid systems. Summary of the Invention

[0005] To address the aforementioned technical problems in existing thin-layer chromatography (TLC) detection of oral liquid formulations, this invention proposes a method and system for detecting impurities in oral liquid components. Based on TLC detection, this invention transforms chromatographic images into calculable curve representations and incorporates background subtraction and target response suppression to highlight non-target responses. This allows for the extraction of abnormal features to form impurity risk assessment results, and triggers retesting and corrective retesting when necessary. This achieves objective identification and consistent control of impurities and interference risks, while also meeting the needs of target component content determination and batch-to-batch consistency management.

[0006] To achieve the above objectives, on the one hand, the present invention provides a method for detecting impurities in oral liquid components, the method comprising: preparing a test solution of the oral liquid to be tested, a reference solution of the target component, and a negative control solution lacking the target medicinal flavor; Thin-layer chromatography was performed on the three solutions and the corresponding chromatographic images were obtained; The image intensity distribution of each solution corresponding to the lane is extracted from the chromatographic image, and the lane image intensity is statistically integrated along the development direction to generate density curves, so as to obtain the test curve, control curve and negative curve corresponding to the three solutions respectively. The first differential fingerprint and the second differential fingerprint are generated in parallel. The first differential fingerprint is a differential curve obtained by background subtraction of the test curve using the negative curve as the background benchmark. The second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve to highlight the non-target response. Based on the first and second differential fingerprints, abnormal features are extracted and impurity risk judgment results are formed. When the impurity risk judgment results do not meet the re-inspection triggering conditions, an impurity detection conclusion is generated based on the impurity risk judgment results; otherwise, a re-inspection is performed, and an impurity detection conclusion is generated based on the re-inspection results.

[0007] In the above scheme, as an optional implementation, the generation of test curve, control curve and negative curve from chromatographic image includes: locating lanes in chromatographic image to determine the lane area corresponding to each solution, integrating the image intensity in the lane area along a direction perpendicular to the development direction to obtain a one-dimensional intensity sequence that varies with the development direction, and performing baseline correction and intensity normalization on the one-dimensional intensity sequence to form a density curve. It also includes using the target peak position of the control curve as an anchor point to align and correct the test curve and the negative curve, so that the target peak positions of each curve correspond to each other.

[0008] In the above scheme, as an optional implementation, extracting abnormal features based on the first differential fingerprint and forming an impurity risk judgment result includes: determining the baseline noise band and key interval in the differential curve corresponding to the first differential fingerprint, and detecting peak structure in the determined region to extract abnormal features, wherein the abnormal features include at least one of new peaks, peak shape anomalies, and key region intensity anomalies. Anomaly degree is generated based on abnormal features, and the anomaly degree is mapped to impurity risk assessment result, wherein the impurity risk assessment result includes at least normal result and abnormal result; When the impurity risk assessment result is normal, the first differential fingerprint is recorded as the qualified character after background subtraction and used to generate the impurity detection conclusion; when the impurity risk assessment result is abnormal, the newly added peak, peak shape deviation or exceeding limit features are output as the re-inspection target features and the re-inspection is triggered.

[0009] In the above scheme, as an optional implementation, the extraction of abnormal features based on the second differential fingerprint and the formation of impurity risk judgment results include: performing scale matching processing on the control curve based on the target peak response of the test curve, and performing compensation subtraction processing on the test curve to suppress the target response based on the scale-matched control curve, so as to obtain the differential curve corresponding to the second differential fingerprint, determining the target peak proximity interval and the preset interest interval in the differential curve, and detecting the peak structure within the interval to extract abnormal features, wherein the abnormal features include at least one of the following: newly added peaks near the target peak, abnormal intensity of non-target peaks, and abnormal peak shape; Anomaly degree is generated based on abnormal features, and the anomaly degree is mapped to impurity risk assessment result, wherein the impurity risk assessment result includes at least normal result and abnormal result; When the impurity risk assessment result is normal, an impurity detection conclusion is generated based on the second differential fingerprint; when the impurity risk assessment result is abnormal, the newly added peak near the target peak or the non-target peak exceeding the limit is output as the re-inspection target feature and the re-inspection is triggered.

[0010] In the above scheme, as an optional implementation, the negative control solution is prepared using the same preparation process as the oral liquid to be tested, but without the target medicinal flavor. The target medicinal flavor includes at least one of motherwort and angelica.

[0011] In the above scheme, as an optional implementation, the re-inspection includes: when the impurity risk determination result is an abnormal result, performing a re-inspection under the same conditions and determining whether the abnormal characteristics corresponding to the abnormal result are reproduced in the re-inspection result; when the abnormal characteristics are reproduced in the re-inspection result under the same conditions, further performing a corrective re-inspection, wherein the corrective re-inspection is performed by changing at least one of the thin-layer chromatography processing conditions, wherein the thin-layer chromatography processing conditions include the developing system, the spotting volume, the sample pretreatment, the color development conditions, and the imaging conditions; based on the results of the re-inspection under the same conditions and the corrective re-inspection, an abnormality attribution conclusion is output, wherein the abnormality attribution conclusion includes abnormal method conditions and suspected impurity abnormalities.

[0012] In the above scheme, as an optional implementation, when the impurity risk judgment result does not trigger a retest or the retest result is normal, a target component content determination step is also included. The step includes: determining the integration interval corresponding to the target peak of the reference curve in the test curve or the first differential fingerprint; integrating the peak intensity within the integration interval to obtain the peak area or integrated intensity of the target peak; establishing a calibration relationship based on the peak area or integrated intensity of the target peak obtained at different concentrations of the target component reference solution; and substituting the peak area or integrated intensity of the target peak of the test curve into the calibration relationship to calculate the target component content.

[0013] In the above scheme, as an optional implementation, a step of constructing a reference template and determining the positional offset range is also included. The step includes: selecting multiple batches of reference curves that have been confirmed to be normal by impurity detection as template samples; aligning each template sample with the target peak as the anchor point and extracting the target peak position parameters; determining the standard position of the target peak and the allowable positional offset range based on the statistical results of the target peak position parameters; before generating the first differential fingerprint and the second differential fingerprint, acquiring the currently detected reference curve and extracting its target peak position parameters, and calculating its positional offset relative to the standard position of the target peak; triggering a re-inspection when the positional offset exceeds the allowable positional offset range, otherwise proceeding to the differential fingerprint generation and impurity risk determination process.

[0014] In the above scheme, as an optional implementation method, it also includes a qualified batch fingerprint database construction and update step, the step of which includes: when the impurity detection conclusion is a normal result or is determined to be a normal result after re-inspection, the control curve, the first differential fingerprint and the second differential fingerprint of the current batch are recorded as qualified fingerprints and stored in the database. Based on multiple batches of qualified fingerprint records in the fingerprint database, statistical summaries are performed on the control curve, the first differential fingerprint, and the second differential fingerprint to construct a qualified batch reference template. Based on the statistical summaries, a consistency threshold parameter is determined. The qualified batch reference template includes at least one of the target peak standard position and the reference morphology parameter of the differential curve. The consistency threshold parameter includes at least one of the allowable position offset range and the allowable anomaly range. For subsequent batches of samples, the target peak position offset is calculated based on its reference curve, and the anomaly is calculated based on its first differential fingerprint and second differential fingerprint. The target peak position offset and anomaly are compared with the corresponding consistency threshold parameters to obtain the consistency judgment result. After qualified fingerprint records are entered into the database, the qualified batch reference template and consistency threshold parameters are updated.

[0015] On the other hand, the present invention also provides an oral liquid component impurity detection system, the system comprising: The sample preparation module is used to prepare the test solution of the oral liquid to be tested, the reference solution of the target component, and the negative control solution lacking the target drug flavor; The thin-layer chromatography imaging module is used to perform thin-layer chromatography processing on three solutions and acquire the corresponding chromatographic images; The curve generation module is used to extract the image intensity distribution of each solution corresponding to the lane from the chromatographic image, and to perform statistical integration of the lane image intensity along the expansion direction to generate density curves, so as to obtain the test curve, control curve and negative curve corresponding to the three solutions respectively. The differential fingerprint generation module is used to generate a first differential fingerprint and a second differential fingerprint in parallel. The first differential fingerprint is a differential curve obtained by subtracting the background of the test curve using the negative curve as a background benchmark. The second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve to highlight the non-target response. The impurity risk assessment module is used to extract abnormal features based on the first differential fingerprint and the second differential fingerprint and form an impurity risk assessment result. The re-inspection control module is used to determine whether the impurity risk assessment result meets the re-inspection triggering conditions. If so, it generates an impurity detection conclusion based on the impurity risk assessment result; otherwise, it performs a re-inspection and generates an impurity detection conclusion based on the re-inspection result.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: 1) By converting chromatographic images into density curves, background subtraction difference curves based on negative curves and target response inhibition difference curves based on control curves are constructed in parallel, so that background interference and non-target responses under complex matrices can be separated and characterized, thereby improving the objectivity and verifiability of impurity risk identification. 2) By mapping the results of abnormal feature extraction, abnormality generation and impurity risk assessment, a quantifiable basis for judgment is formed, reducing reliance on human experience and improving the consistency of judgment. 3) Through a two-level mechanism of re-inspection under the same conditions and corrective re-inspection, as well as the output of anomaly attribution conclusions, the anomaly handling has a closed-loop logic, reducing the cost of repeated re-inspection and improving traceability. 4) By establishing reference templates and allowable positional offset ranges, and continuously updating the qualified batch fingerprint database, the management of batch drift and consistency thresholds can be achieved, thereby improving the stability of long-term operation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for detecting impurities in oral liquid components according to one embodiment of the present invention; Figure 2 This is a structural diagram of an oral liquid component impurity detection system shown in one embodiment of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments of the present invention are only used to explain the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make various equivalent transformations or substitutions to the step sequence, parameter selection, module division, and implementation method of the present invention without departing from the spirit and substance of the present invention, and all such transformations or substitutions should fall within the scope of protection of the present invention. For ease of description, some operations in the following embodiments are given in a functional manner, and their specific implementation can be accomplished by software algorithms, hardware circuits, or a combination of both; without contradiction, the technical features between the various embodiments can be combined or substituted with each other.

[0019] In one embodiment of the present invention, such as Figure 1 As shown, this invention provides a method for detecting impurities in oral liquid components. The method includes: preparing a test solution of the oral liquid to be tested, a reference solution of the target component, and a negative control solution lacking the target medicinal flavor; performing thin-layer chromatography on the three solutions and acquiring corresponding chromatographic images; extracting the image intensity distribution of the corresponding lanes of each solution from the chromatographic images, and statistically integrating the lane image intensity along the development direction to generate density curves, so as to obtain the test curve, control curve, and negative curve corresponding to the three solutions respectively; generating a first differential fingerprint and a second differential fingerprint in parallel. The first differential fingerprint is a differential curve obtained by background subtraction of the test curve using the negative curve as a background benchmark, and the second differential fingerprint is a differential curve obtained by compensating for target response suppression of the test curve based on the control curve to highlight the non-target response; extracting abnormal features based on the first and second differential fingerprints and forming an impurity risk judgment result. When the impurity risk judgment result does not meet the retest triggering condition, an impurity detection conclusion is generated based on the impurity risk judgment result; otherwise, a retest is performed, and an impurity detection conclusion is generated based on the retest result.

[0020] In this embodiment, the test solution is a solution obtained by treating the oral liquid sample to be tested, used for thin-layer chromatography detection. It represents the overall response of the actual production sample, including the target component response, matrix background response, and possible non-target responses (including responses from suspected impurities or abnormal components). The target component reference solution is a standard reference solution containing the target component, used to provide a typical response characterization of the target component under thin-layer chromatography conditions. In a typical application scenario of Compound Yimu Oral Liquid, the target component may be stachydrine derived from Leonurus japonicus. The target component reference solution can be prepared from stachydrine reference standard using a specified solvent to form a control curve and serve as a reference for the target response. The negative control solution is a control solution lacking the target medicinal flavor, used to characterize the background response of the prescription matrix or excipients and other medicinal flavors to the thin-layer chromatography results in the absence of the target medicinal flavor contribution. In this embodiment, the negative control solution can be understood as: a control solution formed by removing the target medicinal flavor, for example, removing Leonurus japonicus and Angelica sinensis from Compound Yimu Oral Liquid, while maintaining the same or comparable preparation process and matrix system as the oral liquid to be tested. This setting allows subsequent background subtraction to have a clear background benchmark, thereby avoiding the subjective uncertainty of relying solely on experience to judge whether something looks like interference or not.

[0021] By setting up the above three solutions in parallel, the test solution provides a true mixed response, the target component reference solution provides a target response reference, and the negative control solution provides a background response reference. The results obtained by the three under the same thin-layer chromatography conditions can be used for subsequent differential fingerprinting construction and impurity risk assessment.

[0022] In this embodiment, thin-layer chromatography (TLC) is performed on three solutions to obtain chromatographic images. This can be achieved using conventional TLC procedures in the art. The purpose is not to limit the specific developing system or chromogenic reagent, but to obtain chromatographic image data that reflects the separation results of each solution on the TLC plate, so as to facilitate subsequent digital processing and curve characterization. Specifically, the TLC process may include: spotting the test solution, the target component reference solution, and the negative control solution at the starting positions of the corresponding lanes on the TLC plate, then developing and separating them under the same developing system, and after chromogenic or visualization processing, imaging the TLC plate to obtain chromatographic images. The chromatographic images can be image data acquired under visible light, ultraviolet light, or other imaging conditions. The acquisition method can be a camera or scanner, etc. The core requirement is that the image can clearly show the spot or band response information of each lane along the development direction, so as to extract the intensity distribution of the lane image. In the application of Compound Motherwort Oral Liquid, this thin-layer chromatography process can be used to: on the one hand, form a control reference based on the response of the target components related to Motherwort (such as stachydrine); on the other hand, form an identification response based on the medicinal flavors such as Angelica sinensis; at the same time, since the compound system has multiple medicinal flavors and excipients in the background, the above image data also carries the background and non-target response information required for subsequent differential fingerprint analysis.

[0023] In this embodiment, the lane image intensity distribution is extracted from the chromatographic image, and the lane image intensity is statistically integrated along the expansion direction to generate a density curve. This curve transforms the lane response in the two-dimensional image into a one-dimensional curve characterization that is easy to calculate and compare. The basic idea is to first obtain the image intensity distribution of each pixel within the lane after determining the lane region in the image for each solution; then, using the expansion direction as the independent variable, the image intensity along the lane width direction perpendicular to the expansion direction is accumulated or integrated to form a one-dimensional intensity sequence that varies with the expansion direction. This one-dimensional intensity sequence exhibits a "peak-valley" structure along the expansion direction, with peak values ​​corresponding to enhanced spot or band-like responses at a certain location in the lane, and valley values ​​corresponding to background or non-response areas. Since this curve is obtained by statistical integration of lane intensity, it is called a density curve in this embodiment, and test curves, control curves, and negative curves are obtained respectively.

[0024] By converting images into density curves, this embodiment transforms the differences in spot position, intensity, and shape that originally relied on visual observation into curve differences that can be directly calculated, providing a unified data foundation for subsequent differential fingerprint construction and anomaly feature extraction.

[0025] In this embodiment, the first differential fingerprint is a differential curve obtained by background subtraction of the test curve using the negative curve as a background benchmark. Its core function is to offset background responses from the formulation matrix, excipients, and non-target medicinal ingredients as much as possible while maintaining the overall information of the test curve. This allows the remaining curve to more concentratedly reflect the new or enhanced responses relative to the background benchmark, thereby facilitating the identification of anomalous features such as new peaks, abnormal peak shapes, or abnormal intensity in key regions. Specifically, the negative curve can be understood as the background curve generated by the formulation system itself under conditions lacking the contribution of the target medicinal ingredient. When certain response structures appear in the test curve, and these structures do not correspond to or are significantly weaker in the negative curve, the first differential fingerprint after background subtraction will highlight these structures more, thereby reducing the masking effect of complex matrix on anomaly identification. Conversely, for background responses that coexist with the test curve and the negative curve and have similar shapes, background subtraction can offset them to a certain extent, reducing the risk of misjudging the background as impurities. It should be understood that the background subtraction is not required to be limited to a certain fixed numerical algorithm. Those skilled in the art can perform differential processing based on curve data to achieve the effect of subtraction with the negative curve as the background benchmark. The result is based on the first differential fingerprint differential curve used for anomaly identification.

[0026] In this embodiment, the second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve, in order to highlight the non-target response. Its core function is that when the target component in the test curve has a strong response, and a weak non-target peak may be superimposed near the target peak, background subtraction alone may not be enough to highlight the abnormal structure near the target peak. Therefore, by introducing the control curve as a reference for the target response and using a compensating subtraction method to suppress the target response, the influence of the target peak can be weakened, thereby highlighting newly added peaks or abnormal non-target peaks near the target peak. Specifically, the control curve provides the typical response morphology of the target component under the current thin-layer chromatography conditions. By performing compensating matching between the control curve and the target response of the test curve, and then subtracting the matched control response from the test curve, the result achieves a differential curve effect where the target response is suppressed and the non-target response is more prominent. Thus, if there are abnormal newly added peaks, peak shoulders, double peaks, abnormal tailing, or abnormal non-target peak intensities in the region near the target peak, the second differential fingerprint will more easily highlight them, avoiding the strong response of the target peak from masking the abnormal peak. Similarly, the compensation deduction process is not limited to a single algorithm implementation. The key lies in its processing purpose and processing result. The target response is suppressed by using the comparison curve as a reference, and the differential curve used for anomaly identification of the second differential fingerprint is output as the standard.

[0027] In this embodiment, extracting anomalous features based on the first and second differential fingerprints to form impurity risk determination results is the key to achieving objective impurity risk identification in this invention. Anomalous features can be obtained from the peak structure and intensity distribution of the differential curve. For example, when a new peak appears in the differential curve that is still significantly present relative to the reference background or target suppression; or when anomalies in peak shape appear that are inconsistent with the expected peak shape, such as obvious broadening, tailing, splitting, or peak shoulders; or when an intensity anomaly exceeding the normal background level appears in the critical interval, these can all be considered as anomalous features. Since the first and second differential fingerprints are constructed for different purposes—the former focuses more on the newly added or enhanced response after background subtraction, while the latter focuses more on the prominence of non-target responses after target response suppression—the parallel use of both can complementaryly detect anomalies from different angles, thereby improving the coverage and reliability of impurity risk identification.

[0028] After obtaining the abnormal features, an impurity risk assessment result can be generated and compared with the re-inspection triggering conditions: if the impurity risk assessment result does not meet the re-inspection triggering conditions, it indicates that no abnormal features requiring further confirmation were observed under the current testing conditions, or the degree of abnormality is insufficient to trigger a re-inspection, and an impurity detection conclusion is generated based on this assessment result; if the impurity risk assessment result meets the re-inspection triggering conditions, a re-inspection is performed to reconfirm the abnormal features, and an impurity detection conclusion is generated based on the re-inspection result. Through the structure of differential fingerprinting parallel processing, abnormal feature extraction, risk assessment, and re-inspection, this embodiment can transform traditional verbal judgment into a verifiable judgment process based on curves and differential results.

[0029] In this embodiment, the retest is used to verify and confirm the abnormal features corresponding to the triggering conditions. The retest can be understood as re-executing the above-mentioned thin-layer chromatography processing, imaging, and curve generation process under comparable detection conditions, and comparing the difference curve obtained from the retest with the abnormal features. When the retest result supports the continued existence of the abnormal features, the abnormality can be regarded as an abnormal signal with stable or repeatable characteristics, and an impurity detection conclusion can be generated accordingly. When the retest result no longer presents the abnormal features, the abnormality can be regarded as an abnormal signal that has not been confirmed by the retest, and a corresponding impurity detection conclusion can be generated in conjunction with the impurity risk assessment result. Therefore, this embodiment, while maintaining a compact process, achieves control from initial detection to retest, ensuring that the impurity conclusion output has a traceable basis for judgment.

[0030] In thin-layer chromatography (TLC), the plate image directly presents the color, shape, and position of spots. However, in complex matrix scenarios such as compound oral liquids, fluctuations in spot brightness, uneven background, and slight positional drift can easily make stable and verifiable comparisons between different lanes difficult. To ensure that subsequent background subtraction and target response suppression are based on a unified quantitative representation, this embodiment converts the response of each lane in the TLC image into a one-dimensional curve representation along the development direction. By using the target peak of the control lane as an anchor point to align the positions of other lanes, spot comparison is transformed into curve comparison, reducing the impact of imaging conditions, colorimetric differences, and slight development deviations on the consistency of judgment.

[0031] In one embodiment of the present invention, for image data after thin-layer chromatography imaging, a method is provided to convert the chromatographic image into a test curve, a control curve, and a negative curve and perform alignment correction, including: locating lanes in the chromatographic image to determine the lane regions corresponding to each solution; extracting a one-dimensional intensity sequence varying along the development direction based on the lane regions and forming a density curve through baseline correction and intensity normalization, thereby obtaining the test curve, control curve, and negative curve; subsequently, using the target peak position of the control curve as an anchor point, performing alignment correction on the test curve and the negative curve to ensure that the three curves correspond consistently at the target peak position, so as to ensure that subsequent differential processing and anomaly identification are established under the same position coordinate system.

[0032] In this embodiment, the chromatographic image is a plate image obtained after thin-layer chromatography processing on the same plate. The image typically contains multiple lanes arranged along the same development direction. To improve the stability of lane positioning and curve generation, the chromatographic image can be first oriented and the effective region can be cropped to ensure that the development direction of the image remains consistent in the image coordinate system, for example, uniformly from bottom to top or from left to right, and to minimize the exclusion of areas unrelated to the lanes, such as plate edges, labels, and reflections. This preprocessing is not limited to a specific algorithm and can be implemented by manually presetting the cropping frame or by software automatically detecting the plate surface boundaries. Its purpose is to provide a stable spatial reference for subsequent statistical integration along the development direction.

[0033] In this embodiment, lane positioning is used to determine the lane regions corresponding to the test sample, reference sample, and negative control in the chromatographic image. A lane region can be understood as the projection area in the image of the strip-shaped development channel formed after the same solution is spotted and developed on a thin-layer plate. Specifically, lane positioning can be achieved using the following approaches: 1. Positioning based on plate layout: In common quality control processes, the three solutions are often spotted on the same plate in a fixed order with consistent relative spacing. The system can directly delineate each lane region in the image according to the preset number of lanes, lane center-to-center spacing, and lane width; 2. Positioning based on image intensity distribution: The image intensity changes are statistically analyzed along a direction perpendicular to the development direction. The difference in overall intensity or texture between the lane region and adjacent blank areas is used to determine the lane centerline and boundary, thus obtaining the lane region; 3. Positioning based on markers or reference lines: If the thin-layer plate or acquisition device provides fixed reference boundaries or markers, the relative positions of the lane regions can be determined accordingly. In practice, the output of lane localization includes at least the area of ​​each lane, for example determined by the lane centerline and lane width, in order to extract image intensity within that area and generate a one-dimensional sequence.

[0034] After obtaining the lane regions, this embodiment characterizes the image intensity within the lane regions using a curve representation along the unfolding direction. Specifically, using the unfolding direction as the position axis of the curve, at each position point, the image intensity on the corresponding lane cross-section (i.e., the direction perpendicular to the unfolding direction) is accumulated or integrated to obtain a one-dimensional intensity sequence that varies with the unfolding direction. The intuitive meaning of this process is to transform the speckled or banded responses in a two-dimensional image into peak structures in a one-dimensional curve. To ensure consistent peak structure orientation—that is, a stronger response results in a higher curve—the meaning of intensity can be uniformly processed according to the imaging method. For example, if a speckle appears brighter in the image, brightness intensity is directly used; if a speckle appears darker, it can be converted to darkness intensity before integration, causing the curve corresponding to enhanced response to rise. Thus, one-dimensional intensity sequences corresponding to the test lane, control lane, and negative lane can be obtained respectively.

[0035] Due to factors such as thin-layer plate noise, uneven color development, gradual illumination, or scanning shadows, one-dimensional intensity sequences often exhibit slowly fluctuating backgrounds that change with location. This embodiment performs baseline correction on the one-dimensional intensity sequence to mitigate the interference of these background fluctuations on peak structure extraction and comparison. Baseline correction can be achieved using conventional baseline estimation and subtraction methods in the art, such as estimating and subtracting the background level in regions without significant peaks, or using a smoothed low-frequency trend as the baseline for correction. The result of baseline correction should make the curve closer to the stable background level in non-response regions, thereby making the peak structure more prominent and facilitating subsequent location of target and anomalous peaks.

[0036] Differences in batches, shooting parameters, or color intensity may lead to scale differences in the overall amplitude of the curves. To ensure comparability of the three curves, this embodiment performs intensity normalization processing on the baseline-corrected one-dimensional intensity sequence to form density curves. The purpose of intensity normalization is to unify the intensity scale of the curves to a comparable range. This can be achieved by methods such as scaling according to the effective intensity range of the curve, scaling according to a certain stable reference interval, or standardizing according to the overall energy of the curve. This processing does not limit the specific normalization strategy, but it should compress the intensity scale differences of the curves in different lanes without changing the peak position information, so as to make subsequent peak structure detection and alignment more stable. After completing baseline correction and intensity normalization, three density curves are obtained: the test curve, the control curve, and the negative curve.

[0037] In actual thin-layer chromatography, even with the same plate operation, slight differences in development distance, slight plate tilt, or solvent front differences can still cause overall peak position shifts. To ensure effective comparison of curves at the same position, this embodiment uses the target peak position of the control curve as an anchor point to align and correct the test curve and negative curve, ensuring that the target peak positions of each curve correspond consistently. Specifically, the target peak position can be determined first in the control curve. The target peak can be determined based on a predetermined identification method for the target component. For example, for the detection of motherwort-related target components in Compound Motherwort Oral Liquid, the target peak can correspond to the typical peak position range of the target component under the thin-layer chromatography conditions; the system can search for peak structures within this range and determine the center position of the target peak. Subsequently, the same target peak search strategy is used in the test curve and negative curve to obtain their respective target peak positions. Alignment correction can be achieved by position translation, local scaling, or equivalent coordinate correction methods to adjust the target peak positions of the test curve and negative curve to be consistent with the target peak position of the control curve, thereby achieving alignment of the three curves at key reference points.

[0038] The key to the above alignment correction is to establish a unified position coordinate using the same anchor point semantics (i.e. the target peak position of the comparison curve), so that subsequent background subtraction or target response suppression processing can be carried out within the same position frame to compare peak structures and avoid mistaking position drift as a new peak or an abnormal peak.

[0039] After completing lane positioning, curve generation, baseline correction, intensity normalization, and anchor point alignment, this embodiment outputs three aligned density curves: the test curve, the control curve, and the negative curve. This output can serve as unified input data for subsequent differential fingerprint generation, abnormal feature extraction, and impurity risk assessment in this invention. This transforms the entire detection process from intuitive image interpretation into a curve-based, structured, and verifiable processing chain, thereby improving the consistency and traceability of judgments in complex oral liquid matrix scenarios.

[0040] In one embodiment of the present invention, addressing the problems of strong matrix background, numerous interfering components, and strong subjectivity in visual interpretation during thin-layer chromatography detection of oral liquids, this embodiment provides an anomaly identification and impurity risk assessment process based on a first differential fingerprint. The process includes: after generating and aligning the test curve, control curve, and negative curve, subtracting the background of the test curve using the negative curve as a background benchmark to obtain the differential curve corresponding to the first differential fingerprint; determining the baseline noise band and key interval in the differential curve corresponding to the first differential fingerprint, and detecting peak structures in the determined regions to extract abnormal features, wherein the abnormal features include at least one of new peaks, abnormal peak shapes, and abnormal intensity in key regions; generating an anomaly degree based on the abnormal features, and mapping the anomaly degree to an impurity risk assessment result, wherein the impurity risk assessment result includes at least normal results and abnormal results; when the impurity risk assessment result is a normal result, recording the first differential fingerprint as a qualified characterization after background subtraction and using it to generate an impurity detection conclusion; when the impurity risk assessment result is an abnormal result, outputting new peaks, peak shape deviations, or exceeding limits as re-inspection target features and triggering re-inspection.

[0041] In this embodiment, the purpose of the first differential fingerprint is to highlight the new or enhanced responses in the test sample relative to the background baseline, and to offset the common background caused by the prescription matrix, excipients, and non-target drug flavors as much as possible. Specifically, when the test curve and the negative curve are in the same coordinate system, for example, the target peak positions correspond, the negative curve is used as the background baseline to perform background subtraction processing on the test curve, and a differential curve that varies with the expansion direction is output. When there is no abnormal response, the differential curve should mainly show fluctuations close to the noise level, while when there are new peaks, changes in peak shape, or local intensity anomalies, detectable peak structure or intensity deviations will appear at the corresponding positions.

[0042] In this embodiment, the baseline noise band is used to characterize the noise level in areas where the difference curve should theoretically not show a valid response or only exhibit random fluctuations, thus providing a reference for subsequent anomaly determination. The baseline noise band can be determined as follows: 1. Based on blank area selection, select areas that do not contain major spot responses in the thin-layer plate unfolding direction as noise bands, such as segments near the starting point but avoiding the spot diffusion area, and stable segments near the solvent front but avoiding frontal anomalies; 2. Based on selection outside the scope of business concern, select segments outside the typical response range of the target odor or target component and with sparse responses as noise bands; 3. Based on low response screening, select a continuous segment with small intensity fluctuations and no obvious peak structure in the difference curve as a noise band. After determining the noise band, the fluctuation amplitude of the difference curve within the noise band can be statistically analyzed to characterize the noise level, such as obtaining the typical fluctuation range, upper fluctuation limit, or stable background threshold of the noise band. This threshold is not limited to a specific statistical method; its function is to provide a quantitative reference for what level constitutes an anomaly.

[0043] In this embodiment, the critical interval is used to focus on the range of differential curves that require special attention, so as to avoid misjudging random fluctuations in irrelevant areas as anomalies, and also to facilitate the location of anomalies to a clearly defined segment that can be re-examined. The critical interval can be determined according to the following principles: 1. Set around a predetermined area of ​​concern. According to the usual practices of quality control of compound oral liquids, the segment where the target component or key medicinal flavor related response is located can be set as the key area of ​​concern. For example, a certain width of the area of ​​concern can be set around the typical response position of the target component; 2. Exclude unstable edge areas. The unstable segments such as the starting diffusion area and the solvent front easily fluctuating area can be removed from the critical interval, and only the middle expansion segment is retained as the main judgment area; 3. Set a fixed segment according to process experience. Under the same product and the same thin-layer system, the critical interval can remain consistent for a long time to match the internal process interpretation framework of the enterprise. Once the critical interval is determined, subsequent peak structure detection and anomaly feature extraction are carried out within this interval or based on this interval, so as to make the anomaly judgment more controllable and verifiable.

[0044] In this embodiment, after determining the baseline noise band and key intervals, peak structure detection is performed on the difference curve to extract feature information for characterizing anomalies. Peak structure detection can be implemented using common curve peak detection methods in the field, such as smoothing and denoising the difference curve, searching for local maxima, and combining parameters such as peak width, peak height, peak area, or peak symmetry to confirm the existence of peak structure.

[0045] Based on this, the abnormal features extracted in this embodiment include at least one of the following: New peak: A response with a clear peak structure that exceeds the noise level within the critical interval, and the peak structure does not conform to the characteristics of random fluctuations in the background; Abnormal peak shape: Significant peak shape deviation in the difference curve, such as significant peak broadening, the appearance of shoulder peaks, the appearance of splitting trends, or abnormal tailing, which can characterize peak shape changes; Abnormal intensity in the critical region: A sustained increase in intensity, enhanced abnormal fluctuations, or local intensity exceeding limits in the difference curve within the critical interval, even if it does not form a typical sharp peak, can be considered an abnormal response.

[0046] To facilitate re-inspection and localization, the output of abnormal features should typically include at least: the location segment where the abnormality is located, the type of abnormality (such as setting a new peak / peak shape abnormality / intensity abnormality), and the degree of abnormality (such as the degree of exceeding the limit of peak height relative to noise level or the length of continuous exceeding the limit).

[0047] This embodiment generates anomaly scores based on the aforementioned abnormal features. Anomaly scores can be understood as a quantitative representation of the degree of abnormality, used to unify multiple types of abnormal features onto a quantifiable scale. The generation of anomaly scores can comprehensively consider at least one of the following factors: the degree to which the abnormal peak exceeds the noise level, the width or area of ​​the abnormal peak, the degree of peak shape deviation, and the length or amplitude of sustained exceedance in the critical interval. After obtaining the anomaly score, it is compared with a preset threshold or judgment rule to output an impurity risk judgment result. The impurity risk judgment result includes at least normal and abnormal results. A normal result is output when the anomaly score does not reach the trigger threshold or does not meet the anomaly judgment rule; an abnormal result is output when the anomaly score reaches the trigger threshold or meets the anomaly judgment rule. The core of this mapping process is to transform the visible anomalies of the difference curve into verifiable judgment outputs, thereby reducing inconsistencies in subjective human judgment.

[0048] When the impurity risk assessment result is normal, this embodiment uses the first differential fingerprint as a qualified characterization after background subtraction to participate in the generation of impurity detection conclusion. This means that no abnormal response requiring further confirmation was found from the perspective of background subtraction, and the corresponding impurity detection conclusion can be directly output.

[0049] When the impurity risk assessment result is abnormal, this embodiment outputs the abnormal features as re-inspection target features and triggers a re-inspection. The re-inspection target features include at least newly added peak position segments, peak shape deviation segments, or intensity exceeding limits segments, enabling the re-inspection to address specific issues and avoid merely repeating the process without pinpointing and confirming the anomaly. After the re-inspection is triggered, subsequent processes can re-acquire the differential curve and check whether the re-inspection target features still appear to support the formation of the final impurity detection conclusion.

[0050] In one embodiment of the present invention, addressing the issue that in the thin-layer chromatography detection of oral liquids, the target component response is strong, easily masks the weak response of neighboring components, and may exhibit co-migration or adjacent interference, this embodiment provides an anomaly identification and impurity risk assessment process that uses the control response as a reference to suppress the target component response in the test response and highlight non-target responses. By performing scale matching between the test curve and the control curve at the target peak, and then performing compensation subtraction processing on the test curve based on the matched control curve to obtain the difference curve corresponding to the second difference fingerprint; then, abnormal features are extracted around the target peak's adjacent region and a preset region of interest to form a judgment output, thereby enabling suspicious minor anomalies near the target peak to be stably captured and located in a curve-like manner, providing a clear verification target for subsequent re-examination. Specifically, the process of extracting abnormal features and forming an impurity risk assessment result based on the second differential fingerprint includes: performing scale matching processing on the control curve based on the target peak response of the test curve, and performing compensation subtraction processing on the test curve to suppress the target response, thereby obtaining the differential curve corresponding to the second differential fingerprint; determining the target peak proximity interval and the preset interest interval in the differential curve, and detecting peak structure within the interval to extract abnormal features, wherein the abnormal features include at least one of the following: newly added peaks near the target peak, abnormal intensity of non-target peaks, and abnormal peak shape; generating an anomaly degree based on the abnormal features, and mapping the anomaly degree to an impurity risk assessment result, wherein the impurity risk assessment result includes at least normal results and abnormal results; when the impurity risk assessment result is a normal result, generating an impurity detection conclusion based on the second differential fingerprint; when the impurity risk assessment result is an abnormal result, outputting the newly added peaks near the target peak or the non-target peak exceeding the limit feature as the re-inspection target feature and triggering a re-inspection.

[0051] In this embodiment, the purpose of the second differential fingerprint is to weaken the dominant response of the target component in the test curve as much as possible without changing the curve position coordinate system, thereby highlighting non-target responses that are obscured by the target peak, such as adjacent weak peaks, shoulder peaks, adjacent small peaks, or local abnormal elevations. Unlike the processing that uses a negative background as a benchmark, the second differential fingerprint emphasizes suppressing the target peak with the control response as a reference, and is especially suitable for scenarios where the target peak is strong and it is necessary to pay more attention to whether there are accompanying abnormalities near the target peak.

[0052] In this embodiment, scale matching is used to eliminate the difference in intensity scale between the test curve and the control curve, making them comparable at the target peak. Specifically, the response characteristics of the target peak can be determined in both the test and control curves first, such as the peak height, peak area, or integral intensity of the target peak interval. Then, the control curve is proportionally adjusted so that its response level at the target peak is close to that of the test curve. The effect of this is that when the test curve is subsequently subtracted, the main response consistent with the target peak is primarily weakened, rather than the overall intensity difference being mistakenly treated as an anomaly.

[0053] To avoid the impact of noise on scale matching, stable peak structure parameters can be selected near the target peak range for matching. The curve can be smoothed or stabilized before matching, but the consistency of the target peak position coordinates will not be changed.

[0054] In this embodiment, the compensation subtraction process for target response suppression involves subtracting from the test curve based on the scale-matched control curve to suppress the target peak. Simultaneously, a compensation mechanism is introduced to reduce the risk of false negative peaks due to over-subtraction or residual target peak interference due to under-subtraction. This can be understood as follows: the subtraction focuses on the response consistent with the target peak, while compensation corrects for potential local baseline shifts, local morphological differences, or slight tailing differences introduced during the subtraction process, making the resulting difference curve more accurately reflect the true existence and extent of non-target responses. The compensation method is not limited to a single algorithm; it can include local baseline recalibration, local intensity fine-tuning, or applying different subtraction weights to the target peak region and non-target regions, as long as the effect is that the target peak is significantly suppressed, while non-target peaks or abnormal rises in the adjacent region are more easily detected. After the compensation subtraction is completed, the difference curve corresponding to the second difference fingerprint is obtained.

[0055] In this embodiment, to make anomaly identification more focused and facilitate re-examination and localization, two areas of interest need to be determined in the differential curve of the second differential fingerprint: the target peak proximity interval and the preset interest interval. The target peak proximity interval is the adjacent range set around the target peak location, used to focus on capturing possible shoulder peaks, adjacent small peaks, and local abnormal rises caused by co-migration near the target peak. The width of this interval can be set based on the peak width, tailing characteristics of the target peak in the thin-layer system, and actual interpretation experience, covering potential interference areas near the target peak while minimizing the introduction of irrelevant noise due to excessive width. The preset interest interval is the interest range set in addition to the target peak proximity interval, based on product process experience or known areas prone to interference, used to capture non-target peak intensity anomalies or morphological anomalies. This interval can be a fixed segment or multiple discrete segments, the purpose of which is to match the enterprise's internal interpretation framework for easily interfered and sensitive areas.

[0056] After determining the aforementioned interval, this embodiment performs peak structure detection on the difference curve within the interval to extract feature information for characterizing anomalies. Peak structure detection can be implemented using common curve peak detection methods in the art, such as smoothing and denoising the difference curve, searching for local maxima, and combining parameters such as peak width, peak height, peak area, or peak symmetry to confirm the existence of peak structure. Based on this, the abnormal features extracted in this embodiment include at least one of the following: New peaks near the target peak: A response exceeding the noise level and possessing a clear peak structure appears in the interval near the target peak, or an additional peak structure inconsistent with the main shape of the target peak appears; Abnormal intensity of non-target peaks: An abnormal response such as continuous local intensity increase, abnormal fluctuation enhancement, or peak intensity exceeding limits appears within the preset interest interval; Abnormal peak shape: The difference curve shows obvious peak shape deviation within the interest interval, such as abnormal broadening, shoulder peak trend, splitting trend, or abnormal tailing.

[0057] To facilitate re-inspection and localization, the output of abnormal features should typically include at least: the location segment where the abnormality is located, the type of abnormality (e.g., a new peak near the target peak, abnormal intensity of a non-target peak, or abnormal peak shape), and the degree of abnormality (e.g., the degree of exceeding the limit of the relative noise level or the duration of exceeding the limit).

[0058] This embodiment generates anomaly score based on the aforementioned anomalous features. Anomaly score can be understood as a quantitative representation of the degree of anomalousness, used to unify different types of anomalous features to a quantifiable scale. The generation of anomaly score can comprehensively consider at least one of the following factors: peak height and area of ​​adjacent newly added peaks, intensity exceeding limits and duration of non-target intervals, and peak shape deviation.

[0059] After obtaining the anomaly level, it is compared with a preset threshold or judgment rule to output an impurity risk judgment result. The judgment result includes at least normal and abnormal results: a normal result is output when the anomaly level does not reach the trigger threshold or does not meet the abnormal judgment rule; an abnormal result is output when the anomaly level reaches the trigger threshold or meets the abnormal judgment rule. When the result is judged as normal, this embodiment generates an impurity detection conclusion based on the second differential fingerprint, indicating that no abnormal response requiring further confirmation was found from the perspective of target response suppression. When the result is judged as abnormal, this embodiment outputs the newly added peak near the target peak or the non-target peak exceeding the limit as the re-inspection target feature and triggers re-inspection, so that the re-inspection can be carried out with a clear location segment and anomaly type for targeted verification, avoiding the re-inspection from being stuck in repetitive operations and unable to locate and confirm the anomaly, thereby supporting the formation of the final impurity detection conclusion.

[0060] In one embodiment of the present invention, to reduce the interference of the compound oral liquid matrix, excipients, and non-target medicinal ingredients on the thin-layer chromatography response, and to provide a verifiable background benchmark for subsequent background subtraction, a negative control solution is set as the control input. The negative control solution is prepared using the same preparation process as the oral liquid to be tested, but without the target medicinal ingredient. The target medicinal ingredient includes at least one of Leonurus japonicus and Angelica sinensis. Specifically, the preparation process of the negative sample is consistent with that of the oral liquid to be tested, including the order of feeding, extraction method, filtrate merging, concentration or non-concentration, volume adjustment, and excipient addition method. Only the raw materials corresponding to the target medicinal ingredient are not added or not added during the feeding stage, to ensure that the negative sample retains the same matrix background and excipient system as the oral liquid to be tested. After the negative sample is prepared, it is prepared using the same sampling and pretreatment methods as the test solution, such as the same quantitative volume, solvent addition, shaking / ultrasounding, centrifugation / filtration processes. This negative control solution is used to form a negative curve in the same thin-layer chromatography process under the same conditions and serves as a background benchmark for subsequent differential processing.

[0061] In one embodiment of the present invention, in order to avoid occasional anomalies caused by factors such as spotting fluctuations, development differences, and differences in color development and imaging in thin-layer chromatography detection being misjudged as impurities, and also to avoid real impurity anomalies being masked by background or condition fluctuations, a two-level re-examination and anomaly attribution mechanism is introduced when outputting abnormal results. First, the test is run again under the same conditions to confirm whether the anomaly is reproducible. Then, if the anomaly is reproducible, at least one of the thin-layer chromatography processing conditions is changed to perform a corrective re-examination to distinguish between anomalies caused by method conditions and anomalies caused by suspected impurities, thereby providing a verifiable basis for the final conclusion. Specifically, the re-inspection includes: when the impurity risk assessment result is abnormal, performing a re-inspection under the same conditions and determining whether the abnormal characteristics corresponding to the abnormal result are reproduced in the re-inspection result; when the abnormal characteristics are reproduced in the re-inspection result under the same conditions, further performing a corrective re-inspection, wherein the corrective re-inspection is performed by changing at least one of the thin-layer chromatography processing conditions, wherein the thin-layer chromatography processing conditions include the developing system, the spotting volume, the sample pretreatment, the color development conditions, and the imaging conditions; based on the results of the re-inspection under the same conditions and the corrective re-inspection, an abnormality attribution conclusion is output, wherein the abnormality attribution conclusion includes abnormal method conditions and suspected impurity abnormalities.

[0062] The same-condition retest is used to verify whether the initial anomaly was caused by an occasional operational deviation. Here, "same conditions" can be understood as: maintaining the same thin-layer chromatography processing conditions and image acquisition conditions as the initial test, such as using the same type of thin-layer plate, the same development system configuration, the same spotting method and spotting volume control method, the same color development method and color development time control method, and the same imaging equipment and imaging parameter control method.

[0063] Reproducibility judgment does not involve inventing a new set of discrimination logic. Instead, it involves verifying the target characteristics of the initial abnormal output during re-examination. These characteristics include the location segment corresponding to the initial abnormality, the type of abnormality (new peak, peak shape deviation, intensity exceeding limits, etc.), and the description of the degree of abnormality. If the re-examination under the same conditions shows the same type of abnormality characteristics again in the corresponding location segment, the abnormality can be considered reproducible. If no corresponding abnormality characteristics appear or only scattered fluctuations inconsistent with the initial result appear, the abnormality can be considered non-reproducible and tends to be classified as a method condition abnormality or an occasional operational abnormality, thus avoiding conclusions about impurities based on non-reproducible fluctuations.

[0064] When the abnormal characteristics can be reproduced by re-examination under the same conditions, this embodiment further performs corrective re-examination. The purpose is not to expand the detection range, but to determine the sensitivity of the abnormality to the method conditions through slight changes in conditions, thereby supporting the attribution of the abnormality. In corrective re-examination, at least one of the following thin-layer chromatography processing conditions can be changed: Developing system: For example, the ratio or preparation method of the developing solvent can be adjusted controllably without changing the detection purpose; Spotting volume: For example, the spotting volume can be appropriately increased or decreased within a comparable range to observe whether the abnormal characteristics change significantly with the loading; Sample pretreatment: For example, the filtration method, clarification method, or extraction / dilution method can be changed to eliminate false peaks or tailing caused by turbidity, particles, or extraction differences; Colorimetric conditions: For example, the colorimetric time window or key control points of the colorimetric method can be adjusted to verify whether the abnormality is caused by uneven colorimetric development or over-development; Imaging conditions: For example, the exposure, contrast, light source uniformity correction method, or imaging distance can be adjusted to verify whether the abnormality is caused by imaging bias.

[0065] The key to corrective review is to change only the selected condition item each time or prioritize changing a single condition item, while keeping other conditions as consistent as possible, so that whether the anomaly changes significantly with the change of the condition is interpretable and verifiable.

[0066] This embodiment forms an anomaly attribution conclusion based on the combined results of same-condition retesting and corrective retesting. This may include: if same-condition retesting does not reproduce the anomalous characteristics, it tends to output a method condition anomaly, such as caused by sporadic spotting, development fluctuations, or imaging fluctuations, to avoid judging non-reproducible anomalies as impurities; if same-condition retesting can reproduce the anomalous characteristics, and corrective retesting shows that the anomalous characteristics significantly weaken, disappear, or undergo morphological changes highly consistent with the condition changes after changing at least one condition, it tends to output a method condition anomaly, indicating that the anomaly is more likely caused by method condition sensitivity; if same-condition retesting can reproduce the anomalous characteristics, and corrective retesting shows that the anomalous characteristics remain stable and exhibit consistent anomalous behavior after changing at least one condition, it tends to output a suspected impurity anomaly, providing a basis for further confirmation.

[0067] In addition, to facilitate subsequent review and traceability, the anomaly attribution conclusions can be accompanied by the following records: the type of change in the condition item used in the re-examination, the corresponding location segment of the abnormal feature, and a description of the changing trend of the abnormal feature, so as to ensure that the conclusions are interpretable.

[0068] In one embodiment of the present invention, to further support quantitative needs in quality control after thin-layer chromatography detection has completed impurity risk identification and verification, when the impurity risk determination does not trigger retesting or the retesting confirms the result as normal, the content of the target component is converted using the obtained curve data without changing the existing thin-layer chromatography processing and imaging workflow. This content determination process uses the target peak of the control curve as a positioning reference, determines the integration interval, integrates the peak intensity to obtain the peak area or integrated intensity that can be used for quantification, and then combines the calibration relationship obtained from different concentrations of reference standards to complete the content conversion, thereby enabling the same detection data to simultaneously support impurity identification and content assessment. Specifically, when the impurity risk assessment result does not trigger a retest or the retest result is normal, the method also includes a target component content determination step. This step includes: determining the integration interval corresponding to the target peak of the control curve in the test curve or the first differential fingerprint; integrating the peak intensity within the integration interval to obtain the peak area or integrated intensity of the target peak; establishing a calibration relationship based on the peak area or integrated intensity of the target peak obtained at different concentrations of the target component reference solution; and substituting the peak area or integrated intensity of the target peak of the test curve into the calibration relationship to calculate the target component content.

[0069] In this embodiment, content determination can be performed based on either the test curve or the first differential fingerprint. The basic logic is as follows: the test curve reflects the overall response of the test sample in the development direction, and is suitable for scenarios where the target peak shape is clear and background interference is controllable; the first differential fingerprint, after background subtraction, can further weaken the common background of the matrix and non-target medicinal flavors, making the integral of the target peak more stable, and is suitable for scenarios where the background fluctuations are obvious, but the target peak can still maintain an integrable shape after background subtraction. In actual implementation, one of them can be selected as the quantitative input according to the signal-to-noise ratio of the curve, provided that the target peak response in the selected curve can still be reliably located and integrated.

[0070] In this embodiment, the integration interval is used to define the integration range of the target peak, so as to stably convert the peak response into a numerical value that can be used for quantification. The determination of the integration interval uses the target peak of the control curve as a positioning reference, and can be implemented as follows: 1) Locate the position of the target peak in the control curve and determine the effective boundary range of the target peak, for example, using the initial rising segment of the peak and the interval falling back to near the baseline as the boundary; 2) Map this boundary range to the same position coordinate system of the test curve or the first differential fingerprint, thereby obtaining the integration interval corresponding to the target peak of the control curve; 3) When the target peak has slight broadening or tailing, the integration interval can be appropriately expanded without introducing interference from adjacent peaks to ensure complete coverage of the peak area; when there are weak responses that may interfere with the target peak, the integration interval is converged first, so that it only covers the main range of the target peak and avoids including adjacent abnormal responses in the integration. The core of the above integration interval determination process is to make the integration range of the target peak repeatable and interpretable, and to avoid the integration interval changing with the operator's subjective adjustment.

[0071] In this embodiment, integrating the peak intensity within the integration interval yields the peak area or integrated intensity of the target peak. The peak intensity here originates from a one-dimensional intensity sequence or density curve generated during curve generation, and its integration result can be understood as the cumulative response of the target peak within that interval. To improve the stability of the integration, it can be ensured that the curve has undergone necessary baseline correction before integration, so that the integration better reflects the peak's own response rather than the overall baseline shift.

[0072] In this embodiment, the calibration relationship is used to map peak area or integrated intensity to target component content. Specifically, it can be established as follows: Prepare target component reference solutions of different concentrations, and obtain the corresponding target peak area or integrated intensity under the same thin-layer chromatography processing and imaging conditions; establish a calibration relationship with reference concentration as input and target peak area or integrated intensity as output, and use it as the basis for subsequent conversion; after the test sample passes the impurity risk assessment and is not triggered for retesting, or is confirmed to be normal after retesting, obtain the target peak area or integrated intensity of the test curve or the first differential fingerprint, and substitute it into the calibration relationship to obtain the target component content.

[0073] To ensure the usability of calibration relationships, commonly used concentration ranges can be covered when establishing calibration relationships, so that the target peak response of the test sample falls within the calibration range, thereby avoiding the uncertainty caused by extrapolation.

[0074] This embodiment emphasizes that the content determination occurs under the condition that the impurity risk assessment has not triggered a retest or the retest result is normal. Its purpose is to further use the same curve data for quantitative conversion after confirming that there are no abnormal responses that need to be reviewed, thereby improving the integrity of the test output without increasing the number of additional test rounds, and making the test results easier for quality trend management and batch-to-batch comparison.

[0075] In one embodiment of the present invention, to reduce the impact of peak position drift caused by differences in thin-layer plates, slight differences in the developing system, environmental factors, and fluctuations in imaging conditions on subsequent differential fingerprinting and anomaly determination in thin-layer chromatography detection, the target peak position of the control curve is first verified for consistency before differential fingerprint generation. This verification does not rely on subjective judgment of a single control result, but rather constructs a control template based on multiple batches of normal results, forming the standard position of the target peak and the allowable position offset range, thereby transforming the reliability of the control peak position into a verifiable pre-judgment condition; when the peak position shift of the current control curve exceeds the allowable range, a re-examination is triggered first to eliminate the influence of method condition drift and avoid mistransmitting peak position drift to differential and anomaly analysis. Specifically, this embodiment includes steps for constructing reference templates and determining positional offset ranges. These steps include: selecting multiple batches of reference curves confirmed to be normal by impurity detection conclusions as template samples; aligning each template sample with the target peak as the anchor point and extracting the target peak position parameters; determining the standard position of the target peak and the allowable positional offset range based on the statistical results of the target peak position parameters; before generating the first and second differential fingerprints, acquiring the currently detected reference curve and extracting its target peak position parameters, and calculating its positional offset relative to the standard position of the target peak; triggering a retest when the positional offset exceeds the allowable positional offset range, otherwise proceeding to the differential fingerprint generation and impurity risk determination process.

[0076] In this embodiment, the template sample refers to a set of control curves from different batches that have been confirmed to be normal, used to characterize the natural fluctuation range of the control peak position of the product under stable process conditions. The selection of the template sample follows two principles: 1) it is derived from confirmed normal test batches to avoid including abnormal peak position drift in the reference range; 2) it covers a certain number of batches to reflect the reasonable changes in the control peak position under different batches, different thin-layer plates, and environmental fluctuations, thereby making the subsequent judgment representative.

[0077] In this embodiment, the purpose of aligning with the target peak as the anchor point is to transform the control curves from different batches into the same positional coordinate system, eliminating the overall translational effects caused by differences in unfolding distance or sampling position, and enabling the target peak position parameters to be statistically analyzed and compared on the same scale. Specifically, the target peak position is first located in each template sample control curve, and then mapped to a unified reference position, ensuring that the target peaks of each template sample are aligned to consistent positional coordinates, thus providing a basis for subsequent extraction of positional parameters and statistical range.

[0078] In this embodiment, the target peak position parameter is used to characterize the position of the target peak on the curve. It can be extracted using methods common in the art and verifiable in the field, such as using the position corresponding to the peak apex as the position parameter, or using the centroid position of the target peak or a stable point near the peak apex as the position parameter. During extraction, the target peak can be located in conjunction with the peak structure detection results to avoid misidentifying noise fluctuations or nearby interference as the target peak position. Once the position parameter is determined, subsequent offset calculations are all based on the same position parameter definition to ensure consistency.

[0079] In this embodiment, the standard position of the target peak is used to indicate the location where the target peak should appear under stable conditions, and the allowable position offset range is used to describe the acceptable offset range of the target peak under normal fluctuations. Specifically, the target peak position parameters of the template samples are statistically summarized to obtain the central tendency and fluctuation range, and the standard position and allowable offset range are determined accordingly. This allowable offset range reflects the natural drift allowable range of multiple batches of normal samples under actual detection conditions, ensuring that subsequent judgments do not rely on arbitrary settings but rather match historical stable performance.

[0080] In this embodiment, before proceeding to differential fingerprint generation, the currently detected control curve is acquired and its target peak position parameters are extracted. These parameters are then compared with the standard position of the target peak to obtain the position offset. The position offset reflects the degree of deviation of the current control peak position from the historical stable center. When the position offset exceeds the allowable range, this embodiment triggers a re-inspection, prioritizing the investigation of peak position drift caused by fluctuations in the development system, sample quantity, color development, and imaging conditions. This prevents peak position anomalies from being propagated to subsequent differential processing and causing misjudgments. When the position offset is within the allowable range, it indicates that the current control peak position is in an acceptable stable state, and the differential fingerprint generation and anomaly detection process can proceed, thereby ensuring the comparability and interpretability of subsequent differential results.

[0081] In one embodiment of the present invention, to address the problem of drift in the judgment criteria caused by batch differences, minor process fluctuations, and equipment status changes in oral liquid thin-layer chromatography detection during long-term operation, a qualified batch fingerprint database construction and dynamic update mechanism is introduced after completing differential fingerprint generation and impurity risk assessment. Confirmed normal test results are precipitated as reusable qualified fingerprint records, and a qualified batch reference template and consistency threshold parameters are statistically formed based on multiple batch qualified records for consistency assessment of subsequent batch results. Simultaneously, the reference template and threshold parameters are updated after new qualified records are added to the database, enabling the judgment criteria to adaptively adjust with the normal evolution of production and testing status, thereby achieving traceable, verifiable, and sustainable judgment criteria. Specifically, the qualified batch fingerprint database construction and updating steps provided in this embodiment include: when the impurity detection conclusion is a normal result or is determined to be a normal result after re-inspection, the control curve, first differential fingerprint, and second differential fingerprint of the current batch are recorded as qualified fingerprints in the database; based on multiple batches of qualified fingerprint records in the fingerprint database, the control curve, first differential fingerprint, and second differential fingerprint are statistically summarized to construct a qualified batch reference template, and a consistency threshold parameter is determined based on the statistical summary result. The qualified batch reference template includes at least one of the target peak standard position and the reference morphology parameter of the differential curve, and the consistency threshold parameter includes at least one of the allowable position offset range and the allowable anomaly range; for subsequent batches of samples, the target peak position offset is calculated based on its control curve, and the anomaly is calculated based on its first differential fingerprint and second differential fingerprint, and the target peak position offset and anomaly are compared with the corresponding consistency threshold parameter to obtain a consistency judgment result; after the qualified fingerprint records are entered into the database, the qualified batch reference template and consistency threshold parameter are updated.

[0082] In this embodiment, a qualified fingerprint record is a structured retention of a single test result, containing at least three types of data objects: a control curve, a first differential fingerprint, and a second differential fingerprint. These data objects are used to describe the comprehensive characterization of the batch under the control response, background subtraction perspective, and target suppression perspective.

[0083] The timing of data entry is limited to when the impurity detection result is normal, or when a re-inspection confirms a normal result. This is done to ensure that the fingerprint database primarily consists of stable, acceptable normal states, avoiding the inclusion of abnormal or unconfirmed states in the reference set, thereby improving the reliability of subsequent templates and thresholds.

[0084] In this embodiment, the qualified batch reference template is used to provide the typical performance that should be presented under normal conditions. Its source is not a single sample, but rather a statistical summary of multiple batches of qualified fingerprint records in the fingerprint database. Specifically, the control curve, the first differential fingerprint, and the second differential fingerprint can be statistically summarized to extract at least one reference element. For example, the target peak standard position: used to describe the typical center position of the target peak of the control curve under normal conditions; the reference morphological parameters of the differential curve: used to describe the typical morphological characteristics of the first or second differential fingerprint under normal conditions, such as the typical fluctuation level of the key interval, the sparsity of the typical peak structure, or the typical morphological boundary, etc. Here, the reference morphological parameters emphasize verifiability and comparability, without limiting specific mathematical forms; their role is to provide a reference benchmark for subsequent consistency judgment.

[0085] In this embodiment, the consistency threshold parameter is used to describe the allowable fluctuation boundary under normal conditions, enabling objective comparison of results from subsequent batches. The threshold parameter includes at least two typical components: Allowable position offset range: used to constrain the acceptable deviation range of the target peak position of the control curve relative to the standard position, thereby avoiding incomparability caused by peak drift; Allowable anomaly range: used to constrain the upper limit or acceptable range of anomaly between the first and second differential fingerprints under normal conditions, thereby avoiding misjudging normal fluctuations as anomalies. The threshold parameter is obtained through statistical summarization of multiple batches of qualified records, ensuring that the threshold reflects historical normal performance rather than being arbitrarily set, thus improving the stability and traceability of the judgment criteria.

[0086] In this embodiment, when performing consistency determination on subsequent batches of samples, at least two benchmark comparisons are conducted, including: 1) calculating the target peak position offset based on the control curve and comparing it with the allowable position offset range to determine whether the current detected peak position status is still within an acceptable consistency range; 2) calculating the anomaly degree based on the first differential fingerprint and the second differential fingerprint and comparing it with the allowable anomaly degree range to determine whether the degree of anomaly under the differential perspective still conforms to the stable performance of the qualified batch. This consistency determination result is used to output a verifiable conclusion on the consistency degree between the current batch and the qualified batch reference template, thereby assisting subsequent release decisions, trend analysis, or review arrangements. The key is to assign changes from different sources (peak position changes and anomaly changes) to their respective threshold scales, avoiding information loss caused by summarizing with a single indicator.

[0087] In this embodiment, when a new qualified fingerprint record is entered into the database, the qualified batch reference template and consistency threshold parameters are updated. The purpose of the update is not to frequently change the caliber, but to enable the reference template and threshold to absorb new qualified sample information while ensuring historical traceability, thereby adapting to long-term drift, batch distribution changes, and equipment status changes of the detection system under normal conditions.

[0088] During the update process, the data caliber can be kept consistent with that in the construction phase: statistical summaries are still performed on the control curve, the first differential fingerprint, and the second differential fingerprint, and then the target peak standard position, reference morphological parameters, and corresponding threshold parameters are updated synchronously, so that the entire library and the judgment benchmark form a closed loop maintenance.

[0089] In another embodiment of the present invention, an oral liquid component impurity detection system is also provided. In this embodiment, the oral liquid component impurity detection system is deployed in the daily quality control scenario of Compound Motherwort Oral Liquid, and is used to realize image acquisition, curve processing, differential fingerprint generation, anomaly identification, and closed-loop control of re-inspection in the thin-layer chromatography detection process, thereby transforming results traditionally relying on visual interpretation into verifiable and traceable detection conclusions. The system can operate using a combined architecture of laboratory acquisition terminal + local computing terminal + fingerprint database, and can also synchronize results to an intranet server when network conditions are available to support multi-terminal verification and data retention. Specifically, as... Figure 2 As shown, the system includes: a sample preparation module for preparing a test solution of the oral liquid to be tested, a reference solution of the target component, and a negative control solution lacking the target medicinal flavor; a thin-layer chromatography imaging module for performing thin-layer chromatography processing on the three solutions and acquiring corresponding chromatographic images; a curve generation module for extracting the image intensity distribution of the corresponding lanes of each solution from the chromatographic images and statistically integrating the lane image intensity along the development direction to generate density curves, so as to obtain the test curve, control curve, and negative curve corresponding to the three solutions respectively; and a differential fingerprint generation module for generating a first differential fingerprint and a second differential fingerprint in parallel. The first differential fingerprint is a differential curve obtained by subtracting the background of the test curve using the negative curve as a background benchmark. The second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve to highlight the non-target response. The impurity risk judgment module is used to extract abnormal features based on the first and second differential fingerprints and form an impurity risk judgment result. The re-inspection control module is used to determine whether the impurity risk judgment result meets the re-inspection triggering condition, generate an impurity detection conclusion based on the impurity risk judgment result, otherwise execute the re-inspection and generate an impurity detection conclusion based on the re-inspection result.

[0090] In this embodiment, the thin-layer chromatography imaging module can be composed of the following hardware: 1) a thin-layer plate imaging device, including a support platform for fixing the position of the thin-layer plate, a light-shielding structure, and a uniform light source component. The light source component can be configured as at least one of visible light uniform illumination and ultraviolet illumination to adapt to the observation method after color development and reduce the interference of ambient light on the image intensity distribution; 2) an imaging device, which can be one or more of an industrial camera, a high-definition camera, or a mobile imaging device, with a fixed bracket to ensure stable shooting distance and angle. During shooting, the position of the thin-layer plate is constrained by a calibration plate or positioning frame so that the geometric relationship of different lanes on the same plate can be reused; 3) a computing terminal, which can be a laboratory workstation or an industrial control computer, used to complete calculation tasks such as curve generation, differential fingerprint generation, and risk assessment, and to interact with the database. The core function of this imaging environment is to ensure that the image intensity changes of the colored spots in the thin-layer chromatography detection of Compound Motherwort Oral Liquid originate as much as possible from the sample itself rather than shooting fluctuations, thereby providing stable input for subsequent curve shaping, differential analysis, and judgment.

[0091] In this embodiment, the system can operate independently offline, with the computing terminal locally storing detection records; it can also operate in an intranet environment, with the computing terminal communicating with the server via a wired or wireless LAN to write detection images, curves, differential fingerprints, anomaly features, and detection conclusions into the database. The database may contain at least the following data tables or datasets: raw dataset (chromatographic images, lane positioning results, intermediate sequences generated from curves); fingerprint dataset (control curves, first differential fingerprint, second differential fingerprint, and their associated batch information); judgment dataset (anomaly features, anomaly degree, impurity risk judgment results, re-inspection trigger records, and final detection conclusions); and permission and audit set (for recording operation accounts, operation times, review records, and data traceability links). Through this organizational method, each detection result can be traced back to the original image and processing process, and statistical summaries of historical normal batches can be supported, providing a data foundation for subsequent stability management.

[0092] In this embodiment, the routine detection process for Compound Motherwort Oral Liquid, which connects the sample preparation module and the thin-layer chromatography imaging module, includes: the test solution is used to reflect the overall response of the oral liquid to be tested; the control solution is used to provide a reference response for the target component; and the negative control solution is used to provide a background response when the target medicinal flavor is absent. After the three are processed and imaged on the same plate, the system enters the data processing and judgment stage.

[0093] The function of the curve generation module is to convert the speckle image into a comparable one-dimensional density curve. In practice, the corresponding lane region is first located in the image, and then the image intensity within the lane region is statistically integrated along the direction perpendicular to the unfolding direction to form a one-dimensional curve that varies with the unfolding direction, so that each lane corresponds to a calculable and verifiable curve representation.

[0094] The differential fingerprint generation module generates two types of differential curves in parallel. The first differential fingerprint is used to subtract background from the test curve using the negative curve as a background benchmark, thereby offsetting the common background caused by the prescription matrix, excipients, and non-target drug flavors as much as possible, making new peaks, peak shape changes, or local intensity anomalies more easily highlighted. The second differential fingerprint is used to perform target response suppression compensation subtraction processing on the test curve based on the control curve, so that after the main response of the target peak is suppressed, the non-target responses near or at other positions of the target peak are more prominent, thereby improving the detectability of adjacent impurities, accompanying peaks, or interfering peaks.

[0095] The impurity risk assessment module extracts abnormal features based on two types of difference curves and generates impurity risk assessment results. Abnormal features may include at least information such as new peaks, abnormal peak shapes, abnormal intensity in key regions, new peaks near the target peak, or abnormal intensity of non-target peaks, and outputs the location segment and type of abnormality so that re-inspection can be performed with a clear objective.

[0096] The re-inspection control module is used to translate the judgment results into process routing. This includes: when the re-inspection trigger conditions are not met, directly generating an impurity detection conclusion based on the impurity risk judgment result; when the re-inspection trigger conditions are met, the system outputs a re-inspection command and records the re-inspection target characteristics, while simultaneously writing the re-inspection results back to the database to form a traceable closed-loop control for re-inspection. Re-inspection can be performed by laboratory personnel again using thin-layer chromatography and imaging under predetermined detection conditions. The system repeatedly curves and performs differential processing on the re-inspection images to determine whether the abnormality still occurs, and forms a final conclusion accordingly.

[0097] In this embodiment, the system output includes at least: impurity risk assessment results, abnormal location segment and abnormality type, whether re-inspection is triggered, and final impurity detection conclusion. The results can be displayed on the laboratory workstation interface or synchronized to the intranet server for quality management personnel to review, and can be associated with batch numbers, production records, or sample retention records to meet the daily quality inspection record retention and traceability needs.

[0098] The technical solutions of the present invention have been described in detail above with reference to the accompanying drawings and embodiments, but the present invention is not limited to the above embodiments. Those skilled in the art can make various modifications, variations, or equivalent substitutions to the above embodiments without departing from the concept of the present invention. For example, adjustments can be made to the implementation algorithm of the image processing steps, the calculation method of abnormal features, the determination method of the threshold, the correction conditions for re-inspection, and the integration form of system modules. As long as the technical means of implementation and the resulting technical effects are the same as or equivalent to the present invention, they should all fall within the protection scope of the present invention. The protection scope defined in the claims shall ultimately prevail.

Claims

1. A method for detecting impurities in oral liquid components, characterized in that, The method includes: preparing a test solution of the oral liquid to be tested, a reference solution of the target component, and a negative control solution lacking the target medicinal flavor; Thin-layer chromatography was performed on the three solutions and the corresponding chromatographic images were obtained; The image intensity distribution of each solution corresponding to the lane is extracted from the chromatographic image, and the lane image intensity is statistically integrated along the development direction to generate density curves, so as to obtain the test curve, control curve and negative curve corresponding to the three solutions respectively. The first differential fingerprint and the second differential fingerprint are generated in parallel. The first differential fingerprint is a differential curve obtained by background subtraction of the test curve using the negative curve as the background benchmark. The second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve to highlight the non-target response. Based on the first and second differential fingerprints, abnormal features are extracted and impurity risk judgment results are formed. When the impurity risk judgment results do not meet the re-inspection triggering conditions, an impurity detection conclusion is generated based on the impurity risk judgment results; otherwise, a re-inspection is performed, and an impurity detection conclusion is generated based on the re-inspection results.

2. The method for detecting impurities in oral liquid components according to claim 1, characterized in that, The process of generating test curves, control curves, and negative curves from chromatographic images includes: locating lanes in the chromatographic images to determine the lane regions corresponding to each solution; integrating the image intensity within the lane regions along a direction perpendicular to the development direction to obtain a one-dimensional intensity sequence that varies with the development direction; and performing baseline correction and intensity normalization on the one-dimensional intensity sequence to form a density curve. It also includes using the target peak position of the control curve as an anchor point to align and correct the test curve and the negative curve, so that the target peak positions of each curve correspond to each other.

3. The method for detecting impurities in oral liquid components according to claim 2, characterized in that, Extracting abnormal features based on the first differential fingerprint and forming an impurity risk judgment result includes: determining the baseline noise band and key interval in the differential curve corresponding to the first differential fingerprint, and detecting peak structure in the determined region to extract abnormal features, wherein the abnormal features include at least one of new peaks, abnormal peak shape, and abnormal intensity in key regions; Anomaly degree is generated based on abnormal features, and the anomaly degree is mapped to impurity risk assessment result, wherein the impurity risk assessment result includes at least normal result and abnormal result; When the impurity risk assessment result is normal, the first differential fingerprint is recorded as the qualified character after background subtraction and used to generate the impurity detection conclusion; when the impurity risk assessment result is abnormal, the newly added peak, peak shape deviation or exceeding limit features are output as the re-inspection target features and the re-inspection is triggered.

4. The method for detecting impurities in oral liquid components according to claim 2, characterized in that, Extracting abnormal features based on the second differential fingerprint and forming an impurity risk judgment result includes: performing scale matching processing on the control curve based on the target peak response of the test curve, and performing compensation subtraction processing on the test curve to suppress the target response based on the scale-matched control curve, so as to obtain the differential curve corresponding to the second differential fingerprint; determining the target peak proximity interval and the preset interest interval in the differential curve; and detecting the peak structure within the interval to extract abnormal features, wherein the abnormal features include at least one of the following: newly added peaks near the target peak, abnormal intensity of non-target peaks, and abnormal peak shape; Anomaly degree is generated based on abnormal features, and the anomaly degree is mapped to impurity risk assessment result, wherein the impurity risk assessment result includes at least normal result and abnormal result; When the impurity risk assessment result is normal, an impurity detection conclusion is generated based on the second differential fingerprint; when the impurity risk assessment result is abnormal, the newly added peak near the target peak or the non-target peak exceeding the limit is output as the re-inspection target feature and the re-inspection is triggered.

5. The method for detecting impurities in oral liquid components according to claim 1, characterized in that, The negative control solution was prepared using the same preparation process as the oral liquid to be tested, but without the target medicinal flavor. The target medicinal flavor includes at least one of motherwort and angelica.

6. The method for detecting impurities in oral liquid components according to claim 3 or 4, characterized in that, The retesting includes: when the impurity risk assessment result is abnormal, performing a retest under the same conditions and determining whether the abnormal characteristics corresponding to the abnormal result are reproduced in the retesting result; when the abnormal characteristics are reproduced in the retesting result under the same conditions, further performing a corrective retesting, wherein the corrective retesting is performed by changing at least one of the thin-layer chromatography processing conditions, wherein the thin-layer chromatography processing conditions include the developing system, the spotting volume, the sample pretreatment, the color development conditions, and the imaging conditions; based on the results of the retesting under the same conditions and the corrective retesting, an abnormality attribution conclusion is output, wherein the abnormality attribution conclusion includes abnormal method conditions and suspected impurity abnormalities.

7. The method for detecting impurities in oral liquid components according to claim 3 or 4, characterized in that, When the impurity risk assessment result does not trigger a retest or the retest result is normal, the method also includes a target component content determination step. The step includes: determining the integration interval corresponding to the target peak of the reference curve in the test curve or the first differential fingerprint; integrating the peak intensity within the integration interval to obtain the peak area or integrated intensity of the target peak; establishing a calibration relationship based on the peak area or integrated intensity of the target peak obtained at different concentrations of the target component reference solution; and substituting the peak area or integrated intensity of the target peak of the test curve into the calibration relationship to calculate the target component content.

8. The method for detecting impurities in oral liquid components according to claim 1, characterized in that, The method also includes steps for constructing reference templates and determining positional offset ranges. These steps include: selecting multiple batches of reference curves that have been confirmed as normal by impurity detection as template samples; aligning each template sample with the target peak as the anchor point and extracting the target peak position parameters; determining the standard position of the target peak and the allowable positional offset range based on the statistical results of the target peak position parameters; before generating the first and second differential fingerprints, acquiring the currently detected reference curve and extracting its target peak position parameters, and calculating its positional offset relative to the standard position of the target peak; triggering a retest when the positional offset exceeds the allowable positional offset range, otherwise proceeding to the differential fingerprint generation and impurity risk determination process.

9. The method for detecting impurities in oral liquid components according to claim 3 or 4, characterized in that, It also includes the steps of constructing and updating the qualified batch fingerprint database, which include: when the impurity detection result is normal or is determined to be normal after re-inspection, the control curve, the first differential fingerprint and the second differential fingerprint of the current batch are recorded as qualified fingerprints and entered into the database. Based on multiple batches of qualified fingerprint records in the fingerprint database, statistical summaries are performed on the control curve, the first differential fingerprint, and the second differential fingerprint to construct a qualified batch reference template. Based on the statistical summaries, a consistency threshold parameter is determined. The qualified batch reference template includes at least one of the target peak standard position and the reference morphology parameter of the differential curve. The consistency threshold parameter includes at least one of the allowable position offset range and the allowable anomaly range. For subsequent batches of samples, the target peak position offset is calculated based on its reference curve, and the anomaly is calculated based on its first differential fingerprint and second differential fingerprint. The target peak position offset and anomaly are compared with the corresponding consistency threshold parameters to obtain the consistency judgment result. After qualified fingerprint records are entered into the database, the qualified batch reference template and consistency threshold parameters are updated.

10. A system for detecting impurities in oral liquid components, characterized in that, The system includes: The sample preparation module is used to prepare the test solution of the oral liquid to be tested, the reference solution of the target component, and the negative control solution lacking the target drug flavor; The thin-layer chromatography imaging module is used to perform thin-layer chromatography processing on three solutions and acquire the corresponding chromatographic images; The curve generation module is used to extract the image intensity distribution of each solution corresponding to the lane from the chromatographic image, and to perform statistical integration of the lane image intensity along the expansion direction to generate density curves, so as to obtain the test curve, control curve and negative curve corresponding to the three solutions respectively. The differential fingerprint generation module is used to generate a first differential fingerprint and a second differential fingerprint in parallel. The first differential fingerprint is a differential curve obtained by subtracting the background of the test curve using the negative curve as a background benchmark. The second differential fingerprint is a differential curve obtained by compensating for the suppression of the target response of the test curve based on the control curve to highlight the non-target response. The impurity risk assessment module is used to extract abnormal features based on the first differential fingerprint and the second differential fingerprint and form an impurity risk assessment result. The re-inspection control module is used to determine whether the impurity risk assessment result meets the re-inspection triggering conditions. If so, it generates an impurity detection conclusion based on the impurity risk assessment result; otherwise, it performs a re-inspection and generates an impurity detection conclusion based on the re-inspection result.

Images