Quality detection method and device of transdermal patch, computer equipment and readable storage medium

Abstract

The invention relates to a transdermal patch quality detection method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: according to the transparency of the transdermal patch and a target detection scene, obtaining an acquisition parameter of a near infrared spectrum of the transdermal patch; according to the acquisition parameters, acquiring sample near infrared spectrums of a plurality of patch samples of the transdermal patch; obtaining the sample piece weight and the sample target component content of each patch sample; according to the sample near infrared spectrum, the sample piece weight and the sample target component content of each patch sample, establishing a piece weight detection model and a component detection model of the transdermal patch; according to the acquisition parameters, acquiring a product near infrared spectrum of a to-be-detected product of the transdermal patch in the target detection scene; and according to the near infrared spectrum of the product, obtaining the product tablet weight and the product target component content of the product to be detected by using the tablet weight detection model and the component detection model. By adopting the method, the quality detection efficiency of the transdermal patch can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of process analysis, and particularly to a method, device, computer equipment, computer-readable storage medium, and computer program product for quality inspection of transdermal patches. Background Art

[0002] With the application of advanced manufacturing technology in the pharmaceutical manufacturing field, continuous pharmaceutical manufacturing technology has emerged. In continuous pharmaceutical manufacturing, using an integrated production line of components, pharmaceutical materials can continuously flow between unit operations to achieve the production of pharmaceuticals. Among them, in order to ensure the product quality of continuous pharmaceutical manufacturing, it is necessary to use process analysis technology to perform real-time quality inspection on the product flow.

[0003] Among them, in the continuous manufacturing of transdermal patches, production instruments usually integrate processes such as feeding of semi-finished adhesive solution, drying of the adhesive solution, and laminating of the backing layer, and can continuously complete the coating, drying, and lamination of transdermal patches to obtain transdermal patch products before cutting and packaging. In related technologies, offline detection methods such as ultraviolet-visible spectroscopy or high-performance liquid chromatography are usually used to detect the quality of transdermal patches. However, these methods have cumbersome detection steps and long time consumption, and are difficult to be applied to real-time quality inspection of the continuous manufacturing process of transdermal patches. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for quality inspection of transdermal patches for the above technical problems.

[0005] In a first aspect, the present application provides a method for quality inspection of transdermal patches, including:

[0006] Obtaining the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario;

[0007] Collecting the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0008] Obtaining the sample weight and the content of the sample target component of each patch sample;

[0009] Establishing a weight detection model and a component detection model of the transdermal patch according to the sample near-infrared spectra, the sample weights, and the content of the sample target components of each patch sample;

[0010] Collecting the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters;

[0011] Based on the near-infrared spectrum of the product, the product tablet weight and the content of the product target component of the product to be detected are obtained by using the tablet weight detection model and the component detection model.

[0012] In one embodiment, the transdermal patch includes a backing layer, an adhesive layer, and a release liner; according to the transparency of the transdermal patch and the target detection scenario, the acquisition parameters of the near-infrared spectrum of the transdermal patch are obtained, including: in the case where the target detection scenario is on-line detection or in-line detection, the acquisition distance parameter indicating non-contact acquisition is obtained, and the scanning parameter is obtained according to the tape running speed of the target detection scenario; if both the adhesive layer and the release liner of the transdermal patch are transparent, the first background configuration parameter indicating setting the acquisition background is obtained; otherwise, the first background configuration parameter indicating not setting the acquisition background is obtained; in the case where the target detection scenario is near-line detection, if both the backing layer and the release liner of the transdermal patch are opaque, the acquisition distance parameter indicating non-contact acquisition is obtained; otherwise, the acquisition distance parameter indicating contact acquisition is obtained; if the backing layer, the adhesive layer, and the release liner are all transparent, the first background configuration parameter indicating setting the acquisition background is obtained; otherwise, the first background configuration parameter indicating not setting the acquisition background is obtained; according to the first background configuration parameter, the acquisition distance parameter, the acquisition position parameter, the scanning parameter, and the second background configuration parameter, the acquisition parameters of the transdermal patch are obtained.

[0013] In one embodiment, the transdermal patch includes a release liner and a backing layer; after establishing the tablet weight detection model and the component detection model of the transdermal patch according to the sample near-infrared spectrum, the sample tablet weight, and the sample target component content of each patch sample, it further includes: in the case where there are differences between the product release liner and the product backing layer of the product to be detected and the sample release liner and the sample backing layer of the patch sample, obtaining a new patch sample using the product release liner and the product backing layer; updating the tablet weight detection model and the component detection model using the new patch sample to obtain the updated tablet weight detection model and the updated component detection model.

[0014] In one embodiment, establishing the tablet weight detection model and the component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weights, and the sample target component contents of the patch samples includes: dividing the sample near-infrared spectra, the sample tablet weights, and the sample target component contents corresponding to each of the patch samples into a calibration set and a test set; obtaining a plurality of candidate processing combinations according to a plurality of spectral preprocessing methods and a plurality of characteristic wavelength screening methods; using each of the candidate processing combinations to process the sample near-infrared spectra in the calibration set to obtain a spectral processing result set corresponding to each of the candidate processing combinations; constructing a candidate tablet weight detection model and a candidate component detection model corresponding to each of the candidate processing combinations according to the calibration set and each of the spectral processing result sets; using the calibration set and the test set to evaluate the accuracy of each of the candidate tablet weight detection models and each of the candidate component detection models, and determining the optimal tablet weight detection model and the optimal component detection model; and obtaining the tablet weight detection model and the component detection model according to the optimal tablet weight detection model and the optimal component detection model.

[0015] In one embodiment, using the calibration set and the test set to evaluate the accuracy of each of the candidate tablet weight detection models and each of the candidate component detection models, and determining the optimal tablet weight detection model and the optimal component detection model includes: for each of the candidate processing combinations, using the calibration set to verify the candidate tablet weight detection model and the candidate component detection model to obtain the cross-validation correlation coefficient and the cross-validation root mean square error of the candidate tablet weight detection model and the candidate component detection model; using the test set to verify the candidate tablet weight detection model and the candidate component detection model to obtain the prediction correlation coefficient and the prediction root mean square error of the candidate tablet weight detection model and the candidate component detection model; screening out the optimal tablet weight detection model from each of the candidate tablet weight detection models according to the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate tablet weight detection models; and screening out the optimal component detection model from each of the candidate component detection models according to the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate component detection models.

[0016] In one embodiment, constructing a candidate tablet weight detection model and a candidate component detection model corresponding to each of the candidate processing combinations according to the calibration set and each of the spectral processing result sets includes: for each of the candidate processing combinations, constructing the candidate tablet weight detection model and the candidate component detection model corresponding to the candidate processing combination based on the improved kernel partial least squares method according to the calibration set and each of the spectral processing result sets.

[0017] Second aspect, the present application also provides a quality detection device for a transdermal patch, including:

[0018] A first acquisition module, configured to obtain acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario;

[0019] A first acquisition module, configured to acquire the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0020] A second acquisition module, configured to obtain the sample tablet weight and the sample target component content of each of the patch samples;

[0021] A model construction module, configured to establish a tablet weight detection model and a component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weight, and the sample target component content of each of the patch samples;

[0022] A second acquisition module, configured to acquire the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters;

[0023] A third acquisition module, configured to obtain the product tablet weight and the product target component content of the product to be detected according to the product near-infrared spectrum by using the tablet weight detection model and the component detection model.

[0024] Third aspect, the present application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0025] Obtain acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario;

[0026] Acquire the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0027] Obtain the sample tablet weight and the sample target component content of each of the patch samples;

[0028] Establish a tablet weight detection model and a component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weight, and the sample target component content of each of the patch samples;

[0029] Acquire the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters;

[0030] Obtain the product tablet weight and the product target component content of the product to be detected according to the product near-infrared spectrum by using the tablet weight detection model and the component detection model.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0032] Obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario;

[0033] Collect the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0034] Obtain the sample weight and the content of the sample target component of each patch sample;

[0035] Establish a weight detection model and a component detection model for the transdermal patch according to the sample near-infrared spectra, the sample weights, and the contents of the sample target components of the patch samples;

[0036] Collect the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters;

[0037] Obtain the product weight and the content of the product target component of the product to be detected according to the product near-infrared spectrum by using the weight detection model and the component detection model.

[0038] In a fifth aspect, the present application further provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0039] Obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario;

[0040] Collect the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0041] Obtain the sample weight and the content of the sample target component of each patch sample;

[0042] Establish a weight detection model and a component detection model for the transdermal patch according to the sample near-infrared spectra, the sample weights, and the contents of the sample target components of the patch samples;

[0043] Collect the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters;

[0044] Obtain the product weight and the content of the product target component of the product to be detected according to the product near-infrared spectrum by using the weight detection model and the component detection model.

[0045] The above quality inspection method, device, computer equipment, computer-readable storage medium and computer program product for the transdermal patch first obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario, then collect the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters, and obtain the sample tablet weight and the content of the sample target component of each patch sample. Then, a tablet weight detection model and a component detection model for the transdermal patch are established based on the sample near-infrared spectra, sample tablet weights and sample target component contents of each patch sample. Then, the product near-infrared spectrum of the product to be detected of the transdermal patch is collected in the target detection scenario according to the acquisition parameters, and based on the product near-infrared spectrum, the product tablet weight and the content of the product target component of the product to be detected are obtained using the tablet weight detection model and the component detection model.

[0046] In this solution, by establishing a quantitative detection model of the product tablet weight, drug component content and near-infrared spectrum of the transdermal patch, the near-infrared spectrum detection technology can be applied to the quality inspection of the transdermal patch, so that rapid and accurate quality inspection of the product can be carried out during the continuous manufacturing process of the transdermal patch, effectively improving the detection effect, saving production time and reducing waste caused by unqualified products. Moreover, according to the transparency of the transdermal patch and the characteristics of the quality inspection scenario during the continuous manufacturing process of the transdermal patch, this solution sets the acquisition parameters of the near-infrared spectrum and applies them to the acquisition of the near-infrared spectra of patch samples and products to be detected, enabling more accurate near-infrared spectra of the products to be detected to be collected in the target detection scenario, and avoiding model errors caused by differences in the acquisition methods of near-infrared spectra between samples and products, which is conducive to obtaining more accurate quality inspection results. Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a schematic flowchart of the quality inspection method for the transdermal patch in an embodiment;

[0049] Figure 2 It is a schematic flowchart of obtaining the acquisition parameters of the near-infrared spectrum of the transdermal patch in an embodiment;

[0050] Figure 3 It is a schematic flowchart of establishing a tablet weight detection model and a component detection model in an embodiment;

[0051] Figure 4Schematic flow chart for determining the optimal tablet weight detection model and the optimal component detection model in one embodiment;

[0052] Figure 5 Schematic flow chart for establishing the tablet weight detection model and the component detection model in another embodiment;

[0053] Figure 6 Schematic diagram of the screening results of characteristic wavelength variables in one embodiment;

[0054] Figure 7 Schematic diagram of the prediction effect of the tablet weight detection model and the component detection model in the calibration set in one embodiment;

[0055] Figure 8 Schematic diagram of the prediction effect of the tablet weight detection model and the component detection model in the test set in one embodiment;

[0056] Figure 9 Schematic diagram of the screening results of characteristic wavelength variables in another embodiment;

[0057] Figure 10 Schematic diagram of the prediction effect of the tablet weight detection model and the component detection model in the calibration set in another embodiment;

[0058] Figure 11 Schematic diagram of the prediction effect of the tablet weight detection model and the component detection model in the test set in another embodiment;

[0059] Figure 12 Structural block diagram of the quality detection device for transdermal patches in one embodiment;

[0060] Figure 13 Internal structure diagram of a computer device in one embodiment. Detailed implementation manners

[0061] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] In one embodiment, as Figure 1 shown, a quality detection method for transdermal patches is provided. In this embodiment, the method is illustrated by taking its application to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0063] Step S101, according to the transparency of the transdermal patch and the target detection scenario, obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch.

[0064] Among them, the method of this embodiment can be applied to detecting the quality of a transdermal patch during the continuous manufacturing process of the transdermal patch. In this step, acquisition parameters for collecting the near-infrared spectrum of the transdermal patch in the target detection scenario can be obtained according to the pre-determined target detection scenario and the transparency of the transdermal patch. Optionally, a Fourier transform near-infrared spectrometer can be used in the embodiments of this application to collect the near-infrared spectrum.

[0065] Among them, during the continuous manufacturing process of the transdermal patch, there can be multiple detection scenarios, such as near-line detection, on-line detection, in-line detection, etc. Among them, near-line detection can be to set up an analytical instrument at a position adjacent to the production equipment for detection, on-line detection can be to divert a part of the patch product during the production process and introduce it into the analytical instrument for detection, and in-line detection can be to directly set the analytical instrument at the position that needs to be monitored in the patch production process for detection.

[0066] Among them, the transparency of the transdermal patch can include transparent or opaque. Exemplarily, the transparency of the transdermal patch can be determined according to the near-infrared spectrum of the transdermal patch. Among them, if the near-infrared spectra of a single-layer transdermal patch and an overlapping multi-layer transdermal patch are the same, it can be determined that the transdermal patch is opaque, otherwise it can be determined that the transdermal patch is transparent.

[0067] Among them, the acquisition parameters for near-infrared spectrum collection can be used to indicate the acquisition method of the near-infrared spectrum, which can include but is not limited to one or more of background configuration parameters, acquisition position parameters, resolution parameters, scanning parameters, acquisition distance parameters. Among them, the background configuration parameters can be used to indicate whether to set an acquisition background, the type of acquisition background to set, etc., the acquisition position parameters can be used to indicate the position of setting the near-infrared spectrum acquisition device, the resolution parameters can be used to indicate the acquisition resolution of the near-infrared spectrum, the scanning parameters can be used to indicate the number of scans of the near-infrared spectrum, etc., and the acquisition distance parameters can be used to indicate whether the near-infrared spectrum is collected in a contact manner or a non-contact manner (exemplarily, contact acquisition can mean placing the transdermal patch on the scanning window of the near-infrared spectrum acquisition device for scanning, and non-contact acquisition can mean setting the transdermal patch at a position at a preset distance from the scanning window of the near-infrared spectrum acquisition device for scanning).

[0068] In this step, corresponding acquisition parameters can be obtained according to different combinations of the transparency of the transdermal patch and the target detection scenario. Exemplarily, when the transdermal patch is opaque, background configuration parameters indicating not to set an acquisition background can be obtained; when the target detection scenario is on-line detection or in-line detection, distance acquisition parameters indicating non-contact acquisition can be obtained.

[0069] Step S102: According to the acquisition parameters, acquire the sample near-infrared spectra of multiple patch samples of the transdermal patch.

[0070] Among them, according to the quality fluctuation range in the production process of the transdermal patch, multiple patch samples of the transdermal patch can be set, which may include patch samples corresponding to multiple different tablet weights and target ingredient contents. Exemplarily, the tablet weight and target ingredient content of the patch samples can be adjusted by coating the transdermal patch with different coating thicknesses of the adhesive solution and adhesive solutions with different target ingredient contents.

[0071] In this step, according to the acquisition parameters determined previously, the acquisition method corresponding to the acquisition parameters can be used to acquire the near-infrared spectra of each patch sample, and the sample near-infrared spectrum corresponding to each patch sample is obtained.

[0072] Step S103: Obtain the sample tablet weight and sample target ingredient content of each patch sample.

[0073] Among them, for each patch sample, the sample tablet weight and sample target ingredient content can be measured using standard methods to establish a database of the sample tablet weight and sample target ingredient content of the transdermal patch.

[0074] Exemplarily, the patch sample can be a strip-shaped semi-finished product of the transdermal patch, where samples of a preset area can be obtained from different positions (such as the front, middle, and rear sections) of the sample, and then the weight of each sample is measured using an analytical balance, and the target ingredient content in each sample is measured using high-performance liquid chromatography. Then, based on the average values of the weights and target ingredient contents of these samples, the sample tablet weight and sample target ingredient content of the patch sample can be obtained.

[0075] Step S104: Establish a tablet weight detection model and an ingredient detection model for the transdermal patch according to the sample near-infrared spectra, sample tablet weights, and sample target ingredient contents of each patch sample.

[0076] Among them, according to the correlation between the sample near-infrared spectra and sample tablet weights of each patch sample, and the correlation between the sample near-infrared spectra and sample target ingredient contents, a tablet weight detection model and an ingredient detection model for the transdermal patch can be established respectively. Among them, the tablet weight detection model can predict the tablet weight of the transdermal patch based on the near-infrared spectrum of the transdermal patch, and the ingredient detection model can predict the target ingredient content of the transdermal patch based on the near-infrared spectrum of the transdermal patch.

[0077] Step S105: According to the acquisition parameters, acquire the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario.

[0078] Among them, the product to be detected of the transdermal patch can be the transdermal patch product produced in the continuous manufacturing process. In this step, according to the acquisition parameters obtained previously, the corresponding acquisition method can be used to perform near-infrared spectroscopy acquisition on the product to be detected, and the product near-infrared spectrum corresponding to the product to be detected can be obtained.

[0079] Step S106, according to the product near-infrared spectrum, use the tablet weight detection model and the component detection model to obtain the product tablet weight and the content of the target component of the product to be detected.

[0080] Among them, according to the acquired product near-infrared spectrum, the product tablet weight of the product to be detected can be obtained by using the previously constructed tablet weight detection model, and the content of the target component of the product to be detected can be obtained by using the component detection model.

[0081] In the above quality detection method of the transdermal patch, by establishing a quantitative detection model of the product tablet weight and drug component content of the transdermal patch and the near-infrared spectrum, the near-infrared spectroscopy detection technology can be applied to the quality detection of the transdermal patch, so that the quality of the product can be quickly and accurately detected during the continuous manufacturing process of the transdermal patch, effectively improving the detection effect, saving production time and reducing the waste caused by unqualified products. Moreover, according to the transparency of the transdermal patch and the characteristics of the quality detection scenario during the continuous manufacturing process of the transdermal patch, the acquisition parameters of the near-infrared spectrum are set and applied to the near-infrared spectrum acquisition of the patch sample and the product to be detected, which can obtain a more accurate near-infrared spectrum of the product to be detected in the target detection scenario and avoid the model error caused by the difference in the acquisition method of the near-infrared spectrum between the sample and the product, which is beneficial to obtaining a more accurate quality detection result.

[0082] In an exemplary embodiment, the transdermal patch may include a backing layer, an adhesive layer, and a release liner; as Figure 2 shown, according to the transparency of the transdermal patch and the target detection scenario, obtaining the acquisition parameters of the near-infrared spectrum of the transdermal patch may include:

[0083] Step S201, if the transdermal patch is transparent, obtain the background configuration parameters for setting the acquisition background; otherwise, obtain the background configuration parameters for not setting the acquisition background.

[0084] Among them, when performing near-infrared spectroscopy detection on a transdermal patch, if the overall transparency of the transdermal patch is transparent, near-infrared light can pass through the transdermal patch. As a result, the near-infrared light will pass through an air segment with an uncertain distance and be diffusely reflected by the environment before entering the acquisition device, thereby affecting the accuracy of the acquired near-infrared spectrum. Based on this, in this step, when the transdermal patch is transparent, background configuration parameters indicating the setting of the acquisition background during the near-infrared spectrum acquisition of the transdermal patch can be obtained. Among them, the acquisition background can be set on the back side of the transdermal patch to provide a stable and low-absorption background for the acquisition of the near-infrared spectrum, avoiding the light path passing through an air segment with an uncertain distance and avoiding the occurrence of environmental diffuse reflection, so as to reduce the influence of the environment on the spectrum. Among them, when the transdermal patch is opaque, near-infrared light will not pass through the transdermal patch. As a result, the reflected light received by the acquisition device mainly comes from the diffuse reflection of the sample itself rather than the reflection from the environment. Therefore, in this case, background configuration parameters indicating that the acquisition background is not set during the near-infrared spectrum acquisition of the transdermal patch can be obtained.

[0085] It can be understood that during the manufacturing process of the transdermal patch, usually a glue solution containing the target component is coated on the anti-adhesive layer to obtain a glue layer, and then the backing layer can be covered on the surface of the glue layer. Thus, in different scenarios, the overall transparency of the transdermal patch can be manifested in different situations.

[0086] Exemplarily, in the case where the target detection scenario is on-line detection or in-line detection, since the raw materials for preparing the transdermal patch usually contain volatile solvents, and the components in the volatile solvents are likely to affect the near-infrared spectrum, therefore, when performing near-infrared spectrum acquisition, near-infrared spectrum acquisition can be performed at the drying completion stage of the transdermal patch, or near-infrared spectrum acquisition can be performed at the film covering completion stage, so as to ensure that the solvent volatilizes completely as much as possible. Among them, if the acquisition is performed at the drying completion stage, the transdermal patch includes an anti-adhesive layer and a glue layer. At this time, if both the anti-adhesive layer and the glue layer are transparent, the transdermal patch is overall transparent, and if at least one of the anti-adhesive layer and the glue layer is opaque, the transdermal patch is opaque; if the acquisition is performed at the film covering completion stage, the transdermal patch includes an anti-adhesive layer, a glue layer, and a backing layer. At this time, if all three are transparent, the transdermal patch is overall transparent, and if at least one of the three is opaque, the transdermal patch is opaque. Among them, it can be understood that in the case where the backing layer is opaque, in order to be able to perform near-infrared spectrum acquisition on the glue layer of the transdermal patch, the acquisition can be performed at the drying completion stage to avoid the influence of the backing layer.

[0087] Exemplarily, in the case where the target detection scenario is near-line detection, in order to ensure complete solvent volatilization, the transdermal patch can be sampled and detected after drying and laminating. Thus, in the near-line detection scenario, the transdermal patch to be detected can include a backing layer, an adhesive layer, and a release liner. Among them, if the backing layer, the adhesive layer, and the release liner are all transparent, the transdermal patch is transparent as a whole, while if at least one of the backing layer, the adhesive layer, and the release liner is opaque, the transdermal patch is opaque.

[0088] Among them, since the temperature at the spot position of the near-infrared probe can usually reach 70°C, when the background configuration parameters indicate setting to collect the background, the thermal deformation of the collected background easily causes a change in the distance between the transdermal patch and the near-infrared collection device, thereby affecting the accuracy of the near-infrared spectrum obtained by collection. Based on this, the material used for collecting the background can be heat-resistant to deformation, with a low absorption rate and a high reflectivity in the near-infrared spectral band. For example, it can be a gold-plated reflector, an aluminum-plated reflector, a frosted stainless steel background plate, etc.

[0089] Step S202, in the case where the target detection scenario is on-line detection or in-line detection, obtain a collection distance parameter indicating non-contact collection, and obtain a scanning parameter according to the running speed of the transdermal patch in the target detection scenario.

[0090] Among them, in the case where the target detection scenario is on-line detection or in-line detection, since near-infrared spectrum collection is performed on the transdermal patch moving on the conveyor belt, there needs to be a certain distance between the near-infrared spectrum collection device and the transdermal patch. Based on this, a collection distance parameter indicating non-contact collection can be obtained. Among them, non-contact collection can refer to a preset distance between the near-infrared spectrum collection device and the transdermal patch. Among them, the preset distance can be determined according to the performance parameters of the near-infrared spectrum collection device.

[0091] Among them, in the case where the target detection scenario is on-line detection or in-line detection, it is necessary to perform near-infrared spectrum collection on the moving transdermal patch. In order to avoid the scanned area covered by one spectral sample being too large, it is necessary to determine the scanning parameter according to the running speed to control the size of the scanned area. Exemplarily, the minimum number of scans can be calculated based on the resolution of the near-infrared spectrum collection device and the running speed of the conveyor belt to maintain the resolution unchanged and ensure the spectral quality, thereby obtaining the scanning parameter. It can be understood that in the case where the target detection scenario is near-line detection, the preset number of scans that can meet this resolution can be used as the scanning parameter of the transdermal patch.

[0092] Optionally, in the case where the target detection scenario is online detection or in-line detection, an acquisition position parameter indicating near-infrared spectrum acquisition in the drying completion section or the film laminating completion section can also be obtained. Among them, the drying completion section can be the part of the transdermal patch production line for transporting the dried but unlaminated transdermal patches, and the film laminating completion section can be the part of the production line for transporting the dried and film laminated transdermal patches. Exemplarily, for online detection, the near-infrared spectrum acquisition device can be set at the position of the conveyor belt diverted from the drying completion section or the film laminating completion section; for in-line detection, the near-infrared spectrum acquisition device can be set at the position of the drying completion section or the film laminating completion section.

[0093] Step S203, in the case where the target detection scenario is near-line detection, if both the backing layer and the release layer of the transdermal patch are opaque, an acquisition distance parameter indicating non-contact acquisition is obtained; otherwise, an acquisition distance parameter indicating contact acquisition is obtained.

[0094] Among them, in the case where the target detection scenario is near-line detection, in order to ensure complete solvent volatilization, an acquisition position parameter indicating that the acquisition device can be set beside the production line can be obtained. In this scenario, sampling and detection can be performed after the transdermal patch is dried and the film lamination is completed. Thus, in the near-line detection scenario, the transdermal patch to be detected can include a backing layer, an adhesive layer, and a release layer.

[0095] Among them, in the case where the target detection scenario is near-line detection, the corresponding acquisition distance parameter can be obtained according to the transparency of the backing layer and the release layer of the transdermal patch. Optionally, when both the backing layer and the release layer are opaque, one of them needs to be removed to scan the adhesive layer containing the target component. Therefore, there needs to be a certain distance between the near-infrared spectrum acquisition device and the transdermal patch, and thus an acquisition distance parameter indicating non-contact acquisition can be obtained. Among them, non-contact acquisition can refer to a preset distance between the near-infrared spectrum acquisition device and the transdermal patch. Among them, the preset distance can be determined according to the performance parameters of the near-infrared spectrum acquisition device. Optionally, when at least one of the backing layer and the release layer is transparent, the transdermal patch can be directly covered on the scanning window of the near-infrared spectrum acquisition device for scanning, so an acquisition distance parameter indicating contact acquisition can be obtained.

[0096] Step S204, according to the background configuration parameter, the acquisition distance parameter, and the scanning parameter, obtain the acquisition parameter of the transdermal patch.

[0097] Among them, according to the background configuration parameter, the acquisition distance parameter, and the scanning parameter obtained in the foregoing process, an acquisition parameter for determining the acquisition method for near-infrared spectrum acquisition of the transdermal patch can be obtained.

[0098] In this embodiment, by finely setting various parameters such as background configuration parameters, acquisition distance, and scanning parameters in near-infrared spectroscopy according to the transparency of the backing layer, adhesive layer, and release liner of the transdermal patch, as well as the characteristics of the near-line detection, on-line detection, or in-line detection scenarios, it can flexibly adapt to various different transdermal patch detection scenarios, which is conducive to obtaining a more accurate detection effect.

[0099] In an exemplary embodiment, the transdermal patch may include a release liner and a backing layer; after establishing a tablet weight detection model and a component detection model for the transdermal patch based on the sample near-infrared spectrum, sample tablet weight, and sample target component content of each patch sample, it may further include: when there are differences between the product release liner and product backing layer of the product to be detected and the sample release liner and sample backing layer of the patch sample, obtaining a new patch sample using the product release liner and product backing layer; updating the tablet weight detection model and the component detection model with the new patch sample to obtain an updated tablet weight detection model and an updated component detection model.

[0100] Among them, since the transdermal patch is a multi-layer structure including a release liner, an adhesive layer, and a backing layer, when there are differences between the product release liner or product backing layer of the product to be detected and the sample release liner and sample backing layer of the patch sample used to establish the tablet weight detection model and the component detection model, the model needs to be adjusted to make it applicable to detecting the product to be detected.

[0101] In this embodiment, when the release liner or the backing layer of the transdermal patch changes, multiple new patch samples using the product release liner and product backing layer of the product to be detected can be obtained, and the sample near-infrared spectrum, sample tablet weight, and sample target component content of each new patch sample can be collected respectively. Thus, these data can be used to maintain and update the established tablet weight detection model and component detection model, obtaining an updated tablet weight detection model that can predict the product tablet weight of the product to be detected, and an updated component detection model that can predict the product target component content of the product to be detected.

[0102] In this embodiment, when the release liner and / or the backing layer of the transdermal patch changes, obtaining transdermal patch samples using the new release liner and / or backing layer and updating the previously constructed tablet weight detection model and component detection model with these samples can make the previously constructed model applicable to quality detection of transdermal patches using the new release liner and / or backing layer, thereby saving the time for re-training and constructing the model, which is conducive to improving the detection efficiency of transdermal patches.

[0103] In an exemplary embodiment, such as Figure 3As shown, based on the sample near-infrared spectra, sample tablet weights, and sample target component contents of each patch sample, a tablet weight detection model and a component detection model for transdermal patches can be established, which may include:

[0104] Step S301: Divide the sample near-infrared spectra, sample tablet weights, and sample target component contents corresponding to each patch sample into a calibration set and a test set.

[0105] Specifically, in this step, the data corresponding to each patch sample can be divided into a calibration set and a test set. Among them, assuming there are n patch samples in total, the calibration set can include the sample near-infrared spectra, sample tablet weights, and sample target component contents corresponding to x patch samples; while the test set can include the sample near-infrared spectra, sample tablet weights, and sample target component contents corresponding to the remaining n - x patch samples. Among them, the calibration set can be used to establish the tablet weight detection model and the component detection model, while the test set is used to verify the established model and does not participate in the modeling.

[0106] Step S302: Obtain multiple candidate processing combinations according to multiple spectral preprocessing methods and multiple characteristic wavelength screening methods.

[0107] Specifically, before constructing the tablet weight detection model and the component detection model using the calibration set, spectral preprocessing and characteristic wavelength screening can be performed on the sample near-infrared spectra corresponding to each patch sample in the calibration set, and the screened characteristic wavelengths can be applied to the subsequent model construction.

[0108] Optionally, common spectral preprocessing methods for near-infrared spectra can include mean centering (MC), Z-score standardization (Z-score), vector normalization (VN), subtract straight line (SSL), eliminate constant offset (ECO), smoothing & derivative, standard normal variate transformation (SNV), detrending, multiplicative scatter correction (MSC), max-min normalization, etc. These spectral preprocessing methods can be used alone or in combination for spectral preprocessing.

[0109] Optionally, common feature wavelength screening methods for near-infrared spectroscopy may include backward interval partial least squares (biPLS), uninformative variable elimination (UVE), Monte Carlo-uninformative variable elimination (MCUVE), competitive adaptive reweighted sampling (CARS), etc.

[0110] In this step, multiple spectral preprocessing methods can be combined with multiple feature wavelength screening methods to obtain multiple candidate processing combinations for near-infrared spectroscopy.

[0111] Step S303: Use each candidate processing combination to process the near-infrared spectra of each sample in the calibration set to obtain a spectral processing result set corresponding to each candidate processing combination.

[0112] Among them, for each candidate processing combination, the spectral preprocessing method and the feature wavelength screening method included therein can be sequentially used to preprocess and screen the feature wavelengths of the near-infrared spectra of each sample in the calibration set to obtain a spectral processing result set corresponding to each candidate processing combination in the calibration set. Among them, the spectral processing result set corresponding to a candidate processing combination may include the spectral processing results of the sample near-infrared spectra corresponding to each patch sample in the calibration set under the corresponding candidate processing combination. The spectral processing results may include the processed spectra obtained by processing the sample near-infrared spectra using the spectral preprocessing method, and the sample feature wavelengths screened using the feature wavelength screening method.

[0113] Step S304: Construct a candidate tablet weight detection model and a candidate component detection model corresponding to each candidate processing combination according to the calibration set and each spectral processing result set.

[0114] Among them, for each candidate processing combination, a candidate tablet weight detection model and a candidate component detection model corresponding to the candidate processing combination can be constructed by using the sample tablet weight, the sample target component content of each patch sample in the calibration set, and the processed spectra and sample feature wavelengths obtained by processing each patch sample under the candidate processing combination, so that multiple candidate tablet weight detection models and multiple candidate component detection models corresponding to multiple different candidate processing combinations can be obtained.

[0115] Step S305: Evaluate the accuracy of each candidate tablet weight detection model and each candidate component detection model using the calibration set and the test set to determine the optimal tablet weight detection model and the optimal component detection model.

[0116] Among them, for the multiple candidate tablet weight detection models and multiple candidate component detection models constructed as described above, the calibration set and the test set can be used to verify their accuracy, and the accuracy indicators corresponding to each model can be obtained. Exemplarily, the accuracy indicators of the model can include the prediction correlation coefficient, the root mean square error of prediction, etc.

[0117] Among them, after obtaining the accuracy indicators corresponding to each model, the candidate tablet weight detection model with the highest corresponding accuracy among the multiple candidate tablet weight detection models can be used as the optimal tablet weight detection model, and the candidate component detection model with the highest corresponding accuracy among the multiple candidate component detection models can be used as the optimal component detection model. Exemplarily, the optimal tablet weight detection model can be the candidate tablet weight detection model with the smallest root mean square error of prediction, and the optimal component detection model can be the candidate component detection model with the smallest root mean square error of prediction. It can be understood that the candidate processing combinations corresponding to the optimal tablet weight detection model and the optimal component detection model can be the same or different.

[0118] Step S306, obtain a tablet weight detection model and a component detection model according to the optimal tablet weight detection model and the optimal component detection model.

[0119] Among them, after obtaining the optimal tablet weight detection model and the optimal component detection model, the optimal tablet weight detection model can be used as the tablet weight detection model, and the optimal component detection model can be used as the component detection model.

[0120] In this embodiment, by obtaining various candidate processing combinations according to different spectral preprocessing and characteristic wavelength screening methods, using each candidate processing combination to process the sample near-infrared spectrum and constructing the corresponding candidate tablet weight detection model and candidate component detection model, and then screening out the optimal tablet weight detection model and the optimal component detection model according to the accuracy of each model as the tablet weight detection model and the component detection model for subsequent application, it is possible to screen out the processing combination with the highest prediction accuracy rate among the processing combinations of various candidate spectral preprocessing methods and characteristic wavelength screening methods. Subsequently, this processing combination can be used to process the product near-infrared spectrum of the product to be detected, and the corresponding model can be used to perform quality detection on the product to be detected, which is beneficial to obtaining more accurate detection results.

[0121] In an exemplary embodiment, as Figure 4 shown, using the calibration set and the test set to evaluate the accuracy of each candidate tablet weight detection model and each candidate component detection model, and determining the optimal tablet weight detection model and the optimal component detection model, includes:

[0122] Step S401: For each candidate processing combination, use the calibration set to verify the candidate tablet weight detection model and the candidate component detection model, and obtain the cross-validation correlation coefficient and the root mean square error of cross-validation of the candidate tablet weight detection model and the candidate component detection model; use the test set to verify the candidate tablet weight detection model and the candidate component detection model, and obtain the prediction correlation coefficient and the root mean square error of prediction of the candidate tablet weight detection model and the candidate component detection model.

[0123] Among them, in order to better verify the accuracy of each candidate tablet weight detection model and each candidate component detection model, and improve the utilization rate of sample data, the cross-validation method can be used to verify the model in this embodiment.

[0124] Specifically, in this step, the K-fold cross-validation method can be used to construct and verify the model based on the calibration set and the test set.

[0125] Among them, the calibration set of the transdermal patch can be further divided into K sample sets, so that these sample sets can be used for K times of model training and verification in the K-fold cross-validation method.

[0126] Among them, during the model training process, K-1 sample sets in the calibration set can be used as the training set, and the sample tablet weight, the content of the sample target component, and the spectral processing results of each patch sample in the training set can be used to train and construct the candidate tablet weight detection model and the candidate component detection model, and the remaining one sample set can be used as the validation set to verify the constructed candidate tablet weight detection model and candidate component detection model. Among them, according to the differences between the tablet weight prediction results and the target component prediction results of the candidate tablet weight detection model and the candidate component detection model obtained in each fold during training on the validation set and the sample tablet weight and the sample target component content of each patch sample in the validation set, the cross-validation correlation coefficient and the root mean square error of cross-validation corresponding to the candidate tablet weight detection model and the candidate component detection model obtained in this fold of training can be calculated. Then, the cross-validation correlation coefficients and the root mean square errors of cross-validation corresponding to the candidate tablet weight detection model and the candidate component detection model obtained in each fold of training can be averaged to obtain the cross-validation correlation coefficient and the root mean square error of cross-validation corresponding to the candidate tablet weight detection model and the candidate component detection model corresponding to this candidate processing combination.

[0127] Meanwhile, after K times of training, all K sample sets in the calibration set and their corresponding spectral processing results can be used to construct the candidate tablet weight detection model and the candidate component detection model, and the test set can be used to verify the constructed candidate tablet weight detection model and candidate component detection model to determine their generalization ability. Among them, according to the differences between the tablet weight prediction results and target component prediction results of the candidate tablet weight detection model and candidate component detection model on the test set and the sample tablet weight and sample target component content of each patch sample in this test set, the prediction correlation coefficient and prediction root mean square error corresponding to the candidate tablet weight detection model and candidate component detection model can be calculated respectively.

[0128] Step S402, according to the cross-validation correlation coefficient, prediction correlation coefficient, cross-validation root mean square error, and prediction root mean square error of each candidate tablet weight detection model corresponding to each candidate processing combination, select the optimal tablet weight detection model from each candidate tablet weight detection model.

[0129] Step S403, according to the cross-validation correlation coefficient, prediction correlation coefficient, cross-validation root mean square error, and prediction root mean square error of each candidate component detection model corresponding to each candidate processing combination, select the optimal component detection model from each candidate component detection model.

[0130] Among them, for the candidate tablet weight detection models and candidate processing combinations corresponding to multiple candidate processing combinations, according to the cross-validation correlation coefficient, prediction correlation coefficient, cross-validation root mean square error, and prediction root mean square error of these models, the optimal tablet weight detection model and the optimal component detection model can be selected from them.

[0131] Exemplarily, when screening the optimal tablet weight detection model and the optimal component detection model, corresponding weights can be set for the cross-validation correlation coefficient, prediction correlation coefficient, cross-validation root mean square error, and prediction root mean square error respectively, and the comprehensive index corresponding to each candidate tablet weight detection model and each candidate component detection model can be calculated by the method of weighted summation. Among them, the comprehensive index can be positively correlated with the cross-validation correlation coefficient and prediction correlation coefficient, and negatively correlated with the cross-validation root mean square error and prediction root mean square error. Therefore, the larger the comprehensive index value of the model, the higher the accuracy of the model can be indicated. Among them, according to the comprehensive index corresponding to each candidate tablet weight detection model and each candidate component detection model, the optimal tablet weight detection model and the optimal component detection model corresponding to the highest comprehensive index can be selected from them respectively.

[0132] In this embodiment, by using the cross-validation method to construct and validate the candidate tablet weight detection model and the candidate component detection model, the sample data can be fully utilized in the model construction. By calculating the cross-validation correlation coefficient, prediction correlation coefficient, cross-validation root mean square error, and prediction root mean square error corresponding to each candidate tablet weight detection model and each candidate component detection model, the performance of the model can be quantitatively evaluated based on these indicators, so that the model with the best performance can be selected to detect the tablet weight and the content of the target component of the product to be detected.

[0133] In an exemplary embodiment, according to the calibration set and each spectral processing result set, candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination are constructed, including: for each candidate processing combination, based on the calibration set and each spectral processing result set, candidate tablet weight detection models and candidate component detection models corresponding to the candidate processing combination are constructed based on the improved kernel partial least squares method.

[0134] Specifically, in this application, for each candidate processing combination, based on the improved kernel partial least squares method, candidate tablet weight detection models and candidate component detection models can be constructed according to the calibration set and the spectral processing result set corresponding to the candidate processing combination.

[0135] Among them, partial least squares regression (PLS) is widely used in the establishment of near-infrared quantitative models. On the basis of partial least squares, kernel partial least squares regression (KPLS) maps sample data to a high-dimensional space through a kernel function during the regression prediction process to achieve non-linear modeling of complex data. Based on kernel partial least squares, the improved kernel partial least squares regression (IKPLS) used in this embodiment can be calculated recursively, so as to avoid high-intensity operations and reduce the operation cost. Specifically, the process of using the improved kernel partial least squares method to construct the model can include:

[0136] The first step is to calculate the covariance matrix X T y, where X is the spectral matrix and y is the sample tablet weight or the content of the sample target component of the transdermal patch; the second step is to let w i = X T y, and perform standardization processing on w i : w i = w i / ‖w i‖; Step 3, calculate a column r of the weight matrix R for calculating the score matrix from the spectral matrix X according to the following formula i : r i = w i (i = 1), Fourth step, calculate the score vector t i (a column of the score matrix T of X): t i = Xr i ; Fifth step, calculate the loading vector p i (a column of the loading matrix P of X): i ; Sixth step, calculate q i (the i-th value of the loading vector q of y): Seventh step: Update the covariance matrix: Return to Step 2 to calculate the next latent variable; Eighth step, after calculating all the principal components, the regression coefficient vector b can be obtained from the following formula: b = Rq. i i i (the i-th value of the loading vector q of y): In this embodiment, by constructing the candidate tablet weight detection model and the candidate component detection model based on the improved kernel partial least squares method, the efficient construction of the model can be realized, which is beneficial to improving the overall efficiency of the quality detection of transdermal patches.

[0137]

[0138] To further illustrate the quality detection method of the transdermal patch of the present application embodiment, the following takes the in-line detection and in-line detection of the batch production process of the transparent transdermal patch with rivastigmine as the target component as an example to illustrate this method.

[0139] Figure 5

[0140] First, please refer to Figure 5 , which is a schematic diagram of the construction process of the tablet weight detection model and the component detection model in this embodiment.

[0140] Among them, before constructing the model, multiple samples of rivastigmine transdermal patches can be collected first. Among them, the tablet weight and the rivastigmine content of the transdermal patch samples can be adjusted by changing the coating thickness during the coating of the transdermal patch and the rivastigmine content in the semi-finished glue solution during the preparation of the samples.

[0141] Among them, for each transdermal patch sample, the acquisition parameters of the corresponding near-infrared spectrum can be determined according to the specific target detection scenario (off-line detection, on-line detection or in-line detection), and the sample near-infrared spectrum corresponding to each sample can be acquired by using the acquisition method corresponding to the acquisition parameters. Among them, according to the near-infrared spectra of the acquired samples, a variety of candidate processing combinations can be obtained according to a variety of spectral preprocessing methods and a variety of characteristic wavelength screening methods, and the near-infrared spectra of the samples corresponding to the samples in the calibration set and the test set are processed respectively by using various candidate processing combinations to obtain the processed near-infrared spectra of the samples.

[0142] Meanwhile, the physicochemical properties of each sample can be determined, including detecting the rivastigmine content (i.e., the content of the sample target component) in each transdermal patch sample by using high performance liquid chromatography (HPLC), and detecting the sample tablet weight of each transdermal patch sample by using an analytical balance.

[0143] Then, after obtaining the near-infrared spectra of the samples corresponding to each transdermal patch sample, the sample target component content and the sample tablet weight, the abnormal samples can be removed by using the principal component analysis method, and the remaining samples can be divided into a calibration set and a test set by using the Kennard-Stone algorithm. Then, according to a variety of spectral preprocessing methods and a variety of characteristic wavelength screening methods, a variety of candidate processing combinations can be obtained, and the near-infrared spectra of the samples corresponding to the samples in the calibration set are processed respectively by using various candidate processing combinations to obtain the corresponding spectral processing result sets. Subsequently, by using the calibration set and its corresponding spectral processing result sets corresponding to each candidate processing combination, candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination can be constructed by using the improved kernel partial least squares method, and the K-fold cross-validation method is used to perform internal verification of the candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination by using the calibration set to calculate the cross-validation correlation coefficient and the root mean square error of cross-validation corresponding to each candidate tablet weight detection model and candidate component detection model respectively, and the external verification of the candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination is performed by using the test set to calculate the prediction correlation coefficient and the root mean square error of prediction corresponding to each candidate tablet weight detection model and candidate component detection model respectively.

[0144] Finally, according to the cross-validation correlation coefficient, prediction correlation coefficient, root mean square error of cross-validation and root mean square error of prediction corresponding to each candidate tablet weight detection model and each candidate component detection model respectively, the optimal tablet weight detection model and the optimal component detection model can be screened respectively and used as the tablet weight detection model and the component detection model of the transdermal patch respectively.

[0145] In an exemplary embodiment, the above method can be applied to the in-line detection of the mass production process of rivastigmine transparent transdermal patches.

[0146] Among them, 25 batches of samples with a coating thickness of 0.2 mm, 0.25 mm, 0.3 mm, 0.35 mm, 0.4 mm and a rivastigmine content of 80%, 90%, 100%, 110%, 120% can be prepared, and 10 cm of transdermal patch samples are collected from these 25 batches of samples 2 as samples for establishing the model.

[0147] Among them, a Fourier transform near-infrared spectroscopy instrument can be installed near the mass production area to provide in-line near-infrared detection. Among them, since the backing layer, adhesive layer, and anti-adhesive layer of the transdermal patch are all transparent, background configuration parameters indicating the setting of the acquisition background can be obtained, and since the target detection scenario is in-line detection, acquisition distance parameters indicating contact acquisition can be obtained. Based on the background configuration parameters and the acquisition distance parameters, in this embodiment, the transdermal patch can be placed on the scanning window of the Fourier transform near-infrared spectrometer and then the gold-plated reflection block can be covered on the transdermal patch for scanning. At the same time, according to the accuracy requirements of in-line detection, the acquisition parameters can be determined as a resolution of 16 cm -1 , 64 scans, and the scanning range is 12000 - 4000 cm -1 .

[0148] Among them, for each transdermal patch sample, the sample near-infrared spectra corresponding to each sample can be collected by the acquisition method corresponding to the above acquisition parameters. At the same time, the rivastigmine content (i.e., the sample target component content) in each transdermal patch sample can be detected by high-performance liquid chromatography, and the sample tablet weight of each transdermal patch sample can be detected by an analytical balance.

[0149] Among them, after obtaining the sample near-infrared spectra, sample target component contents, and sample tablet weights corresponding to each transdermal patch sample, the candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination can be constructed according to the above method, and the optimal tablet weight detection model and the optimal component detection model can be selected from them as the tablet weight detection model and component detection model of the transdermal patch.

[0150] Among them, the candidate processing combinations corresponding to the tablet weight detection model and the component detection model selected in this embodiment are all to preprocess the near-infrared spectra by mean centering and use the competitive adaptive reweighted sampling method to screen the characteristic wavelength variables. Among them, the screening results of the characteristic wavelength variables for the tablet weight can be as shown in Figure 6 part (a), and the screening results of the characteristic wavelength variables for the target component content can be as shown in Figure 6As shown in part (b), where the abscissa represents the wavelength range, the ordinate represents the absorbance, and the vertical line is the selected wavelength variable.

[0151] Among them, the prediction effect of the tablet weight detection model constructed in this embodiment in the calibration set can be as Figure 7 shown in part (a), and the prediction effect of the component detection model in the calibration set can be as Figure 7 shown in part (b). In the figure, the abscissa represents the measured value, the ordinate represents the predicted value assigned to the sample by the model during internal cross-validation. The circles in the figure represent the samples participating in the modeling, and the slashes in the figure represent the regression line. The closer the sample is to the regression line, the better the quantitative detection model effect. Among them, the cross-validation correlation coefficient of the tablet weight detection model is R 2 = 0.99756, and the root mean square error of cross-validation is RMSECV = 0.00098; the cross-validation correlation coefficient of the component detection model is R 2 = 0.99558, and the root mean square error of cross-validation is RMSECV = 0.26934.

[0152] Among them, the prediction effect of the tablet weight detection model constructed in this embodiment in the test set can be as Figure 8 shown in part (a), and the prediction effect of the component detection model in the test set can be as Figure 8 shown in part (b). In the figure, the abscissa represents the measured value of the test set, the ordinate represents the model predicted value, the slash represents the regression line, and the circles represent the predicted samples. The closer the sample is to the regression line, the better the model prediction effect, and the closer the predicted value is to the true value. Among them, the prediction correlation coefficient of the tablet weight detection model is R 2 = 0.99271, and the root mean square error of prediction is RMSEP = 0.00161; the prediction correlation coefficient of the component detection model is R 2 = 0.99237, and the root mean square error of prediction is RMSEP = 0.35887. It can be seen that the tablet weight detection model and the component detection model constructed in this embodiment have relatively accurate predictions for the test set.

[0153] Based on this, in this embodiment, the product near-infrared spectra of the rivastigmine transdermal patch products to be detected can be collected in the near-line detection scenario according to the same acquisition parameters, and the tablet weight detection model and the component detection model can be used to detect the tablet weight and the content of the product target component of the product to be detected.

[0154] In an exemplary embodiment, the quality detection method of the transdermal patch of the present application can be used to perform in-line monitoring on the continuous production process of a transparent transdermal patch with rivastigmine as the target component.

[0155] Among them, according to the expected quality fluctuation range of the rivastigmine transdermal patch, samples with rivastigmine contents of 80%, 100%, and 120% of the standard content and coating thicknesses of 0.225 mm, 0.25 mm, 0.275 mm, 0.3 mm, and 0.325 mm can be prepared respectively. At the same time, according to the variation range of the web travel speed of the production line that requires in-line monitoring, the web travel speeds for collecting spectra can be set to 0.2 m / min, 0.3 m / min, and 0.5 m / min. Among them, this variation range of the web travel speed can be based on the length of the drying oven of the continuous coating production line for the transdermal patch product and the output requirements of continuous production. If the web travel speed is too fast, the transdermal patch product cannot be fully dried; if the web travel speed is too slow, the output requirements of continuous production cannot be met.

[0156] Among them, since the transdermal patch is transparent and the target detection scenario is in-line detection, background configuration parameters indicating the setting of the acquisition background, acquisition position parameters indicating near-infrared spectrum acquisition in the drying completion section, acquisition distance parameters indicating non-contact acquisition, and scanning parameters that can be obtained according to the web travel speed of the target detection scenario can be obtained. Specifically, in this embodiment, the probe of the Fourier transform near-infrared spectroscopy instrument can be installed between the cloth drying completion section and the backlining film laminating section of the transdermal patch production line to collect the near-infrared spectrum of the product that has been dried but not yet laminated, and collect the sample section corresponding to the spectrum. Among them, a polytetrafluoroethylene white board can be used as the background of the spectrum scanning section to avoid the influence of the production line material on the spectrum, and the required sample amount can be obtained according to the web travel speed of the production line and the reference data, and the acquisition parameters are determined as a resolution of 16 cm -1 , the number of scans is 20 times, and the scanning range is 12000 - 4000 cm -1 . Among them, when collecting the spectrum, the production line web travel can be started first to ensure its stable web travel speed, then the near-infrared spectrum scanning can be started and the position where the scanning starts can be recorded, and the position where the scanning stops can be recorded when the scanning stops. Then the area between the two positions is the sample section corresponding to the spectrum.

[0157] Among them, one 10 cm 2 transdermal patch sample can be cut from each of the front, middle, and rear positions of the sample section of the transdermal patch. The tablet weight and rivastigmine content of the transdermal patch sample are detected by an analytical balance and high performance liquid chromatography, and the average tablet weight and rivastigmine content of the three transdermal patch samples are used as the sample tablet weight and the sample target component content of this near-infrared spectrum scanning sample section to establish a database.

[0158] Among them, similar to the process of near-line detection, after obtaining the sample near-infrared spectra, sample target component contents, and sample tablet weights corresponding to each transdermal patch sample, candidate tablet weight detection models and candidate component detection models corresponding to each candidate processing combination can be constructed according to the above method, and the optimal tablet weight detection model and the optimal component detection model can be selected therefrom as the tablet weight detection model and the component detection model of the transdermal patch.

[0159] Among them, the candidate processing combination corresponding to the tablet weight detection model selected in this embodiment is to preprocess the near-infrared spectrum using vector normalization and use the competitive adaptive reweighted sampling method to screen for characteristic wavelength variables, while the candidate processing combination corresponding to the component detection model is to preprocess the near-infrared spectrum using a combination of standard normal variate transformation and detrending and use the competitive adaptive reweighted sampling method to screen for characteristic wavelength variables. Among them, the screening results of the characteristic wavelength variables for the tablet weight can be as Figure 9 shown in part (a) of Figure 9 , and the screening results of the characteristic wavelength variables for the target component content can be as

[0160] shown in part (b) of Figure 10 , where the abscissa represents the wavelength range, the ordinate represents the absorbance, and the vertical line is the selected wavelength variable. Figure 10 Among them, the prediction effect of the tablet weight detection model constructed in this embodiment in the calibration set can be as 2 shown in part (a) of 2 , and the prediction effect of the component detection model in the calibration set can be as

[0161] shown in part (b) of Figure 11 , where the abscissa represents the measured value, the ordinate represents the predicted value given by the model during internal cross-validation, the circles in the figure represent the samples participating in the modeling, and the diagonal line in the figure represents the regression line. The closer the sample is to the regression line, the better the quantitative detection model effect. Among them, the cross-validation correlation coefficient of the tablet weight detection model is R Figure 11 =0.99835, and the root mean square error of cross-validation is RMSECV = 0.00070; the cross-validation correlation coefficient of the component detection model is R 2 =0.94802, and the root mean square error of cross-validation is RMSECV = 0.96745.

[0161] Among them, the prediction effect of the tablet weight detection model constructed in this embodiment in the test set can be as Figure 11 shown in part (a) of Figure 11 , and the prediction effect of the component detection model in the test set can be as 2= 0.99204, the predicted root mean square error is RMSEP = 0.00112; the predicted correlation coefficient of the component detection model is R 2 = 0.87670, the predicted root mean square error is RMSEP = 1.10731. It can be seen that the tablet weight detection model and the component detection model constructed in this embodiment have relatively accurate predictions for the test set.

[0162] Based on this, in this embodiment, the product near-infrared spectrum of the rivastigmine transdermal patch product to be detected can be collected in the scenario of in-line detection according to the same acquisition parameters, and the tablet weight detection model and the component detection model can be used to detect the tablet weight and the content of the product target component of the product to be detected.

[0163] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily have to be sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0164] Based on the same inventive concept, an embodiment of the present application also provides a quality detection device for a transdermal patch for implementing the quality detection method of the transdermal patch involved above. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the quality detection device for a transdermal patch provided below can refer to the limitations on the quality detection method of the transdermal patch in the above text, and will not be repeated here.

[0165] In an exemplary embodiment, as Figure 12 shown, a quality detection device 1200 for a transdermal patch is provided, including:

[0166] A first acquisition module 1201, configured to obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario.

[0167] A first acquisition module 1202, configured to acquire the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters;

[0168] A second acquisition module 1203, configured to acquire the sample tablet weight and the sample target component content of each patch sample.

[0169] A model construction module 1204 is configured to establish a tablet weight detection model and a component detection model for the transdermal patch according to the sample near-infrared spectra, sample tablet weights, and sample target component contents of the patch samples.

[0170] A second acquisition module 1205 is configured to acquire the product near-infrared spectrum of the product to be detected of the transdermal patch according to the acquisition parameters in the target detection scenario.

[0171] A third obtaining module 1206 is configured to obtain the product tablet weight and the product target component content of the product to be detected by using the tablet weight detection model and the component detection model according to the product near-infrared spectrum.

[0172] In an exemplary embodiment, the first obtaining module 1201 is further configured to: if the transdermal patch is transparent, obtain background configuration parameters for setting the acquisition background; otherwise, obtain background configuration parameters for not setting the acquisition background; in the case where the target detection scenario is online detection or in-line detection, obtain an acquisition distance parameter indicating non-contact acquisition, and obtain a scanning parameter according to the running speed of the target detection scenario; in the case where the target detection scenario is near-line detection, if both the backing layer and the anti-adhesive layer of the transdermal patch are opaque, obtain an acquisition distance parameter indicating non-contact acquisition; otherwise, obtain an acquisition distance parameter indicating contact acquisition; and obtain the acquisition parameters of the transdermal patch according to the background configuration parameters, the acquisition distance parameter, and the scanning parameter.

[0173] In an exemplary embodiment, the transdermal patch includes an anti-adhesive layer and a backing layer; the apparatus further includes: a sample addition module, configured to obtain a new patch sample using the product anti-adhesive layer and the product backing layer in the case where there are differences between the product anti-adhesive layer and the product backing layer of the product to be detected and the sample anti-adhesive layer and the sample backing layer of the patch sample; and a model update module, configured to update the tablet weight detection model and the component detection model using the new patch sample to obtain the updated tablet weight detection model and the updated component detection model.

[0174] In an exemplary embodiment, the model construction module 1204 is further configured to: divide the sample near-infrared spectra, the sample tablet weights, and the sample target component contents corresponding to each of the patch samples into a calibration set and a test set; obtain a plurality of candidate processing combinations according to a plurality of spectral preprocessing methods and a plurality of characteristic wavelength screening methods; use each of the candidate processing combinations to process the sample near-infrared spectra in the calibration set to obtain a spectral processing result set corresponding to each of the candidate processing combinations; construct a candidate tablet weight detection model and a candidate component detection model corresponding to each of the candidate processing combinations according to the calibration set and each of the spectral processing result sets; use the calibration set and the test set to evaluate the accuracy of each of the candidate tablet weight detection models and each of the candidate component detection models, and determine an optimal tablet weight detection model and an optimal component detection model; and obtain the tablet weight detection model and the component detection model according to the optimal tablet weight detection model and the optimal component detection model.

[0175] In an exemplary embodiment, the model construction module 1204 is further configured to: for each of the candidate processing combinations, use the calibration set to verify the candidate tablet weight detection model and the candidate component detection model to obtain the cross-validation correlation coefficient and the cross-validation root mean square error of the candidate tablet weight detection model and the candidate component detection model; use the test set to verify the candidate tablet weight detection model and the candidate component detection model to obtain the prediction correlation coefficient and the prediction root mean square error of the candidate tablet weight detection model and the candidate component detection model; and screen out the optimal tablet weight detection model from each of the candidate tablet weight detection models according to the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate tablet weight detection models; and screen out the optimal component detection model from each of the candidate component detection models according to the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate component detection models.

[0176] In an exemplary embodiment, the model construction module 1204 is further configured to: for each of the candidate processing combinations, construct the candidate tablet weight detection model and the candidate component detection model corresponding to the candidate processing combination based on the partial least squares method according to the calibration set and each of the spectral processing result sets.

[0177] Each module in the above-mentioned quality detection device for transdermal patches can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or stored in a memory in the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0178] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structural diagram may be as shown in Figure 13 . The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a quality detection method for a transdermal patch. The display unit of the computer device is used to form a visually visible picture, which may be a display screen, a projection device, or a virtual reality imaging device. The display screen may be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device may be a touch layer covered on the display screen, or may be a button, a trackball, or a touchpad provided on the housing of the computer device, or may also be an external keyboard, a touchpad, or a mouse, etc.

[0179] Those skilled in the art can understand that Figure 13 the structure shown in is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0180] In an embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0181] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0182] In an embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0183] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0184] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0185] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.

[0186] The above embodiments only express several implementation manners of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of this application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several deformations and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application should be subject to the appended claims.

Claims

1. A quality inspection method for a transdermal patch, characterized in that, The method includes: Obtaining the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario; Collecting the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters; Obtaining the sample tablet weight and the content of the sample target component of each patch sample; Establishing a tablet weight detection model and a component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weights, and the content of the sample target components of each patch sample; Collecting the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters; Obtaining the product tablet weight and the content of the product target component of the product to be detected by using the tablet weight detection model and the component detection model according to the product near-infrared spectrum.

2. The method according to claim 1, characterized in that, The obtaining the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario includes: If the transdermal patch is transparent, obtaining background configuration parameters for setting an acquisition background; otherwise, obtaining background configuration parameters without setting the acquisition background; In the case where the target detection scenario is on-line detection or in-line detection, obtaining an acquisition distance parameter indicating non-contact acquisition, and obtaining a scanning parameter according to the tape running speed of the target detection scenario; In the case where the target detection scenario is near-line detection, if both the backing layer and the release liner of the transdermal patch are opaque, obtaining an acquisition distance parameter indicating non-contact acquisition; otherwise, obtaining an acquisition distance parameter indicating contact acquisition; Obtaining the acquisition parameters of the transdermal patch according to the background configuration parameters, the acquisition distance parameters, and the scanning parameters.

3. The method according to claim 1, wherein The transdermal patch includes a release liner and a backing layer; After establishing the tablet weight detection model and the component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weights, and the content of the sample target components of each patch sample, it further includes: In the case where there are differences between the product release liner and the product backing layer of the product to be detected and the sample release liner and the sample backing layer of the patch sample, obtaining new patch samples using the product release liner and the product backing layer; Updating the tablet weight detection model and the component detection model using the new patch samples to obtain the updated tablet weight detection model and the updated component detection model.

4. The method according to claim 1, wherein The establishing the tablet weight detection model and the component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weights, and the content of the sample target components of each patch sample includes: Dividing the sample near-infrared spectra, the sample tablet weights, and the content of the sample target components corresponding to each patch sample into a calibration set and a test set; Obtaining multiple candidate processing combinations according to multiple spectral preprocessing methods and multiple characteristic wavelength screening methods; Processing the sample near-infrared spectra in the calibration set using each candidate processing combination to obtain a spectral processing result set corresponding to each candidate processing combination; Based on the calibration set and each of the spectral processing result sets, construct candidate tablet weight detection models and candidate component detection models corresponding to each of the candidate processing combinations; Use the calibration set and the test set to evaluate the accuracy of each of the candidate tablet weight detection models and each of the candidate component detection models, and determine the optimal tablet weight detection model and the optimal component detection model; Based on the optimal tablet weight detection model and the optimal component detection model, obtain the tablet weight detection model and the component detection model.

5. The method according to claim 4, wherein The step of using the calibration set and the test set to evaluate the accuracy of each of the candidate tablet weight detection models and each of the candidate component detection models, and determining the optimal tablet weight detection model and the optimal component detection model includes: For each of the candidate processing combinations, use the calibration set to verify the candidate tablet weight detection model and the candidate component detection model, and obtain the cross-validation correlation coefficient and the cross-validation root mean square error of the candidate tablet weight detection model and the candidate component detection model; use the test set to verify the candidate tablet weight detection model and the candidate component detection model, and obtain the prediction correlation coefficient and the prediction root mean square error of the candidate tablet weight detection model and the candidate component detection model; Based on the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate tablet weight detection models, screen out the optimal tablet weight detection model from each of the candidate tablet weight detection models; Based on the cross-validation correlation coefficient, the prediction correlation coefficient, the cross-validation root mean square error, and the prediction root mean square error of each of the candidate component detection models, screen out the optimal component detection model from each of the candidate component detection models.

6. The method according to claim 4, wherein The step of based on the calibration set and each of the spectral processing result sets, construct candidate tablet weight detection models and candidate component detection models corresponding to each of the candidate processing combinations includes: For each of the candidate processing combinations, based on the calibration set and each of the spectral processing result sets, construct the candidate tablet weight detection model and the candidate component detection model corresponding to the candidate processing combination based on the improved kernel partial least squares method.

7. A quality inspection device for a transdermal patch, characterized in that, The device includes: A first acquisition module, configured to obtain the acquisition parameters of the near-infrared spectrum of the transdermal patch according to the transparency of the transdermal patch and the target detection scenario; A first acquisition module, configured to acquire the sample near-infrared spectra of multiple patch samples of the transdermal patch according to the acquisition parameters; A second acquisition module, configured to obtain the sample tablet weight and the sample target component content of each of the patch samples; A model construction module, configured to establish a tablet weight detection model and a component detection model of the transdermal patch according to the sample near-infrared spectra, the sample tablet weight, and the sample target component content of each of the patch samples; A second acquisition module, configured to acquire the product near-infrared spectrum of the product to be detected of the transdermal patch in the target detection scenario according to the acquisition parameters; A third acquisition module, configured to obtain the product tablet weight and the product target component content of the product to be detected by using the tablet weight detection model and the component detection model according to the product near-infrared spectrum.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 6.

Images