Rapid detection system for medicine quality

Abstract

The invention discloses a rapid drug quality detection system, and relates to the technical field of drug quality detection. The system integrates a multi-modal fusion detection front end, comprises an image acquisition assembly, a spectral analysis component, a thermal imaging auxiliary module and an image recognition module, and realizes comprehensive detection of appearance, components and thermal state characteristics of drugs. Through microscopic imaging, light source accurate regulation and control and imaging task scheduling, the detection precision and efficiency are improved. The intelligent self-adaptive detection middle end is combined with a physical and chemical detection platform, and deep analysis is carried out on suspected problem drugs. And the cloud data processing and remote monitoring rear end ensures real-time uploading, intelligent analysis and remote operation and maintenance of data, and constructs a drug full-life-cycle traceability system.

Description

Technical Field

[0001] The invention belongs to the technical field of drug quality detection, and in particular relates to a rapid drug quality detection system. Background Art

[0002] Traditional drug quality testing methods often rely on manual sampling and laboratory testing, a time-consuming process that carries the risk of missed detections and misjudgments. Advances in technology, particularly image processing, spectral analysis, thermal imaging, and artificial intelligence, have provided new solutions for drug quality testing. However, existing technologies are mostly limited to a single detection modality, making it difficult to fully reflect drug quality. With the rapid development of the pharmaceutical market and increasing regulatory requirements, traditional testing methods are no longer able to meet the demands of modern drug quality control.

[0003] Furthermore, drug quality issues can arise during production, storage, and transportation due to a variety of factors, such as packaging damage, ingredient changes, and microbial contamination. These issues can pose a serious threat to patient health. Therefore, developing a rapid, accurate, and comprehensive drug quality testing system is of great practical significance and application value. Summary of the Invention

[0004] The purpose of the present invention is to provide a drug quality rapid detection system, which solves the existing problems.

[0005] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention is a drug quality rapid detection system, comprising

[0007] Multimodal fusion detection front-end, which has:

[0008] The image acquisition component consists of multiple imaging devices distributed around the drug delivery track. Each imaging device is equipped with a device that can switch light sources of different wavelengths to capture macroscopic and microscopic images of the drug's appearance.

[0009] Spectral analysis components, including near-infrared spectroscopy and Raman spectroscopy analyzers, measure the deviation of the active ingredient content and trace impurities of the drug based on the interaction characteristics of light and matter during the short period of time when the drug passes through a specific detection area;

[0010] Thermal imaging auxiliary module, which monitors thermal changes during drug testing and assists in determining drug quality stability based on thermal characteristics;

[0011] The image recognition module is built based on a neural network and is trained using a large amount of drug appearance data. It can quickly identify drug conditions, including but not limited to damaged packaging, unclear labels, and abnormal tablet appearance, and preliminarily infer internal quality risks of drugs based on image features.

[0012] As a preferred technical solution of the present invention, the image acquisition component further includes:

[0013] The microscopic imaging module is activated when the drug arrives at the preset workstation, focusing on the key and subtle parts of the drug and accurately capturing tiny defects on the outer shell of tablets and capsules;

[0014] The light source control module monitors the light source status in real time through light sensors and adjusts immediately if any abnormality is detected. For special drugs, the light source parameters can be precisely adjusted as needed to ensure detection accuracy.

[0015] The imaging task scheduling mechanism assigns macro imaging and micro imaging to different processing units for parallel execution. It decides whether to start micro imaging based on the preliminary inspection results of the drug appearance, and conducts microscopic spot checks on some drugs according to specific rules to improve efficiency.

[0016] As a preferred technical solution of the present invention, the spectrum analysis component further includes:

[0017] The environmental monitoring module collects environmental parameters of the detection area in real time, adjusts the spectrum acquisition mode according to environmental changes, and reduces the interference of the environment on the accuracy of the spectral data;

[0018] The drug dosage form adaptation library pre-stores the light-related characteristic data of various drug dosage forms. During detection, it quickly matches the corresponding spectral acquisition strategy based on the dosage form to overcome the influence of differences in the physical state of the drugs.

[0019] As a preferred technical solution of the present invention, the multimodal fusion detection front end further comprises:

[0020] The fusion algorithm unit assigns weights to different modal data based on the key points of drug quality testing, and judges drug quality by integrating the key features of each modality. The weight allocation is dynamically adjusted according to the drug quality risk assessment model;

[0021] The fusion model takes multimodal data as input and outputs drug quality judgment results through neural network calculations. It uses a large number of labeled samples to optimize model parameters, learn the best fusion method, and continuously correct the internal parameters of the model through the backpropagation algorithm to improve judgment accuracy.

[0022] As a preferred technical solution of the present invention, the workflow of the multimodal fusion detection front end is as follows:

[0023] When a drug enters the inspection area, the image acquisition component is triggered. The surrounding imaging device automatically switches the light source band according to the characteristics of the drug, and simultaneously collects macro and micro images of the drug from different angles. The macro image is used to preliminarily judge the integrity of the drug packaging and the clarity of the label, while the micro image is used to capture subtle textures and wrinkle details.

[0024] At the same time, the spectrum analyzer in the spectrum analysis unit is activated. When the drug passes through a specific window for a short period of time, it emits light of a specific wavelength toward the drug and receives reflected or scattered light signals. Based on the principle of interaction between light and matter, the chemical composition of the drug is determined, quickly and accurately determining the deviation in the content of the active ingredient and the presence of trace impurities.

[0025] The thermal imaging module monitors thermal changes during the entire drug testing process with set accuracy, capturing heat release or absorption signals caused by drug deterioration or abnormal chemical reactions in real time to assist in determining drug quality stability.

[0026] The image recognition module processes the collected images of drug appearance in real time in the background. Based on the pre-trained neural network model, it uses the convolution layer to extract image features, reduces the data volume through dimensionality reduction through the pooling layer, and then integrates the features through the fully connected layer to output the classification results of drug appearance problems. Based on the image features, it preliminarily predicts the internal quality risks of the drugs. The entire recognition process is completed in a very short time, ensuring the efficiency of detection.

[0027] As a preferred technical solution of the present invention, it also includes an intelligent adaptive detection mid-end, specifically including:

[0028] Physical testing platform: Laser contour scanning acquires shape data. The testing unit consists of an adaptive gripping and conveying device and an intelligent control system. This unit can quickly adjust gripping and conveying parameters based on the shape characteristics of the drug, adapting to drugs of different shapes and textures. High-precision sensors are used to monitor drug mass distribution, density, and mechanical strength.

[0029] Chemical detection platform: Relying on an integrated chemical workstation, it automatically matches the optimal pretreatment and detection scheme according to drug identification information, and uses a multi-element micro-chemical sensor array to conduct parallel and rapid detection of key chemical indicators of drugs.

[0030] As an optimal technical solution of the present invention, the sample selection of the intelligent adaptive detection mid-end is based on the multimodal fusion detection front-end results. Medicines with damaged appearance, questionable internal quality or abnormal ingredients are marked as samples to be inspected, and regular inspection samples are extracted according to a certain proportion.

[0031] As a preferred technical solution of the present invention, the chemical detection platform also includes: a reagent management system, which ensures the storage conditions of reagents through temperature and humidity control storage chambers, monitors the status of reagents using identification tracking technology, and plans the order of reagent retrieval according to the detection tasks to avoid reagent waste and cross contamination.

[0032] As a preferred technical solution of the present invention, it also includes cloud data processing and remote monitoring backend:

[0033] Cloud data management: Data collected from front-end and mid-end devices is uploaded to the cloud in real time via high-speed networks. Based on cloud computing architecture, big data, and machine learning platforms, data is cleaned and mined, and algorithms are run to improve the accuracy of identifying quality anomalies. This builds a traceability chain for the entire drug life cycle, connecting enterprises and regulatory databases.

[0034] Network redundancy is ensured by adopting multi-network access technology, automatically switching to the backup link when the main network fails, and deploying intelligent traffic control equipment at network nodes to ensure real-time transmission of key data;

[0035] Cloud computing resource management introduces an intelligent containerized management platform to elastically allocate resources based on the system's real-time load and plan capacity expansion in advance using predictive algorithms.

[0036] Remote operation and maintenance support: equip the detection system with a remote communication module. Operators and experts can monitor the system in real time through mobile devices or web pages, quickly diagnose and locate faults, and use virtual augmented reality technology to remotely assist in repairs, shortening the fault repair time.

[0037] The present invention has the following beneficial effects:

[0038] The rapid drug quality detection system of the present invention significantly improves the efficiency and accuracy of drug quality testing. It comprehensively judges drug quality through multimodal fusion technology, effectively identifying damaged packaging, unclear labeling, abnormal appearance, and internal quality issues. The intelligent adaptive detection mid-end performs targeted testing on suspected problematic drugs, reducing false positives and missed detections. The cloud data processing and remote monitoring back-end enables real-time data transmission, intelligent analysis, and remote operation and maintenance, reducing operation and maintenance costs, improving the overall reliability and traceability of the system, and providing strong support for drug quality supervision.

[0039] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is an architecture diagram of the system in the present invention;

[0042] Figure 2 It is a flow chart of the multimodal fusion detection front end in the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0044] like Figure 1-2 The present invention provides a drug quality rapid detection system, comprising:

[0045] Multimodal fusion detection front-end, which has:

[0046] The image acquisition component consists of multiple imaging devices distributed around the drug delivery track. Each imaging device is equipped with a device that can switch light sources of different wavelengths to capture macroscopic and microscopic images of the drug's appearance.

[0047] Spectral analysis components, including near-infrared spectroscopy and Raman spectroscopy analyzers, measure the deviation of the active ingredient content and trace impurities of the drug based on the interaction characteristics of light and matter during the short period of time when the drug passes through a specific detection area;

[0048] Thermal imaging auxiliary module, which monitors thermal changes during drug testing and assists in determining drug quality stability based on thermal characteristics;

[0049] The image recognition module is built based on a neural network and is trained using a large amount of drug appearance data. It can quickly identify drug conditions, including but not limited to damaged packaging, unclear labels, and abnormal tablet appearance, and preliminarily infer internal quality risks of drugs based on image features.

[0050] A specific application embodiment of the present invention is as follows:

[0051] 1. System construction and debugging

[0052] (1) Multimodal fusion detection front-end

[0053] Hardware Installation:

[0054] Multiple imaging devices are precisely installed on the conveyor belts in the pharmaceutical production workshop, evenly distributed around the track to ensure that images of the drug's appearance can be captured from all angles. These imaging devices utilize high-resolution, high-frame-rate industrial cameras paired with a custom-designed, switchable light source covering multiple wavelengths, including ultraviolet, visible, and near-infrared, to accommodate the inspection needs of diverse drug characteristics.

[0055] Near-infrared and Raman spectrometers are installed on either side of the conveyor belt in specific detection areas, ensuring that drugs are accurately detected within a very short time as they pass through. The analyzers are connected to a high-precision optical signal collector and data transmission lines to ensure accurate and real-time data collection.

[0056] The thermal imager is mounted above the conveyor belt. The camera's shooting angle and focal length are adjusted to clearly cover the entire inspection area, and the resolution is set to 0.05°C to keenly capture thermal changes in the pharmaceuticals. The thermal imager is also equipped with dedicated heat dissipation and temperature control devices to ensure stable operation.

[0057] Software configuration and training:

[0058] The image recognition module builds a neural network model based on deep learning frameworks (such as TensorFlow or PyTorch). It collects hundreds of thousands of images of the appearance of drugs produced by the company across different batches and storage conditions. These images include both normal drugs and those with various cosmetic defects. These images are then annotated for model training. After thousands of rounds of iterative training, the model has achieved an accuracy rate of over 95% in identifying issues such as damaged drug packaging, unclear labels, and abnormal tablet appearance. It can also initially infer internal quality risks of drugs based on image features. For example, for tablets with subtle surface texture changes, it can predict the risk of internal fragmentation with an 80% accuracy rate.

[0059] The fusion algorithm unit writes algorithm code based on the drug quality risk assessment model. It quantifies factors such as the importance of drug ingredients, the severity of appearance defects, and the level of abnormal heat in thermal imaging into specific parameters. After inputting these parameters into the model, the weights of different modal data are calculated in real time. For example, for a key cardiovascular drug, if the appearance only has minor scratches, but spectral analysis shows that the active ingredient content deviation is close to the critical value, the spectral data weight is set to 0.6, the image data weight is 0.2, and the thermal imaging data weight is 0.2.

[0060] (2) Intelligent Adaptive Detection Mid-End

[0061] Physical testing platform construction:

[0062] A laser contour scanner is installed above a specific station on the conveyor belt, precisely calibrating its position and scanning angle to ensure rapid and accurate acquisition of the drug's three-dimensional contour data. The adaptive gripping and conveying device connected to the scanner utilizes a high-precision electric robotic arm and intelligent conveyor belt. The gripping area of ​​the robotic arm features a variety of adaptive fixtures tailored to common drug shapes, such as round, oval, and square. A gripping force sensor and path adjustment motor are integrated within the robotic arm, enabling rapid adjustments to the gripping force (with an accuracy of up to 0.03N) and conveying path (with an accuracy of up to 0.08mm) based on the drug's shape within 0.2 seconds.

[0063] High-precision sensors, including microgravity sensors, force sensors, and torque sensors, are deployed around the physical inspection area. These sensors are connected to a data acquisition card via shielded cables and then to an intelligent control system. Powdered drug inspection stations are additionally equipped with airflow control devices and electrostatic adsorption devices to ensure uniform dispersion of the powdered drugs during testing, facilitating accurate assessment of moisture agglomeration, mixing uniformity, and other conditions.

[0064] Chemical testing platform construction:

[0065] The integrated chemical workstation is placed in the clean area of ​​the workshop, and is divided into functional areas, including a drug pretreatment area, a reagent storage area, and a detection and analysis area. In the drug pretreatment area, grinding equipment of various specifications, an intelligent temperature control system, and an ultrasonic auxiliary device are installed. The parameters of the grinding equipment can be automatically adjusted according to dozens of preset grinding parameter combinations, and the temperature control system has an accuracy of ±0.5°C. The reagent storage area is equipped with a temperature-controlled and humidity-controlled storage cabin, which is equipped with multi-layer shelves and a radio frequency identification (RFID) tag reader to monitor the status of the reagents in real time. The detection and analysis area integrates an array of 15 micro chemical sensors based on different principles. The sensors are connected to signal amplifiers and data collectors to ensure parallel and rapid detection of key chemical indicators such as heavy metal content in drug solutions (detection limit as low as ppb level), trace impurities, and changes in bioactive ingredients.

[0066] Sample selection and reagent management system configuration:

[0067] A sample selection program was developed and connected to the data output interface of the multimodal fusion detection front-end. Based on the front-end detection results, drug samples were marked according to pre-set dynamic sampling rules. For example, during the initial production phase of a newly developed tablet drug, if any minor appearance anomalies or significant deviations in spectral analysis were observed, the tablet would be immediately marked as a sample for inspection, and the regular sampling rate would be increased to 10%.

[0068] The reagent management system software is integrated with temperature and humidity-controlled storage chambers and RFID tag readers. A database records all reagent information, including name, specification, storage conditions, and expiration date. When a test task is initiated, the system automatically searches for the optimal pretreatment and testing plan based on the drug's identification information. It also intelligently schedules reagent retrieval based on inventory availability to avoid reagent waste and cross-contamination.

[0069] (3) Cloud data processing and remote monitoring backend

[0070] Cloud data management system construction:

[0071] Deploy a cloud computing server cluster in the company's data center, utilizing high-performance server hardware with ample CPU, memory, and storage resources. Install a big data processing platform based on Hadoop and Spark, as well as machine learning frameworks based on Scribd-Learn and TensorFlow. Data collected at the front-end and mid-end is uploaded to cloud data storage nodes in real time via the 5G network. Data is initially encrypted and verified during upload to ensure data integrity and security.

[0072] We developed a drug quality model optimization algorithm, which is activated daily to perform clustering and regression analysis on massive amounts of drug data stored in the cloud. For example, a cluster analysis of data from tablet production in a specific production area during the summer of the previous year revealed an increasing trend in the deviation of the active ingredient content in some drugs due to high humidity. The system then issued an early warning and guided the company to adjust humidity control parameters in summer production workshops, effectively mitigating quality risks.

[0073] Develop an interface program that connects with the pharmaceutical manufacturer's ERP system and the pharmaceutical regulatory authority's database to achieve seamless data exchange. When quality issues arise, quality management personnel can quickly access detailed information such as the raw material supplier, production workshop, production date, and testing personnel of the problematic drug through the company's internal ERP system. Regulatory authorities can also access relevant data in real time for oversight and inspection.

[0074] Network redundancy assurance and cloud computing resource management:

[0075] In terms of network architecture, 5G is used as the primary network, with wired Ethernet as the backup link. Intelligent traffic control devices are deployed on network switches and routers, and a network switching program is written. When the 5G network signal strength falls below a set threshold or a fault occurs, the network switching program automatically switches data transmission to wired Ethernet within 5 seconds. Simultaneously, the intelligent traffic control devices prioritize the real-time transmission of critical data, such as drug test results and fault alarms, keeping critical data transmission delays to less than 0.5 seconds.

[0076] We introduced the Kubernetes intelligent containerized management platform to containerize and deploy various detection module applications in the cloud. We developed a resource monitoring and allocation program to monitor the system's CPU, memory, storage, and other resource usage in real time. Using a predictive algorithm based on historical data, business growth trends, and seasonal fluctuations, we planned capacity expansion plans in advance. For example, before the peak pharmaceutical production season, based on historical data, we predicted a 30% increase in detection workloads. The system automatically allocated more resources to key modules, such as the multimodal fusion detection front-end and the intelligent adaptive detection mid-end, a week in advance to ensure smooth system operation.

[0077] Remote operation and maintenance support system construction:

[0078] Equip the inspection system with a remote communication module, install a dedicated monitoring app on the mobile devices (phones and tablets) of operators and experts, and develop a monitoring and management platform on the web. The app and web platform connect to the inspection system via 5G network or VPN to achieve real-time monitoring.

[0079] The fault diagnosis program is integrated into the detection system's control software. When a system fault occurs, the program completes the preliminary fault location within 30 seconds through a comprehensive analysis of various aspects of information, including the operating status of each system component, data transmission status, and algorithm execution results. It then pushes detailed fault alert information (including fault code, possible cause, and a diagram of the fault location) to the pre-determined technician's mobile phone, and simultaneously uploads fault-related system logs and key data to the cloud. Technicians remotely access the detection system via a mobile app or webpage, utilizing virtual reality (VR) / augmented reality (AR) technology to remotely view the actual scene through head-mounted display devices, conduct virtual operation demonstrations, and guide on-site operators in troubleshooting. This reduces the average fault repair time by over 50%.

[0080] 3. Actual testing process

[0081] (1) Drug online testing

[0082] Drugs are continuously transported from the production line to the inspection system's conveyor belt, first entering the multimodal fusion detection front end. When the drug triggers the sensor, the image acquisition component quickly activates. The surrounding imaging device automatically switches the light source band based on the drug's characteristics, synchronously capturing macro and micro images of the drug's appearance from different angles. For example, for a white tablet drug, to highlight the subtle texture of the tablet's surface, the imaging device automatically switches to the near-ultraviolet light band, clearly capturing subtle scratches as small as 0.005mm on the tablet surface; for capsule drugs, it switches to the visible light band to accurately capture details such as the color uniformity and tiny pores of the capsule shell.

[0083] At the same time, the near-infrared spectrometer and Raman spectrometer in the spectral analysis component are activated. When the drug passes through a specific window for a short period of time (about 1 second), light of a specific wavelength is emitted to the drug and reflected or scattered light signals are received. The chemical composition of the drug is determined based on the principle of interaction between light and matter. The near-infrared spectrometer accurately determines that the deviation of the active ingredient content of the drug is -0.08%, and the Raman spectrometer detects that the trace impurity content is 2ppm, both within the qualified range. The thermal imaging module monitors the thermal changes of the drug testing process with a resolution of 0.05°C throughout the process. No abnormal hot areas are found, indicating that the quality and stability of the drug are good. The image recognition module processes the collected drug appearance images in real time in the background, and outputs the classification results of the drug appearance problems within 0.3 seconds, judging that the drug packaging is complete, the label is clear, and there are no obvious appearance abnormalities.

[0084] (2) Intelligent adaptive sampling

[0085] The detection results of the multimodal fusion detection front-end are transmitted in real time to the sample selection program of the intelligent adaptive detection mid-end. Since the appearance and ingredients of the drug are normal, it is marked as a routine inspection sample according to the routine inspection ratio (assuming it is 3%).

[0086] The drug then enters the physical inspection platform, where laser profile scanning instantly captures the drug's three-dimensional contour data. Based on this data, the adaptive gripping and conveying device rapidly adjusts the gripping force and conveying path within 0.2 seconds, smoothly transporting the drug to the inspection station. High-precision sensors monitor the drug's mass distribution, density, and mechanical strength in real time. For this tablet, the density is uniform, with no significant mass unevenness. The mechanical strength meets standards, and no signs of fragility are observed.

[0087] The drug then enters the chemical testing platform. The integrated chemical workstation automatically matches the optimal pretreatment and testing plan based on the drug identification information, efficiently completing the drug grinding process within 5 seconds, and controlling the grinding particle size to 60 mesh. Using an intelligent temperature control system and ultrasonic auxiliary device, the drug dissolution temperature is precisely controlled at 30°C to ensure complete dissolution of the drug. The introduced micro-chemical sensor array performs parallel rapid testing of key chemical indicators such as heavy metal content, trace impurities, and changes in bioactive ingredients in the drug solution. The detection time for a single chemical indicator does not exceed 20 seconds, and the entire chemical testing process takes 40 seconds, which is more than 65% shorter than traditional methods. The test results show that all chemical indicators are qualified.

[0088] (3) Data upload and remote monitoring

[0089] After drugs are tested at the multimodal fusion detection front-end and intelligent adaptive detection mid-end, all test data is uploaded to the cloud-based data management system in real time via the 5G network. The cloud-based system cleans and mines the data, automatically running a machine learning-based drug quality model optimization algorithm daily to continuously improve the system's accuracy in identifying drug quality anomalies. For example, a cluster analysis of tablet drug testing data for a week revealed a slight upward trend in the active ingredient content of drugs produced by a certain production team during a certain period. The system issued an early warning, allowing the company to promptly inspect and adjust the team's production process, preventing the escalation of quality issues.

[0090] Operators and experts can monitor the testing system's operational status in real time via a mobile app or web platform, including information such as drug testing progress, component status, and test result statistics. If a system malfunction occurs, such as abnormal spectrum analyzer data or a conveyor belt jam, the fault diagnosis program will initially locate the fault within 30 seconds and send detailed fault alert information to technicians. Using VR / AR technology, technicians can remotely assist on-site operators in troubleshooting, ensuring continuous testing.

[0091] IV. Implementation Effect Evaluation

[0092] After introducing the rapid drug quality detection system of the present invention, the pharmaceutical company's drug defective rate was reduced from the original 0.5% to below 0.1%, greatly improving product quality, reducing recalls and complaints caused by quality problems, and maintaining the company's brand reputation and market competitiveness.

[0093] Due to the significant improvement in testing efficiency, the drug sampling work that originally took several hours or even days can now complete the comprehensive testing of a batch of drugs within half an hour, meeting the company's rapid sampling needs for large-scale production and effectively ensuring the efficient operation of the production line.

[0094] The construction of cloud-based data management and remote monitoring backend enables enterprises to issue early warnings of potential quality risks 1-2 weeks in advance and take timely preventive measures. At the same time, it realizes the traceability of the entire life cycle of drugs, provides strong support for quality problem investigation and responsibility determination, and has been highly recognized by drug regulatory authorities.

[0095] In summary, the rapid drug quality detection system of the present invention has demonstrated excellent performance in actual production applications, providing a solid technical guarantee for the high-quality development of the modern pharmaceutical industry.

[0096] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0097] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A drug quality rapid detection system, characterized in that: include Multimodal fusion detection front end, which has: An image acquisition component, which is composed of a plurality of imaging devices distributed around the drug delivery track, each of which is equipped with a device capable of switching light sources of different wavelengths, and is used to acquire macroscopic and microscopic images of the drug appearance; Spectral analysis components, including near-infrared spectrometers and Raman spectrometers, can measure the deviation of the active ingredient content and trace impurities of the drug based on the interaction characteristics between light and matter during the short period of time when the drug passes through a specific detection area; Thermal imaging auxiliary module, which monitors thermal changes during drug testing and assists in judging the stability of drug quality based on thermal characteristics; The image recognition module is built based on a neural network and is trained using a large amount of drug appearance data. It can quickly identify drug conditions, including but not limited to damaged packaging, unclear labels, and abnormal appearance of tablets, and preliminarily infer internal quality risks of drugs based on image features.

2. A drug quality rapid detection system according to claim 1, characterized in that: The image acquisition component also includes: The microscopic imaging module is activated when the drug arrives at the preset station, focusing on the key and subtle parts of the drug and accurately capturing tiny defects on the outer shell of tablets and capsules; The light source control module monitors the light source status in real time through light sensors, and adjusts immediately if an abnormality is found. For special drugs, the light source parameters can be precisely adjusted as needed to ensure detection accuracy; The imaging task scheduling mechanism assigns macro imaging and micro imaging to different processing units for parallel execution. It decides whether to start micro imaging based on the preliminary inspection results of the drug appearance, and conducts microscopic spot checks on some drugs according to specific rules to improve efficiency.

3. A drug quality rapid detection system according to claim 1, characterized in that: The spectrum analysis component also includes: Environmental monitoring module collects environmental parameters of the detection area in real time, adjusts the spectrum acquisition mode according to environmental changes, and reduces the interference of the environment on the accuracy of spectral data; The drug dosage form adaptation library pre-stores the light-related characteristic data of various drug dosage forms. During testing, it quickly matches the corresponding spectral acquisition strategy based on the dosage form to overcome the influence of differences in the physical state of the drugs.

4. A drug quality rapid detection system according to claim 1, characterized in that: The multimodal fusion detection front end also has: The fusion algorithm unit assigns weights to different modal data based on the key points of drug quality detection, and judges the drug quality by combining the key features of each modality. The weight allocation is dynamically adjusted according to the drug quality risk assessment model; The fusion model takes multimodal data as input, outputs drug quality judgment results through neural network calculations, optimizes model parameters using a large number of labeled samples, learns the best fusion method, and continuously corrects the internal parameters of the model through the back-propagation algorithm to improve judgment accuracy.

5. A drug quality rapid detection system according to claim 1, characterized in that: The workflow of the multimodal fusion detection front end is as follows: When a drug enters the inspection area, the image acquisition component is triggered to work. The surrounding imaging device automatically switches the light source band according to the characteristics of the drug, and simultaneously collects macro and micro appearance images of the drug from different angles. The macro image is used to preliminarily judge the integrity of the drug packaging and the clarity of the label, and the micro image is used to capture subtle textures and wrinkle details. At the same time, the spectrometer in the spectral analysis component is activated. When the drug passes through a specific window for a short period of time, it emits light of a specific wavelength to the drug and receives reflected or scattered light signals. Based on the principle of interaction between light and matter, the chemical composition of the drug is determined, and the deviation of the active ingredient content and trace impurities are quickly and accurately obtained. The thermal imaging module monitors the thermal changes of the drug testing process with set accuracy throughout the process, and captures the heat release or absorption signals caused by drug deterioration and abnormal chemical reactions in real time, to assist in judging the stability of drug quality; The image recognition module processes the collected images of drug appearance in real time in the background. Based on the pre-trained neural network model, it uses the convolution layer to extract image features, reduces the data volume through dimensionality reduction through the pooling layer, and then integrates the features through the fully connected layer to output the classification results of drug appearance problems. The internal quality risks of drugs are preliminarily predicted based on the image features. The entire recognition process is completed in a very short time to ensure the efficiency of detection.

6. A drug quality rapid detection system according to claim 1, characterized in that: It also includes intelligent adaptive detection mid-end, including: Physical testing platform: Laser profile scanning obtains shape data. The testing unit is composed of an adaptive clamping and conveying device and an intelligent control system. It can quickly adjust the clamping and conveying parameters according to the shape characteristics of the drug, adapt to drugs of different shapes and textures, and use high-precision sensors to monitor drug quality distribution, density and mechanical strength related data; Chemical detection platform: Relying on an integrated chemical workstation, it automatically matches the best pretreatment and detection scheme according to drug identification information, and uses a multi-element micro-chemical sensor array to perform parallel and rapid detection of key chemical indicators of drugs.

7. A drug quality rapid detection system according to claim 6, characterized in that: The sample selection of the intelligent adaptive detection mid-end is based on the multimodal fusion detection front-end results. Medicines with damaged appearance, questionable internal quality or abnormal ingredients are marked as samples to be inspected, and regular inspection samples are drawn according to a certain ratio.

8. A drug quality rapid detection system according to claim 6, characterized in that: The chemical detection platform also includes: a reagent management system, which ensures reagent storage conditions by controlling the temperature and humidity of the storage chamber, monitors the reagent status using identification tracking technology, and plans the order of reagent retrieval according to the detection task to avoid reagent waste and cross contamination.

9. A drug quality rapid detection system according to claim 1, characterized in that: The system also includes cloud data processing and remote monitoring backend: Cloud data management: upload the data collected by the front-end and mid-end to the cloud in real time through high-speed networks. Based on cloud computing architecture and big data and machine learning platforms, clean and mine data, run algorithms to improve the accuracy of quality anomaly identification, and build a drug life cycle traceability link to connect enterprises and regulatory databases. Network redundancy protection, using multi-network access technology, automatically switching to the backup link when the main network fails, and deploying intelligent traffic control equipment at network nodes to ensure real-time transmission of key data; Cloud computing resource management, introducing an intelligent containerized management platform, elastically allocating resources based on the real-time load of the system, and planning capacity expansion in advance in combination with prediction algorithms; Remote operation and maintenance support: equip the detection system with a remote communication module. Operators and experts can monitor the system in real time through mobile or web terminals, quickly diagnose and locate faults, and use virtual augmented reality technology to remotely assist in repairs, thus shortening the fault repair time.

Images