Deep learning-based clinical laboratory sample storage system
By adopting deep learning technology in the laboratory sample storage system, using sensors to collect and clean sample information, and building a deep learning model for analysis, the problems of data transmission stability and feature information utilization are solved, and the efficiency of sample management and abnormal detection is improved .
Patent Information
- Application Number
- CN202411506657.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-28
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional laboratory sample storage systems are difficult to ensure the stability and integrity of data transmission, and are difficult to make full use of sample feature information for in-depth analysis, resulting in low sample management and abnormal detection efficiency.
A laboratory sample storage system based on deep learning was designed, including sample collection module, data processing module and deep learning module. Sample information is collected through sensors, data cleaning and feature extraction are carried out, and deep learning models are constructed using convolutional neural networks. Analyze and store.
It realizes stable, complete transmission and in-depth analysis of data, improves the accuracy and efficiency of sample management, and can promptly discover potential abnormal samples.
Smart Images

Figure CN119025719A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical data storage, and in particular to a laboratory sample storage system based on deep learning. Background Art
[0002] With the continuous development of medical technology, the laboratory department plays an increasingly important role in disease diagnosis and treatment. The laboratory department needs to process a large number of samples every day. How to efficiently and accurately store and manage these samples has become an urgent problem to be solved. On the one hand, in the sample storage system of the laboratory department, a large amount of medical data needs to be transmitted between different devices, such as from testing instruments to storage systems. It is necessary to solve how to ensure the stability of data transmission, avoid data loss or transmission interruption due to network fluctuations, equipment failures, etc., and ensure that data can enter the storage system for processing in a timely and complete manner. On the other hand, traditional storage systems find it difficult to conduct in-depth analysis of samples. It is impossible to make full use of various characteristic information of samples, such as image features, test results, storage conditions, etc., to perform more accurate classification, anomaly detection and storage location prediction. This makes it difficult to find specific samples quickly and accurately during sample management, and it is also difficult to detect potential abnormal samples in advance. Summary of the invention
[0003] In order to solve the above technical problems, the present invention is implemented through the following technical solutions: A laboratory sample storage system based on deep learning, comprising: The sample collection module is used to collect sample information from the laboratory through sensors, and extract sample association information and feature information based on the sample information; The data processing module is connected to the sample collection module and is used to clean, convert the format and mark the abnormalities of the collected sample association information and feature information to obtain standard feature data and abnormal marking data, so as to facilitate the subsequent deep learning model training and reasoning; A deep learning module, connected to the data processing module, is used to adopt a convolutional neural network architecture to build a deep learning model through historical sample data, and use the deep learning model to analyze the sample information; The sample storage module is used to store sample information and analysis results obtained by the deep learning model for subsequent information query and management; The user interface module is used to set the operation interface for users to input sample information, view analysis results and perform system settings.
[0004] Preferably, the sample collection module includes: The sample identification unit is used to collect sample information from the laboratory through sensors, identify the type and source of the sample information, and obtain an information data packet; An information input unit, used to receive information data packets and input the information data packets of the samples into a database of the sample storage system for storage; The feature extraction unit is used to receive the information data packet obtained by the sample identification unit and extract the feature information of the sample from the information data packet, wherein the feature information includes the color, shape and smell of the sample.
[0005] Preferably, the process of the sensor collecting sample information is as follows: Optical sensors, pressure sensors and RFID sensors are deployed in the sample collection area of the laboratory. Optical sensors are then used to detect the color and transparency characteristics of the samples. When the samples enter the collection area of the optical sensors and are irradiated by light, the optical sensors detect the intensity and wavelength of the reflected light to obtain the appearance characteristics of the samples. Then, the sealing of the sample container is detected using a pressure sensor. A proximity sensor is installed in each storage compartment of the sample storage cabinet. When the sample container is placed in the storage compartment, the proximity sensor senses the entry of the object and immediately activates the pressure sensor to detect the sealing of the sample container and obtain the sealing status information. The unique identification of the sample is then identified using a radio frequency identification sensor. When a sample with an RFID tag approaches the collection area of the sensor, the radio frequency identification sensor reads the information in the tag to obtain the tag identification information.
[0006] Preferably, the detailed process of obtaining the information data packet is as follows: Combining the appearance feature information of the sample collected by the optical sensor, the sealing state information collected by the pressure sensor, and the label identification information collected by the radio frequency identification sensor to obtain integrated information; Analyze the characteristics and identification of the sample based on the integrated information, and identify the sample type and sample source: obtain the sample type by analyzing the color characteristics, shape characteristics and composition characteristics in the appearance characteristic information of the sample; The source of the sample is obtained by reading the collection location information and the inspection department information in the tag identification information obtained from the RFID tag; The identified sample type and sample source are then packaged together with the integrated information of the sample to generate an information data packet, where the information data packet includes: sample number, type, source, and collection time; sample feature information collected by the sensor: appearance characteristics, container sealing and component analysis; other information: patient information and test items.
[0007] Preferably, the process by which the information input unit inputs the information data packet into the database of the sample storage system for storage is as follows: Based on the WiFi network coverage inside the laboratory, WiFi is set up for communication; based on the server where the information entry unit and the database of the sample storage system are located, a WiFi network card is installed and the network is configured, the network name and password are set, and each device is connected to the WiFi network. Then, identity authentication is performed by means of a user name and password, and the information entry unit is connected to the WiFi network of the sample storage system. Communication protocol header information and tail information are added to the information data packet through encapsulation. The encapsulated information data packet is entered into the database of the sample storage system based on real-time transmission, and an error detection and retransmission mechanism is set. A cyclic redundancy check is used to detect whether an error occurs in the data transmission process. If an error is found, a retransmission is automatically requested; Then, the receiving program on the sample storage system database server listens to the port, waits to receive the information data packet from the information entry unit, parses the received information data packet, removes the communication protocol header information and tail information, obtains the original information data packet, and verifies and cleans the information data packet, the verification includes checking the integrity of the data packet, whether the format is correct, and obtains a complete data packet. If the information data packet is incomplete or the format is wrong, the information data packet is discarded, and a retransmission notification is generated, and the retransmission notification is sent to the information entry unit to request retransmission; After the complete data packet is stored in the database, a confirmation message is generated to inform the sender that the data has been successfully stored, and the confirmation message is sent to the information entry unit, so that the information entry unit can know the results of data transmission and storage for subsequent processing or error handling. Through the above detailed storage process, it can be ensured that the information data packet is safely and accurately transmitted to the database of the sample storage system via wireless communication for storage, providing a reliable data basis for subsequent sample management and analysis. In addition, by transmitting the sample information to the sample storage system in the form of a complete data packet, rather than the frequent transmission of scattered individual data, the data is transmitted centrally while eliminating abnormal, missing and invalid data, thereby retaining the integrity of the data.
[0008] Preferably, the data processing module includes: A data cleaning unit is used to clean the data based on sample association information and feature information, remove noise and incomplete data in the data, and obtain sample feature data; The data processing unit is used to obtain the sample feature data obtained in the data cleaning unit, and convert the sample feature data into a standard format to obtain standard feature data for the input of the model; The anomaly detection unit is used to receive standard feature data, identify and mark abnormal samples in the standard feature data, and obtain abnormal marked data to prevent the influence of erroneous data on model training and reasoning; the function of this module ensures the accuracy and consistency of the data, and provides reliable data support for subsequent deep learning models.
[0009] Preferably, the deep learning module includes: The training module uses a convolutional neural network architecture to perform feature learning and pattern recognition through historical sample data to train a deep learning model; The inference module analyzes the newly collected sample information based on the trained deep learning model and combines standard feature data, and generates analysis results, which include the classification results and prediction information of the samples. The design of this module enables the sample storage system to have efficient real-time analysis capabilities and provide timely diagnostic support after sample collection.
[0010] Preferably, the process of building the deep learning model is as follows: Collect historical sample data from the laboratory database, including sample images, test results, and storage conditions; then clean the collected historical sample data to remove noise, outliers, and incomplete data; then annotate the cleaned sample data to determine the category and features of the sample; use data enhancement technology to rotate, flip, crop, and scale image data, and perform random perturbations and interpolation operations on numerical data; clarify the tasks of the deep learning model in the laboratory sample storage system, that is, the task requirements, which include sample classification, anomaly detection, and storage location prediction; For image data, the visual geometry group network architecture is used for feature extraction; For the test result data, a network architecture combining convolutional neural network and fully connected neural network is used for feature extraction. First, the convolutional neural network is used to process the image data and extract image features. Then, the image features and the test result data are input into the fully connected neural network for further analysis and prediction. This is used to combine the information of the image and the test result to improve the accuracy of the model. For storage condition data, a decision tree architecture is used to classify and predict different characteristics of storage conditions, which is convenient for understanding and interpretation; the above three architectures are integrated to form a convolutional neural network; the parameters of the convolutional neural network are determined according to the size, complexity and computing resources of the sample data, and the parameters of the convolutional neural network are appropriately adjusted and optimized according to the task requirements; Then, the preprocessed historical sample data is divided into a training set and a test set; wherein, the division is carried out according to the following ratio: 70% of the data is used as a training set, and 30% of the data is used as a test set; and the stochastic gradient descent algorithm is selected to train the convolutional neural network; Input the training set data into the convolutional neural network, calculate the output result through forward propagation, and then calculate the error between the predicted result and the true label according to the loss function; Use the optimization algorithm to back-propagate the error and update the network's weights and bias parameters to reduce the value of the loss function; repeat this process until the loss function converges or reaches the preset number of training times to obtain a deep learning model.
[0011] Preferably, during the training process, the deep learning model is evaluated through test set data, the accuracy, recall rate, and F1 value of the deep learning model are calculated, and the corresponding evaluation results are obtained to understand the performance and generalization ability of the model; then, according to the evaluation results, the structure and parameters of the model are adjusted, such as increasing or decreasing the number of network layers, adjusting the convolution kernel size, changing the learning rate, etc., to improve the performance of the model.
[0012] Preferably, when the model training achieves satisfactory performance, the trained deep learning model is deployed to the laboratory sample storage system; Save the model's weights and structure information as files, and then load the model into the system for real-time sample analysis and storage management; After the model is deployed, continue to collect new sample data and optimize the model online. Use incremental learning technology to update the model parameters to adapt to new data without retraining the entire model. At the same time, you can also optimize the model's performance, such as using model compression technology to reduce the size and computational complexity of the model, and improve the model's running speed and efficiency.
[0013] The present invention provides a laboratory sample storage system based on deep learning, which has the following beneficial effects: 1. The deep learning-based laboratory sample storage system transmits sample information to the sample storage system in the form of complete data packets, rather than the frequent transmission of scattered individual data. While transmitting the data in a centralized manner, it also eliminates abnormal, missing and invalid data, thereby retaining the integrity of the data. At the same time, by verifying and cleaning the information data packets, it can verify in advance whether the information data packets have defects such as incomplete data or incorrect format before entry, so as to retransmit them in time, ensure the integrity and accuracy of data transmission, and avoid data loss or transmission interruption due to network fluctuations, equipment failure, etc.; and by confirming the feedback effect brought by the information, it can provide information feedback on the data that has been successfully entered, ensuring that the data can enter the storage database in a timely and complete manner, thereby ensuring the stability of medical data transmission.
[0014] 2. This deep learning-based laboratory sample storage system improves data quality and diversity by pre-processing historical sample data such as cleaning, labeling and enhancement, which helps to improve the performance of deep learning models; it uses appropriate network architectures for feature extraction for different types of data, fully leveraging the advantages of various architectures and improving the effect and accuracy of feature extraction; it reasonably divides the training set and test set, and selects appropriate optimization algorithms for model training, ensuring the accuracy and generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 This is a flowchart of a deep learning-based laboratory sample storage system of the present invention. DETAILED DESCRIPTION
[0016] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.
[0017] like Figure 1 As shown, the present invention provides a technical solution: a deep learning-based laboratory sample storage system, including a sample collection module, which is used to collect sample information from the laboratory through a sensor, and extract sample association information and feature information based on the sample information, thereby achieving comprehensive acquisition of sample data and providing a rich data basis for subsequent analysis; The data processing module is connected to the sample collection module and is used to clean, convert the format and mark abnormalities of the collected sample association information and feature information to obtain standard feature data and abnormal marking data; it improves the data quality and availability to facilitate the subsequent deep learning model training and reasoning; A deep learning module is connected to the data processing module and is used to adopt a convolutional neural network architecture to build a deep learning model through historical sample data, and use the deep learning model to analyze sample information, which can mine potential patterns and rules in the samples and improve the accuracy and depth of sample analysis; The sample storage module is used to store sample information and analysis results obtained by the deep learning model for subsequent information query and management, which improves work efficiency and data traceability; The user interface module is used to set the operation interface for users to input sample information, view analysis results and perform system settings, which improves the system's usability and user experience.
[0018] The sample collection module includes: a sample identification unit, which is used to collect sample information from the laboratory through sensors, and identify the type and source of the sample information to obtain an information data packet, which helps to accurately understand the properties and background of the sample and provide accurate information for subsequent processing; an information entry unit, which is used to receive information data packets and enter the sample information data packets into the database of the sample storage system for storage, thereby realizing digital storage of sample data for easy management and query; a feature extraction unit, which is used to receive the information data packets obtained by the sample identification unit and extract the feature information of the sample from the information data packets. The feature information includes the color, shape and smell of the sample, which provides key input data for the deep learning model and helps to improve the accuracy of the analysis.
[0019] The process of the sensor collecting sample information is as follows: Optical sensors, pressure sensors and RFID sensors are deployed in the sample collection area of the laboratory. Optical sensors are then used to detect the color and transparency characteristics of the samples. When the samples enter the collection area of the optical sensors and are irradiated by light, the optical sensors detect the intensity and wavelength of the reflected light to obtain the appearance characteristics of the samples. Then, the sealing of the sample container is detected using a pressure sensor. A proximity sensor is installed in each storage compartment of the sample storage cabinet. When the sample container is placed in the storage compartment, the proximity sensor senses the entry of the object and immediately activates the pressure sensor to detect the sealing of the sample container and obtain the sealing status information. The unique identification of the sample is then identified using a radio frequency identification sensor. When a sample with an RFID tag approaches the collection area of the sensor, the radio frequency identification sensor reads the information in the tag to obtain the tag identification information.
[0020] It should be further explained that, in the specific implementation process, the comprehensive use of optical sensors, pressure sensors and radio frequency identification sensors can comprehensively detect the appearance characteristics, container sealing and unique identification of the samples, thereby improving the integrity and accuracy of the sample information; the pressure sensor is triggered by the proximity sensor to detect the sealing of the sample container, thereby improving the timeliness and accuracy of the detection and helping to ensure the safe storage of the samples.
[0021] By collecting sample information through a variety of sensors, we can comprehensively obtain various information such as the sample's appearance characteristics, container sealing, unique identification, etc., thereby improving the accuracy and completeness of the sample information; the sample identification unit can accurately identify the type and source of the sample information, providing a clear direction for subsequent analysis and management.
[0022] The detailed process of obtaining the information data packet is as follows: Combining the appearance feature information of the sample collected by the optical sensor, the sealing state information collected by the pressure sensor, and the label identification information collected by the radio frequency identification sensor to obtain integrated information; Analyze the characteristics and identification of the sample based on the integrated information, and identify the sample type and sample source: obtain the sample type by analyzing the color characteristics, shape characteristics and composition characteristics in the appearance characteristic information of the sample; The source of the sample is obtained by reading the collection location information and the inspection department information in the tag identification information obtained from the RFID tag; The identified sample type and sample source are then packaged together with the integrated information of the sample to generate an information data packet, where the information data packet includes: sample number, type, source, and collection time; sample feature information collected by the sensor: appearance characteristics, container sealing and component analysis; other information: patient information and test items; through the above process, the sensor can effectively collect sample information from the laboratory and generate an information data packet containing sample type and source information, providing basic data support for subsequent sample storage and management.
[0023] The process by which the information input unit enters the information data packet into the database of the sample storage system for storage is as follows: Based on the WiFi network coverage inside the laboratory, WiFi is set up for communication; based on the server where the information entry unit and the database of the sample storage system are located, a WiFi network card is installed and the network is configured, the network name and password are set, and each device is connected to the WiFi network. Then, identity authentication is performed by means of a user name and password, and the information entry unit is connected to the WiFi network of the sample storage system. Communication protocol header information and tail information are added to the information data packet through encapsulation. The encapsulated information data packet is entered into the database of the sample storage system based on real-time transmission, and an error detection and retransmission mechanism is set. A cyclic redundancy check is used to detect whether an error occurs in the data transmission process. If an error is found, a retransmission is automatically requested; Then, the receiving program on the sample storage system database server listens to the port, waits to receive the information data packet from the information entry unit, parses the received information data packet, removes the communication protocol header information and tail information, obtains the original information data packet, and verifies and cleans the information data packet, the verification includes checking the integrity of the data packet, whether the format is correct, and obtains a complete data packet. If the information data packet is incomplete or the format is wrong, the information data packet is discarded, and a retransmission notification is generated, and the retransmission notification is sent to the information entry unit to request retransmission; After the complete data packet is stored in the database, a confirmation message is generated to inform the sender that the data has been successfully stored, and the confirmation message is sent to the information entry unit, so that the information entry unit can know the results of data transmission and storage for subsequent processing or error handling. Through the above detailed storage process, it can be ensured that the information data packet is safely and accurately transmitted to the database of the sample storage system for storage through wireless communication, providing a reliable data basis for subsequent sample management and analysis, and by transmitting the sample information to the sample storage system in the form of a complete data packet, rather than the frequent transmission of scattered individual data, the data is centrally transmitted while eliminating abnormal, missing and invalid data, retaining the integrity of the data; at the same time, by verifying and cleaning the information data packet, it is possible to verify in advance whether the information data packet has defects such as incomplete data or incorrect format before entry, so as to retransmit it in time, ensure the integrity and accuracy of data transmission, and avoid data loss or transmission interruption due to network fluctuations, equipment failures, etc.; and through the feedback effect brought by the confirmation information, information feedback can be provided for the data that has been successfully entered, ensuring that the data can enter the storage database in a timely and complete manner, ensuring the stability of medical data transmission.
[0024] Wireless communication and identity authentication technologies are used to securely and accurately transmit information data packets to the database for storage, ensuring the reliability and confidentiality of the data. Error detection and retransmission mechanisms are set up to ensure the integrity and accuracy of data transmission, providing a reliable data foundation for subsequent sample management and analysis.
[0025] The data processing module includes: The data cleaning unit is used to clean the data based on the sample association information and feature information, remove the noise and incomplete data in the data, and obtain the sample feature data; removing the noise and incomplete data improves the data quality and reduces the influence of interference factors on subsequent analysis; The data processing unit is used to obtain the sample feature data obtained in the data cleaning unit, and convert the sample feature data into a standard format to obtain standard feature data for the input of the model; The anomaly detection unit is used to receive standard feature data, identify and mark abnormal samples in the standard feature data, and obtain abnormal marked data to prevent the influence of erroneous data on model training and reasoning; it helps to find problem samples in time and improve the security and reliability of sample management; The functions of this module ensure the accuracy and consistency of the data, providing reliable data support for subsequent deep learning models.
[0026] The deep learning modules include: The training module uses a convolutional neural network architecture to perform feature learning and pattern recognition through historical sample data to train a deep learning model; The inference module analyzes the newly collected sample information based on the trained deep learning model and combined with standard feature data, and generates analysis results, which include the classification results and prediction information of the samples; The design of this module enables the sample storage system to have efficient real-time analysis capabilities and provide timely diagnostic support after sample collection.
[0027] It should be further explained that, in the specific implementation process, the training module trains the deep learning model through the convolutional neural network architecture, which can make full use of historical sample data for feature learning and pattern recognition, thereby improving the accuracy and generalization ability of the model; and the reasoning module analyzes the newly collected sample information based on the trained model to generate analysis results, providing a scientific basis for sample classification, prediction and management.
[0028] The deep learning module adopts a convolutional neural network architecture, which performs feature learning and pattern recognition through historical sample data. It can explore the potential rules and patterns in the samples and improve the accuracy and depth of sample analysis. It can also use appropriate network architectures for feature extraction for different types of data, giving full play to the advantages of deep learning and improving the effect of feature extraction.
[0029] The process of building a deep learning model is as follows: Collect historical sample data from the laboratory database, including sample images, test results, and storage conditions; then clean the collected historical sample data to remove noise, outliers, and incomplete data; then annotate the cleaned sample data to determine the category and features of the sample; use data enhancement technology to rotate, flip, crop, and scale image data, and perform random perturbations and interpolation operations on numerical data; clarify the tasks of the deep learning model in the laboratory sample storage system, that is, the task requirements, which include sample classification, anomaly detection, and storage location prediction; For image data, the visual geometry group network architecture is used for feature extraction; for detection result data, a network architecture combining convolutional neural network and fully connected neural network is used for feature extraction; first, the convolutional neural network is used to process the image data and extract image features, and then the image features and the detection result data are input into the fully connected neural network for further analysis and prediction; it is used to combine the information of the image and the detection result to improve the accuracy of the model; for storage condition data, a decision tree architecture is used, and classification and prediction are performed according to different characteristics of the storage conditions. The decision tree architecture is used to classify and predict different characteristics of the storage condition data, which is convenient for understanding and explanation; the above three architectures are integrated to form a convolutional neural network; the parameters of the convolutional neural network are determined according to the size, complexity and computing resources of the sample data, and the parameters of the convolutional neural network are appropriately adjusted and optimized according to the task requirements; Then, the preprocessed historical sample data is divided into a training set and a test set; the division is carried out according to the following ratio: 70% of the data is used as a training set and 30% of the data is used as a test set; and the stochastic gradient descent algorithm is selected to train the convolutional neural network; the training set data is input into the convolutional neural network, the output result is calculated through forward propagation, and then the error between the predicted result and the true label is calculated according to the loss function; Use the optimization algorithm to back-propagate the error and update the network's weights and bias parameters to reduce the value of the loss function; repeat this process until the loss function converges or reaches the preset number of training times to obtain a deep learning model.
[0030] It should be further explained that, in the specific implementation process, the historical sample data was pre-processed by cleaning, labeling and enhancing, which improved the data quality and diversity, and helped to improve the performance of the deep learning model; the appropriate network architecture was used for feature extraction for different types of data, which gave full play to the advantages of various architectures and improved the effect and accuracy of feature extraction; the training set and test set were reasonably divided, and the appropriate optimization algorithm was selected for model training, which ensured the accuracy and generalization ability of the model.
[0031] During the training process of the deep learning model, the deep learning model is evaluated through the test set data, the accuracy, recall rate, and F1 value of the deep learning model are calculated, and the corresponding evaluation results are obtained to understand the performance and generalization ability of the model; then according to the evaluation results, the structure and parameters of the model are adjusted, such as increasing or decreasing the number of network layers, adjusting the convolution kernel size, changing the learning rate, etc., to improve the performance of the model.
[0032] It should be further explained that, in the specific implementation process, the deep learning model is evaluated through the test set data, which can timely understand the performance and problems of the model, and provide a basis for the adjustment and optimization of the model; the model structure and parameters are adjusted according to the evaluation results, which improves the accuracy and adaptability of the model and ensures that the model can better meet the actual application needs.
[0033] When the model training achieves satisfactory performance, the trained deep learning model is deployed to the laboratory sample storage system; Save the model's weights and structure information as files, and then load the model into the system for real-time sample analysis and storage management; After the model is deployed, continue to collect new sample data and optimize the model online. Use incremental learning technology to update the model parameters to adapt to new data without retraining the entire model. At the same time, you can also optimize the model performance, such as using model compression technology to reduce the size and computational complexity of the model, and improve the running speed and efficiency of the model. Through the above steps, you can build a deep learning model in the laboratory sample storage system based on deep learning to achieve efficient storage and management of samples.
[0034] Deploying the trained deep learning model into the system enables real-time sample analysis and storage management, improving work efficiency and management level; optimizing the model online enables it to continuously adapt to new data and changing situations, maintaining the accuracy and effectiveness of the model.
[0035] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.
Claims
1. A laboratory sample storage system based on deep learning, characterized in that: include: The sample collection module is used to collect sample information from the laboratory through sensors, and extract sample association information and feature information based on the sample information; The data processing module is connected to the sample collection module and is used to perform data cleaning, format conversion and abnormal marking on the collected sample association information and feature information to obtain standard feature data and abnormal marking data; A deep learning module, connected to the data processing module, is used to adopt a convolutional neural network architecture to build a deep learning model through historical sample data, and use the deep learning model to analyze the sample information; The sample storage module is used to store sample information and analysis results obtained by the deep learning model for subsequent information query and management; The user interface module is used to set the operation interface for users to input sample information, view analysis results and perform system settings.
2. According to claim 1, a deep learning-based laboratory sample storage system is characterized by: The sample collection module comprises: The sample identification unit is used to collect sample information from the laboratory through sensors, identify the type and source of the sample information, and obtain an information data packet; An information input unit, used to receive information data packets and input the information data packets of the samples into a database of the sample storage system for storage; The feature extraction unit is used to receive the information data packet obtained by the sample identification unit and extract the feature information of the sample from the information data packet, wherein the feature information includes the color, shape and smell of the sample.
3. The deep learning-based laboratory sample storage system according to claim 2, characterized in that: The process of collecting sample information by the sensor is as follows: Optical sensors, pressure sensors and RFID sensors are deployed in the sample collection area of the laboratory. Optical sensors are then used to detect the color and transparency characteristics of the samples. When the samples enter the collection area of the optical sensors and are irradiated by light, the optical sensors detect the intensity and wavelength of the reflected light to obtain the appearance characteristics of the samples. Then, the sealing of the sample container is detected using a pressure sensor. A proximity sensor is installed in each storage compartment of the sample storage cabinet. When the sample container is placed in the storage compartment, the proximity sensor senses the entry of the object and immediately activates the pressure sensor to detect the sealing of the sample container and obtain the sealing status information. The unique identification of the sample is then identified using a radio frequency identification sensor. When a sample with an RFID tag approaches the collection area of the sensor, the radio frequency identification sensor reads the information in the tag to obtain the tag identification information.
4. The deep learning-based laboratory sample storage system according to claim 3, characterized in that: The detailed process of obtaining the information data packet is as follows: Combining the appearance feature information of the sample collected by the optical sensor, the sealing state information collected by the pressure sensor, and the label identification information collected by the radio frequency identification sensor to obtain integrated information; Analyze the characteristics and identification of the sample based on the integrated information, and identify the sample type and sample source: obtain the sample type by analyzing the color characteristics, shape characteristics and composition characteristics in the appearance characteristic information of the sample; The source of the sample is obtained by reading the collection location information and the inspection department information in the tag identification information obtained from the RFID tag; The identified sample type and sample source are then packaged together with the integrated information of the sample to generate an information data package.
5. The deep learning-based laboratory sample storage system according to claim 4, characterized in that: The process of the information input unit inputting the information data packet into the database of the sample storage system for storage is as follows: Based on the WiFi network coverage inside the laboratory, WiFi is set up for communication; based on the server where the information entry unit and the database of the sample storage system are located, a WiFi network card is installed and the network is configured, the network name and password are set, and each device is connected to the WiFi network. Then, identity authentication is performed by means of a user name and password, and the information entry unit is connected to the WiFi network of the sample storage system. Communication protocol header information and tail information are added to the information data packet through encapsulation. The encapsulated information data packet is entered into the database of the sample storage system based on real-time transmission, and an error detection and retransmission mechanism is set. A cyclic redundancy check is used to detect whether an error occurs in the data transmission process. If an error is found, a retransmission is automatically requested; Then, the receiving program on the sample storage system database server listens to the port, waits to receive the information data packet from the information entry unit, parses the received information data packet, removes the communication protocol header information and tail information, obtains the original information data packet, verifies and cleans the information data packet, obtains the complete data packet, and discards the information data packet if it is incomplete or has a wrong format, generates a retransmission notification, and sends the retransmission notification to the information entry unit to request retransmission; After the complete data packet is stored in the database, confirmation information is generated and sent to the information entry unit. Through the above detailed storage process, it can be ensured that the information data packet is safely and accurately transmitted to the database of the sample storage system through wireless communication for storage, providing a reliable data basis for subsequent sample management and analysis.
6. The deep learning-based laboratory sample storage system according to claim 5, characterized in that: The data processing module comprises: A data cleaning unit is used to clean the data based on sample association information and feature information, remove noise and incomplete data in the data, and obtain sample feature data; A data processing unit is used to obtain the sample characteristic data obtained in the data cleaning unit, and convert the sample characteristic data into a standard format to obtain standard characteristic data; The anomaly detection unit is used to receive standard feature data, and identify and mark abnormal samples in the standard feature data to obtain abnormal mark data.
7. The deep learning-based laboratory sample storage system according to claim 6, characterized in that: The deep learning module includes: The training module uses a convolutional neural network architecture to perform feature learning and pattern recognition through historical sample data to train a deep learning model; The inference module analyzes the newly collected sample information based on the trained deep learning model and combines the standard feature data, and generates analysis results, which include the classification results and prediction information of the samples.
8. The deep learning-based laboratory sample storage system according to claim 7, characterized in that: The construction process of the deep learning model is as follows: Collect historical sample data from the laboratory department's database, including sample images, test results, and storage conditions; then clean the collected historical sample data to remove noise, outliers, and incomplete data; then annotate the cleaned sample data to determine the category and characteristics of the sample; Use data enhancement technology to rotate, flip, crop and scale image data, and perform random perturbations and interpolation operations on numerical data; clarify the tasks of deep learning models in the sample storage system of the laboratory department, that is, task requirements, which include sample classification, anomaly detection and storage location prediction; For image data, the visual geometry group network architecture is used for feature extraction; For the test result data, a network architecture combining convolutional neural network and fully connected neural network is used for feature extraction. First, the convolutional neural network is used to process the image data and extract image features. Then, the image features and the test result data are input into the fully connected neural network for further analysis and prediction. This is used to combine the information of the image and the test result to improve the accuracy of the model. For storage condition data, a decision tree architecture is used to classify and predict based on different characteristics of storage conditions. The above three architectures are integrated to form a convolutional neural network. The parameters of the convolutional neural network are determined based on the size, complexity and computing resources of the sample data, and the parameters of the convolutional neural network are appropriately adjusted and optimized based on task requirements. Then, the preprocessed historical sample data is divided into a training set and a test set; wherein, the division is carried out according to the following ratio: 70% of the data is used as a training set, and 30% of the data is used as a test set; and the stochastic gradient descent algorithm is selected to train the convolutional neural network; Input the training set data into the convolutional neural network, calculate the output result through forward propagation, and then calculate the error between the predicted result and the true label according to the loss function; Use the optimization algorithm to back-propagate the error and update the network's weights and bias parameters to reduce the value of the loss function; This process is repeated until the loss function converges or the preset number of training times is reached to obtain a deep learning model.
9. The deep learning-based laboratory sample storage system according to claim 8, characterized in that: During the training process of the deep learning model, the deep learning model is evaluated through the test set data, the accuracy, recall rate, and F1 value of the deep learning model are calculated, and the corresponding evaluation results are obtained; then the structure and parameters of the model are adjusted according to the evaluation results.
10. A laboratory sample storage system based on deep learning according to claim 9, characterized in that: When the model training achieves satisfactory performance, the trained deep learning model is deployed to the laboratory sample storage system; Save the model's weights and structure information as files, and then load the model into the system for real-time sample analysis and storage management; After the model is deployed, continue to collect new sample data and optimize the model online.
Citation Information
Patent Citations
Inventory management system for medical laboratory
CN116130069A
Medical data mining method and system fused with deep learning
CN117292840A
Medical examination process optimization system
CN118072927A
Blood test sample storage management method and system
CN118571443A
Cited By
Specimen management system and management method of intelligent access cabinet
CN119339905A
