Optimization Method and System for Medical Residue Detection Management Based on Artificial Intelligence

Abstract

The present invention discloses an optimization method and system for medical residue detection management based on artificial intelligence. The method includes data collection, data preprocessing, residue location classification, residue tracking management, and residue detection management. The present invention relates to the technical field of medical residue management, specifically referring to an optimization method and system for medical residue detection management based on artificial intelligence. This solution adopts a comprehensive detection management method that combines residue location classification and residue tracking management. By organically combining location classification and tracking management, it can real-time track the status changes, processing procedures, and transfer paths of medical residues, thereby greatly improving management efficiency and accuracy. An integrated convolutional transformer model combined with a pre-trained detection model is used for residue location classification. A reinforcement learning method that combines an improved gated recurrent unit and a graph neural network is used for residue tracking management.

Description

Technical Field

[0001] The present invention relates to the technical field of medical residue management, and specifically refers to an optimized method and system for detecting and managing medical residues based on artificial intelligence. Background Art

[0002] The optimized method and system for detecting and managing medical residues based on artificial intelligence aims to efficiently and accurately detect and manage medical waste through advanced artificial intelligence technologies. The system can, through deep learning, computer vision, and data analysis technologies, monitor residues in the medical process in real time, such as surgical instruments, drug containers, etc., to ensure that they can be identified and properly processed in a timely manner. By optimizing the detection algorithm, the system can automatically identify potential residues, improve the accuracy and efficiency of medical waste management, reduce the risk of human negligence and omission, ensure the cleanliness and environmental safety of the hospital, and comply with relevant regulatory requirements. In addition, the system can also optimize the processing flow according to historical data and real-time situations, provide an intelligent waste management solution for medical institutions, reduce operating costs, and improve the overall service quality.

[0003] However, in the existing optimization of medical residue detection and management, there are technical problems such as insufficient real-time monitoring and tracking capabilities, low dynamic decision-making efficiency, and unclear information on the cross-regional transfer of various residues; in the existing residue positioning and classification methods, there are technical problems such as low classification accuracy, difficulty in adapting to different environmental changes, inability to efficiently handle multiple types of medical waste, and the traditional classification methods often rely on a single model and are difficult to maintain stable performance in a dynamically changing environment; in the existing residue tracking and management methods, there are technical problems such as inability to dynamically adjust the processing strategy, lack of effective cross-regional tracking capabilities, and poor timeliness of decision-making. The existing tracking and management systems are often difficult to handle complex and changing actual scenarios, especially in the cross-regional dependence and sequential decision-making aspects during the tracking process, there are obvious deficiencies. Summary of the Invention

[0004] In view of the above situation, to overcome the defects of the prior art, the present invention provides an optimization method and system for medical residue detection management based on artificial intelligence. In the existing optimization of medical residue detection management, there are technical problems such as insufficient real-time monitoring and tracking capabilities, low dynamic decision-making efficiency, and unclear cross-regional transfer information of various residues. This solution creatively adopts a comprehensive detection management method that combines residue positioning and classification with residue tracking management, achieving full coverage from the positioning, classification to tracking management of medical residues. By organically combining positioning classification and tracking management, it is possible to real-time track the status changes, processing processes, and transfer paths of medical residues, thereby greatly improving management efficiency and accuracy, realizing intelligent and automated management of medical residues, and reducing the risks of manual intervention and omissions. In the existing residue positioning and classification methods, there are technical problems such as low classification accuracy, difficulty in adapting to different environmental changes, and inability to efficiently handle multiple types of medical waste, as well as the fact that traditional classification methods often rely on a single model and are difficult to maintain stable performance in a dynamically changing environment. This solution creatively adopts an integrated convolutional transformer model combined with a pre-trained detection model for residue positioning and classification, achieving high-precision positioning and classification in complex environments. By integrating the pre-trained detection model and the convolutional transformer structure, this solution can effectively combine the spatial information and temporal information of image features, improving the robustness and adaptability of the model. Especially in the highly variable and uncertain environment of medical residues, the model can achieve more accurate object positioning and type classification, significantly improving the accuracy of waste identification. In the existing residue tracking management methods, there are technical problems such as the inability to dynamically adjust processing strategies, lack of effective cross-regional tracking capabilities, and poor timeliness of decision-making. Existing tracking management systems often have difficulty dealing with complex and changing actual scenarios, especially in terms of cross-regional dependencies and temporal decision-making during the tracking process. This solution creatively adopts a reinforcement learning method that combines an improved gated recurrent unit and a graph neural network for residue tracking management, realizing dynamic optimization and precise tracking of the entire process of medical residue treatment. By combining reinforcement learning for strategy optimization, this solution can effectively propagate information between different states and dynamically adjust tracking strategies according to real-time data, improving the system's autonomous decision-making ability and response speed, and ensuring the efficiency and safety of medical residue management.

[0005] The technical solution adopted by the present invention is as follows: The optimization method for medical residue detection management based on artificial intelligence provided by the present invention includes the following steps:

[0006] Step S1: Data collection;

[0007] Step S2: Data preprocessing;

[0008] Step S3: Residue positioning and classification;

[0009] Step S4: Residue tracking management;

[0010] Step S5: Residue detection management.

[0011] Further, in Step S1, the data collection is used to collect multi-source data related to medical residues. Specifically, through camera data collection, data is collected by monitoring the operating room, ward, and waste treatment area, and through the medical system, image data of the medical equipment and waste container area is obtained in real time. Through sensors, temperature, humidity, and air quality environment data are collected, and through label tracking, dynamic data of medical residues is collected to obtain the original dataset for medical residue detection management;

[0012] The original dataset for medical residue detection management specifically includes camera data, environmental data, and label tracking dynamic data.

[0013] Further, in Step S2, the data preprocessing is used to clean, denoise, and format the collected data. Specifically, data preprocessing operations are performed on the original dataset for medical residue detection management to obtain the optimized dataset for medical residue detection management. The specific steps are as follows:

[0014] Step S21: Data cleaning, specifically, data cleaning is performed by removing invalid data and filling data to obtain cleaned and optimized data;

[0015] Step S22: Denoising optimization, specifically, image denoising is performed on the image data, sensor data denoising is performed on the sensing environment data, and signal denoising is performed on the label tracking data to obtain denoised and optimized data;

[0016] Step S23: Data standardization, specifically, resolution and format normalization are performed on the image data, Z-score standardization is performed on the sensor data, and time alignment standardization is performed on the label tracking data to obtain standardized data;

[0017] Step S24: Data augmentation, specifically, data comprehensive augmentation is performed through image data augmentation and sensing data augmentation to obtain augmented and optimized data;

[0018] Step S25: Data integration, specifically, multi-source data fusion is performed on the augmented and optimized data to obtain integrated and optimized data;

[0019] Step S26: Feature engineering optimization, specifically, manual feature screening and extraction are performed on the integrated and optimized data to obtain the optimized dataset for medical residue detection management.

[0020] Further, in step S3, the residue location and classification is used to locate and classify medical residues by using an artificial intelligence algorithm, identify different types of waste, and mark their locations. Specifically, based on the optimized medical residue detection management dataset, an integrated convolutional transformer model combined with a pre-trained detection model is used to perform residue location and classification to obtain residue location and classification reference data. The specific steps are as follows:

[0021] Step S31: Construct a pre-trained detection model. Specifically, construct a pre-trained integrated detection model, and through transfer learning and pre-training integration methods, adjust the weights of the target detection model for the integrated detection model in combination with the optimized medical residue detection management dataset to obtain a pre-trained detection model;

[0022] The integrated detection model specifically refers to an integrated model of YOLOv8 and Faster-R-CNN;

[0023] Step S32: Construct an improved convolutional transformer model. Specifically, construct a standard convolutional neural network and a transformer model, and through the introduction of a bidirectional attention mechanism and a multi-scale feature hierarchical transformation module, perform hierarchical improvement of the convolutional transformer model to obtain an improved convolutional transformer model;

[0024] Step S33: Construct an integrated classifier. Specifically, introduce a comprehensive integration method combining an adaptive integration method and a dynamic weighted integration, and use a standard extreme gradient boosting model as the model classifier to construct a process classifier;

[0025] Step S34: Residue location and classification. Specifically, through the pre-trained detection model, the improved convolutional transformer model, and the integrated classifier, perform model training and model use for residue location and classification to obtain residue location and classification reference data.

[0026] Further, in step S4, the residue tracking management is used to track the entire process of medical residues from generation, treatment to destruction. Specifically, based on the optimized medical residue detection management dataset and the residue location and classification reference data, a reinforcement learning method combining an improved gated recurrent unit and a graph neural network is used to perform residue tracking management to obtain residue tracking management reference data;

[0027] Step S41: Construct an improved gated recurrent sub-network. Specifically, construct a standard bidirectional gated recurrent unit, and construct a long short-term dependence improvement module after the output layer of the bidirectional gated recurrent unit. The long short-term dependence improvement module specifically includes a single-layer standard long short-term memory layer and a single-layer standard gated recurrent unit layer, and through the construction of the long short-term dependence improvement module, update the output features of the gated recurrent sub-network to obtain updated temporal feature data;

[0028] Step S42: Construct a graph neural subnet. Specifically, construct a dynamic tracking status graph of medical residues, and through constructing a standard graph neural subnet, combine the updated temporal feature data to perform spatio-temporal feature extraction to obtain spatio-temporal feature data;

[0029] Step S43: Construct a reinforcement learning model. Specifically, construct state space parameters and action space parameters, combine with the improvement of the reward function, and based on the spatio-temporal feature data, construct a reinforcement learning model to obtain reinforcement learning decision output data;

[0030] The improvement of the reward function specifically includes a medical residue treatment efficiency parameter, a medical residue treatment safety parameter, and a medical residue utilization rate parameter;

[0031] Step S44: Train the residue tracking management model. Specifically, through the improved gated recurrent subnet, the graph neural subnet, and the reinforcement learning model, perform model training to obtain a residue tracking management model;

[0032] Step S45: Residue tracking management. Specifically, use the residue tracking management model, based on the medical residue detection management optimization dataset and the residue location classification reference data, perform residue tracking management to obtain residue tracking management reference data.

[0033] Furthermore, in step S5, the residue detection management is used to perform continuous monitoring and management of medical residues. Specifically, by relying on the residue location classification reference data, read the image detection information of medical residues, and combine with the residue tracking management reference data, read the tracking detection information of residues. By combining the image detection information and the tracking detection information, perform comprehensive detection management of medical residues, and construct a rule-based real-time monitoring system for residues, automatically generate alarms and assist management personnel in recording, generating, and executing the cleaning tasks of medical residues to obtain residue comprehensive detection management reference data.

[0034] The medical residue detection management optimization system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, a residue location classification module, a residue tracking management module, and a residue detection management module;

[0035] The data acquisition module is used for data acquisition. Through data acquisition, an original dataset for medical residue detection management is obtained, and the original dataset for medical residue detection management is sent to the data preprocessing module;

[0036] The data preprocessing module is used for data preprocessing. Through data preprocessing, an optimized dataset for medical residue detection management is obtained, and the optimized dataset for medical residue detection management is sent to the residue location and classification module and the residue tracking management module;

[0037] The residue location and classification module is used for residue location and classification. Through residue location and classification, reference data for residue location and classification is obtained, and the reference data for residue location and classification is sent to the residue tracking management module and the residue detection management module;

[0038] The residue tracking management module is used for residue tracking management. Through residue tracking management, reference data for residue tracking management is obtained, and the reference data for residue tracking management is sent to the residue detection management module;

[0039] The residue detection management module is used for residue detection management. Through residue detection management, reference data for comprehensive residue detection management is obtained.

[0040] The beneficial effects achieved by the present invention using the above solution are as follows:

[0041] (1) Aiming at the technical problems existing in the existing optimization of medical residue detection management, such as insufficient real-time monitoring and tracking capabilities, low dynamic decision-making efficiency, and unclear cross-regional transfer information of various residues, this solution creatively adopts a comprehensive detection management method combining residue location and classification and residue tracking management, realizing comprehensive coverage from the location, classification to tracking management of medical residues. By organically combining location classification and tracking management, it can real-time track the status changes, processing processes and transfer paths of medical residues, thus greatly improving management efficiency and accuracy, realizing intelligent and automated management of medical residues, and reducing the risks of manual intervention and omissions;

[0042] (2) Aiming at the technical problems existing in the existing residue location and classification methods, such as low classification accuracy, difficulty in adapting to different environmental changes, inability to efficiently handle multiple types of medical waste, and the fact that traditional classification methods often rely on a single model and are difficult to maintain stable performance in a dynamically changing environment, this solution creatively adopts an integrated convolutional transformer model combined with a pre-trained detection model for residue location and classification, realizing high-precision location and classification in complex environments. By integrating the pre-trained detection model and the convolutional transformer structure, this solution can effectively combine the spatial information and temporal information of image features, improve the robustness and adaptability of the model. Especially in the highly variable and uncertain environment of medical residues, the model can achieve more accurate object location and type classification, significantly improving the accuracy of waste identification;

[0043] (3) In the existing residue tracking management methods, there are problems such as the inability to dynamically adjust processing strategies, the lack of effective cross-region tracking capabilities, and poor timeliness of decision-making. Existing tracking management systems often struggle to handle complex and changing actual scenarios, especially in cross-region dependencies and sequential decision-making during the tracking process, where there are obvious deficiencies. This solution creatively adopts a reinforcement learning method that combines an improved gated recurrent unit and a graph neural network for residue tracking management, achieving dynamic optimization and precise tracking of the entire process of medical residue treatment. By combining reinforcement learning for policy optimization, this solution can effectively disseminate information between different states, dynamically adjust tracking strategies based on real-time data, improve the system's autonomous decision-making ability and response speed, and ensure the efficiency and safety of medical residue management. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flowchart showing the optimization method for medical residue detection and management based on artificial intelligence provided by the present invention;

[0045] Figure 2 It is a schematic diagram of the optimization system for medical residue detection and management based on artificial intelligence provided by the present invention;

[0046] Figure 3 It is a flowchart showing the data preprocessing in step S2;

[0047] Figure 4 It is a flowchart showing the residue location classification in step S3;

[0048] Figure 5 It is a flowchart showing the residue tracking management in step S4.

[0049] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the present invention.

[0052] Example 1. Refer to Figure 1 , the optimization method for medical residue detection management based on artificial intelligence provided by the present invention includes the following steps:

[0053] Step S1: Data collection;

[0054] Step S2: Data preprocessing;

[0055] Step S3: Residue positioning and classification;

[0056] Step S4: Residue tracking management;

[0057] Step S5: Residue detection management.

[0058] By performing the above operations, aiming at the technical problems existing in the existing optimization of medical residue detection management, such as insufficient real-time monitoring and tracking capabilities, low dynamic decision-making efficiency, and unclear information on the cross-regional transfer of various residues, this solution creatively adopts a comprehensive detection management method combining residue positioning and classification with residue tracking management, achieving comprehensive coverage from the positioning, classification to the tracking management of medical residues. By organically combining positioning and classification with tracking management, it can real-time track the status changes, processing processes and transfer paths of medical residues, thus greatly improving the management efficiency and accuracy, realizing the intelligent and automated management of medical residues, and reducing the risks of manual intervention and omission.

[0059] Example 2. Refer to Figure 1 and Figure 2 , in step S1, the data collection is used to collect multi-source data related to medical residues. Specifically, through camera data collection, data is collected by monitoring the operating rooms, wards, and waste treatment areas, and through the medical system, image data of medical equipment and waste container areas is obtained in real time. Through sensors, temperature, humidity, and air quality environment data are collected, and through label tracking, dynamic data of medical residues is collected to obtain the original dataset for medical residue detection management;

[0060] The original dataset for medical residue detection management specifically includes camera data, environmental data, and label tracking dynamic data.

[0061] Example 3. Refer toFigure 1 , Figure 2 and Figure 3 , this embodiment is based on the above embodiment. In step S2, the data preprocessing is used to clean, denoise, and format the collected data. Specifically, data preprocessing operations are performed on the original medical residue detection management dataset to obtain an optimized medical residue detection management dataset, which specifically includes the following steps:

[0062] Step S21: Data cleaning. Specifically, data cleaning is performed by removing invalid data and filling data to obtain cleaned and optimized data.

[0063] Step S22: Denoising optimization. Specifically, image denoising is performed on image data, sensor data denoising is performed on sensing environment data, and signal denoising is performed on label tracking data to obtain denoised and optimized data.

[0064] Step S23: Data standardization. Specifically, resolution and format normalization are performed on image data, Z-score standardization is performed on sensor data, and time alignment standardization is performed on label tracking data to obtain standardized data.

[0065] Step S24: Data augmentation. Specifically, data comprehensive augmentation is performed through image data augmentation and sensing data augmentation to obtain augmented and optimized data.

[0066] Step S25: Data integration. Specifically, multi-source data fusion is performed on the augmented and optimized data to obtain integrated and optimized data.

[0067] Step S26: Feature engineering optimization. Specifically, artificial feature screening and extraction are performed on the integrated and optimized data to obtain an optimized medical residue detection management dataset.

[0068] Embodiment Four. Refer to Figure 1 , Figure 2 and Figure 4 , this embodiment is based on the above embodiment. In step S3, the residue location and classification is used to locate and classify medical residues using an artificial intelligence algorithm, identify different types of waste, and mark their locations. Specifically, based on the optimized medical residue detection management dataset, an integrated convolutional transformer model combined with a pre-trained detection model is used to perform residue location and classification to obtain residue location and classification reference data, which specifically includes the following steps:

[0069] Step S31: Construct a pre-trained detection model. Specifically, a pre-trained integrated detection model is constructed, and through transfer learning and pre-training integration methods, the weights of the target detection model of the integrated detection model are adjusted in combination with the optimized medical residue detection management dataset to obtain a pre-trained detection model.

[0070] The integrated detection model specifically refers to the integrated model of YOLOv8 and Faster-R-CNN;

[0071] Step S32: Construct an improved convolutional transformer model. Specifically, construct a standard convolutional neural network and a transformer model, and through the introduction of a bidirectional attention mechanism and a multi-scale feature hierarchical transformation module, perform hierarchical improvement of the convolutional transformer model to obtain an improved convolutional transformer model;

[0072] The bidirectional attention mechanism is used to simultaneously focus on the local features and global features of the image through information propagation in two directions;

[0073] The multi-scale feature hierarchical transformation module is used to optimize the classification and localization of multi-scale residues through the combination of local perception and global perception;

[0074] The calculation formula for performing hierarchical improvement of the convolutional transformer model is:

[0075] ;

[0076] In the formula, F TBF is the feature output of the improved convolutional transformer model, MSTF(·) is the operation function of the multi-scale feature hierarchical transformation module, Bi-Att(·) is the operation function of the bidirectional attention mechanism, Transformer(·) is the operation function of the transformer model, and F Conv is the feature output of the standard convolutional neural network;

[0077] Step S33: Construct an integrated classifier. Specifically, introduce a comprehensive integration method that combines an adaptive integration method and dynamic weighted integration, and use a standard extreme gradient boosting model as the model classifier to construct a process classifier;

[0078] The comprehensive integration method constructs a dynamic integration with adaptive weights for model output, and the calculation formula is:

[0079] ;

[0080] In the formula, P F is the comprehensive integration output of the process classifier, k is the total number of classifiers, i is the classifier index, Per(·) is the classifier performance evaluation function, M i is the symbolic representation of the i-th classifier model, and P i is the classification result of the i-th classifier model;

[0081] Step S34: Residue localization and classification. Specifically, through the pre-trained detection model, the improved convolutional transformer model, and the integrated classifier, perform model training and model use for residue localization and classification to obtain residue localization and classification reference data.

[0082] By performing the above operations, in the existing residue localization and classification methods, there are technical problems such as low classification accuracy, difficulty in adapting to different environmental changes, and inability to efficiently handle various medical wastes, and traditional classification methods often rely on a single model and are difficult to maintain stable performance in a dynamically changing environment. This solution creatively uses an integrated convolutional transformer model combined with a pre-trained detection model for residue localization and classification, achieving high-precision localization and classification in complex environments. By integrating the pre-trained detection model and the convolutional transformer structure, this solution can effectively combine the spatial information and temporal information of image features, improving the robustness and adaptability of the model. Especially in the highly variable and uncertain environment of medical residues, the model can achieve more accurate object localization and type classification, significantly improving the accuracy of waste identification.

[0083] Example 5, refer to Figure 1 、 Figure 2 and Figure 5 , based on the above example, in step S4, the residue tracking management is used to track the whole process of medical residues from generation, treatment to destruction. Specifically, according to the optimized dataset of medical residue detection management and the reference data of residue localization and classification, a reinforcement learning method combining an improved gated recurrent unit and a graph neural network is used for residue tracking management to obtain the reference data of residue tracking management;

[0084] Step S41: Construct an improved gated recurrent sub-network. Specifically, construct a standard bidirectional gated recurrent unit, and construct a long short-term memory improvement module after the output layer of the bidirectional gated recurrent unit. The long short-term memory improvement module specifically includes a single-layer standard long short-term memory layer and a single-layer standard gated recurrent unit layer. By constructing the long short-term memory improvement module, the output feature of the gated recurrent sub-network is updated to obtain updated temporal feature data;

[0085] The calculation formula of the updated temporal feature data is:

[0086] ;

[0087] In the formula, is the updated temporal feature data improved by long short-term memory, a t is the updated temporal weight, BiGRU(·) is the operation function of the bidirectional gated recurrent unit, x t is the original data input, used to represent the optimized dataset of medical residue detection management and the reference data of residue localization and classification, h t-1 is the output temporal feature data of the standard bidirectional gated recurrent unit at the previous moment, h tis the output timing feature data of the standard bidirectional gated recurrent unit at time t, where t is the time index, and f(x t ) is the output feature data of the long short-term dependence improvement module;

[0088] Step S42: Construct a graph neural subnet, specifically by constructing a dynamic tracking state graph of medical residues and constructing a standard graph neural subnet, and combining the updated timing feature data to perform spatio-temporal feature extraction to obtain spatio-temporal feature data;

[0089] Step S43: Construct a reinforcement learning model, specifically by constructing state space parameters and action space parameters, and combining reward function improvement, and based on the spatio-temporal feature data, construct a reinforcement learning model to obtain reinforcement learning decision output data;

[0090] The improvement of the reward function specifically includes a medical residue treatment efficiency parameter, a medical residue treatment safety parameter, and a medical residue utilization parameter;

[0091] The calculation formula for the improvement of the reward function is:

[0092] ;

[0093] In the formula, R t is the improved reward function, c1 is the weight of medical residue treatment efficiency, Efficiency(t) is the medical residue treatment efficiency parameter, c2 is the weight of medical residue treatment safety, Risk(t) is the medical residue treatment safety parameter, c3 is the weight of medical residue utilization, and ResourceUtilization(t) is the medical residue utilization parameter;

[0094] Step S44: Train the residue tracking management model, specifically by using the improved gated recurrent subnet, the graph neural subnet, and the reinforcement learning model to perform model training to obtain the residue tracking management model;

[0095] Step S45: Residue tracking management, specifically using the residue tracking management model, and based on the medical residue detection management optimization dataset and the residue location classification reference data, perform residue tracking management to obtain residue tracking management reference data.

[0096] By performing the above operations, in the existing residue tracking management method, there are problems such as the inability to dynamically adjust the processing strategy, the lack of effective cross-regional tracking ability, and poor timeliness of decision-making. Existing tracking management systems often struggle to handle complex and changing actual scenarios, especially in terms of cross-regional dependencies and sequential decision-making during the tracking process. This solution creatively adopts a reinforcement learning method that combines an improved gated recurrent unit and a graph neural network for residue tracking management, achieving dynamic optimization and precise tracking of the entire process of medical residue treatment. By combining reinforcement learning for policy optimization, this solution can effectively spread information between different states, dynamically adjust the tracking strategy according to real-time data, improve the system's autonomous decision-making ability and response speed, and ensure the efficiency and safety of medical residue management.

[0097] Example Six, refer to Figure 1 and Figure 2 , based on the above example, in step S5, the residue detection management is used for continuous monitoring and management of medical residues. Specifically, by relying on the residue positioning and classification reference data, the image detection information of medical residues is read, and in combination with the residue tracking management reference data, the tracking detection information of residues is read. By combining the image detection information and the tracking detection information, comprehensive detection management of medical residues is carried out, and a rule-based real-time monitoring system for residues is constructed to automatically generate alarms and assist management personnel in recording, generating, and executing the cleaning tasks of medical residues, obtaining residue comprehensive detection management reference data.

[0098] Example Seven, refer to Figure 1 and Figure 2 , based on the above example, the medical residue detection management optimization system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, a residue positioning and classification module, a residue tracking management module, and a residue detection management module;

[0099] The data acquisition module is used for data acquisition. Through data acquisition, the original dataset of medical residue detection management is obtained, and the original dataset of medical residue detection management is sent to the data preprocessing module;

[0100] The data preprocessing module is used for data preprocessing. Through data preprocessing, the optimized dataset of medical residue detection management is obtained, and the optimized dataset of medical residue detection management is sent to the residue positioning and classification module and the residue tracking management module;

[0101] The residue positioning and classification module is used for residue positioning and classification. Through residue positioning and classification, residue positioning and classification reference data are obtained, and the residue positioning and classification reference data are sent to the residue tracking management module and the residue detection management module;

[0102] The residue tracking management module is used for residue tracking management. Through residue tracking management, residue tracking management reference data are obtained, and the residue tracking management reference data are sent to the residue detection management module;

[0103] The residue detection management module is used for residue detection management. Through residue detection management, residue comprehensive detection management reference data are obtained.

[0104] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0105] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0106] The above describes the present invention and its implementation manners. This description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural modes and embodiments without creative efforts without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. The medical residue detection management optimization method based on artificial intelligence is characterized by: The method comprises the following steps: Step S1: Data collection, used to collect multi-source data related to medical residues, specifically, to collect data from operating rooms, wards, and waste treatment areas through video data collection, and to obtain image data of medical equipment and waste container areas in real time through the medical system, to collect temperature, humidity, and air quality environmental data through sensors, and to collect dynamic data of medical residues through tag tracking, to obtain the original data set for medical residue detection management; Step S2: data preprocessing to obtain a medical residue detection management optimization data set; Step S3: residue location and classification, used to locate and classify medical residues using artificial intelligence algorithms, identify different types of waste, and mark their locations. Specifically, based on the medical residue detection management optimization data set, an integrated convolution transformer model combined with a pre-trained detection model is used to perform residue location and classification to obtain residue location and classification reference data, specifically including the following steps: Step S31: construct a pre-trained detection model; Step S32: construct an improved convolution transformer model; Step S33: construct an integrated classifier; Step S34: residue location and classification; The calculation formula of the improved convolution transformer model is: F TBF =MSTF((Bi-Att(Transformer(F Conv ))); In the formula, F TBF is the feature output of the improved convolution transformer model, MSTF(·) is the multi-scale feature level transformation module operation function, Bi-Att(·) is the bidirectional attention mechanism operation function, Transformer(·) is the transformer model operation function, and F Conv is the feature output of a standard convolutional neural network; Step S4: residue tracking management, used to track the entire process of medical residues from generation, processing to destruction, specifically, based on the medical residue detection management optimization data set and the residue positioning classification reference data, a reinforcement learning method combining an improved gated recurrent unit and a graph neural network is used to perform residue tracking management to obtain residue tracking management reference data, including the following steps: Step S41: constructing an improved gated recurrent subnet, specifically constructing a standard bidirectional gated recurrent unit, and constructing a long-short-term dependency improvement module after the output layer of the bidirectional gated recurrent unit, wherein the long-short-term dependency improvement module specifically includes a single-layer standard long-short-term memory layer and a single-layer standard gated recurrent unit layer, and by constructing the long-short-term dependency improvement module, the output feature of the gated recurrent subnet is updated to obtain updated time series feature data; The calculation formula for updating the timing characteristic data is: h′ t =(1-a t )·BiGRU(x t ,h t-1 ,h t )+a t ·f(x t ); In the formula, h′ t It is the updated time series feature data after long-term and short-term dependency improvement, a t is the update time sequence weight, BiGRU(·) is the bidirectional gated recurrent unit operation function, x t is the raw data input, used to represent the medical residue detection management optimization dataset and the residue location classification reference data, h t-1 is the output timing characteristic data of the standard bidirectional gated recurrent unit at the previous moment, h t is the output time series feature data of the standard bidirectional gated recurrent unit at time t, t is the time index, f(x t ) is the output feature data of the long-term and short-term dependency improvement module; Step S42: constructing a graph neural subnet, specifically by constructing a medical residue dynamic tracking state graph, and by constructing a standard graph neural subnet, combining the updated time series feature data, performing spatiotemporal feature extraction, and obtaining spatiotemporal feature data; Step S43: constructing a reinforcement learning model, specifically, constructing a reinforcement learning model by constructing state space parameters and action space parameters, combining with reward function improvement, and according to the spatiotemporal feature data to obtain reinforcement learning decision output data; The reward function improvement specifically includes a medical residue processing efficiency parameter, a medical residue processing safety parameter, and a medical residue utilization rate parameter; Step S44: training the residue tracking management model, specifically, performing model training through the improved gated recurrent subnetwork, the graph neural subnetwork and the reinforcement learning model to obtain the residue tracking management model; Step S45: residue tracking management, specifically using the residue tracking management model, performing residue tracking management according to the medical residue detection management optimization data set and the residue location classification reference data, and obtaining residue tracking management reference data; Step S5: Residue detection management, obtaining residue comprehensive detection management reference data.

2. The medical residue detection management optimization method based on artificial intelligence according to claim 1 is characterized in that: In step S2, the data preprocessing is used to clean, denoise and format the collected data, specifically, to perform data preprocessing operations on the original data set of medical residue detection management to obtain an optimized data set of medical residue detection management.

3. The method for optimizing medical residue detection management based on artificial intelligence according to claim 2 is characterized in that: In step S3, the residue location and classification is used to locate and classify medical residues using an artificial intelligence algorithm, identify different types of waste, and mark their locations. Specifically, based on the medical residue detection management optimization data set, an integrated convolution transformer model combined with a pre-trained detection model is used to perform residue location and classification to obtain residue location and classification reference data, which specifically includes the following steps: Step S31: constructing a pre-trained detection model, specifically constructing a pre-trained integrated detection model, and adjusting the weight of the target detection model for the integrated detection model in combination with the medical residue detection management optimization data set through transfer learning and pre-trained integration methods, to obtain a pre-trained detection model; The integrated detection model specifically refers to an integrated model of YOLOv8 and Faster-R-CNN; Step S32: constructing an improved convolution transformer model, specifically constructing a standard convolution neural network and a transformer model, and improving the hierarchy of the convolution transformer model by introducing a bidirectional attention mechanism and a multi-scale feature hierarchical transformation module to obtain an improved convolution transformer model; Step S33: constructing an integrated classifier, specifically introducing a comprehensive integration method combining an adaptive integration method and a dynamic weighted integration method, using a standard extreme gradient boosting model as a model classifier, and constructing a process classifier; Step S34: residue positioning and classification, specifically, performing model training and model use of residue positioning and classification through the pre-trained detection model, the improved convolution transformer model and the integrated classifier to obtain residue positioning and classification reference data.

4. The method for optimizing medical residue detection management based on artificial intelligence according to claim 3 is characterized in that: In step S5, the residue detection management is used to continuously monitor and manage medical residues, specifically by reading the image detection information of medical residues based on the residue positioning classification reference data, and reading the tracking detection information of the residues in combination with the residue tracking management reference data. By combining the image detection information and the tracking detection information, comprehensive detection management of medical residues is performed, and a rule-based real-time monitoring system for residues is constructed, which automatically generates alarms and assists management personnel in recording, generating and executing medical residue cleaning tasks, thereby obtaining comprehensive residue detection management reference data.

5. A medical residue detection management optimization system based on artificial intelligence, used to implement the medical residue detection management optimization method based on artificial intelligence as described in any one of claims 1 to 4, characterized in that: It includes a data acquisition module, a data preprocessing module, a residue positioning and classification module, a residue tracking and management module, and a residue detection and management module.

6. The medical residue detection management optimization system based on artificial intelligence according to claim 5 is characterized by: The data acquisition module is used for data acquisition, and obtains the original data set of medical residue detection management through data acquisition, and sends the original data set of medical residue detection management to the data preprocessing module; The data preprocessing module is used for data preprocessing, and obtains a medical residue detection management optimization data set through data preprocessing, and sends the medical residue detection management optimization data set to the residue positioning classification module and the residue tracking management module; The residue positioning and classification module is used for residue positioning and classification, obtains residue positioning and classification reference data through residue positioning and classification, and sends the residue positioning and classification reference data to the residue tracking management module and the residue detection management module; The residue tracking management module is used for residue tracking management, obtains residue tracking management reference data through residue tracking management, and sends the residue tracking management reference data to the residue detection management module; The residue detection management module is used for residue detection management, and through the residue detection management, residue comprehensive detection management reference data is obtained.

Images