A method for AI-assisted verification of medical supplies entering and leaving the warehouse and for nurses to hand over duties during shift changes.

Abstract

This invention provides an AI-assisted method for verifying the entry and exit of medical supplies and facilitating nurse shift handover, relating to the fields of hospital nursing management and medical supplies information technology. This AI-assisted system for verifying the entry and exit of medical supplies and facilitating nurse shift handover includes a data storage module for storing basic material data and business records, a terminal interaction module for human-computer interaction, an AI visual entry and exit verification module, a shift handover intelligent generation module, and an entry / exit / shift handover linkage verification module. The AI ​​visual entry and exit verification module collects multi-view images of medical supplies to be entered or exited, and extracts the category, quantity, specifications, and expiration date information of the supplies through an intelligent recognition model adapted to medical packaging features. By using AI vision to batch verify supplies, connect the data link between entry / exit and shift handover, automatically generate handover documents and link risk verification, and implement full-link archiving and traceability, the system effectively simplifies nursing documentation work and improves nursing management efficiency and material control accuracy.

Description

Technical Field

[0001] This invention relates to the field of hospital nursing management and medical supplies information technology, specifically to an AI-assisted method for verifying the entry and exit of medical supplies and for nurses to hand over shifts. Background Technology

[0002] In current clinical nursing work, the verification of medical supplies entering and leaving the warehouse and the handover of nurses are the two most repetitive administrative tasks every day, directly affecting nursing efficiency and the safety of medical supplies management. In terms of supplies management, most hospitals still rely on barcode scanning of each item as the core method of entry and exit, supplemented by manual record-keeping for verification. Even in institutions that have introduced SPD (Supply, Processing, and Distribution) supplies management systems, they still depend on manual scanning of each item. Batch inventory checks are time-consuming and have a high error rate. Nursing staff need to spend a significant amount of time completing record-keeping, expiration date checks, and documentation, significantly reducing their direct nursing time. Furthermore, manual scanning is prone to errors and omissions, leading to discrepancies between records and actual supplies, and the loss of near-expiration supplies, creating management risks. The traceability of high-value consumables and emergency supplies also suffers from gaps.

[0003] In terms of nurse shift handover, existing electronic handover systems are generally independent of the materials management system. Their function focuses on transmitting patient condition information and cannot automatically synchronize details of materials entering and leaving the warehouse, inventory balances, and any abnormal statuses. When nurses complete the handover, they need to log into the materials system separately to verify the ledgers and then manually copy the data into the handover record. This involves a large workload of repetitive cross-system data entry, redundant and cumbersome paperwork, resulting in low overall handover efficiency and a high risk of information omissions. Existing technical solutions have the following main shortcomings: First, the level of automation in inventory verification is insufficient, relying heavily on manual barcode scanning. This leads to low efficiency in batch inventory verification, heavy administrative burden on nursing staff, and a failure to free up nursing manpower through technological means. Second, the data in inventory management and shift handover processes are fragmented, with the two systems not interconnected, preventing automatic data flow. Manual transcription reduces efficiency and is prone to data errors. Third, the automated generation capability of handover documents is weak. Inventory handover information still needs to be manually compiled and filled out, and there is a lack of linkage verification mechanisms at the handover stage. Inventory anomalies cannot be proactively alerted at the handover point, and risk transmission lacks closed-loop control. Fourth, the entire inventory traceability chain is incomplete. Inventory records and handover confirmation records are stored in a scattered manner, requiring cross-system retrieval for subsequent accountability, resulting in high management costs and making it difficult to form a coherent inventory flow accountability system. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an AI-assisted method for verifying the entry and exit of medical supplies and for nurses during shift handover. This method solves the problems of low verification efficiency, fragmented data, cumbersome paperwork, lack of linkage verification, and incomplete traceability chain in existing methods for verifying the entry and exit of medical supplies and for nurses during shift handover.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an AI-assisted medical supplies inbound / outbound verification and nurse shift handover system, comprising a data storage module for storing basic material data and business records, a terminal interaction module for human-computer interaction, an AI visual inbound / outbound verification module, a shift handover intelligent generation module, and an inbound / outbound-shift handover linkage verification module, characterized in that: The AI ​​visual inbound / outbound verification module is used to collect multi-view images of medical supplies to be inbound / outbound. It extracts the category, quantity, specifications and expiration date information of the supplies through an intelligent recognition model adapted to the characteristics of medical packaging. It automatically matches the preset inbound / outbound documents to complete batch verification and updates the material ledger simultaneously. During the recognition process, it performs intelligent completion calculations for stacked and obscured items and automatically marks items with insufficient confidence as requiring further verification. The intelligent shift handover generation module is used to intelligently collect data on inbound and outbound flow, real-time inventory balance, and material anomaly warning data during the current shift according to the shift time dimension at preset shift handover nodes. After automatically removing redundant operation records, it matches the clinical handover template to generate structured electronic shift handover documents. The inbound / outbound-shift handover linkage verification module is used to automatically initiate the consistency verification of accounts and physical inventory during the handover confirmation process. Based on historical consumption data, it intelligently determines the risk level of inventory deviation and highlights high-risk deviations, near-expiry warning materials, and unrecorded requisition records. After electronic confirmation by both parties, the ledger is locked and the handover file is associated and archived simultaneously.

[0006] Preferably, the AI ​​visual inbound / outbound verification module calculates the comprehensive identification confidence level of a single type of material using the following formula, and determines whether to trigger manual verification based on the confidence level value:

[0007] The constraints are + + =1 and all weight coefficients are greater than 0; In the formula, The comprehensive identification confidence level for a single type of material, with a value ranging from 0 to 1; The confidence score for matching the appearance features of material packaging is calculated by the cosine similarity between the appearance feature vector and the samples in the material coding database. The confidence score for material code identification is calculated from the average confidence score of the code identification output by the character recognition model; The confidence level for expiration date inkjet printing recognition is calculated from the average confidence level of the characters for expiration date and production date. , These are the weighting coefficients corresponding to appearance characteristics, coding characteristics, and expiration date characteristics, which can be adjusted according to the asset value level of medical products.

[0008] Preferably, the AI ​​visual inbound / outbound verification module calculates the total quantity of materials in stacked / occluded scenarios using the following formula, and achieves intelligent grading and labeling of expiration dates using the following formula: The formula for quantity completion calculation is as follows:

[0009] In the formula, This represents the estimated total quantity of this type of material; The visible quantity of materials that can be fully identified in multi-view images; The projected area of ​​the stacked occlusion region obtained through image segmentation; The standard projected area of ​​a single package for this type of material is pre-stored in the material basic coding database; The stacking density correction coefficient is obtained by training on stacked samples of medical supplies of the same type, and its value ranges from 0.7 to 1.2. This is the rounding function; The formula for calculating shelf life grading is:

[0010] In the formula, This is the percentage of remaining validity period, with a value ranging from 0 to 1; The expiration date marked on the packaging of the supplies; The current system date; The total number of days of validity for this type of material is pre-stored in the material basic code database; System basis The value is automatically classified into four levels of early warning status: normal, near expiration, close to expiration, and expired.

[0011] Preferably, the intelligent shift handover generation module calculates the effective weight of a single business record using the following formula, filters valid business data based on the effective weight, and eliminates invalid and redundant records:

[0012] The constraints are + + =1; In the formula, The effective weight for a single business record, with a value ranging from 0 to 1; Assuming a weight for business type, formal requisition, return, and inventory operations are assigned a value of 1.0, while test and debugging operations are assigned a value of 0.1. The value is assigned to the document status weight: 1.0 for documents that have been reviewed and confirmed, and 0 for documents that have been voided or cancelled. The weight is assigned to the shift, with a value of 1.0 for documents falling entirely within the shift's time period, and a value calculated based on the percentage of time periods for documents spanning multiple shifts. , , The weighting coefficients are for three dimensions: business type, document status, and shift matching, and can be adjusted according to the department handover rules.

[0013] Preferably, the inbound / outbound-shift handover linkage verification module calculates the inventory deviation risk score using the following formula, and classifies the deviation risk level based on the score:

[0014] In the formula, This is a deviation risk score; a higher score indicates a higher risk of anomaly. This represents the current book inventory quantity; The physical inventory quantity obtained from AI visual inventory; This represents the average daily consumption of this type of material over the past 30 days. This represents the standard deviation of the average daily consumption of this type of material over the past 30 days. System basis The risk level is divided into three levels: low, medium, and high. High-risk items are highlighted and strongly highlighted in the handover documents.

[0015] Preferably, the AI ​​vision entry and exit verification module supports manual review and correction. Items with a comprehensive recognition confidence level lower than a preset threshold are automatically pushed to the terminal interaction module. After manual confirmation and correction, the corrected samples are included in the model training dataset to achieve iterative optimization of the recognition model. The structured electronic handover document includes a general nursing handover area and a dedicated materials handover area. The dedicated materials handover area automatically loads four types of details: items issued during the shift, items returned to the warehouse, inventory discrepancies, and expiration date reminders. It also supports handover personnel to add text notes, and all modification operations are fully logged and stored in the data storage module.

[0016] An AI-assisted method for verifying the entry and exit of medical supplies and facilitating nurse shift handover includes the following steps: S1. Basic Data Configuration: Enter the hospital's unified material coding database, departmental shift rules, handover templates, and expiration date warning thresholds to complete standardized basic data storage; S2, AI Batch Inbound / Outbound Verification: Collects multi-view images of medical supplies to be inbound / outbound, extracts the category, quantity, specifications and expiration date information of the supplies through an intelligent recognition model adapted to medical packaging, performs intelligent completion calculation for obscured items, automatically compares with inbound / outbound documents, and automatically updates the material ledger after verification. S3 Intelligent generation of handover documents: When the preset handover time node is reached, the system intelligently filters the valid material flow data of the shift according to the shift time dimension, automatically removes invalid operation records, and matches the handover template after classifying and summarizing according to business type to generate structured electronic handover documents. S4. Linked Verification and Handover Confirmation: During the document verification process between the two parties, the system automatically triggers linked verification of the consistency between accounts and actual inventory. It intelligently determines the risk level of inventory deviation based on historical consumption patterns, highlights high-risk anomalies, and completes electronic signature confirmation after both parties confirm that there are no errors. S5. Data Archiving and Traceability: After the handover is completed, the system automatically links and archives the handover documents, verification logs and the corresponding inbound and outbound records for the corresponding period, forming a complete chain of responsibility for material flow and handover, supporting full-process traceability and query.

[0017] Preferably, in step S2, the AI ​​batch inbound / outbound verification specifically includes three calculation steps: confidence calculation, occlusion completion, and expiration date classification. First, calculate the overall identification confidence level for a single type of material, using the following formula: The constraints are + + =1; In the formula To comprehensively identify confidence levels, Match confidence levels to packaging appearance. To encode the confidence level, To determine the confidence level of the inkjet printing for the expiration date, , , These are the corresponding weighting coefficients; Manual review is triggered when the confidence level falls below a preset threshold. The formula for calculating the quantity of stacked and obscured materials is as follows: In the formula To estimate the total quantity, For the visible quantity, To cover the area, This is the standard area for a single package. This is a stacking density correction factor; Simultaneously calculate the remaining validity period percentage coefficient, the formula is as follows: ,in accordance with The value is divided into four levels of validity period status and updated synchronously to the material ledger.

[0018] Preferably, in step S3, the intelligent generation of handover documents first performs intelligent filtering of valid data using a formula. Calculate the effective weight of a single business record, with the following constraints: + + =1; In the formula For effective weights, Weights based on business type. For document status weights, Weights are assigned to each shift. , , These are the weight coefficients for the corresponding dimensions; After removing redundant records with effective weights below the preset threshold, the core data is categorized and summarized into four dimensions: requisition, return, inventory, and anomaly, and automatically filled into the corresponding fields of the handover template.

[0019] Preferably, in step S4, the linkage verification step is performed using a formula. Calculate the inventory deviation risk score; In the formula For deviation risk score, This refers to the book inventory quantity. This refers to the physical inventory quantity. This represents the average daily consumption over the past 30 days. This represents the standard deviation of daily average consumption. Based on the scores, the risk levels are divided into low, medium and high. High-risk deviations, near-expiry materials and unrecorded requisition records are highlighted in the handover documents to ensure that both parties have a full understanding of the risk status of the materials.

[0020] This invention provides an AI-assisted method for verifying the entry and exit of medical supplies and facilitating shift handover among nurses. It offers the following advantages: 1. This invention uses an AI vision-based inventory verification module to collect multi-view images of medical supplies, combined with comprehensive confidence calculation, stacking occlusion completion and expiration date classification algorithms, and a manual review and iterative optimization mechanism to achieve rapid and accurate verification of batch supplies. This significantly reduces the transactional burden of manual scanning, effectively frees up nursing manpower, and improves the efficiency and accuracy of inventory verification.

[0021] 2. This invention establishes a data link between the management of medical supplies entering and leaving the warehouse and the handover of nurses, enabling real-time synchronization of entry and exit ledgers, inventory balances, and abnormal data. Combined with automatic aggregation logic at the shift level, it achieves the effect of integrated flow of data for both types of business, eliminating redundant cross-system data entry and avoiding data deviation problems caused by manual transcription.

[0022] 3. This invention uses an effective weighted filtering mechanism in the intelligent shift handover generation module to remove invalid records, summarizes them according to business categories to generate standardized handover documents, and, in conjunction with the account-to-actual linkage verification and risk classification early warning at the handover node, achieves the effect of automatic document generation and proactive anomaly control, simplifies nursing documentation work, and strengthens the risk control capability of the handover process.

[0023] 4. This invention links and archives inbound and outbound records, verification logs, and handover documents, and, in conjunction with a full-process operation traceability and electronic signature confirmation mechanism, achieves full-chain traceability of material flow and handover responsibilities, solves the traceability breakpoint problem caused by decentralized storage, reduces management costs, and builds a coherent material responsibility management system. Attached Figure Description

[0024] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a flowchart of the steps of the present invention; Figure 3 This is a data comparison chart for the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1: like Figure 1-3As shown, this embodiment of the invention provides an AI-assisted medical supplies inbound / outbound verification and nurse shift handover system, including a data storage module for storing basic material data and business records, a terminal interaction module for human-computer interaction, an AI visual inbound / outbound verification module, a shift handover intelligent generation module, and an inbound / outbound-shift handover linkage verification module. The AI ​​visual inbound / outbound verification module is used to collect multi-view images of medical supplies to be inbound / outbound, extract the category, quantity, specifications, and expiration date information of the supplies through an intelligent recognition model adapted to medical packaging features, automatically match preset inbound / outbound documents to complete batch verification, and synchronously update the material ledger. During the recognition process, intelligent completion calculation is performed for stacked and obscured items, and items with insufficient confidence are automatically marked for review. The AI ​​visual inbound / outbound verification module calculates the comprehensive recognition confidence of a single type of material using the following formula, and determines whether to trigger manual review based on the confidence value: The constraints are + + =1 and all weight coefficients are greater than 0; where, The comprehensive identification confidence level for a single type of material, with a value ranging from 0 to 1; The confidence score for matching the appearance features of material packaging is calculated by the cosine similarity between the appearance feature vector and the samples in the material coding database. The confidence score for material code identification is calculated from the average confidence score of the code identification output by the character recognition model; The confidence level for expiration date inkjet printing recognition is calculated from the average confidence level of the characters for expiration date and production date. , These are the weighting coefficients for appearance features, coding features, and expiration date features, which can be adjusted according to the asset value level of medical products. The AI-powered visual inventory verification module calculates the total quantity of materials in stacked or obscured scenarios using the following formula, and intelligently labels expiration dates using the following formula: The quantity completion calculation formula is: In the formula, This represents the estimated total quantity of this type of material; The visible quantity of materials that can be fully identified in multi-view images; The projected area of ​​the stacked occlusion region obtained through image segmentation; The standard projected area of ​​a single package for this type of material is pre-stored in the material basic coding database; The stacking density correction coefficient is obtained by training on stacked samples of medical supplies of the same type, and its value ranges from 0.7 to 1.2. This is a rounding function; the formula for calculating the validity period classification is: In the formula, This is the percentage of remaining validity period, with a value ranging from 0 to 1; The expiration date marked on the packaging of the supplies; The current system date; The total shelf life days of this type of material are pre-stored in the material basic code database; the system then... The value is automatically classified into four warning levels: normal, near expiration, close to expiration, and expired. The AI ​​vision inventory verification module supports manual review and correction. Items with a comprehensive recognition confidence level below a preset threshold are automatically pushed to the terminal interaction module. After manual confirmation and correction, the corrected samples are included in the model training dataset to achieve iterative optimization of the recognition model. The structured electronic handover document has a general nursing handover area and a dedicated material handover area. The dedicated material handover area automatically loads four types of details: on-duty requisition, return, inventory discrepancies, and expiration date reminders. It supports handover personnel to add text notes, and all modification operations are fully recorded and stored in the data storage module. Specifically: When performing medical supplies entry verification, the system uses the multi-view acquisition unit of the AI ​​vision entry and exit verification module to capture images of the boxes of consumables to be entered from both the front and side. For routine low-value consumables such as disposable syringes and infusion sets, the system presets the weighting coefficients for appearance features to be 0.6, coding features to be 0.25, and expiration date features to be 0.15. The system obtains the comprehensive identification confidence of a single type of material by weighted summation of the confidence levels of the three types of features. When the comprehensive confidence level is lower than the preset threshold of 0.85, the system automatically marks the corresponding material item and pushes it to the terminal interaction module for manual review.

[0027] For stacked packages of consumables, the system identifies the projected area of ​​the obscured area through image segmentation, combines it with the standard projected area of ​​a single package pre-stored in the material code library in the data storage module, and uses the stacking density correction coefficient obtained by training with samples of this type of consumables to calculate the quantity of materials in the obscured area after rounding. The quantity is then added to the visible quantity to obtain the estimated total quantity of the entire batch of materials. This is then matched one by one with the inbound documents to complete the batch verification. Once the verification is correct, the material ledger data is automatically updated.

[0028] When the preset day / night shift handover point is reached, the intelligent handover generation module filters all material flow records within the shift duration according to the shift length. It calculates the effective weight of each record through a weighted calculation based on three dimensions: business type, document status, and shift matching. It automatically removes invalid records such as test data with a weight lower than 0.4 and voided documents. After summarizing the remaining requisition, return, and inventory data by category, it matches the clinical handover template to generate a structured electronic handover document that includes a general nursing handover area and a material-specific handover area. The material handover area automatically loads the shift's requisition details, return records, inventory discrepancies, and expiration date reminders, and supports the handover personnel to add text notes.

[0029] During the handover verification process, the inbound / outbound / shift handover linkage verification module automatically retrieves the latest AI inventory physical inventory data and book inventory data. It calculates the deviation risk score by combining the average daily consumption and consumption standard deviation of this type of consumable over the past 30 days. The risk is divided into three levels: low, medium, and high based on the score. High-risk deviation items are highlighted in the document. At the same time, the system automatically classifies the expiration status into four levels: normal, near expiration, expiring, and expired based on the ratio of the remaining valid days of the materials to the total expiration period. Near-expiring and expired materials are included in the warning scope. After both parties confirm that there are no errors, they complete the electronic signature. The system simultaneously locks the shift log and archives the shift handover document, verification log, and inbound / outbound records. All manual modifications and remarks are recorded and stored throughout the process.

[0030] Example 2: The intelligent shift handover generation module is used to intelligently collect data on inbound and outbound transactions, real-time inventory levels, and material anomaly warnings during the current shift at preset shift handover nodes, categorized by shift time. After automatically removing redundant operation records, it matches the clinical handover template to generate structured electronic handover documents. The module calculates the effective weight of each business record using the following formula, filters valid business data based on the effective weight, and removes invalid and redundant records: The constraints are + + =1; where, The effective weight for a single business record, with a value ranging from 0 to 1; Assuming a weight for business type, formal requisition, return, and inventory operations are assigned a value of 1.0, while test and debugging operations are assigned a value of 0.1. The value is assigned to the document status weight: 1.0 for documents that have been reviewed and confirmed, and 0 for documents that have been voided or cancelled. The weight is assigned to the shift, with a value of 1.0 for documents falling entirely within the shift's time period, and a value calculated based on the percentage of time periods for documents spanning multiple shifts. , , The weighting coefficients are for three dimensions: business type, document status, and shift matching, and can be adjusted according to the department handover rules.

[0031] Specifically, this embodiment is applied to the management of high-value medical consumables in and out of the emergency department and shift handover. For high-value medical items such as interventional consumables and hemostatic materials, the AI ​​vision in-and-out verification module adjusts the weight configuration to 0.5 for coding features, 0.3 for expiration date features, and 0.2 for appearance features. Priority is given to ensuring the accuracy of material coding and expiration date information. The overall confidence threshold is raised to 0.9. Items below the threshold are forcibly pushed to manual review to ensure that the accounts and actual stock of high-value consumables are completely matched.

[0032] For stacking scenarios involving boxed high-value consumables, a stacking density correction factor of 0.95 is used to minimize quantity estimation errors. After the inventory verification is completed, the system synchronously updates the dedicated ledger for high-value consumables, recording the expiration date and circulation information of each item. During shift handover, the intelligent shift handover generation module lists high-value consumables as a separate handover area, and the weight of the document status dimension in the effective weight calculation is increased to 0.5 to ensure that only officially approved documents are included in the handover scope.

[0033] The inbound / outbound and shift handover linkage verification module implements strict deviation judgment rules for high-value consumables. Any discrepancy between the account and the actual situation is judged as a high-risk mandatory warning. Both parties to the handover must check and confirm one by one before signing. The identification samples that have been manually confirmed and corrected are automatically included in the model training dataset to continuously optimize the adaptation accuracy of the identification model. All handover operations and correction records are synchronously archived to the data storage module to form a complete traceability link for high-value consumables.

[0034] Example 3: The inbound / outbound / shift handover linkage verification module automatically initiates consistency verification between inventory records and physical inventory during the handover confirmation process. Based on historical consumption data, it intelligently determines the risk level of inventory deviations, highlighting high-risk deviations, near-expiry warning materials, and unrecorded requisition records. After electronic confirmation by both parties, the module simultaneously locks the ledger and archives the handover documents. The inbound / outbound / shift handover linkage verification module calculates the inventory deviation risk score using the following formula and classifies the deviation risk level based on the score: In the formula, This is a deviation risk score; a higher score indicates a higher risk of anomaly. This represents the current book inventory quantity; The physical inventory quantity obtained from AI visual inventory; This represents the average daily consumption of this type of material over the past 30 days. This represents the standard deviation of the average daily consumption of this type of material over the past 30 days; the system is based on... The risk level is divided into three levels: low, medium, and high. High-risk items are highlighted and strongly highlighted in the handover documents.

[0035] Specifically, this embodiment is applied to the handover management of emergency supplies in the intensive care unit. For emergency medicines, disposable resuscitation equipment, and other supplies, the expiration date feature weight of the AI ​​vision inbound / outbound verification module is increased to 0.35. The expiration date classification threshold corresponds to the tightening of emergency supplies management requirements. If the remaining expiration date is less than 0.2%, it is judged as nearing its expiration date, ensuring that emergency supplies are always in a valid and usable state. Due to the large fluctuations in the consumption of emergency supplies, when the inbound / outbound-handover linkage verification module calculates the deviation risk score, it automatically corrects the fluctuation tolerance range by combining the consumption data of the past 30 days, avoiding invalid warnings caused by regular consumption fluctuations. At the same time, emergency requisition records that have not been recorded are automatically marked and pushed to the handover documents, ensuring that both parties to the handover have a full understanding of the changes in supplies.

[0036] When handover documents are generated, emergency supplies are categorized separately and placed at the top. In the effective weight calculation, the business type dimension weight is tilted towards emergency requisition business to ensure the complete presentation of core business data. All abnormal items must be marked and confirmed by both parties before the handover and archiving can be completed. The system simultaneously locks the ledger and links the files, forming a complete traceability link for the flow of materials, ensuring the management safety and handover efficiency of emergency supplies.

[0037] Example 4: An AI-assisted method for verifying the entry and exit of medical supplies and facilitating nurse shift handover, comprising the following steps: S1. Basic Data Configuration: Enter the hospital's unified material coding database, departmental shift rules, handover templates, and expiration date warning thresholds to complete standardized basic data storage; S2. AI Batch Inbound / Outbound Verification: Collects multi-view images of medical supplies to be inbound / outbound. Using an intelligent recognition model adapted to medical packaging, it extracts the category, quantity, specifications, and expiration date information of the supplies. It performs intelligent completion calculations for obscured items and automatically compares them with inbound / outbound documents. Once verification is successful, it automatically updates the material ledger. AI batch inbound / outbound verification specifically includes three calculation stages: confidence level calculation, obscuration completion, and expiration date classification. First, it calculates the comprehensive recognition confidence level for a single type of material, using the formula: The constraints are + + =1; where To comprehensively identify confidence levels, Match confidence levels to packaging appearance. To encode the confidence level, To determine the confidence level of the inkjet printing for the expiration date, , , The corresponding weighting coefficients; manual review is triggered when the confidence level is lower than the preset threshold; quantity estimation is performed for stacked and obscured materials, using the following formula: In the formula To estimate the total quantity, For the visible quantity, To cover the area, This is the standard area for a single package. This is the stacking density correction factor; simultaneously, the remaining shelf life percentage factor is calculated using the following formula: ,in accordance with The value is classified into four levels of validity period status and updated synchronously to the material ledger; S3. Intelligent Generation of Shift Handover Documents: Upon reaching the preset shift handover time, the system intelligently filters valid material flow data for the current shift based on shift time, automatically removes invalid operation records, and matches the handover template after categorizing and summarizing by business type to generate structured electronic shift handover documents. The intelligent generation of shift handover documents first performs intelligent filtering of valid data using formulas. Calculate the effective weight of a single business record, with the following constraints: + + =1; where For effective weights, Weights based on business type. For document status weights, Weights are assigned to each shift. , , The corresponding dimension weight coefficients are used; after removing redundant records with effective weights below the preset threshold, the core data is categorized and summarized according to four dimensions: requisition, return, inventory, and exception, and the corresponding fields in the handover template are automatically filled in. S4. Linked Verification and Handover Confirmation: During the document verification process between the two parties, the system automatically triggers a linked verification of the consistency between accounts and physical inventory. It intelligently determines the risk level of inventory deviation based on historical consumption patterns, highlighting high-risk anomalies. After both parties confirm that everything is correct, they complete the electronic signature confirmation. The linked verification process uses a formula... Calculate the inventory deviation risk score; where For deviation risk score, This refers to the book inventory quantity. This refers to the physical inventory quantity. This represents the average daily consumption over the past 30 days. The standard deviation of daily consumption is used; based on the score, the risk levels are divided into low, medium and high. High-risk deviations, near-expiry materials and unrecorded requisition records are highlighted in the handover documents to ensure that both parties to the handover have a full understanding of the risk status of the materials. S5. Data Archiving and Traceability: After the handover is completed, the system automatically links and archives the handover documents, verification logs and the corresponding inbound and outbound records for the corresponding period, forming a complete chain of responsibility for material flow and handover, supporting full-process traceability and query.

[0038] Specifically, this embodiment is applied to the daily inbound and outbound verification and shift handover process of low-value medical consumables in general internal medicine wards. The implementation process begins with basic data configuration: the hospital's unified low-value consumable coding library, the shift rules for the three-shift rotation in internal medicine wards, the clinical handover template, and the routine expiration date warning threshold are entered into the system and stored in a standardized manner. When carrying out material warehousing operations, nursing staff place boxes of disposable syringes, infusion sets, and other consumables to be warehoused at the data collection station. The system initiates the AI ​​batch inbound and outbound verification process, collecting multi-view images of the consumables to be verified. Using a recognition model adapted to medical packaging features, the system extracts the category, quantity, specifications, and expiration date information of various consumables. For low-value consumables, a weighting coefficient of 0.6 for appearance features, 0.25 for coding features, and 0.15 for expiration date features are set. A weighted calculation is performed to obtain the comprehensive identification confidence level for a single type of consumable. When the confidence level is lower than the preset threshold of 0.85, a manual review process is automatically triggered. For stacked packages of consumables, the system obtains the projected area of ​​the obscured area through image segmentation. Combining the standard projected area of ​​a single package pre-stored in the database with the stacking density correction coefficient obtained through sample training, the system estimates the quantity of consumables in the obscured area. This quantity is then added to the visible quantity to obtain the estimated total quantity of the entire batch of consumables. After comparing each item with the warehousing document and confirming that there are no errors, the system automatically updates the material ledger. At the same time, it calculates the remaining expiration rate coefficient of various consumables, classifies them into four levels of status according to rules: normal, near expiration, expiring, and expired, and marks them synchronously in the ledger. When the preset shift handover time arrives, the system automatically initiates the intelligent generation process for handover documents. It filters all material flow records within the current shift, calculating the effective weight of each record based on three dimensions: business type, document status, and shift matching. It automatically removes invalid and redundant records such as test data with a weight below 0.4 and voided documents. The remaining effective data is categorized and summarized into four dimensions: requisition, return, inventory, and anomalies. This data is automatically filled into the corresponding fields of the handover template, generating a structured electronic handover document containing both a general nursing handover area and a dedicated material handover area. The system also allows handover personnel to add textual notes. During document verification by both parties, the system triggers a linked verification process, retrieving the latest AI inventory data for physical and book inventory. It calculates a deviation risk score by combining the average daily consumption and standard deviation of the consumables over the past 30 days, classifying risks into low, medium, and high levels. High-risk deviation items, near-expiry consumables, and unrecorded requisition records are highlighted in the document. After both parties confirm that all content is correct, they complete the electronic signature. After the handover is completed, the system executes the data archiving and traceability process, automatically linking and archiving the handover documents, verification logs and the corresponding inbound and outbound records for the corresponding period, forming a complete chain of responsibility for material flow and handover, supporting subsequent full-process traceability and query, and storing all manual modifications and remarks throughout the process.

[0039] This embodiment is applied to the inbound and outbound verification and shift handover management of high-value medical consumables in the emergency department. During implementation, basic data configuration is first completed, including inputting a dedicated coding library for high-value consumables, the emergency department's two-shift system rules, a special handover template for high-value consumables, and a strict expiration date warning threshold. Simultaneously, the overall confidence threshold is increased to 0.9 to adapt to the management requirements of high-value materials. When performing high-value consumable inbound verification, nursing staff place boxed interventional consumables, hemostatic materials, and other high-value items in the collection area. The system initiates an AI-powered batch inbound and outbound verification process, collecting multi-view images and adjusting the weight configuration. The coding feature weight is set to 0.5, the expiration date feature weight to 0.3, and the appearance feature weight to 0.2, prioritizing the accuracy of material coding and expiration date information identification. A weighted calculation of the overall identification confidence score is performed, and items below the threshold are forcibly pushed for manual review. Only after confirmation can the item be entered into the accounting system. For stacking scenarios involving boxed high-value consumables, a stacking density correction factor of 0.95 is used to reduce quantity estimation errors. After verification, the dedicated ledger for high-value consumables is updated synchronously, recording the expiration date and circulation information for each item. At shift handover points, the system executes an intelligent handover document generation process, filtering the circulation records of high-value consumables during the shift. When calculating the effective weight, the weight of the document status dimension is increased to 0.5 to ensure that only officially approved documents are included in the handover scope. High-value consumables are listed separately in the handover area, and the details of inbound and outbound shipments and inventory balances are summarized by category to generate a dedicated handover document. During the verification process by both parties, the system initiates a linked verification, applying a zero-deviation judgment rule to high-value consumables. Any discrepancies between the ledger and the actual inventory are judged as high-risk and forcibly highlighted. Both parties must verify and confirm each item before completing the electronic signature. Manually verified and corrected identification samples are automatically included in the model training dataset to continuously optimize the accuracy of the identification model. After the handover is completed, the system will uniformly link and archive all high-value consumables' inbound and outbound records, handover documents, verification logs, and manual correction records to form a complete end-to-end traceability system for high-value consumables, meeting the requirements for refined management of high-value materials.

[0040] This embodiment is applied to the inbound and outbound verification and shift handover management of emergency supplies in the intensive care unit. Before implementation, basic data configuration is completed, including the input of a dedicated coding library for emergency medicines and resuscitation equipment, the shift rules for the four-shift rotation in the ICU, and a priority handover template for emergency supplies. At the same time, the expiration date warning threshold is tightened to adapt to the strict management standards for emergency supplies. When performing inbound and outbound verification of emergency supplies, the system initiates an AI batch inbound and outbound verification process. After collecting multi-view images, the weight of the expiration date feature is increased to 0.35 to strengthen the recognition priority of expiration date information. The weighted comprehensive confidence score is calculated and the corresponding review mechanism is triggered. When calculating the remaining expiration rate coefficient, the threshold for determining near-expiration is adjusted to a remaining rate below 0.2, ensuring that emergency supplies are always in a valid and usable state. Given the frequent and urgent requisitions of emergency supplies, the system adjusts the stacking density coefficient according to the packaging characteristics of the supplies during quantity completion calculations. Simultaneously, emergency requisition records that have not followed the standard procedures are automatically marked and temporarily stored. At shift handover points, the system executes an intelligent handover document generation process, filtering emergency supply circulation data for the shift and prioritizing emergency requisition-related business when calculating effective weights. This ensures complete presentation of core business data, placing emergency supplies at the top of the handover document, and separately listing requisition details, inventory balance, and expiration status. During the verification process by both parties, the system initiates a linked verification process. When calculating the deviation risk score, considering the large fluctuations in emergency supply consumption over the past 30 days, the system automatically corrects the deviation tolerance range using the consumption standard deviation, avoiding invalid warnings caused by regular consumption fluctuations. Simultaneously, incomplete emergency requisition records are automatically pushed to the handover document for highlighting, ensuring that both parties have a comprehensive understanding of any changes in supplies. All anomalies must be marked and confirmed by both parties before electronic signatures can be completed. After the handover is completed, the system will link and archive the emergency supplies' inbound and outbound records, handover confirmation information, verification logs, and anomaly handling records, forming a clear chain of responsibility for the flow and handover of supplies, ensuring the safety of emergency supplies management and the efficiency of handover.

[0041] Table 1: Comparison of Core Features of Overall Technical Solutions ; Table 2: Comparison of Efficiency in Medical Supplies Inbound and Outbound Verification ; Table 3: Comparison of Nursing Shift Handover Efficiency ; Table 4: Comparison of Inventory Deviation and Risk Management Capability ; Table 5: Comparison of Adaptability to Different Clinical Scenarios ; Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-assisted medical supplies inbound / outbound verification and nurse shift handover system, comprising a data storage module for storing basic supplies data and business records, a terminal interaction module for human-computer interaction, an AI vision inbound / outbound verification module, a shift handover intelligent generation module, and an inbound / outbound-shift handover linkage verification module, characterized in that: The AI ​​visual inbound / outbound verification module is used to collect multi-view images of medical supplies to be inbound / outbound. It extracts the category, quantity, specifications and expiration date information of the supplies through an intelligent recognition model adapted to the characteristics of medical packaging. It automatically matches the preset inbound / outbound documents to complete batch verification and updates the material ledger simultaneously. During the recognition process, it performs intelligent completion calculations for stacked and obscured items and automatically marks items with insufficient confidence as requiring further verification. The intelligent shift handover generation module is used to intelligently collect data on inbound and outbound flow, real-time inventory balance, and material anomaly warning data during the current shift according to the shift time dimension at preset shift handover nodes. After automatically removing redundant operation records, it matches the clinical handover template to generate structured electronic shift handover documents. The inbound / outbound-shift handover linkage verification module is used to automatically initiate the consistency verification of accounts and physical inventory during the handover confirmation process. Based on historical consumption data, it intelligently determines the risk level of inventory deviation and highlights high-risk deviations, near-expiry warning materials, and unrecorded requisition records. After electronic confirmation by both parties, the ledger is locked and the handover file is associated and archived simultaneously.

2. The AI-assisted medical supplies entry and exit verification and nurse shift handover system according to claim 1, characterized in that: The AI ​​visual inbound / outbound verification module calculates the comprehensive identification confidence level of a single type of material using the following formula, and determines whether to trigger manual verification based on the confidence level value: The constraints are + + =1 and all weight coefficients are greater than 0; In the formula, The comprehensive identification confidence level for a single type of material, with a value ranging from 0 to 1; The confidence score for matching the appearance features of material packaging is calculated by the cosine similarity between the appearance feature vector and the samples in the material coding database. The confidence score for material code identification is calculated from the average confidence score of the code identification output by the character recognition model; The confidence level for expiration date inkjet printing recognition is calculated from the average confidence level of the characters for expiration date and production date. , These are the weighting coefficients corresponding to appearance characteristics, coding characteristics, and expiration date characteristics, which can be adjusted according to the asset value level of medical products.

3. The AI-assisted medical supplies entry and exit verification and nurse shift handover system according to claim 2, characterized in that: The AI ​​visual inventory verification module calculates the total quantity of materials in stacked and obscured scenarios using the following formula, and uses the following formula to achieve intelligent grading and labeling of expiration dates: The formula for quantity completion calculation is as follows: In the formula, This represents the estimated total quantity of this type of material; The visible quantity of materials that can be fully identified in multi-view images; The projected area of ​​the stacked occlusion region obtained through image segmentation; The standard projected area of ​​a single package for this type of material is pre-stored in the material basic coding database; The stacking density correction coefficient is obtained by training on stacked samples of medical supplies of the same type, and its value ranges from 0.7 to 1.

2. This is the rounding function; The formula for calculating shelf life grading is: In the formula, This is the percentage of remaining validity period, with a value ranging from 0 to 1; The expiration date marked on the packaging of the supplies; The current system date; The total number of days of validity for this type of material is pre-stored in the material basic code database; System basis The value is automatically classified into four levels of early warning status: normal, near expiration, close to expiration, and expired.

4. The AI-assisted medical supplies entry and exit verification and nurse shift handover system according to claim 1, characterized in that: The intelligent shift handover generation module calculates the effective weight of a single business record using the following formula, and filters valid business data based on the effective weight, eliminating invalid and redundant records: The constraints are + + =1; In the formula, The effective weight for a single business record, with a value ranging from 0 to 1; Assuming a weight for business type, formal requisition, return, and inventory operations are assigned a value of 1.0, while test and debugging operations are assigned a value of 0.

1. The value is assigned to the document status weight: 1.0 for documents that have been reviewed and confirmed, and 0 for documents that have been voided or cancelled. The weight is assigned to the shift, with a value of 1.0 for documents falling entirely within the shift's time period, and a value calculated based on the percentage of time periods for documents spanning multiple shifts. , , The weighting coefficients are for three dimensions: business type, document status, and shift matching, and can be adjusted according to the department handover rules.

5. The AI-assisted medical supplies entry and exit verification and nurse shift handover system according to claim 1, characterized in that: The inbound / outbound-shift handover linkage verification module calculates the inventory deviation risk score using the following formula, and classifies the deviation risk level based on the score: In the formula, This is a deviation risk score; a higher score indicates a higher risk of anomaly. This represents the current book inventory quantity; The physical inventory quantity obtained from AI visual inventory; This represents the average daily consumption of this type of material over the past 30 days. This represents the standard deviation of the average daily consumption of this type of material over the past 30 days. System basis The risk level is divided into three levels: low, medium, and high. High-risk items are highlighted and strongly highlighted in the handover documents.

6. The AI-assisted medical supplies entry and exit verification and nurse shift handover system according to claim 1, characterized in that: The AI ​​vision inventory verification module supports manual review and correction. Items with a comprehensive recognition confidence level lower than a preset threshold are automatically pushed to the terminal interaction module. After manual confirmation and correction, the corrected samples are included in the model training dataset to achieve iterative optimization of the recognition model. The structured electronic handover document includes a general nursing handover area and a dedicated materials handover area. The dedicated materials handover area automatically loads four types of details: items issued during the shift, items returned to the warehouse, inventory discrepancies, and expiration date reminders. It also supports handover personnel to add text notes, and all modification operations are fully logged and stored in the data storage module.

7. An AI-assisted method for verifying the entry and exit of medical supplies and for nurse shift handover, using an AI-assisted system for verifying the entry and exit of medical supplies and for nurse shift handover as described in any one of claims 1 to 6, characterized in that, Includes the following steps: S1. Basic Data Configuration: Enter the hospital's unified material coding database, departmental shift rules, handover templates, and expiration date warning thresholds to complete standardized basic data storage; S2, AI Batch Inbound / Outbound Verification: Collects multi-view images of medical supplies to be inbound / outbound, extracts the category, quantity, specifications and expiration date information of the supplies through an intelligent recognition model adapted to medical packaging, performs intelligent completion calculation for obscured items, automatically compares with inbound / outbound documents, and automatically updates the material ledger after verification. S3 Intelligent generation of handover documents: When the preset handover time node is reached, the system intelligently filters the valid material flow data of the shift according to the shift time dimension, automatically removes invalid operation records, and matches the handover template after classifying and summarizing according to business type to generate structured electronic handover documents. S4. Linked Verification and Handover Confirmation: During the document verification process between the two parties, the system automatically triggers linked verification of the consistency between accounts and actual inventory. It intelligently determines the risk level of inventory deviation based on historical consumption patterns, highlights high-risk anomalies, and completes electronic signature confirmation after both parties confirm that there are no errors. S5. Data Archiving and Traceability: After the handover is completed, the system automatically links and archives the handover documents, verification logs and the corresponding inbound and outbound records for the corresponding period, forming a complete chain of responsibility for material flow and handover, supporting full-process traceability and query.

8. The method for AI-assisted verification of medical supplies entering and leaving the warehouse and for nurses to hand over shifts, as described in claim 7, is characterized in that: In step S2, the AI ​​batch inbound / outbound verification specifically includes three calculation steps: confidence calculation, occlusion completion, and validity period classification. First, calculate the overall identification confidence level for a single type of material, using the following formula: The constraints are + + =1; In the formula To comprehensively identify confidence levels, Match confidence levels to packaging appearance. To encode the confidence level, To determine the confidence level of the inkjet printing for the expiration date, , , These are the corresponding weighting coefficients; Manual review is triggered when the confidence level falls below a preset threshold. The formula for calculating the quantity of stacked and obscured materials is as follows: In the formula To estimate the total quantity, For the visible quantity, To cover the area, This is the standard area for a single package. This is a stacking density correction factor; Simultaneously calculate the remaining validity period percentage coefficient, the formula is as follows: ,in accordance with The value is divided into four levels of validity period status and updated synchronously to the material ledger.

9. The method for AI-assisted verification of medical supplies entering and leaving the warehouse and for nurses to hand over shifts, as described in claim 7, is characterized in that: In step S3, the intelligent generation of handover documents first performs intelligent filtering of valid data using a formula. Calculate the effective weight of a single business record, with the following constraints: + + =1; In the formula For effective weights, Weights based on business type. For document status weights, Weights are assigned to each shift. , , These are the weight coefficients for the corresponding dimensions; After removing redundant records with effective weights below the preset threshold, the core data is categorized and summarized into four dimensions: requisition, return, inventory, and anomaly, and automatically filled into the corresponding fields of the handover template.

10. The method for AI-assisted verification of medical supplies entering and leaving the warehouse and for nurses to hand over shifts, as described in claim 7, is characterized in that: In step S4, the linkage verification step is performed using the formula... Calculate the inventory deviation risk score; In the formula For deviation risk score, This refers to the book inventory quantity. This refers to the physical inventory quantity. This represents the average daily consumption over the past 30 days. This represents the standard deviation of daily average consumption. Based on the scores, the risk levels are divided into low, medium and high. High-risk deviations, near-expiry materials and unrecorded requisition records are highlighted in the handover documents to ensure that both parties have a full understanding of the risk status of the materials.

Images