Intelligent distribution and inventory monitoring system for surgical consumables

Abstract

The invention discloses an intelligent distribution and inventory monitoring system for surgical consumables. The intelligent distribution and inventory monitoring system comprises a data interface gateway, a central processing server and an intelligent consumable storage unit, operation schedules and unstructured text data in an HIS system, an ORIS system and an EMR system are integrated through a data interface gateway, an NLP analysis module in a central processing server is used for extracting requirements of'operation-special consumables', and a consumable consumption probability distribution model based on multi-dimensional features and a surgeon doctor preference coefficient are combined; using a dynamic target inventory calculation engine to generate a target inventory level that varies with time according to a formula; the inventory strategy module automatically generates an electronic replenishment instruction according to comparison with real-time inventory; the intelligent storage unit realizes accurate tracking of consumable delivery and warehousing through RFID (Radio Frequency Identification Device); and the closed-loop feedback module continuously optimizes model parameters by using actual consumption data. According to the system and the method, prospective inventory management based on future operation requirements can be realized, and the accuracy of consumable supply, the inventory turnover rate and the operation safety guarantee capability are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology and intelligent logistics management technology, specifically, it relates to an intelligent distribution and inventory monitoring system for surgical consumables. Background Technology

[0002] Against the backdrop of modern medical informatization, the operating room, as the core and crucial link in hospital operations, directly impacts the overall quality of medical services and patient safety through its management efficiency, resource allocation rationality, and operational safety. Surgical consumables, as the basic material foundation of the surgical procedure, are diverse in type and value, with some having strict expiration dates. The scientific and intelligent level of their management system has become an important indicator of a hospital's refined management capabilities. Traditional medical consumable management relies heavily on manual inventory and paper-based document processing. This method is not only inefficient and costly in terms of manpower, but also prone to problems such as distorted inventory information, expired consumables, and untimely emergency requisitions due to human intervention, posing a potential threat to the efficient and orderly surgical process.

[0003] To overcome the drawbacks of purely manual management, existing technologies have widely adopted automated or semi-automated management solutions based on the Internet of Things (IoT), with high-value consumable cabinets using Radio Frequency Identification (RFID) or barcode / QR code technologies being a typical example. Specifically, such systems typically attach a unique electronic identifier (such as an RFID tag or barcode) to the individual packaging of each surgical consumable and deploy dedicated smart storage cabinets around the operating room. When consumables are received, the inventory data is initialized and updated through batch scanning. When medical staff retrieve consumables according to surgical needs, the reader / writer inside the smart cabinet automatically senses the retrieved consumable tag and deducts the corresponding inventory quantity from the backend database in real time. This technology largely achieves real-time and visual inventory information, significantly improving the accuracy and efficiency of inventory checks and effectively tracing the usage path of consumables. Compared to traditional management models, this represents a significant technological advancement. Its core inventory replenishment logic is usually based on a preset static safety stock threshold model. That is, the system pre-sets a fixed minimum inventory threshold for each type of consumable i. When the system monitors the real-time inventory of the consumable Meet the conditions When the time is right, a replenishment request is automatically triggered to the central warehouse or supplier. At a certain stage of technological development, this reactive management model based on static thresholds has indeed effectively solved the basic stockout problem caused by information delays.

[0004] However, with the rapid development of surgical techniques, the popularization of personalized and minimally invasive surgery, and the increasingly stringent cost control requirements of hospital operations, some inherent characteristics of the aforementioned technical solutions at the principle level have gradually revealed their inherent limitations in addressing new challenges. The fundamental contradiction lies in the profound mismatch between a passive replenishment logic based on static, isolated thresholds and a highly dynamic surgical demand scenario with inherent correlations and a certain degree of predictability. The reason for this is that the consumption of consumables in the operating room is not a stationary, random process conforming to a standard Poisson distribution, but rather a demand flow with significant time-varying characteristics closely related to the determined surgical schedules within a future period. Traditional static thresholds... The settings are often based solely on historical average consumption rates and a fixed safety margin. They cannot sense or respond to upcoming "demand peaks" or "demand troughs" determined by specific surgical types. For example, if multiple large joint replacement surgeries with high demand for a particular orthopedic implant are scheduled in the coming days, and the static threshold for that implant... If the safety thresholds are set too conservatively, the system will be unable to anticipate this predictable demand surge, potentially triggering replenishment alerts during peak surgical periods. However, by then, logistical delays may prevent a timely response, leading to surgical delays. Conversely, to avoid such extreme situations, administrators are typically forced to set safety thresholds for all consumables. Inventory levels are generally set at a high level, but this inevitably leads to excessive stockpiling of a large number of non-standard, high-value consumables, tying up a significant amount of working capital, increasing warehousing and management costs, and raising the risk of consumables becoming obsolete due to exceeding their expiration date. This management strategy is essentially a crude model of "trading high inventory for low risk," which cannot achieve a dynamic and precise match between inventory levels and actual surgical needs.

[0005] Furthermore, existing technological models have failed to fully exploit the rich data value contained in surgical scheduling information. Theoretically, an ideal inventory monitoring system should not have an optimal inventory level that is a constant scalar, but rather a dynamically changing function over time, relating to future time windows. Surgical plan Directly related. A more precise expected inventory level. It can be modeled as:

[0006]

[0007] in The representative is the first one derived from historical data statistics. Type of surgery for the first The expected consumption of these consumables is ∑, which is the sum of the expected consumption of all scheduled surgeries within the future cycle. This takes into account the safety stock portion, which accounts for the uncertainty of consumption. Clearly, the simple existing technology... The decision-making rule completely ignores the crucial, forward-looking predictive term in the formula. This renders the system's decision-making base completely devoid of foresight, leaving it only able to passively respond to inventory depletion events that have already occurred. This lack of a model-level framework leads to the system's "shortsightedness" in resource allocation, preventing the formation of a forward-looking and intelligent allocation strategy based on predictive demand analysis, resulting in a paradoxical dilemma where inventory backlog and potential shortage risks coexist.

[0008] Therefore, how to design a system that can deeply integrate real-time surgical scheduling data, build and apply a dynamic, surgical type-based consumable consumption prediction model to replace the traditional static threshold management, and realize intelligent, forward-looking early warning and precise automatic allocation of inventory for each type of consumable, so as to fundamentally solve the resource mismatch problem caused by the rigidity of the existing technology management model, has become a key challenge and an urgent technical problem for those skilled in the art. Summary of the Invention

[0009] The technical problem this invention aims to solve is to overcome the fundamental defects of existing surgical consumables inventory management systems, which rely on static safety stock thresholds, resulting in a lack of foresight, rigid inventory strategies, and low resource allocation efficiency. Existing technologies can only passively respond to past inventory depletion events and cannot effectively link with dynamically changing future consumable demands determined by surgical scheduling, thus leading to a contradiction between inventory backlog and potential shortage risks.

[0010] To achieve the above-mentioned objectives, this invention provides an intelligent allocation and inventory monitoring system for surgical consumables. It aims to construct and apply a consumable consumption prediction model based on multi-dimensional features by deeply integrating surgical scheduling data from the hospital information system, in order to generate a target inventory level that is dynamically adjusted over time. This will enable a shift from passive replenishment to a forward-looking and predictive resource allocation model, significantly improving the precision of consumable management and the responsiveness of the supply chain.

[0011] To achieve the above objectives, the technical solution adopted by the present invention is: an intelligent distribution and inventory monitoring system for surgical consumables, the system comprising: a data interface gateway, a central processing server, and at least one intelligent consumable storage unit that establishes a bidirectional data communication connection with the central processing server.

[0012] The data interface gateway is physically a rack-mounted server deployed within the hospital's internal network security domain. It is configured with an application programming interface (API) conforming to medical information exchange standards. Its core function is to establish a secure and persistent data link with the Hospital Information System (HIS), Operating Room Information System (ORIS), and Electronic Medical Record System (EMR). Internally, the data interface gateway contains a data extraction, transformation, and loading (ETL) protocol stack. This stack is strictly configured to actively poll and retrieve structured surgical scheduling data from the HIS and ORIS at preset time intervals (e.g., every 30 minutes). The structured data includes at least the surgical session number for a specific future time window (e.g., the scheduled surgical date and time), the surgeon's identification, the operating room number, and a standardized surgical procedure code (e.g., based on the International Classification of Diseases (ICD-9-CM-3)). Simultaneously, the data interface gateway is also authorized to access unstructured or semi-structured text data in the EMR system associated with the aforementioned surgical procedures. Specifically, this refers to the electronic surgical record filled out by the surgeon before the operation, which contains details of the surgical plan and the requirements for special instruments and consumables. The data interface gateway performs preliminary formatting processing on the acquired structured and unstructured data, encapsulates it into a unified data object, and transmits it in real time to the central processing server through a channel encrypted with Transport Layer Security (TLS 1.3).

[0013] The central processing server, serving as the computing and decision-making core of the system of this invention, is a high-performance computing server cluster deployed in a data center or a virtual server group instantiated on cloud infrastructure. The central processing server integrates a relational database management system and several collaborative core software functional modules, including: a surgical information deep analysis module, a consumable consumption baseline model library, a dynamic target inventory calculation engine, an inventory strategy and allocation instruction generation module, and a closed-loop feedback and model self-optimization module.

[0014] Furthermore, the surgical information deep analysis module integrates a natural language processing (NLP)-based entity recognition and relation extraction submodule. This submodule employs a "transformer-based bidirectional encoder representation" (BERT) language model architecture, pre-trained on biomedical text corpora (e.g., PubMed abstracts and public clinical trial reports) and fine-tuned with surgical record text. When a data object is received from the data interface gateway, the surgical information deep analysis module first directly stores the structured surgical scheduling data into the relational database of the central processing server. Subsequently, the NLP submodule processes the unstructured electronic surgical record text, performing a sequence labeling task to identify and extract predefined entity types from the text. These entity types include: non-standard consumable names, specific specifications, brand preferences, and quantity units. After entity recognition, the submodule further performs a relation extraction task, associating the identified consumable entities with the corresponding surgical session numbers to form structured "surgery-special consumable requirement" triples. Through this process, the surgical information deep analysis module transforms the raw, mixed-structure surgical information into a fully structured, machine-readable list of surgical requirements, providing complete and unambiguous data input for subsequent accurate requirement prediction.

[0015] Furthermore, the consumable consumption baseline model library is a dedicated data structure built on top of the relational database of the central processing server. This model library represents each type of managed consumable in the system. and every standardized surgical procedure Store and maintain a corresponding consumable consumption probability distribution model. This model is not merely a scalar of expected consumption value, but rather a measure of consumption quantity. A complete statistical description. As one specific implementation, this probability distribution model is parameterized as a normal distribution. , where the mean Representative techniques For consumables Historical average consumption and variance This represents the degree of uncertainty or volatility in the consumption amount. Furthermore, the model library stores a preference correction coefficient vector for each surgeon k. This vector quantifies a particular physician's relative average consumable usage habits when performing a particular surgery (e.g., (This indicates that the dosage is too high). These model parameters , The initial value was obtained through batch statistical analysis of historical surgical consumption data of the hospital, and was continuously updated by the closed-loop feedback and model self-optimization module during system operation.

[0016] Furthermore, the dynamic target inventory calculation engine is the core decision-making module of this invention. This engine operates within a configurable prediction time window. (For example, =7 days) as the cycle, for each type of consumable in inventory Calculate its current time. Dynamic target inventory level The calculation process strictly follows the following deterministic mathematical formula:

[0017]

[0018] in:

[0019] In the future time window All scheduled surgical events A set of.

[0020] and Representing surgical events The corresponding surgical procedure code and the surgeon's identification.

[0021] and These are the mean and variance of consumable consumption for this procedure, retrieved from the consumable consumption baseline model library.

[0022] These are preference correction coefficients retrieved from the model library for the surgeon in question.

[0023] It is a precise weighted sum of the total certain demand for all scheduled surgeries within the future forecast window, which constitutes the main part of the target inventory.

[0024] It is the standard deviation of the total future deterministic demand calculated based on the principle of the additivity of variances.

[0025] Z is a safety factor that directly corresponds to the preset inventory service level, and its value is determined by the inverse cumulative distribution function of the standard normal distribution. For example, to achieve a service level of 97.5% (i.e., a stockout probability of 2.5%), Z is set to 1.96.

[0026] It is a time window based on historical data statistics. Internal consumables The expected consumption of emergency or unplanned surgeries is used to cover unplanned needs.

[0027] It is an absolute minimum basic inventory used to cope with extreme situations such as supply chain disruptions. Its value is set as the amount of the consumable consumed during the average replenishment cycle.

[0028] The dynamic target inventory calculation engine executes the above formula in each calculation cycle to generate a new target inventory for each type of consumable that reflects the precise future demand. The value is a function that changes dynamically over time, rather than a fixed threshold.

[0029] Furthermore, the inventory strategy and allocation instruction generation module, based on the output of the dynamic target inventory calculation engine, formulates and executes specific inventory operations. This module continuously receives real-time inventory data from all intelligent consumable storage units. At each decision-making moment, this module performs checks on each type of consumable. Perform a comparison operation: The module is triggered if and only if this condition is met. Upon triggering, the module first calculates the required replenishment quantity. Then, it will query the central database for the in-transit inventory of that consumable. Ultimately, it generates a standardized electronic replenishment instruction. This instruction is a structured data packet (e.g., in JSON format) that explicitly includes: a unique identifier for the consumable, the calculated net replenishment quantity, and so on. The instruction includes the urgency level of the demand (based on the time elapsed since the first required surgery) and the physical location code of the target storage unit. This instruction is automatically pushed to the hospital's central warehouse management system or supply chain management (SCM) platform to initiate the physical replenishment process.

[0030] The intelligent consumable storage unit is a physical device deployed in the operating room area. Its cabinet is constructed of medical-grade 304 stainless steel and is internally divided into multiple independent storage compartments. Each storage compartment has an integrated UHF (Ultra-High Frequency) Radio Frequency Identification (RFID) reader antenna at its opening, and all antennas are connected to a central RFID reader module. This module operates in the 920MHz to 925MHz frequency band and complies with the EPCglobal UHF Class 1 Gen 2 / ISO 18000-6C communication protocol. Each surgical consumable is affixed or bound with a passive UHF RFID tag before being stored. The tag stores the consumable's Globally Unique Trade Item Number (GTIN), batch number, and expiration date. When a consumable is placed into or removed from a storage compartment, the corresponding antenna immediately senses the RFID tag's entry or exit and transmits the tag information to an industrial-grade single-board computer embedded in the cabinet. This computer runs a real-time operating system, responsible for processing RFID data and updating the inventory records stored in local non-volatile memory in real time. Meanwhile, the computer, through its integrated Ethernet or Wi-Fi module, reports every inventory change event (including operator identity, timestamp, consumable information, and change type) to the central processing server in real time, thereby ensuring the timeliness and accuracy of inventory information in the central database. The cabinet door of the storage unit is equipped with an electronic lock and access control is achieved through a user authentication terminal integrating a near-field communication (NFC) card reader and fingerprint recognition module, ensuring the traceability of operations.

[0031] As a key technical feature of this invention, the system also includes a closed-loop feedback and model self-optimization module. The function of this module is to ensure that the parameters in the consumable consumption baseline model library continuously evolve with the accumulation of practical data, thereby improving the accuracy of predictions. Its workflow is as follows: After a surgery is completed, the central processing server generates an actual consumable consumption list for that surgery session based on the precise consumable requisition records collected from the intelligent consumable storage unit during the surgery. Subsequently, the closed-loop feedback and model self-optimization module is triggered, comparing this actual consumption list with the preoperative demand predicted by the model. For each consumable i and corresponding surgical procedure j, the module uses the principle of Bayesian update to adjust the model parameters. and Specifically, if historical consumption data is modeled as observations from a normal distribution, this module utilizes new observations (i.e., the actual consumption of this surgery) and combines them with the prior distribution (i.e., the previous distribution). and The new posterior distribution parameters are calculated and used to update the model library. Similarly, by analyzing the deviation of a specific surgeon k's consumption data from the overall average over a long period, this module periodically recalibrates its preference correction coefficient vector. This continuous, automated model optimization process enables the predictive capabilities of this invention to be adaptive, dynamically capturing the long-term evolution of consumable consumption patterns caused by advancements in medical technology, changes in clinical pathways, or the development of doctors' personal habits.

[0032] This invention also provides a method for intelligent allocation and inventory monitoring of surgical consumables based on the above system, the method comprising the following steps:

[0033] Step S1: Through the data interface gateway, periodically obtain surgical scheduling data and related unstructured surgical record text covering a future preset time window from the hospital information system, operating room information system and electronic medical record system.

[0034] Step S2: Using the surgical information deep analysis module in the central processing server, the acquired data is processed, the structured data is stored in the database, and natural language processing technology is used to extract specific consumable requirements from unstructured text to form a complete and structured list of future surgical needs.

[0035] Step S3: Based on the demand list generated in step S2, the dynamic target inventory calculation engine retrieves the consumable consumption probability distribution model parameters and surgeon preference coefficients corresponding to each surgery from the consumable consumption baseline model library.

[0036] Step S4: The dynamic target inventory calculation engine executes its internally fixed deterministic mathematical formulas to calculate the dynamic target inventory level of each type of consumable i in the inventory at the current moment. .

[0037] Step S5: The inventory strategy and allocation instruction generation module obtains the real-time inventory quantity from all intelligent consumable storage units. And compare it with the dynamic target inventory level calculated in step S4. Perform a comparison item by item.

[0038] Step S6: For those that satisfy For consumables subject to certain conditions, the inventory strategy and allocation instruction generation module calculates the required replenishment quantity and generates a standardized electronic replenishment instruction that includes consumable identifier, replenishment quantity, urgency level, and target location, and automatically sends it to the upstream supply chain management system.

[0039] Step S7: After the surgery is completed, the closed-loop feedback and model self-optimization module collects the actual consumable consumption data of the surgery and updates the corresponding model parameters in the consumable consumption baseline model library based on this data, so as to realize the continuous self-optimization of the prediction model.

[0040] The beneficial effects of this invention are:

[0041] First, it achieves a fundamental shift from passive inventory management to proactive demand forecasting. This invention no longer relies on fixed, lagging safety stock thresholds, but instead generates dynamic inventory targets that precisely match future demand by directly analyzing future surgical schedules, giving inventory management unprecedented predictability.

[0042] Second, it significantly improves inventory turnover and reduces capital tied up. Because inventory levels are precisely calculated and dynamically adjusted based on actual, near-term surgical needs, it avoids general overstocking to cope with uncertain demand peaks. This is especially true for high-value, long-cycle, and non-standard consumables, which can significantly reduce inventory holding costs and the risk of expiration and disposal.

[0043] Third, it improves the reliability and security of the surgical supply chain. The system can identify and warn of potential shortages of consumables due to intensive surgical scheduling in advance, and automatically trigger a proactive replenishment process to ensure that all surgical consumables are available in advance and in sufficient quantities, fundamentally eliminating surgical delays or cancellations caused by negligence in consumable management.

[0044] Fourth, a data-driven continuous optimization closed loop has been constructed. Through closed-loop feedback and model self-optimization mechanisms, the system can continuously learn and evolve from actual consumption data. Its predictive model becomes increasingly accurate over time, demonstrating strong adaptability and ensuring the long-term effectiveness of the system in the ever-changing medical environment. Attached Figure Description

[0045] Figure 1 This is a system block diagram of an intelligent distribution and inventory monitoring system for surgical consumables according to the present invention.

[0046] The attached diagram is labeled as follows: 10, Data Interface Gateway; 20, Central Processing Server; 21, Surgical Information Deep Analysis Module; 22, Consumable Consumption Baseline Model Library; 23, Dynamic Target Inventory Calculation Engine; 24, Inventory Strategy and Allocation Instruction Generation Module; 25, Closed-Loop Feedback and Model Self-Optimization Module; 30, Intelligent Consumable Storage Unit; S1, Data Acquisition Step; S2, Data Parsing Step; S3, Model Retrieval Step; S4, Target Calculation Step; S5, Inventory Comparison Step; S6, Instruction Generation Step; S7, Model Optimization Step. Detailed Implementation

[0047] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit the scope of protection of the invention.

[0049] Reference Figure 1 The system block diagram shown illustrates an intelligent allocation and inventory monitoring system for surgical consumables. This system aims to achieve proactive and dynamic management of consumable inventory levels through in-depth analysis and modeling of future surgical scheduling data. In a specific embodiment, the system is deployed in a tertiary-level general hospital with over 2,000 beds and an annual surgical volume exceeding 50,000. The physical and logical architecture of the entire system revolves around a central processing server 20. This server interacts with the hospital's existing information infrastructure through a data interface gateway 10 and communicates in real-time with multiple intelligent consumable storage units 30 deployed in the surgical area, forming a complete closed loop of information collection, intelligent decision-making, physical execution, and data feedback.

[0050] Specifically, the data interface gateway 10 is physically manifested as a Dell PowerEdge R740 2U rack server deployed in the secure isolation zone of the hospital's internal data center. This server is equipped with two Intel Xeon Gold 6248R processors, 256GB of DDR4 error-correcting memory, and a RAID 10 array consisting of four 2TB NVMe solid-state drives, designed to provide a high-throughput and low-latency hardware foundation for data processing. Its network interface is a dual-port 10Gb SFP+ fiber optic network card, ensuring a high-speed and stable connection with the hospital's core switch. The server runs CentOS 8, on which an application programming interface (API) service based on Python 3.8 and conforming to the HL7 FHIR standard is deployed. At the core of this service is a data extraction, transformation, and loading (ETL) protocol stack, configured as a background daemon process, which is woken up and executed every 30 minutes via a system-level cron job scheduler. During execution, the protocol stack proactively initiates predefined SQL queries to the Hospital Information System (HIS) database (typically Oracle or SQL Server) to retrieve all confirmed surgical schedules for the next 14 calendar days. The retrieved data fields are strictly limited, including at least a unique identifier for the surgical session, the planned surgery date and timestamp, the surgeon's employee ID, the operating room physical number, and a standardized surgical procedure code based on the International Classification of Diseases (ICD-9-CM-3). Simultaneously, the protocol stack, through an authorized service account, accesses the file servers of the Operating Room Information System (ORIS) and Electronic Medical Record System (EMR) to obtain unstructured informed consent forms or surgical record texts associated with the aforementioned surgical sessions, completed pre-operatively by the surgeon. These texts typically contain detailed requirements for specific brands, models, or non-standard consumables. After acquiring structured and unstructured data, the data interface gateway 10 uses the Pandas data processing library for preliminary cleaning and formatting. For example, it unifies the time format from different systems to the ISO 8601 standard and encapsulates all relevant information into a unified JSON data object. Finally, this data object is pushed to the central processing server 20 through a secure channel established using the TLS 1.3 standard and the AES-256-GCM encryption suite.

[0051] The central processing server 20 is the computing and decision-making hub of the technical solution of this invention. In this embodiment, it is not a single physical server, but a hybrid computing cluster composed of three physical servers and virtual machines on a private cloud platform. The physical servers are responsible for hosting the core relational database, while the elastic and scalable virtual machines are used to run various computationally intensive application modules. The database system uses PostgreSQL 13 and is configured with a high-availability architecture of master-slave replication to ensure data security and service continuity. Around this database, the central processing server 20 deploys several collaborative core software functional modules, including a surgical information deep analysis module 21, a consumable consumption baseline model library 22, a dynamic target inventory calculation engine 23, an inventory strategy and allocation instruction generation module 24, and a closed-loop feedback and model self-optimization module 25. These modules together constitute a complete processing chain from data input to decision output.

[0052] Furthermore, the core function of the surgical information deep parsing module 21 is to transform the mixed-structure data received from the data interface gateway 10 into fully structured data that can be used for subsequent quantitative analysis. For structured surgical scheduling information, this module directly parses it and writes it into the corresponding data table in the database. Its key technical challenge lies in processing unstructured electronic surgical record text. To this end, this module integrates an entity recognition and relation extraction submodule based on natural language processing (NLP). The core technology of this submodule is a language model based on "transformer-based bidirectional encoder representation" (BERT). This model is first initialized using BERT-base-uncased weights pre-trained on biomedical literature corpora (such as PubMed abstracts), and then subjected to targeted domain-adaptive fine-tuning using more than 50,000 anonymized historical surgical record texts accumulated within the hospital. The finely tuned model is deployed as a service that provides a RESTful API through the Flask framework. Upon receiving a surgical record text, the API performs a sequence labeling task to precisely identify and extract predefined entity types from the text. These include product names of consumables (e.g., "Johnson & Johnson Ethicon absorbable sutures"), specific specifications (e.g., "4-0 Vijo"), brand preferences (e.g., "specify the use of imported brands"), and required quantities (e.g., "two sets as spares"). After entity identification, the module further performs relation extraction, firmly binding the identified consumable entities with the unique identifier of the currently processed surgical session, forming a structured "surgery-special consumable requirement" triple record, which is then stored in a dedicated table in the database. Through this process, the precise requirements previously implicit in the doctor's natural language description are made explicit and structured, providing a complete and unambiguous data foundation for subsequent inventory requirement calculations.

[0053] As the foundation of the predictive capability of this invention, the consumable consumption baseline model library 22 is not a standalone software module, but rather a set of carefully designed data tables and views built upon the relational database of the central processing server 20. This model library covers every type of consumable managed by the system. (Uniquely identified by its Global Trade Item Number (GTIN)) and each standardized surgical procedure The combination of (identified by ICD-9-CM-3 codes) maintains a corresponding consumable consumption probability distribution model. This model aims to quantitatively describe the performance of techniques. At that time, consumables are consumed quantity The statistical regularity. In a preferred embodiment, the probability distribution model is parameterized as a normal distribution. The mean here This represents the average consumption calculated based on historical data, while the variance... This represents the degree of dispersion or uncertainty of the consumption around the mean. These two parameters are stored in a core parameter table in the database. Furthermore, to personalize the predictive model for each surgeon, the model library also includes a physician preference coefficient table for each surgeon k and each type of consumable. Store a preference correction coefficient This coefficient reflects the doctor's (k) use of consumables. This refers to the systematic bias relative to the average level of all physicians; for example, a value of 1.1 indicates that the physician typically uses 10% more of that type of consumable than the average. These model parameters... , The initial values ​​were calculated through large-scale batch statistical analysis of consumable consumption data from approximately 120,000 surgeries performed in the hospital over the past three years, prior to the system's launch. These parameters are not static but are periodically updated based on new data by the closed-loop feedback and model self-optimization module 25 during system operation.

[0054] The dynamic target inventory calculation engine 23 is the core innovation of this invention, distinguishing it from traditional inventory management systems. It is a high-performance computing program written in C++, executed every 6 hours via a cron job scheduler. The engine's mission is to calculate the dynamic target inventory level that each type of consumable i in the inventory should maintain at the current time t. Here, Δt is a configurable prediction time window, set to 7 days in this embodiment. The calculation process strictly follows a built-in deterministic mathematical formula that integrates multiple dimensions, including future deterministic demand, demand uncertainty, unplanned contingency needs, and basic security reserves. The specific formula is as follows:

[0055]

[0056] In this formula, the first term It is a weighted sum of the expected consumable needs for all scheduled surgeries within the next 7 days. The engine will iterate through the surgical list for the next 7 days. For each surgery It will be based on its technique and the surgeon The corresponding average consumption is retrieved from the consumable consumption baseline model library 22. Doctor preference coefficient Multiplying these values ​​yields the personalized expected consumption of consumable i for that surgery. Finally, the expected consumption for all surgeries is summed. (Second term) This is the safety stock component, used to address the uncertainty of forecasts. It's based on the additive property of variances of independent random variables in statistics, and it adds the variance of consumption from all future surgeries. (The summation is performed using the squared values ​​of the doctor preference coefficients, and then the square root is taken to obtain the standard deviation of the future total demand. This standard deviation is multiplied by a safety factor Z, the value of which is determined by the preset inventory service level. For example, to achieve a service level of 97.5% (i.e., allowing a 2.5% stockout risk), the Z value is taken as 1.96 by consulting the inverse cumulative distribution function of the standard normal distribution. (Third term) This is the inventory level set to cover unplanned emergency surgery needs. Its value is derived by analyzing emergency surgery consumable consumption data from the past 24 months, taking the 95th percentile consumption of that consumable within a 7-day period. (Item 4) This is an absolute minimum basic inventory level used to mitigate the risk of supply chain disruptions such as delayed deliveries from suppliers. Its value is typically set at 1.5 times the amount of the consumable consumed during the average replenishment cycle. Through rigorous calculations in each calculation cycle, the engine generates a target inventory value for each consumable that dynamically changes with future surgical schedules, thereby replacing the static and unchanging safety stock threshold in traditional management models.

[0057] Subsequently, the inventory strategy and allocation instruction generation module 24 is responsible for converting the calculated dynamic target inventory level into specific physical inventory operation instructions. This module is an event-driven microservice that continuously monitors two types of events: the new target inventory value generated by the dynamic target inventory calculation engine 23, and real-time inventory change information reported from all intelligent consumable storage units 30. At each decision point, the core logic of this module is to perform a simple comparison for each type of consumable i: ,in This is the current physical inventory level obtained from the intelligent consumables storage unit. The module's replenishment logic is triggered only when the actual inventory is lower than the dynamic target inventory. After triggering, the module first calculates the theoretical replenishment demand. Next, it queries the in-transit inventory information table in the central database to obtain the amount of inventory that has been ordered but has not yet arrived. Ultimately, the module calculated the net replenishment quantity to be... This generates a standardized electronic replenishment instruction. The instruction uses JSON format, and its data structure is strictly defined, including a unique instruction ID, a generation timestamp, the consumable's GTIN code, the consumable name, the calculated net replenishment quantity, the urgency level (e.g., "high" if the consumable is needed in surgery within 48 hours), and the physical location code of the target smart consumable storage unit. An example instruction is as follows: {"instructionId": "PO-20231027-00123", "timestamp": "2023-10-27T08:00:15Z", "itemId": "GTIN-01234567890123", "itemName": "Absorbable Hemostatic GelatinSponge", "quantity": 25, "urgencyLevel": "HIGH", "destination": "OR-C-CABINET-04"}. The instruction is automatically pushed to the hospital's central warehouse management system (WMS) or supply chain management (SCM) platform via API, thereby seamlessly initiating the physical picking and distribution process.

[0058] The system's physical execution layer consists of intelligent consumable storage units 30 deployed in the operating room preparation area and each operating room. These cabinets are constructed from a single piece of 1.5 mm thick medical-grade SUS 304 stainless steel sheet, bent and welded, with a brushed finish. Their dimensions are 1800 mm (height) x 1200 mm (width) x 600 mm (depth). The interior of each cabinet is divided into multiple independent storage compartments of varying sizes to accommodate consumable packaging of different sizes. Each compartment's opening has an integrated Laird S9028PCL UHF circularly polarized RFID reader antenna. All antennas within the cabinet are connected via RF coaxial cables to a central Impinj Speedway R420 series RFID reader module. The reader operates within the 920MHz to 925MHz standard range in China and strictly adheres to the EPCglobal UHF Class 1 Gen 2 / ISO 18000-6C communication protocol. Correspondingly, all consumables managed by the system are affixed or bound with a passive UHF RFID tag by central warehouse staff before being stored. This tag uses a Higgs-9 chip from Alien Technology and is encapsulated in a biocompatible PET plastic shell. The tag's EPC storage area contains key information such as the consumable's GTIN, production batch number, and expiration date. When a tagged consumable is placed in or removed from a storage compartment by medical personnel, the antenna corresponding to that compartment immediately senses the presence or absence of the tag signal and transmits the read tag EPC code to the reader module. The cabinet houses an industrial-grade single-board computer, specifically a Raspberry Pi Compute Module 4 (4GB RAM, 32GB eMMC storage), serving as the control core. It runs a customized embedded Linux operating system based on the Yocto project. The software on this computer parses data streams from RFID readers, updates inventory records stored in local non-volatile memory in real time, and packages each inventory change event (including operator ID, precise timestamp, consumable EPC code, and change type "inbound" or "outbound") into data packets via its integrated Ethernet interface, reporting them to the central processing server 20 via the MQTT protocol. To ensure authorized and traceable operations, the cabinet door is equipped with a 12V DC electromagnetic lock with a 280kg suction force. Its operation is controlled by a user authentication terminal integrating a Korean Suprema BioMini Slim 2 fingerprint recognition module and an NXP PN532 NFC card reader. Medical personnel must swipe their work cards or verify their fingerprints to open the door and access consumables.

[0059] As a key component for achieving long-term effectiveness and adaptability, the system also integrates a closed-loop feedback and model self-optimization module 25. This module utilizes precise and real consumption data generated during daily operation to continuously calibrate and optimize the statistical model parameters in the consumable consumption baseline model library 22. Its workflow is designed as an automated background batch processing task, executed daily at midnight. When a surgery is marked as "completed" in the system, the central processing server 20 aggregates all consumable outbound records associated with the operating room from all intelligent consumable storage units 30 during the surgery (from the patient's entry to exit from the operating room), generating a precise list of actual consumable consumption down to the individual item level. The closed-loop feedback and model self-optimization module 25 is then triggered, comparing this actual consumption list with the pre-operative model-predicted demand. For each consumable i consumed in the surgery, the module uses a Bayesian update principle to adjust the consumption model parameters of procedure j for consumable i. and Specifically, this module will update the model parameters before ( and The parameters of the posterior distribution are considered as those of a normal-gamma prior distribution. The actual consumption of this surgery is used as new observational evidence. The module uses this evidence and the prior distribution to calculate the parameters of the posterior distribution using Bayes' theorem. Then, the expectation and variance of this posterior distribution are used as the new, updated parameters. and The value is recorded and written back to the database. Similarly, this module performs long-term aggregated analysis of the systematic deviation between the consumption data of a specific surgeon k when performing similar surgeries and the average consumption level of all surgeons, and periodically (e.g., quarterly) recalculates and updates its preference correction coefficient vector. This continuous, data-driven model optimization process enables the predictive capabilities of this invention to automatically adapt to long-term changes in consumable consumption patterns caused by medical technology innovation, changes in clinical pathways, or the evolution of doctors' personal operating habits, ensuring the accuracy and timeliness of the system's predictions.

[0060] Based on the above system architecture, the present invention also provides a method for intelligent allocation and inventory monitoring of surgical consumables, which strictly follows the following steps in sequence:

[0061] Step S1, Data Acquisition Step: The data interface gateway 10 actively retrieves surgical scheduling data and related surgical record texts for the next 14 days from the HIS, ORIS and EMR systems through API calls and database queries at 30-minute intervals.

[0062] Step S2, data parsing step: After receiving the data, the surgical information deep parsing module 21 in the central processing server 20 directly stores the structured data into the database and starts its NLP submodule to process the unstructured text, extracting special consumable requirements from it, and finally forming a complete, machine-readable list of future surgical requirements.

[0063] Step S3, Model Retrieval Step: At the beginning of each calculation cycle, the dynamic target inventory calculation engine 23, based on the demand list generated in step S2, iterates through each surgery within the next 7 days and retrieves the consumable consumption probability distribution model parameters corresponding to each surgery from the consumable consumption baseline model library 22. and ) and the surgeon's preference adjustment coefficient .

[0064] Step S4, Target Calculation Step: The dynamic target inventory calculation engine 23 executes its internally fixed deterministic mathematical formulas, comprehensively considering the expected demand, uncertainty, emergency demand, and basic reserves of all future surgeries, to calculate the dynamic target inventory level of each consumable i in the inventory at the current moment. .

[0065] Step S5, Inventory Comparison Step: The inventory strategy and allocation instruction generation module 24 continuously obtains the real-time aggregated inventory levels from all smart consumable storage units 30. And compare it with the dynamic target inventory level calculated in step S4. Perform item-by-item, real-time comparisons.

[0066] Step S6, Instruction Generation Step: For any condition that is met... The consumables inventory strategy and allocation instruction generation module 24 is immediately triggered. It calculates the net replenishment quantity and generates a standardized electronic replenishment instruction containing consumable identifier, replenishment quantity, urgency level and target location. This instruction is automatically pushed to the upstream central warehouse management system.

[0067] Step S7, Model Optimization Step: After the surgery is completed each day, the closed-loop feedback and model self-optimization module 25 will collect accurate and real consumable consumption data of all the surgeries completed that day, and based on this new data, use the Bayesian update method to update the corresponding model parameters in the consumable consumption baseline model library 22, thereby realizing the continuous self-learning and evolution of the prediction model.

[0068] Example 1

[0069] This embodiment aims to illustrate the operation of the system of the present invention in a specific scenario. The scenario is set as follows: at 8:00 AM on a certain Monday, the system starts to calculate the dynamic target inventory for the next 7 days (Δt=7 days) for a specific high-value consumable - "Zimmer Biomet artificial hip joint femoral stem prosthesis, model Trabecular Metal Taper, size 12" (hereinafter referred to as "femoral stem XYZ").

[0070] The system input data is as follows:

[0071] Surgical schedules obtained from HIS / ORIS show that two total hip replacement surgeries (ICD-9-CM-3 code 81.51) are planned to use this type of femoral stem within the next 7 days. Surgery A, performed by Dr. Li, is scheduled for Wednesday; Surgery B, performed by Dr. Wang, is scheduled for Friday.

[0072] The surgical information deep analysis module 21 has processed the electronic medical records of two surgeries and found no special quantity requirements for the femoral stem XYZ.

[0073] Retrieved from Consumable Consumption Baseline Model Library 22:

[0074] The model parameters for surgical procedure 81.51 and femoral stem XYZ are: average consumption. =1.0, variance =0.05 2 =0.0025.

[0075] Doctor Preference Rating: Dr. Li Li = 1.0 (standard dosage), Dr. Wang Wang = 0.9 (There is a tendency to choose a smaller size during the operation based on the actual situation, so the confirmed usage rate of this size is slightly low).

[0076] System preset parameters: Inventory service level is 97.5%, corresponding to a safety factor Z=1.96.

[0077] Historical data analysis shows the expected emergency room demand for this consumable within 7 days. =0.1 units. The replenishment cycle for this consumable is 3 days, with an average consumption of 0.4 units during this period, therefore the basic inventory... =1.5 × 0.4 = 0.6.

[0078] Real-time inventory obtained from smart consumables storage unit 30 =1.

[0079] Dynamic target inventory calculation engine 23 performs calculations:

[0080] Calculate the expected total amount of future certain demand:

[0081] indivual

[0082] Calculate the total variance of future deterministic demand:

[0083]

[0084] Calculate the total standard deviation:

[0085]

[0086] Calculate the safety stock related to uncertainty:

[0087] indivual

[0088] Calculate the final dynamic target inventory level:

[0089] =1.9 + 0.132 + 0.1 + 0.6 = 2.732 = 3

[0090] The system rounds up and sets the target inventory to 3.

[0091] Inventory strategy and allocation instruction generation module 24 execution decision:

[0092] Comparison: Real-time inventory =1, Dynamic Target Inventory =3.

[0093] If 1 < 3, replenishment is triggered.

[0094] Check in-transit inventory =0.

[0095] Calculate the net replenishment quantity: (3−1)−0=2 units.

[0096] Generate and send a replenishment instruction, requesting the central warehouse to replenish 2 femoral stem XYZ units to the lower-level warehouse.

[0097] Comparative Example 1

[0098] This comparative example uses the traditional, static threshold-based reorder point (ROP) method to manage the same scenario in Example 1.

[0099] The management parameters are set as follows:

[0100] Based on historical data from the past six months, the average weekly consumption of femoral stem XYZ is 1.5 units, which means the average daily consumption is approximately 0.21 units.

[0101] The replenishment lead time is 3 days.

[0102] Safety stock is set to a fixed value, typically to cover the average consumption over a week, i.e., SS = 1.5 units, rounded up to 2 units.

[0103] Inventory strategy calculation:

[0104] Calculate the demand during the lead time: Demand during LT = 0.21 × 3 = 0.63.

[0105] Calculate the reorder point: ROP = Demand during LT + SS = 0.63 + 2 = 2.63, rounded down to 3.

[0106] The ordering strategy is as follows: when inventory falls below 3 units, order up to the maximum inventory level. Assume the maximum inventory level (MaxStock) is set to 4 units to cover the needs of one replenishment cycle plus safety stock.

[0107] At the same time on Monday morning at 8:00 AM, the real-time inventory was 1 unit.

[0108] Decision-making process:

[0109] Comparison: Real-time inventory =1 unit, and the reorder point ROP =3 units.

[0110] If 1 < 3, replenishment is triggered.

[0111] Calculate the replenishment quantity: Order Quantity = Max Stock − =4−1=3.

[0112] The system generates an order to replenish 3 femoral stems XYZ.

[0113] Comparative analysis: The system of this invention accurately calculates that only 2 additional units are needed to meet the certain demand and various risks for the next week, while the traditional method orders 3, resulting in an excess of 1 unit of inventory buildup and capital tied up. More importantly, if no surgeries are scheduled for the next week, the dynamic target inventory of the system of this invention will be significantly reduced to [a lower value]. + =0.1 + 0.6 = 0.7 (i.e., the target is 1). Since the real-time inventory is 1, the system will not trigger any replenishment, thus avoiding unnecessary inventory holding. In contrast, the traditional ROP method is decoupled from future actual demand; it will trigger replenishment as long as the inventory is below the static threshold, lacking intelligence and precision.

[0114] Data Comparison

[0115] To further quantify the beneficial effects of the technical solution of this invention, we conducted a 6-month simulation and data comparison of the present invention (Example 1) and the traditional ROP method (Comparative Example 1) in an orthopedic surgery center of a hospital, targeting 30 types of high-value implantable consumables. The results are shown in the table below:

[0116] sheet

[0117]

[0118] Data shows that, through deep integration and dynamic calculation with surgical scheduling, this invention can significantly reduce inventory levels and capital occupation while ensuring the safety of surgical supply, greatly improve inventory turnover efficiency, and almost completely eliminate planned surgical consumable shortages caused by poor management. At the same time, it effectively reduces the risk of high-value consumables expiring and being scrapped, demonstrating outstanding economic and management benefits.

[0119] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A smart distribution and inventory monitoring system for surgical consumables, characterized in that, include: A data interface gateway (10) is configured to establish a data link with the hospital information system, the operating room information system and the electronic medical record system to periodically extract surgical scheduling data and related surgical record text covering a future preset time window; A central processing server (20) is connected in communication with the data interface gateway (10) to receive and process the surgical scheduling data and surgical record text. The core function of the central processing server (20) is to generate a dynamic target inventory level that is dynamically adjusted over time and matches the future surgical needs for each type of consumable managed by the system based on the data, and to formulate an inventory allocation strategy based on the dynamic target inventory level. as well as At least one intelligent consumable storage unit (30) establishes a bidirectional data communication connection with the central processing server (20). The intelligent consumable storage unit (30) is configured to monitor the physical inventory of various consumables stored inside it in real time and report the real-time inventory data to the central processing server (20) so that the central processing server (20) can compare the real-time inventory with the dynamic target inventory level and trigger an inventory allocation operation.

2. The system according to claim 1, characterized in that, The central processing server (20) integrates multiple core software functional modules that work together, including: A surgical information deep parsing module (21) is used to transform the hybrid structured data obtained from the data interface gateway (10) into a fully structured surgical requirement list; A consumable consumption baseline model library (22) is used to store and maintain the corresponding historical consumption statistical model parameters for each combination of consumable and surgical procedure; A dynamic target inventory calculation engine (23) is used to calculate the dynamic target inventory level based on the structured surgical demand list and by calling the parameters in the consumable consumption baseline model library (22); An inventory strategy and allocation instruction generation module (24) is used to compare the real-time inventory level with the dynamic target inventory level and generate standardized electronic replenishment instructions when preset conditions are met; and A closed-loop feedback and model self-optimization module (25) is used to continuously update and optimize the model parameters in the consumable consumption baseline model library (22) based on the actual consumable consumption data after the operation.

3. The system according to claim 2, characterized in that, The surgical information deep analysis module (21) integrates an entity recognition and relation extraction submodule based on natural language processing. This submodule adopts a "transformer-based bidirectional encoder representation" (BERT) language model architecture that has been pre-trained on a biomedical text corpus and fine-tuned on surgical record text. It is configured to perform sequence labeling tasks on the surgical record text to identify and extract predefined entity types from the text. These entity types include at least non-standard consumable names, specific specifications, brand preferences, and quantity units. After completing entity recognition, the submodule further performs relation extraction tasks to associate the identified consumable entities with the corresponding surgical session number, thereby transforming the original unstructured text information into a structured "surgery-special consumable requirements" record and incorporating it into the surgical requirements list.

4. The system according to claim 2, characterized in that, The consumable consumption baseline model library (22) is a dedicated data structure built on top of a relational database within the central processing server (20); this model library is for each type of managed consumable in the system. and every standardized surgical procedure Store and maintain a corresponding consumable consumption probability distribution model. The probability distribution model is parameterized as a normal distribution. , where the mean Representative techniques For consumables Historical average consumption, variance This represents the degree of uncertainty or volatility in the consumption amount; the model library also stores a preference correction coefficient vector for each surgeon k. This vector is used to quantify a particular doctor's consumable usage habits relative to the average level when performing surgery.

5. The system according to claim 2, 3 or 4, characterized in that, The dynamic target inventory calculation engine (23) calculates the dynamic target inventory level of each consumable i in the inventory at the current time t by executing the following deterministic mathematical formula, with a configurable prediction time window Δt as the period. : in: For future time windows The set of all scheduled surgical events s within the range; and Representing surgical events The corresponding surgical procedure code and the surgeon's identification; , The mean, variance, and preference correction coefficient of consumable consumption for the surgical procedure and the surgeon are retrieved from the consumable consumption baseline model library (22). First item This constitutes a weighted sum of the total deterministic demand for all scheduled surgeries within the future forecast window; Second item The safety stock is configured to cope with demand uncertainty, where Z is the safety factor corresponding to the preset inventory service level; Third item It is based on historical data statistics and is presented in a time window. Internal consumables Expected consumption of emergency or unplanned surgeries; Fourth item It is the absolute minimum basic inventory set to cope with the risk of supply chain disruptions.

6. The system according to claim 2, characterized in that, The inventory strategy and allocation instruction generation module (24) is configured to continuously receive real-time inventory data from all smart consumable storage units (30). At each decision point, this module performs a comparison operation for each consumable i: if and only if When the conditions are met, this module is triggered; after being triggered, the module calculates the required replenishment quantity. And check the in-transit inventory of the consumable. This ultimately generates a standardized electronic replenishment instruction, whose data content explicitly includes the unique identifier of the consumable and the calculated net replenishment quantity. The system automatically pushes the urgency level of the demand and the physical location code of the target storage unit to the upstream warehouse management system or supply chain management platform.

7. The system according to claim 2, characterized in that, The workflow of the closed-loop feedback and model self-optimization module (25) includes: after a surgery is completed, the central processing server (20) generates an actual consumable consumption list for the surgery based on the accurate consumable requisition records collected from the intelligent consumable storage unit (30) during the surgery; subsequently, the closed-loop feedback and model self-optimization module (25) compares the actual consumption list with the preoperative model-based predicted demand, and performs a specific analysis on each type of consumable consumed during the surgery. and corresponding surgical procedures The method employs the Bayesian update principle, using the actual consumption of this surgery as the new observation value, and combining it with the model parameters before the update. and As a prior distribution, a new posterior distribution parameter is calculated, and the corresponding model parameter in the consumable consumption baseline model library (22) is updated with the posterior distribution parameter; at the same time, this module also periodically recalibrates its preference correction coefficient vector by long-term aggregation analysis of the deviation between the consumption data of a specific surgeon k and the overall average level. .

8. The system according to claim 1, characterized in that, The data interface gateway (10) has a data extraction, conversion and loading protocol stack embedded inside. The protocol stack is configured to actively poll and pull structured surgical scheduling data from the hospital information system and operating room information system at a preset time period. The structured data includes at least the surgical session number, the scheduled surgical date and time, the surgeon's identity, the operating room number and the standardized surgical procedure code within a specific future time window. At the same time, the data interface gateway (10) is also authorized to access unstructured or semi-structured text data in the electronic medical record system associated with the surgical session. The data interface gateway (10) performs preliminary formatting on the acquired structured and unstructured data, encapsulates it into a unified data object, and transmits it to the central processing server (20) through a channel encrypted based on the transport layer security protocol.

9. The system according to claim 1, characterized in that, The cabinet of the intelligent consumable storage unit (30) is divided into multiple independent storage compartments. Each storage compartment has an integrated ultra-high frequency radio frequency identification (RFID) reader antenna at its opening. All antennas are connected to a central RFID reader module. Each surgical consumable is affixed or bound with a passive ultra-high frequency RFID tag before being stored. The tag stores the unique identification information of the consumable. When a tagged consumable is placed into or removed from a storage compartment, the corresponding antenna immediately senses the entry or exit of the RFID tag and transmits the tag information to the industrial-grade single-board computer embedded in the cabinet. This computer is responsible for updating the local inventory records in real time and reporting each inventory change event, including the operator's identity, timestamp, consumable information, and change type, to the central processing server (20) in real time. The cabinet door of the intelligent consumable storage unit (30) is equipped with an electronic lock and access control is achieved through a user authentication terminal that integrates a near field communication (NFC) card reader and a fingerprint recognition module.

10. A method for intelligent allocation and inventory monitoring of surgical consumables, characterized in that, Includes the following steps: Data acquisition step (S1): Through a data interface gateway (10), periodically acquire surgical scheduling data and related unstructured surgical record text covering a future preset time window from the hospital information system, operating room information system and electronic medical record system; Data parsing step (S2): The surgical information deep parsing module (21) in a central processing server (20) is used to process the acquired data, store the structured data in the database, and use natural language processing technology to extract special consumable requirements from unstructured text to form a complete and structured list of future surgical requirements. Model retrieval and target calculation steps (S3, S4): The dynamic target inventory calculation engine (23) in the central processing server (20) retrieves the consumable consumption probability distribution model parameters and surgeon preference coefficients corresponding to each surgery from a consumable consumption baseline model library (22) based on the demand list, and executes the internally fixed deterministic mathematical formula to calculate the dynamic target inventory level of each consumable in the inventory at the current moment. Inventory comparison and instruction generation steps (S5, S6): The inventory strategy and allocation instruction generation module (24) in the central processing server (20) obtains the real-time inventory from all smart consumable storage units (30) and compares it with the dynamic target inventory level item by item. For consumables that meet the condition that the actual inventory is lower than the target inventory, the required replenishment quantity is calculated and a standardized electronic replenishment instruction is generated and automatically sent to the upstream supply chain management system. as well as Model optimization step (S7): After the operation is completed, the closed-loop feedback and model self-optimization module (25) in the central processing server (20) collects the actual consumable consumption data of the operation and updates the corresponding model parameters in the consumable consumption baseline model library (22) based on this data, so as to realize the continuous self-optimization of the prediction model.

Images