Medical material intelligent supply chain management scheme and system based on multi-mode AI

Abstract

The invention belongs to the technical field of medical material management, and particularly relates to a medical material intelligent supply chain management scheme and system based on multi-modal AI. According to the invention, through multi-modal data fusion, full-life-cycle and full-link tracking of materials is realized, information islands are eliminated, management transparency is improved, unexpected consumption and physical damage risks are identified in an early stage, a large amount of overstock or clinical supply interruption is avoided, operation cost is reduced, and based on a dynamic loss threshold, replenishment and cross-warehouse allocation can be realized just in need. The inventory turnover rate and the supply guarantee rate are considered, environment monitoring and validity period information are processed, it is ensured that temperature and humidity sensitive materials are stored under the optimal condition, losses caused by expiration or environment abnormity are reduced, the waste rate and the excessive stockpiling risk are comprehensively reduced, meanwhile, timely availability of key materials is clinically guaranteed, and the service life of the materials is prolonged. And the economic benefit and safety of the medical material supply chain are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical material management, and particularly relates to a medical material intelligent supply chain management scheme and system based on multi-modal AI. BACKGROUND

[0002] With the continuous improvement of the efficiency and safety of the supply chain in the medical industry, hospitals and medical institutions have higher requirements for material management. In a typical medical material supply chain, the whole process of materials from procurement, warehousing, distribution to use and scrap involves multiple links and multiple data types, including physical properties, batch information, procurement and use records, warehouse environment monitoring data, etc. The current medical material supply chain has core pain points such as process fragmentation, high inventory cost, high regulatory risk, and low coordination efficiency. Traditional management relies on manual operation, resulting in consumable waste, long turnover cycle, and difficulty in meeting the DRG / DIP medical insurance cost control requirements. Although existing technical solutions such as RFID intelligent cabinets and SPD systems achieve single-point digitization, the hardware data is isolated, the system compatibility is poor, and there is a lack of multi-source perception fusion and dynamic decision-making capability, which cannot support whole-chain lean management.

[0003] The existing system mainly relies on manual inventory or simple inventory difference alarms, and it is difficult to quantify and warn in time about unexpected consumption and physical damage of materials. Often, problems can only be found after inventory or loss reporting, which has caused resource waste and safety hazards. There is a lack of a unified multi-source data fusion platform, resulting in poor end-to-end traceability of the whole life cycle of materials. Many hospitals still use fixed replenishment cycles or quantitative replenishment methods based on historical average usage, which cannot dynamically combine current consumption, forecast demand, and material status for intelligent replenishment scheduling, easily leading to overstocking or stockout situations, increasing operating costs and affecting clinical safety. SUMMARY

[0004] The purpose of the present application is to provide a medical material intelligent supply chain management scheme and system based on multi-modal AI, which can realize accurate quantification and real-time warning of material consumption through deep fusion and intelligent analysis of multi-source material information, generate adaptive replenishment and maintenance strategies, and improve safety and operational efficiency.

[0005] The technical solutions adopted by the present application are as follows: A medical material intelligent supply chain management scheme based on multi-modal AI, comprising: Obtaining physical characteristic data, identity data and business flow data of medical materials, and generating material whole life cycle information; Obtaining clinical demand information and inventory state data, and generating current material consumption; acquiring the material taking weight and the material billing record based on the material whole life cycle information, and generating the material unexpected consumption based on the material taking weight and the material billing record; acquiring the material storage state image data based on the material whole life cycle information, and acquiring the material physical damage risk based on the material storage state image data; acquiring the material effective period data and the material storage environment information based on the material whole life cycle information, and predicting the material timeliness loss based on the material effective period data and the material storage environment information; acquiring the material comprehensive loss based on the material current loss, the material unexpected consumption, the material physical damage risk and the material timeliness loss, and acquiring the dynamic loss threshold based on the historical consumption data, determining whether the material comprehensive loss exceeds the dynamic loss threshold, and executing the corresponding management strategy based on the determination result.

[0006] In a preferred solution, the step of acquiring the physical feature data, the identity data and the business flow data of the medical material, and generating the material whole life cycle information, comprises: acquiring the physical feature data of the medical material, wherein the physical feature data comprises weight, volume and appearance; acquiring the identity data of the medical material, wherein the identity data comprises material category and material effective period data; acquiring the business flow data of the medical material, wherein the business flow data comprises purchase order, material billing record and department consumption record; associating and integrating the physical feature data, the identity data and the business flow data of each medical material to generate the material whole life cycle information.

[0007] In a preferred solution, the step of acquiring the clinical demand information and the inventory state data, and generating the material current loss, comprises: acquiring the clinical demand information and the inventory state data; acquiring the corresponding clinical demand value based on the clinical demand information; acquiring the previous inventory value and the current inventory value based on the inventory state data; acquiring the material current loss based on the clinical demand value, the previous inventory value and the current inventory value.

[0008] In a preferred solution, the step of acquiring the material taking weight and the material billing record based on the material whole life cycle information, and generating the material unexpected consumption based on the material taking weight and the material billing record, comprises: acquiring the material taking weight and the material billing record based on the material whole life cycle information; acquiring the material billing taking weight based on the material billing record; The taking weight of the material is based on the material charging and the taking weight of the material, and the taking weight is obtained to obtain the deviation weight, and is marked as the unexpected consumption of the material.

[0009] In a preferred embodiment, the step of obtaining the material storage state image data based on the material life cycle information, and obtaining the physical damage risk of the material based on the material storage state image data, comprises: obtaining the material storage state image data based on the material life cycle information; obtaining the material storage image of multiple perspectives based on the material storage state image data, and constructing a plane rectangular coordinate system in the material storage image of each perspective; obtaining the multiple material edge profile inflection point coordinates in the material storage image of each perspective based on the plane rectangular coordinate system; obtaining the material shape image standard area of each perspective; generating the material storage deformation based on the multiple material edge profile inflection point coordinates in the material storage image of each perspective and the corresponding material shape image standard area, and marking as the physical damage risk.

[0010] In a preferred embodiment, the step of obtaining the material effective period data and the material storage environment information based on the material life cycle information, and predicting the material timeliness loss based on the material effective period data and the material storage environment information, comprises: obtaining the material effective period data and the material storage environment information based on the material life cycle information; obtaining the standard temperature effective time and the stored time of the medical material based on the material effective period data; obtaining the standard storage temperature, the actual average storage temperature and the temperature coefficient based on the material storage environment information; obtaining the material timeliness loss of the medical material based on the standard temperature storage effective time, the stored time, the standard storage temperature, the actual average storage temperature and the temperature coefficient.

[0011] In a preferred embodiment, the step of obtaining the material comprehensive loss based on the material current loss, the unexpected consumption of the material, the physical damage risk of the material and the material timeliness loss, obtaining the dynamic loss threshold based on the historical consumption data, determining whether the material comprehensive loss exceeds the dynamic loss threshold, and executing the corresponding management strategy based on the determination result, comprises: obtaining the material weight, wherein the material weight comprises the current loss weight, the unexpected weight of the material, the physical damage weight of the material and the timeliness weight of the material; obtaining the material comprehensive loss based on the material current loss, the unexpected consumption of the material, the physical damage risk of the material, the material timeliness loss and the material weight; obtaining the dynamic loss threshold based on the historical consumption data; determining whether the material comprehensive loss exceeds the dynamic loss threshold; If the comprehensive consumption of the material does not exceed the dynamic consumption threshold, it is determined that the material consumption is normal, and a replenishment strategy table is obtained, wherein the replenishment strategy table includes a plurality of material comprehensive consumption intervals and a material replenishment strategy corresponding to each material comprehensive consumption interval, a corresponding material replenishment strategy is obtained from the replenishment strategy table based on the material comprehensive consumption corresponding to the material comprehensive consumption interval, and medical material replenishment is carried out based on the material capture strategy; If the comprehensive consumption of the material exceeds the dynamic consumption threshold, it is determined that the material consumption is abnormal, a consumption exceeding value of the material comprehensive consumption exceeding the dynamic consumption threshold is obtained, and a management strategy table is obtained, wherein the management strategy table includes a plurality of consumption exceeding value intervals and an alarm strategy and an alarm replenishment strategy corresponding to each consumption exceeding value interval, a corresponding alarm maintenance strategy and an alarm replenishment strategy are obtained from the management strategy table based on the consumption exceeding value interval corresponding to the consumption exceeding value, the medical material is processed based on the alarm maintenance strategy, and the medical material replenishment is carried out based on the alarm replenishment strategy.

[0012] In a preferred scheme, the step of obtaining the dynamic consumption threshold based on the historical consumption data comprises: obtaining a standard consumption threshold and a standard consumption value; obtaining a plurality of historical consumption values corresponding to the historical consumption data; obtaining the dynamic consumption threshold based on the standard consumption threshold, the standard consumption value and the plurality of historical consumption values.

[0013] The application also provides a medical material intelligent supply chain management system based on multi-modal AI, which is used for the above-mentioned medical material intelligent supply chain management scheme based on multi-modal AI, and comprises: A life cycle module is configured to obtain physical characteristic data, identity data and business flow data of the medical material, and generate material full life cycle information. A current consumption module is configured to obtain clinical demand information and inventory state data, and generate current consumption of the material. An unexpected consumption module is configured to obtain material taking weight and material billing records based on the material full life cycle information, and generate unexpected consumption of the material based on the material taking weight and the material billing records. A physical damage module is configured to obtain material storage state image data based on the material full life cycle information, and obtain material physical damage risk based on the material storage state image data. A material shelf life module is configured to obtain material expiration date data and material storage environment information based on the material full life cycle information, and predict material shelf life consumption based on the material expiration date data and the material storage environment information. The material management module obtains comprehensive consumption of the material based on current consumption of the material, unexpected consumption of the material, physical damage risk of the material and time-sensitive consumption of the material, obtains a dynamic consumption threshold based on historical consumption data, determines whether the comprehensive consumption of the material exceeds the dynamic consumption threshold, and executes a corresponding management strategy based on the determination result.

[0014] Also, a medical material intelligent supply chain management terminal based on multi-modal AI includes: One or more processors; A storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement a medical material intelligent supply chain management scheme based on multi-modal AI.

[0015] The technical effects achieved by the present application are: The present application, in. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The present application provides a method flowchart; Figure 2 The present application provides a system module diagram. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0018] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0019] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In a preferred embodiment" appearing in different places in this specification does not mean the same embodiment, nor is it an independent or alternative embodiment.

[0020] Thirdly, the present application is described in detail in conjunction with the schematic diagram, and in the detailed description of the embodiments of the present application, the schematic diagram is only an example and should not limit the scope of protection of the present application.

[0021] Please refer to the accompanying Figure 1 As shown in the accompanying drawings, a medical material intelligent supply chain management scheme based on multi-modal AI is provided, including: S1, acquire physical feature data, identity data and business flow data of medical supplies, and generate whole life cycle information of the supplies; S2, acquire clinical demand information and inventory state data, and generate current consumption of the supplies; S3, acquire the weight of the supplies taken and the charging record of the supplies based on the whole life cycle information of the supplies, and generate unexpected consumption of the supplies based on the weight of the supplies taken and the charging record of the supplies; S4, acquire storage state image data of the supplies based on the whole life cycle information of the supplies, and acquire physical damage risk of the supplies based on the storage state image data of the supplies; S5, acquire expiration date data and storage environment information of the supplies based on the whole life cycle information of the supplies, and predict time-sensitive consumption of the supplies based on the expiration date data and the storage environment information of the supplies; S6, acquire comprehensive consumption of the supplies based on the current consumption of the supplies, the unexpected consumption of the supplies, the physical damage risk of the supplies and the time-sensitive consumption of the supplies, acquire a dynamic consumption threshold based on historical consumption data, determine whether the comprehensive consumption of the supplies exceeds the dynamic consumption threshold, and execute corresponding management strategies based on the determination result.

[0022] As in the above steps S1 to S6, the physical characteristics (such as weight, volume and appearance), identity (such as batch number, production date and material category) and business flow data (such as purchase order, material billing record and department consumption record) of the materials are collected by using radio frequency identification (RFID), bar code, sensor and other technologies, and are fused to build an electronic file (material full life cycle information) of each material, to realize end-to-end traceability from production, transportation to final disposal. Based on the latest clinical demand (such as doctor / nurse initiated application) and real-time inventory status (such as inventory checking data of the warehouse system), the current consumption of the material is calculated, the normal consumption level is reflected, the "taking weight" and "billing record" (such as the weight corresponding to the actual taking weight and billing amount) are extracted from the full life cycle information, the unexpected consumption index is calculated, the storage state is photographed regularly or in real time, combined with visual AI detection (such as object detection and defect recognition network based on deep learning), hidden dangers such as packaging damage, deformation and leakage are identified, and physical damage risk is output. According to the material expiration date data (production / expiry date) and warehouse environment monitoring information (temperature), the timeliness consumption caused by expiration and environmental abnormalities is predicted. The current consumption, unexpected consumption, physical damage risk and timeliness consumption are fused according to the weight to obtain the comprehensive consumption index of the material. The threshold is dynamically calculated based on the historical consumption fluctuation data, and it is judged whether the comprehensive consumption is beyond the normal range. For the case beyond the threshold, the warning or management strategy is triggered: such as automatic replenishment, inventory review, storage adjustment, environment adjustment or waste loss, etc. Through multi-modal data fusion, the full life cycle and full link tracking of the material is realized, the information silos are eliminated, the management transparency is improved, the unexpected consumption and physical damage risk are identified early, the large accumulation or clinical supply interruption is avoided, the operation cost is reduced, the dynamic consumption threshold is used to realize the demand replenishment and cross-warehouse allocation, the inventory turnover rate and supply guarantee rate are considered, the environment monitoring and expiration date information are included in the processing, the storage of temperature and humidity sensitive materials under the best conditions is ensured, the loss caused by expiration or environmental abnormalities is reduced, the waste rate and overstocking risk are comprehensively reduced, and the timely availability of key materials for clinical use is ensured, the economic benefit and safety of the medical material supply chain are improved.

[0023] In a preferred embodiment, the step of acquiring physical characteristic data, identity data and business flow data of the medical material, and generating material full life cycle information, comprises: S101, acquiring physical characteristic data of the medical material, wherein the physical characteristic data includes weight, volume and appearance; S102, acquiring identity data of the medical material, wherein the identity data includes material category and material expiration date data; S103, acquiring business flow data of the medical material, wherein the business flow data includes purchase order, material billing record and department consumption record; S104, the physical characteristic data, identity data and business flow data of each medical material are associated and integrated to generate material life cycle information.

[0024] As in the above steps S101 to S104, each piece of medical material is weighed, volume measured and appearance photographed by using embedded sensors (such as electronic scales, volume measuring instruments, 3D scanners or cameras), and the collected weight, length x width x height and image information are uploaded to the material management platform in real time. In the warehouse or intelligent storage cabinet link, the material is attached or injected with a unique identifier (RFID tag, barcode or two-dimensional code), records the category code, production batch, production / expiry date and other key information, queries and verifies the identifier in combination with the enterprise resource planning (ERP) system or barcode management system, ensures that the material category, specification and shelf life are accurate, obtains the purchase order information from the procurement system, including the supplier, order quantity and expected arrival time, obtains the billing records corresponding to the material warehouse in and out from the financial and billing systems to reflect the actual use cost, and grabs the material consumption log of each department from the hospital information system (HIS) or department management system, including the use time, amount and using department. The above three types of data are indexed by the unique identifier of the material and stored in the middle database according to the time line, a complete track from warehouse entry, inspection, distribution, use, loss reporting and recovery is constructed, a graph database or time series database model is used to support multi-dimensional query and traceability according to the material, batch or department dimensions, and the data is cleaned, de-duplicated and synchronized to ensure information consistency and real-time performance. The three-dimensional holographic record of the identity, physics and business of each piece of medical material can be achieved, the source, shelf life and use of the material can be traced at any time, the traceability and compliance are greatly improved, the multi-source data is uniformly associated and managed in the same platform, the information silos and manual input errors are eliminated, the management decisions are based on reliable data, the whole process of the material from warehouse entry to use is monitored, the wrong and missing use and expired loss are reduced, the operation cost is reduced, and the clinical use and drug safety are ensured.

[0025] In a preferred embodiment, the step of obtaining clinical demand information and inventory status data and generating the current consumption of the material includes: S201, obtaining clinical demand information and inventory status data; S202, obtaining the corresponding clinical demand value based on the clinical demand information; S203, obtaining the corresponding last period inventory value and current inventory value based on the inventory status data; S204, obtaining the current consumption of the material based on the clinical demand value, the last period inventory value and the current inventory value.

[0026] As in the above steps S201 to S204, the clinical demand information (from the department application sheet or electronic medical order, pushed by the hospital information system (HIS) in real time) and the inventory state data (obtained through the inventory checking system or the intelligent storage cabinet, including the last period (the end of the previous statistical period) and the current inventory) are captured in real time or at a fixed time, the clinical demand information is summarized, the number of material application and the number of approved applications of each department in the same statistical period are counted, and a clinical demand value is aggregated to reflect the real clinical rigid demand. The last period inventory value (taking the inventory snapshot at the end of the previous statistical period) and the current inventory value (taking the real-time inventory at the current time or at the end of the period) are obtained from the inventory database or the storage cabinet database. The clinical demand value, the last period inventory value and the current inventory value are comprehensively calculated to obtain the current consumption of materials, and the calculation formula of the current consumption of materials is , wherein U represents the current consumption of materials, L represents the clinical demand value, represents the last period inventory value, represents the current inventory value. The clinical demand and the inventory change are integrated to realize multi-dimensional quantification of material consumption, avoid deviation caused by single data, provide reliable basis for replenishment, allocation and inventory optimization, accurately monitor the consumption rate, help to reduce operating costs and waste, and at the same time guarantee the continuous supply of key materials in clinical practice.

[0027] In a preferred embodiment, the step of obtaining the material taking weight and the material billing record based on the material full life cycle information, and generating the unexpected consumption of materials based on the material taking weight and the material billing record, comprises: S301, obtaining the material taking weight and the material billing record based on the material full life cycle information; S302, obtaining the material billing taking weight based on the material billing record; S303, obtaining the taking deviation weight based on the material billing taking weight and the material taking weight, and marking it as the unexpected consumption of materials.

[0028] As in steps S301 to S303 described above, the complete life cycle record of the material is retrieved in the middle platform database or graph database indexed by the unique identification of the material, the material taking weight (recorded by the sensor (such as an electronic scale) or smart cart scanning of the lower computer, real-time record of the weight change at each time of taking), and the material billing record (extracted from the financial system or billing subsystem, including the billing time, billing unit price, and billing weight). The billing taking weight is extracted from the billing record. If the same batch of material generates multiple billing in a statistical period, the billing taking weights are accumulated to obtain the total billing taking weight of the period. Combined with the actual taking weight reported by the sensor, the difference between the material billing taking weight and the material taking weight is calculated, which is the taking deviation weight. The cross comparison of the taking weight and the billing weight can timely find abnormality such as missing billing (taking more than billing) or overbilling (billing more than taking), enhance the consistency of accounts and reality, quickly locate the specific taking time, operator or system interface point for the deviated link, support post-event auditing and responsibility identification, and include non-expected consumption into key monitoring indicators to help managers evaluate the hidden loss caused by improper operation, missing accounting, or theft and waste.

[0029] In a preferred embodiment, the step of obtaining the material storage state image data based on the material full life cycle information and obtaining the material physical damage risk based on the material storage state image data comprises: S401, obtaining material storage state image data based on material full life cycle information; S402, obtaining a plurality of perspective material storage images based on the material storage state image data, and constructing a plane rectangular coordinate system in each perspective material storage image; S403, obtaining a plurality of material edge profile inflection point coordinates in each perspective material storage image based on the plane rectangular coordinate system; S404, obtaining a material shape image standard area of each perspective; S405, generating material storage deformation based on the plurality of material edge profile inflection point coordinates in each perspective material storage image and the corresponding material shape image standard area, and marking as a physical damage risk.

[0030] As in the above steps S401 to S405, with the unique identification of the material as the index, the high-definition images of the material in each storage link (such as warehousing, shelving, allocation, inventory) are pulled from the material management platform or image repository, ensuring coverage of the storage state at different time points and scenes. For each set of storage scene images, select at least two or more mutually orthogonal or semi-orthogonal perspectives (such as front, side, top view), and on each image, construct a planar rectangular coordinate system (coordinate origin and unit length determined by calibration objects or checkerboard calibration board) according to the shooting device calibration results, so that subsequent deformation measurement has spatial consistency. Using image processing algorithms (such as Canny operator + Hough transform or deep learning-based instance segmentation network), accurately detect the outer contour of each piece of material in the coordinate system, and extract several key corner coordinates to form a polygonal contour representation of the material at that perspective. Corresponding to the ideal shape or undamaged state of the material, the standard contour area is predefined and stored in the same coordinate system (i.e. the projected area of the material when it is perfectly packaged and correctly placed). The material storage deformation is calculated by integrating the multiple material shape image area coordinates in the material storage images of each perspective and the corresponding material shape image standard area, and is marked as a physical damage risk. The calculation formula of the material storage deformation is , where P represents the material storage deformation, k represents the number of material shape image areas in the material storage images of multiple perspectives, k = 1, 2, 3…t, represents the material shape image area in the kth perspective of the material storage image, represents the material shape image standard area corresponding to the kth perspective of the material storage image, where the calculation of the material shape image area can be obtained using the area formula, for example, the shoelace theorem. Taking the shoelace theorem as an example, the calculation formula of the material shape image area is , where represents the material shape image area, h represents the number of multiple material edge contour corner coordinates in the material storage image of each perspective, h = 1, 2, 3…m, represents the x-axis coordinate point of the hth material edge contour corner in the material storage image of each perspective, represents the x-axis coordinate point of the h+1th material edge contour corner in the material storage image of each perspective, represents the y-axis coordinate point of the hth material edge contour corner in the material storage image of each perspective, and The y-axis coordinate point of the h+1th material edge profile inflection point in the material storage image of each view angle is represented, and when h takes the value of m, m+1 is represented as 1. The multi-view and coordinate system-based profile inflection point extraction and area comparison avoid misjudgment caused by occlusion or perspective distortion under a single view angle. The whole process can be automatically executed in an intelligent warehouse system, real-time monitoring of the material packaging and stacking state is not necessary for manual visual inspection, and the efficiency is significantly improved. Through deformation quantization monitoring, timely alarm can be given when the material packaging integrity is damaged at the initial stage, i.e., when a small indentation or deformation just appears, to avoid material loss, quality risk or clinical accidents caused by serious damage.

[0031] In a preferred embodiment, the step of obtaining material validity period data and material storage environment information based on material life cycle information, predicting material timeliness loss based on material validity period data and material storage environment information, comprises: S501, obtaining material validity period data and material storage environment information based on material life cycle information; S502, obtaining standard temperature validity time and stored time of the medical material based on the material validity period data; S503, obtaining standard storage temperature, actual average storage temperature and temperature coefficient based on the material storage environment information; S504, obtaining material timeliness loss of the medical material based on standard temperature storage validity time, stored time, standard storage temperature, actual average storage temperature and temperature coefficient.

[0032] In the above steps S501 to S504, the validity period data (production / expiry date, used to calculate the shelf life length under standard storage) and storage environment information (temperature monitoring records of the warehouse or transportation vehicle, including temperature values at each time point) are pulled from the life cycle database with the material unique identifier as the index. A standard temperature validity time (maximum number of days available under standard storage temperature (e.g., 4°C) specified by the manufacturer or regulatory agency) is pre-set in the database, and the stored time (cumulative storage days from the date of material storage or last review) is extracted. The standard storage temperature is extracted from the pre-set database, and the temperature mean value in the monitored time period, i.e., the actual average storage temperature, is obtained. The storage temperature coefficient of the medical material (usually taking a value of 1.5-2.5, used to describe the reaction acceleration multiple when the temperature increases by 10°C) is extracted from the pre-set database. The standard temperature storage validity time, stored time, standard storage temperature, actual average storage temperature and temperature coefficient are comprehensively calculated to obtain the material timeliness loss of the medical material. The calculation formula of the material timeliness loss is , wherein represents the material timeliness loss, represents the standard temperature storage validity time, is expressed as the stored time, is expressed as the actual average storage temperature, is expressed as the standard storage temperature, Q is expressed as the temperature coefficient, is expressed as the actual remaining storage effective time, the accelerated influence of the actual temperature deviation on the stability of chemical and biological products is quantified as an equivalent time through the temperature coefficient, the inaccuracy of simply evaluating the expiration risk according to the calendar days is avoided, the consumption ratio can be continuously updated according to real-time environmental data, the materials that will exceed the standard are identified in advance, the clinical supply interruption and expiration loss are reduced, the equivalent storage days and the consumption ratio report of each batch of materials can be output, the regulatory compliance and audit requirements are met, and the environment and consumption track can be accurately reproduced when a complaint occurs.

[0033] In a preferred embodiment, the step of obtaining the comprehensive consumption of the materials based on the current consumption of the materials, the unexpected consumption of the materials, the physical damage risk of the materials and the timeliness consumption of the materials, and obtaining the dynamic consumption threshold based on the historical consumption data, determining whether the comprehensive consumption of the materials exceeds the dynamic consumption threshold, and executing the corresponding management strategy based on the determination result, includes: S601, obtaining the material weight, wherein the material weight includes the current consumption weight, the unexpected weight of the materials, the physical damage weight of the materials and the timeliness weight of the materials; S602, obtaining the comprehensive consumption of the materials based on the current consumption of the materials, the unexpected consumption of the materials, the physical damage risk of the materials, the timeliness consumption of the materials and the weight of the materials; S603, obtaining the dynamic consumption threshold based on the historical consumption data; S604, determining whether the comprehensive consumption of the materials exceeds the dynamic consumption threshold; If the comprehensive consumption of the materials does not exceed the dynamic consumption threshold, it is determined that the consumption of the materials is normal, a restocking strategy table is obtained, wherein the restocking strategy table includes a plurality of comprehensive consumption intervals of the materials and a corresponding restocking strategy of the materials for each comprehensive consumption interval of the materials, a corresponding restocking strategy of the materials is obtained from the restocking strategy table based on the comprehensive consumption interval of the materials corresponding to the comprehensive consumption of the materials, and medical materials are restocked based on the restocking strategy. If the comprehensive consumption of the materials exceeds the dynamic consumption threshold, it is determined that the consumption of the materials is abnormal, a consumption exceeding value of the comprehensive consumption of the materials exceeding the dynamic consumption threshold is obtained, a management strategy table is obtained, wherein the management strategy table includes a plurality of consumption exceeding value intervals and a corresponding alarm strategy and alarm restocking strategy for each consumption exceeding value interval, a corresponding alarm maintenance strategy and alarm restocking strategy are obtained from the management strategy table based on the consumption exceeding value interval corresponding to the consumption exceeding value, medical materials are processed based on the alarm maintenance strategy, and medical materials are restocked based on the alarm restocking strategy.

[0034] As in the above steps S601 to S604, the material weight, i.e. the current consumption weight, the material unexpected weight, the material physical damage weight and the material timeliness weight, is read from the pre-set database or the AI model training result. The sum of each weight is usually normalized to 1 to ensure that the comprehensive consumption is in a comparable scale. According to the weight, the material current consumption, the material unexpected consumption, the material physical damage risk and the material timeliness consumption are weighted and fused to generate a single material comprehensive consumption. The calculation formula of the material comprehensive consumption is , wherein Z represents the material comprehensive consumption, represents the current consumption weight, U represents the material current consumption, represents the material unexpected weight, F represents the material unexpected consumption, represents the material physical damage weight, P represents the material storage deformation, represents the material timeliness weight, represents the material timeliness consumption. The upper limit of the current normal range is obtained based on the historical consumption data to obtain a dynamic consumption threshold which can be automatically adjusted according to the seasonal and department demand fluctuations. If the material comprehensive consumption does not exceed the dynamic consumption threshold, it is determined that the material consumption is normal. From the replenishment strategy table, the corresponding replenishment frequency and replenishment amount (such as regular small batch replenishment, regular large batch replenishment, etc.) are obtained according to the pre-defined consumption interval (such as 0-5, 5-8). The replenishment instruction is automatically generated and issued to the supply chain system according to the table strategy to ensure that the inventory is maintained at a safe level. If the material comprehensive consumption exceeds the dynamic consumption threshold, it is determined that the material consumption is abnormal. From the management strategy table, the corresponding alarm strategy (SMS, email, billboard warning, on-site verification) and alarm replenishment strategy (such as emergency allocation, priority replenishment, limit use) are obtained according to the consumption exceeding value interval (such as 0-3, 3-6, 6-8). Alarm maintenance (such as starting inventory, manual review or package inspection) is performed, and at the same time, an urgent replenishment instruction is issued to ensure that the risk material is replenished in time or isolated. The multi-dimensional loss factors are fused into a single measurable index to avoid single index blind area and realize accurate grasp of the overall health status of the material. The dynamic threshold mechanism can automatically adjust according to the seasonal and event-driven demand fluctuations to effectively reduce the false positive and false negative rates. According to the different consumption levels, regular replenishment or emergency disposal can be flexibly selected to improve the supply chain efficiency. While ensuring the sufficiency of clinical materials, the inventory holding cost and operation risk are maximally reduced through fine grading replenishment and early warning.

[0035] In a preferred embodiment, the step of obtaining a dynamic consumption threshold based on historical consumption data comprises: S6031, obtaining a standard consumption threshold and a standard consumption value; S6032, obtaining a plurality of corresponding historical consumption values based on historical consumption data; S6033, obtaining a dynamic loss threshold value based on the standard loss threshold value, the standard consumption value and the plurality of historical consumption values.

[0036] As in the above steps S6031 to S6033, the standard loss threshold value (the highest loss allowed by the material under ideal environment and normal business conditions) and the standard consumption value (the normal consumption amount of the material (such as 1000 units per month) in the same statistical period) are defined in advance in the pre-set database or provided by experts and suppliers, the historical consumption values are extracted from the warehouse management system or the time series database according to the same statistical period (day / week / month), the standard loss threshold value, the standard consumption value and the plurality of historical consumption values are comprehensively calculated to obtain a dynamic loss threshold value, and the calculation formula of the dynamic loss threshold value is , wherein J represents the dynamic loss threshold value, represents the standard loss threshold value, represents the standard consumption value, and i represents the number of the plurality of historical consumption values, i = 1, 2, 3…n, represents the i-th historical consumption value, the dynamic threshold value automatically fluctuates up and down with the historical consumption level, can adapt to short-term demand fluctuations of seasonality or special events (such as large-scale operations and peak periods), reduces the false alarm rate, and the standard loss threshold value and the consumption value can be flexibly set by the hospital or the supplier according to the characteristics of the material, thereby improving the adaptability.

[0037] Please refer to the accompanying Figure 2 , the application also provides a medical material intelligent supply chain management system based on a multi-modal AI, which is used for the above-mentioned medical material intelligent supply chain management scheme based on a multi-modal AI, and comprises: a life cycle module, configured to obtain physical characteristic data, identity data and business flow data of the medical material, and generate whole life cycle information of the material; a current loss module, configured to obtain clinical demand information and inventory state data, and generate current loss of the material; an unexpected consumption module, configured to obtain material taking weight and material billing records based on the whole life cycle information of the material, and generate unexpected consumption of the material based on the material taking weight and the material billing records; a physical damage module, configured to obtain material storage state image data based on the whole life cycle information of the material, and obtain physical damage risk of the material based on the material storage state image data; a material shelf life module, configured to obtain material expiration date data and material storage environment information based on the whole life cycle information of the material, and predict material shelf life loss based on the material expiration date data and the material storage environment information; The material management module obtains comprehensive consumption of the material based on current consumption, unexpected consumption, physical damage risk and time-sensitive consumption of the material, and obtains a dynamic consumption threshold based on historical consumption data, determines whether the comprehensive consumption of the material exceeds the dynamic consumption threshold, and executes a corresponding management strategy based on the determination result.

[0038] The life cycle module obtains physical characteristic data (such as weight, volume, appearance), identity data (such as category, expiration date) and business flow data (such as purchase order, use record, billing record) of the medical material, performs data fusion on the multi-source information, and generates a full life cycle information portrait of each item of material. The current consumption module collects clinical demand information and inventory state data (such as last period and current inventory), and calculates the current consumption of the material based on these data, which is used to evaluate whether the current material is abnormally consumed. The unexpected consumption module obtains the unexpected consumption of the material by reading the use weight record and the billing record and comparing the deviation between the two. The physical damage module collects storage state image data (multi-angle) of the material, constructs a coordinate system on the image, extracts edge corner coordinates, and judges whether the material has physical damage risk such as compression damage and deformation in combination with the standard contour area. The material time limit module predicts the remaining effective use time of the material in the current environment in combination with the effective period data and real-time temperature and humidity environment data. The material management module fuses the aforementioned four types of consumption indicators, calculates the comprehensive consumption value of the material, dynamically generates a consumption threshold based on historical consumption trends, judges whether the current comprehensive consumption is abnormal, if normal, looks up the table to obtain the optimal replenishment strategy, automatically pushes the replenishment task, if abnormal, calculates the exceeding degree, triggers a hierarchical alarm and replenishment maintenance strategy, and performs abnormal traceability analysis, forming a full-chain data perception and feedback mechanism from warehousing, storage, use, consumption, abnormality and replenishment, realizing closed-loop control of medical material management, introducing image recognition, sensor collection, structured form, time series data and other multi-modal data sources, realizing comprehensive perception and recognition of the state of the material, and comprehensively calculating the consumption by using multi-factor weighting in combination with historical behavior data to dynamically set a management threshold, so that the management strategy is more personalized, precise and adaptive. Based on the dynamic consumption threshold, the demand replenishment and cross-warehouse allocation can be realized, the inventory turnover rate and the supply guarantee rate are taken into account, the environmental monitoring and expiration date information are included in the processing, the storage of temperature and humidity sensitive materials under the best conditions is ensured, the consumption caused by expiration or environmental abnormalities is reduced, the waste rate and excessive inventory risk are reduced, and the timely availability of key materials for clinical use is ensured, and the economic benefit and safety of the medical material supply chain are improved. The medical material intelligent supply chain management terminal based on multi-modal AI comprises: one or more processors; a storage device having one or more programs stored thereon; When one or more programs are executed by one or more processors, the one or more processors implement the multimodal AI-based medical material intelligent supply chain management scheme.

[0039] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application. The structures, devices and operation methods not specifically described and explained in the present application are implemented according to the conventional means in the art, unless otherwise specified and limited.

Claims

1. A multimodal AI-based intelligent supply chain management solution for medical supplies, characterized by: include: Obtain physical characteristic data, identification data, and business flow data of medical supplies, and generate information on the entire life cycle of the supplies; Obtain clinical demand information and inventory status data, and generate current material consumption; Obtain material usage weight and material billing records based on the full life cycle information of materials, and generate unexpected material consumption based on the material usage weight and material billing records; Obtain material storage status image data based on the material's full life cycle information, and obtain material physical damage risk based on the material storage status image data; Obtain material validity period data and material storage environment information based on the material life cycle information, and predict material timeliness loss based on the material validity period data and material storage environment information; The comprehensive material loss is obtained based on the current material loss, unexpected material consumption, material physical damage risk and material time-sensitive loss, and the dynamic loss threshold is obtained based on historical consumption data. It is determined whether the comprehensive material loss exceeds the dynamic loss threshold, and the corresponding management strategy is executed based on the judgment result.

2. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps to obtain the physical characteristic data, identification data, and business flow data of medical supplies and generate the full life cycle information of the supplies include: Obtaining physical characteristic data of medical supplies, where the physical characteristic data includes weight, volume, and appearance; Obtain identification data of medical supplies, including the type and expiration date of the supplies; Obtain business flow data for medical supplies, including purchase orders, material billing records, and department consumption records; The physical characteristic data, identity data and business flow data of each medical supply are associated and integrated to generate information on the entire life cycle of the supply.

3. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps to obtain clinical demand information and inventory status data and generate current material consumption include: Obtain clinical demand information and inventory status data; Obtain corresponding clinical demand values ​​based on clinical demand information; Obtain the corresponding previous period inventory value and current inventory value based on inventory status data; Obtain current material consumption based on the previous inventory value and current inventory value based on clinical demand value.

4. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps for obtaining material usage weight and material billing records based on the material life cycle information and generating unexpected material consumption based on the material usage weight and material billing records include: Obtain material usage weight and material billing records based on the full life cycle information of materials; Obtain material billing weight based on material billing records; Obtain the material withdrawal deviation weight based on the material billing weight and the material withdrawal weight, and mark it as unexpected material consumption.

5. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps of obtaining material storage status image data based on the material full life cycle information and obtaining material physical damage risk based on the material storage status image data include: Acquire material storage status image data based on the material life cycle information; Acquire material storage images of multiple perspectives based on the material storage status image data, and construct a plane rectangular coordinate system in the material storage images of each perspective; Obtaining the coordinates of multiple material edge contour inflection points in the material storage image at each viewing angle based on a plane rectangular coordinate system; Obtain the standard area of ​​the material form image at each viewing angle; Based on the coordinates of multiple material edge contour inflection points in the material storage image at each perspective and the corresponding standard area of ​​the material shape image, the material storage deformation is generated and marked as physical damage risk.

6. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps of obtaining material validity period data and material storage environment information based on the material life cycle information and predicting material time-sensitive consumption based on the material validity period data and material storage environment information include: Obtain material validity data and material storage environment information based on the material life cycle information; Obtain the standard temperature validity period and storage time of medical supplies based on the material validity period data; Obtain standard storage temperature, actual average storage temperature and temperature coefficient based on material storage environment information; Obtain the timeliness loss of medical supplies based on the effective storage time at standard temperature, storage time, standard storage temperature, actual average storage temperature and temperature coefficient.

7. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 1 is characterized in that: The steps of obtaining comprehensive material loss based on current material loss, unexpected material loss, material physical damage risk, and material time-sensitive loss, obtaining a dynamic loss threshold based on historical consumption data, determining whether the comprehensive material loss exceeds the dynamic loss threshold, and executing corresponding management strategies based on the determination result include: Obtain material weights, including current loss weight, unexpected material weight, physical damage weight, and timeliness weight. Obtain comprehensive material loss based on current material loss, unexpected material consumption, material physical damage risk, material timeliness loss, and material weight; Get dynamic consumption thresholds based on historical consumption data; Determine whether the comprehensive material consumption exceeds the dynamic consumption threshold; If the comprehensive material loss does not exceed the dynamic loss threshold, the material loss is determined to be normal, and a replenishment strategy table is obtained. The replenishment strategy table includes multiple comprehensive material loss intervals and the corresponding material replenishment strategy for each comprehensive material loss interval. Based on the comprehensive material loss interval corresponding to the comprehensive material loss, the corresponding material replenishment strategy is obtained from the replenishment strategy table, and medical material replenishment is performed based on the material capture strategy. If the comprehensive material consumption exceeds the dynamic consumption threshold, the material consumption is determined to be abnormal, and the consumption excess value of the comprehensive material consumption exceeding the dynamic consumption threshold is obtained, and a management strategy table is obtained, wherein the management strategy table includes multiple consumption excess value intervals and the alarm strategy and alarm replenishment strategy corresponding to each consumption excess value interval. Based on the consumption excess value interval corresponding to the consumption excess value, the corresponding alarm maintenance strategy and alarm replenishment strategy are obtained from the management strategy table, the medical supplies are processed based on the alarm maintenance strategy, and the medical supplies are replenished based on the alarm replenishment strategy.

8. The multimodal AI-based intelligent supply chain management solution for medical supplies according to claim 7 is characterized in that: The steps for obtaining a dynamic consumption threshold based on historical consumption data include: Get the standard consumption threshold and standard consumption value; Obtain corresponding multiple historical consumption values ​​based on historical consumption data; A dynamic consumption threshold is obtained based on a standard consumption threshold, a standard consumption value, and a plurality of historical consumption values.

9. A multimodal AI-based intelligent supply chain management system for medical supplies, applied to the multimodal AI-based intelligent supply chain management solution for medical supplies according to any one of claims 1 to 8, characterized in that: include: The life cycle module is used to obtain the physical characteristics data, identification data and business flow data of medical supplies and generate information on the entire life cycle of the supplies; The current consumption module is used to obtain clinical demand information and inventory status data and generate the current consumption of materials; The unexpected consumption module obtains the material usage weight and material billing records based on the material life cycle information, and generates unexpected material consumption based on the material usage weight and material billing records; The physical damage module obtains material storage status image data based on the material's full life cycle information, and obtains the material's physical damage risk based on the material storage status image data; The material validity module obtains material validity period data and material storage environment information based on the material life cycle information, and predicts material timeliness loss based on the material validity period data and material storage environment information; The material management module obtains the comprehensive material loss based on the current material loss, unexpected material consumption, material physical damage risk and material time-sensitive loss, and obtains the dynamic loss threshold based on historical consumption data, determines whether the comprehensive material loss exceeds the dynamic loss threshold, and executes the corresponding management strategy based on the judgment result.

10. A multimodal AI-based intelligent supply chain management terminal for medical supplies, characterized by: include: one or more processors; a storage device having one or more programs stored thereon; When one or more programs are executed by one or more processors, the one or more processors implement the multimodal AI-based intelligent supply chain management solution for medical supplies as described in any one of claims 1 to 8.

Images