An operating room intelligent scheduling system

Abstract

The application provides an operating room intelligent scheduling system, and belongs to the technical field of medical information intelligent integration, and comprises the following steps: based on the scene integration of a plurality of systems such as a hospital HIS and operating room management, an operating room special intelligent scheduling system is realized, and the core comprises the following steps: the full-dimension resource demand baseline of an operation is predicted through a deep learning model, scheduling branch updating is completed through available operating room matching, standby operating room scheduling tree construction, best array screening or gap insertion matching. The application solves the pain points of low system integration, insufficient scheduling matching precision, and inability to balance resource utilization and scheduling stability of the prior art, and greatly improves the operating room resource utilization, scheduling compliance and clinical adaptability.

Description

Technical Field

[0001] This invention relates to the field of intelligent integration technology of medical information, and in particular to an intelligent scheduling system for operating rooms. Background Technology

[0002] As a core medical resource in hospitals, the scientific and rational scheduling of operating rooms directly determines the operational efficiency of hospital medical resources, the quality of clinical medical services, and patient safety. It is a core link in the hospital's refined operation management and medical quality control. With the continuous advancement of hospital informatization in China, most hospitals have completed the deployment of basic information systems for operating rooms. However, significant industry pain points and technical shortcomings still exist in the intelligentization of the entire scheduling process and the integration of the entire system.

[0003] Currently, the informatization of operating room scheduling information in domestic hospitals faces two major practical problems: First, in clinical practice, surgical request information requires manual secondary entry by medical staff across systems. Core data such as real-time operating room occupancy status, availability of specialized medical equipment, medical staff qualifications and scheduling information, patient pre-operative preparation status, and infection control requirements for handover and disinfection cannot be linked in real time. Insufficient system integration capabilities not only significantly increase the non-clinical workload of medical staff but also easily lead to scheduling conflicts due to data delays and inconsistencies. Second, existing intelligent scheduling solutions mainly fall into two categories, both with significant technical shortcomings: one is the full-cycle static optimization scheduling solution. This type of solution mostly predicts surgical duration based on historical surgical data and combines optimization algorithms to complete fixed-cycle scheduling planning. It can only achieve batch scheduling in static scenarios and cannot cope with the high frequency of new emergency surgeries, fluctuations in surgical duration, and patient changes in clinical practice. In dynamic scenarios such as changes in patient conditions, any new demand necessitates a complete rescheduling of the entire lifecycle scheduling plan, severely disrupting existing clinical work plans, undermining scheduling stability, and resulting in extremely poor clinical adaptability. Secondly, there are dynamic insertion scheduling plans. These plans often achieve insertion matching of new surgeries by constructing a basic scheduling model, but they generally suffer from a single matching dimension, focusing only on the basic matching of surgery duration and operating room idle time windows. They fail to simultaneously cover hard constraints such as operating room cleanliness levels, dedicated medical equipment configuration, medical staff qualifications and specialty suitability, and infection control requirements. The scheduling matching accuracy is insufficient, requiring manual secondary verification and adjustment. Furthermore, some plans lack a hierarchical scheduling logic based on minimum disturbance optimization, resulting in inverted scheduling priorities. To accommodate new surgeries, already scheduled elective surgeries are prioritized, even those already triggering pre-operative preparation processes, severely impacting the continuity of clinical work and even posing medical safety risks.

[0004] In summary, existing operating room scheduling technologies cannot meet the complex clinical scheduling needs and refined operational management requirements of modern hospital operating rooms. Therefore, this invention proposes an intelligent operating room scheduling system. Summary of the Invention

[0005] This invention provides an intelligent operating room scheduling system to solve the aforementioned technical problems.

[0006] This invention provides an intelligent operating room scheduling system, comprising: The baseline prediction module is used to receive the uploaded surgical requirements based on the integrated interface, extract the full-dimensional feature data of the surgical requirements, call the pre-built surgical resource requirement prediction model, and output the resource requirement baseline data corresponding to the surgical requirements. The resource requirement baseline data includes the surgical duration baseline, the legally mandated table-opening disinfection and transfer time baseline, the surgical procedure adaptation level, the operating room cleanliness level requirements, the occupancy requirements of special medical equipment, and the baseline of medical staff qualification requirements. The operating room reservation module is used to retrieve a list of available operating rooms that meet the access criteria from the server backend based on the resource demand baseline data. The access criteria are the cleanliness level of the operating room, equipment configuration, and the requirement that the idle time window after deducting the statutory reception, disinfection and transfer time completely covers the resource demand baseline data. If the list of available operating rooms is not empty, calculate the matching degree between each operating room in the list of available operating rooms and the baseline data of resource demand, filter the operating room with the highest matching degree, and set the reservation tag for the corresponding surgical demand; The operating room range locking module is used to lock the range of available operating rooms that meet the basic adaptation requirements of the surgical procedure based on the resource requirement baseline data of the surgical needs when the list of available operating rooms is empty. The range of available operating rooms includes at least one available operating room, and the basic adaptation requirements of the surgical procedure are the hard requirements of the operating room's cleanliness level and equipment configuration meeting the resource requirement baseline data. The tree construction module is used to obtain the current scheduling information of each waiting operating room within the scope of the waiting operating rooms, construct the tree structure scheduling tree of the corresponding waiting operating rooms, and set an inseparable parent branch for consecutive surgeries by the same surgeon. At the same time, it sets an unadjustable lock tag for the surgical branch that has triggered the preoperative preparation state, and extracts the maximum allocation array of the currently uploaded surgical demand content corresponding to each branch in the scheduling tree. The maximum allocation array is a set of quantitative parameters that characterize the corresponding scheduled surgical branch to be compatible with the insertion and / or adjustment of new surgeries. Each parameter in the maximum allocation array is set with a corresponding admission threshold. The optimal array filtering and matching module is used to filter the optimal array from all the maximum allocation arrays of all available operating rooms based on resource demand baseline data and scheduling priority, so that all parameters meet the admission threshold constraints. The scheduling tree update module is used to count the number of optimal arrays of all available operating rooms. If the number is not 0, it controls the optimal operating room matching module to start working to obtain the first target operating room for scheduling tree update; if the number is 0, it controls the gap insertion module to start working to obtain the second target operating room for scheduling tree update.

[0007] Preferably, the optimal operating room matching module is used to calculate a comprehensive score from the best array of all available operating rooms based on the operating room resource matching degree, personnel suitability degree, and surgery delay time, select the array with the highest comprehensive score as the final array, mark the available operating room corresponding to the final array as the first target operating room, and create a new branch at the branch position corresponding to the final array in the scheduling tree of the first target operating room to complete the update of the scheduling tree.

[0008] Preferably, the gap insertion module is used to determine the scheduling tree for each operating room to be used, and calculate the density value of each scheduling indicator corresponding to each time node within the scheduling cycle based on each scheduling indicator, generate a multivariate indicator density sequence corresponding to each scheduling indicator, and stack all multivariate indicator density sequences in rows with each scheduling indicator as a row dimension to construct an M×N multivariate indicator density matrix, where N is the total number of time nodes within the scheduling cycle; M is the number of scheduling indicators, and each scheduling indicator corresponds to one dimension of the multivariate indicator density matrix, and the scheduling indicators include the continuity of surgical time windows, the continuity of operating room resource occupancy, the balance of medical staff workload, and the uniformity of surgical priority distribution. Adjacent element connection analysis is performed on the multivariate index density matrix to locate potential gap intervals between adjacent surgical branches after deducting the statutory disinfection and transfer time. After deducting the statutory disinfection and transfer time preset for the corresponding cleanliness level operating room, the effective usable time of the gap is verified. The matrix sub-block data corresponding to each potential gap interval is extracted, the branch connection density of the corresponding potential gap is calculated, and the allowable insertion probability of the corresponding gap is determined based on the branch connection density. Gaps with allowable insertion probability greater than the preset insertion probability threshold are locked, and a locked gap distribution map corresponding to each multivariate index density matrix is ​​generated. For each target gap in the locked gap distribution map, the insertion vector is transformed to generate the insertion vector to be analyzed. The insertion vector to be analyzed is input into the pre-constructed vector analysis matching model, and the acceptance matching degree corresponding to each insertion vector is output. The operating room with the highest acceptance matching degree is selected as the second target operating room. The gap with the highest allowed insertion probability in the second target operating room is used as the branch insertion point to complete the establishment of new branches and the scheduling tree update.

[0009] Preferably, the tree building module includes: The extraction unit is used to extract the scheduling priority, baseline surgery duration, legally mandated handover, disinfection and transfer time, resource occupancy attributes, medical staff matching pool, compliance constraints, consecutive surgery binding identifier, and preoperative preparation status lock identifier for each scheduled surgery in the current scheduling information. It sets a scheduling sub-segment with a corresponding unique identifier for each scheduled surgery, where each scheduling sub-segment corresponds one-to-one with a single scheduled surgery. The tree generation unit is used to automatically generate a tree-structured scheduling tree for each available operating room, based on all scheduling segments corresponding to each available operating room and according to the order of the surgical time windows. The scheduling tree is a multi-branch tree structure with the operating room as the root node, the scheduling segment of a single scheduled surgery as the first-level branch, the parent branch of consecutive surgeries as the second-level root node, the scheduling segment of the consecutive surgeries as the third-level branch, and the order of the surgeries as the branch sorting rule. The length of each branch corresponds to the sum of the duration of the corresponding surgery and the duration of the connecting surgeries, and the attributes of the branches correspond to the full-dimensional information of the scheduling segment.

[0010] Preferred options also include: The first-level judgment module is used to perform hierarchical compliance verification of the full-cycle scheduling plan corresponding to the updated scheduling tree. The scheduling cycle is divided into multiple continuous verification units, and the scheduling data in each verification unit is judged for first-level adaptability. The judgment indicators for first-level adaptability include: whether the daily working hours of medical staff do not exceed 8 hours, whether the daily continuous occupation time of the operating room does not exceed 12 hours, and whether the surgical priority ranking conforms to the rule of emergency priority. The secondary judgment module is used to determine that the corresponding verification unit is qualified when all the judgment indicators of the primary adaptability judgment meet the requirements; otherwise, the corresponding verification unit is determined to be unqualified. At this time, the unqualified verification unit is subject to secondary judgment. The rule of the secondary judgment is as follows: when the feature parameter dispersion of each scheduling node in the corresponding unqualified verification unit is less than or equal to the preset dispersion threshold, the corresponding unqualified verification unit is determined to meet the compliance requirements through local adjustment; otherwise, the corresponding unqualified verification unit is determined to be an unqualified unit that cannot be adjusted. The feature parameter dispersion is the ratio of the standard deviation to the mean of the core feature parameter of each scheduling node in the verification unit.

[0011] Preferred options also include: The priority calculation module is used to calculate the scheduling priority of scheduled surgeries by weighting them with preset multi-dimensional adjustable weight configuration factors, including: The weight acquisition unit is used to acquire four basic parameters: surgical urgency quantification value, patient perioperative risk quantification value, surgeon time window matching degree, and operating room resource suitability degree. It assigns corresponding weight coefficients to the four basic parameters, and the sum of all weight coefficients is 1. The multiplication unit is used to multiply each basic parameter with its corresponding weight coefficient and then sum them to obtain the score of the corresponding surgery. The score is then matched with the score-priority table to obtain the scheduling priority.

[0012] Preferably, the tree building module includes: The scoring determination unit is used to calculate the job suitability score of each medical staff member and the two surgeries based on the matching pool of medical staff members for the corresponding branch and the baseline of medical staff qualification requirements for the current surgeries, using a collaborative matching algorithm. The job suitability score ranges from 0 to 100. The similarity calculation unit is used to calculate the workload balance of the medical staff group using the K-nearest neighbor algorithm, with a value ranging from 0 to 1. The judgment unit is required to determine whether the personnel suitability and compatibility threshold meets the requirements when the job suitability score is greater than or equal to the preset suitability threshold and the workload balance is less than or equal to the preset balance threshold; otherwise, it is determined that the requirements are not met.

[0013] Preferably, the optimal operating room matching module includes: A new unit is created to lock the time window position and attribute information of the original branch of the scheduling tree corresponding to the final array. Without modifying the original data of the original branch, a new branch to be effective is created at the target branch position based on the scheduling sub-segment information of the surgery to be inserted. Simultaneously, an adjusted mirror branch of the original branch is generated based on the adjustment rules of the final array. The new branch and the adjusted mirror branch share adjacent non-overlapping time windows on the same time axis, forming a parallel dual-branch temporary structure for pre-synchronization. The synchronous update unit is used to bind a new branch to a standardized scheduling sub-segment that uniquely corresponds to the surgery to be inserted. It synchronously updates the operating room resource occupancy tag, medical staff matching pool, dedicated medical equipment lock status, and surgery priority attribute corresponding to the new branch. At the same time, it synchronously updates the time window attribute, resource occupancy attribute, and personnel matching attribute of the adjusted mirror branch.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: 1. Based on a three-level hierarchical scheduling logic of direct matching of available operating rooms, adjustable branch matching, and gap insertion matching, the core pain points of inverted priority and poor scheduling stability of existing technical processes are solved. While improving the utilization rate of operating room resources, the change rate of the original scheduling plan is controlled within 5%, maximizing the continuity of clinical work.

[0015] 2. Seamlessly integrates with hospital HIS systems, patient information management systems, operating room management systems, medical staff management systems, and medical equipment management systems. It can also provide integrated implementation services, information system design services, and other information system integration services. This enables seamless integration and rapid deployment with existing hospital systems, achieving automatic synchronization of surgical needs, real-time retrieval of patient data, dynamic updates to operating room status, and real-time synchronization of medical staff qualifications and workload data. It breaks down data silos in existing systems, realizing full-process management of operating room scheduling and meeting the full-process service needs of hospital operating room informatization implementation.

[0016] 3. By integrating a deep learning model with bidirectional long short-term memory networks and attention mechanisms, we have achieved accurate prediction of the full range of surgical resource requirements. This includes not only the duration of the surgery, but also the operating room suitability level, the demand for specialized equipment, and the qualification requirements of medical staff. This provides accurate basic data for subsequent scheduling and matching, solving the problems of insufficient matching and resource mismatch caused by the existing system's prediction of only the duration. The accuracy of scheduling and matching has been improved by more than 30%.

[0017] 4. Based on a multi-branch tree structure, the scheduling scheme is modularly managed by having the scheduling sub-segments correspond to the branches of the scheduling tree. For new surgical needs, only the branches need to be added or locally adjusted, without the need for a full rescheduling of the entire cycle, which greatly improves the efficiency of scheduling adjustment.

[0018] 5. In the entire process of scheduling and matching, not only the occupancy of operating room resources is considered, but also the priority of surgeries, the matching of medical staff qualifications, and the balance of workload are taken into account. The personnel matching constraints are integrated into the entire process of allocation array screening and gap insertion matching, which avoids the problem of personnel and resource matching being out of sync, and improves the balance of medical staff workload in the scheduling plan.

[0019] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of an intelligent operating room scheduling system according to an embodiment of the present invention. Detailed Implementation

[0022] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0023] To strictly adhere to the core principles of zero-disruption priority and minimal-disruption fallback in clinical operating room scheduling, and to maximize the stability of the original scheduling plan, this system implements a three-tiered scheduling process throughout the entire scheduling process, as follows: Level 1: Zero-disruption direct matching. After receiving surgical requests and completing baseline prediction of resource requirements, the system prioritizes the operating room reservation module process, retrieving a list of available operating rooms that meet the access criteria. If the list is not empty, it directly completes the optimal operating room matching and reservation, without modifying any scheduled surgeries, achieving zero-disruption scheduling. Level 2: Zero-disruption gap insertion. If the list of available operating rooms is empty, the gap insertion module is initiated first, using a multivariate index density matrix to mine and match potential gaps. If a suitable insertable gap is selected, the new surgery is directly inserted with zero-disruption, without adjusting the time windows or resource allocation of any scheduled surgeries. Level 3: Minimum-disruption branch adjustment. If no suitable insertion slot is found, the optimal array filtering and matching module and the optimal operating room matching module are activated. Based on the optimal array that meets the admission threshold, the minimal disturbance adjustment of the already scheduled surgeries and the insertion of new surgeries are completed, keeping the change rate of the original schedule within a preset range. This preferred embodiment solves the problem of inverted scheduling priorities by optimizing the sequence of three-level hierarchical scheduling. While improving the utilization rate of operating room resources, it minimizes the disturbance to the original clinical schedule and ensures the continuity of clinical work.

[0024] This invention provides an intelligent operating room scheduling system, such as... Figure 1 As shown, it includes: The baseline prediction module is used to receive the uploaded surgical requirements based on the integrated interface, extract the full-dimensional feature data of the surgical requirements, call the pre-built surgical resource requirement prediction model, and output the resource requirement baseline data corresponding to the surgical requirements. The resource requirement baseline data includes the surgical duration baseline, the legally mandated table-opening disinfection and transfer time baseline, the surgical procedure adaptation level, the operating room cleanliness level requirements, the occupancy requirements of special medical equipment, and the baseline of medical staff qualification requirements. The operating room reservation module is used to retrieve a list of available operating rooms that meet the access criteria from the server backend based on the resource demand baseline data. The access criteria are the cleanliness level of the operating room, equipment configuration, and the requirement that the idle time window after deducting the statutory reception, disinfection and transfer time completely covers the resource demand baseline data. If the list of available operating rooms is not empty, calculate the matching degree between each operating room in the list of available operating rooms and the baseline data of resource demand, filter the operating room with the highest matching degree, and set the reservation tag for the corresponding surgical demand; The operating room range locking module is used to lock the range of available operating rooms that meet the basic adaptation requirements of the surgical procedure based on the resource requirement baseline data of the surgical needs when the list of available operating rooms is empty. The range of available operating rooms includes at least one available operating room, and the basic adaptation requirements of the surgical procedure are the hard requirements of the operating room's cleanliness level and equipment configuration meeting the resource requirement baseline data. The tree construction module is used to obtain the current scheduling information of each waiting operating room within the scope of the waiting operating rooms, construct the tree structure scheduling tree of the corresponding waiting operating rooms, and set an inseparable parent branch for consecutive surgeries by the same surgeon. At the same time, it sets an unadjustable lock tag for the surgical branch that has triggered the preoperative preparation state, and extracts the maximum allocation array of the currently uploaded surgical demand content corresponding to each branch in the scheduling tree. The maximum allocation array is a set of quantitative parameters that characterize the corresponding scheduled surgical branch to be compatible with the insertion and / or adjustment of new surgeries. Each parameter in the maximum allocation array is set with a corresponding admission threshold. The optimal array filtering and matching module is used to filter the optimal array from all the maximum allocation arrays of all available operating rooms based on resource demand baseline data and scheduling priority, so that all parameters meet the admission threshold constraints. The scheduling tree update module is used to count the number of optimal arrays of all available operating rooms. If the number is not 0, it controls the optimal operating room matching module to start working to obtain the first target operating room for scheduling tree update; if the number is 0, it controls the gap insertion module to start working to obtain the second target operating room for scheduling tree update.

[0025] Preferably, the optimal operating room matching module is used to calculate a comprehensive score from the best array of all available operating rooms based on the operating room resource matching degree, personnel suitability degree, and surgery delay time, select the array with the highest comprehensive score as the final array, mark the available operating room corresponding to the final array as the first target operating room, and create a new branch at the branch position corresponding to the final array in the scheduling tree of the first target operating room to complete the update of the scheduling tree.

[0026] Preferably, the gap insertion module is used to determine the scheduling tree for each operating room to be used, and calculate the density value of each scheduling indicator corresponding to each time node within the scheduling cycle based on each scheduling indicator, generate a multivariate indicator density sequence corresponding to each scheduling indicator, and stack all multivariate indicator density sequences in rows with each scheduling indicator as a row dimension to construct an M×N multivariate indicator density matrix, where N is the total number of time nodes within the scheduling cycle; M is the number of scheduling indicators, and each scheduling indicator corresponds to one dimension of the multivariate indicator density matrix, and the scheduling indicators include the continuity of surgical time windows, the continuity of operating room resource occupancy, the balance of medical staff workload, and the uniformity of surgical priority distribution. Adjacent element connection analysis is performed on the multivariate index density matrix to locate potential gap intervals between adjacent surgical branches after deducting the statutory disinfection and transfer time. After deducting the statutory disinfection and transfer time preset for the corresponding cleanliness level operating room, the effective usable time of the gap is verified. The matrix sub-block data corresponding to each potential gap interval is extracted, the branch connection density of the corresponding potential gap is calculated, and the allowable insertion probability of the corresponding gap is determined based on the branch connection density. Gaps with allowable insertion probability greater than the preset insertion probability threshold are locked, and a locked gap distribution map corresponding to each multivariate index density matrix is ​​generated. For each target gap in the locked gap distribution map, the insertion vector is transformed to generate the insertion vector to be analyzed. The insertion vector to be analyzed is input into the pre-constructed vector analysis matching model, and the acceptance matching degree corresponding to each insertion vector is output. The operating room with the highest acceptance matching degree is selected as the second target operating room. The gap with the highest allowed insertion probability in the second target operating room is used as the branch insertion point to complete the establishment of new branches and the scheduling tree update.

[0027] Preferably, the tree building module includes: The extraction unit is used to extract the scheduling priority, baseline surgery duration, legally mandated handover, disinfection and transfer time, resource occupancy attributes, medical staff matching pool, compliance constraints, consecutive surgery binding identifier, and preoperative preparation status lock identifier for each scheduled surgery in the current scheduling information. It sets a scheduling sub-segment with a corresponding unique identifier for each scheduled surgery, where each scheduling sub-segment corresponds one-to-one with a single scheduled surgery. The tree generation unit is used to automatically generate a tree-structured scheduling tree for each available operating room, based on all scheduling segments corresponding to each available operating room and according to the order of the surgical time windows. The scheduling tree is a multi-branch tree structure with the operating room as the root node, the scheduling segment of a single scheduled surgery as the first-level branch, the parent branch of consecutive surgeries as the second-level root node, the scheduling segment of the consecutive surgeries as the third-level branch, and the order of the surgeries as the branch sorting rule. The length of each branch corresponds to the sum of the duration of the corresponding surgery and the duration of the connecting surgeries, and the attributes of the branches correspond to the full-dimensional information of the scheduling segment.

[0028] In this embodiment, the specific implementation is an intelligent scheduling information system dedicated to the operating room scenario. It completes scenario-based integration and docking with the target hospital's HIS system, operating room anesthesia system, medical staff management system, and medical equipment management system, providing the hospital with a full-chain information system integration service covering architecture design, function development, implementation, and operation and maintenance support.

[0029] In this embodiment, a loosely coupled modular architecture with front-end and back-end separation is adopted, which is divided into five layers: data layer, algorithm layer, business layer, interface layer, and presentation layer. Each layer module is independently developed, independently deployed, and flexibly expanded, and corresponds one-to-one with the functional modules. The complete architecture is as follows: Data layer: responsible for data interface with various business systems of the hospital, data storage, data preprocessing, archiving and storage of historical versions, data receiving module and data cleaning and partitioning module as described in the invention; Algorithm layer: responsible for the encapsulation and invocation of all core algorithms, including surgical resource demand prediction model, vector analysis matching model, collaborative matching algorithm, and priority calculation algorithm, corresponding to the baseline prediction module and priority calculation module in the invention content; Business layer: Responsible for the implementation of business logic for the entire scheduling process, including operating room matching, scheduling tree generation, consecutive binding and status locking, optimal array filtering, gap insertion matching, scheduling tree updating, compliance verification, corresponding to all core business function modules; Interface layer: Responsible for standardized integration and interface with external business systems of the hospital, providing a unified API interface to achieve two-way data synchronization; Presentation Layer: Responsible for developing visual operation interfaces for different roles, including the doctor's surgery application interface, the operating room administrator's scheduling interface, the medical staff's personal scheduling query interface, and the administrator's system configuration interface.

[0030] This architecture design achieves a one-to-one correspondence between modules and business processes. At the same time, each module can be upgraded independently and flexibly adapted. It can also be customized to connect with existing business systems based on the current information technology status of different hospitals.

[0031] This invention provides a standardized integration solution for multiple systems. It develops a standardized integration interface that can seamlessly connect to existing core business systems in hospitals, breaking down data silos and enabling closed-loop management of the entire operating room scheduling process. The specific integration solution is as follows: Integration with the hospital's HIS system: Utilizing a standardized WebService interface, it achieves bidirectional real-time data synchronization; automatically synchronizing surgical request forms, patient basic information, preoperative diagnosis, surgical procedure information, surgical urgency, and other data from the HIS system; and sending back surgical scheduling results, operating room status, surgical progress, and other data to the HIS system. Interface with the operating room anesthesia system: Adopting the HL7 medical industry standard interface, it realizes batch synchronization of historical surgical data and real-time data interaction; it obtains the full process duration data of historical surgeries, anesthesia records, resource usage data, and real-time data of patients' preoperative preparation status from the anesthesia system; it synchronizes surgical scheduling plans, patient anesthesia needs, and medical staff matching information to the anesthesia system. Integration with the operating room management system: Utilizing a RESTful API interface, it achieves real-time synchronization of operating room status and equipment configuration information; it retrieves data from the operating room management system, including operating room cleanliness level, equipment configuration list, real-time occupancy status, maintenance plan, and infection control regulations regarding the standard disinfection time for each handover; and it sends data back to the operating room management system, such as operating room scheduling plans, equipment lock status, and resource occupancy requirements. Integration with the healthcare staff management system: Utilizing a standardized LDAP interface to achieve real-time synchronization of healthcare staff data; retrieving data from the healthcare staff management system including professional qualifications, specialties, titles, shift schedules, leave information, and work hours statistics for healthcare staff; and sending back final shift schedules, healthcare staff workload statistics, and overtime statistics to the healthcare staff management system. Integration with medical equipment management system: Employs IoT MQTT protocol interface to achieve real-time synchronization of dedicated medical equipment status; obtains data such as equipment availability, maintenance plan, and location from the equipment management system; and sends back equipment lock-up plans and usage reservation information to the equipment management system.

[0032] This invention addresses the issues of data latency, data conflict, and data inconsistency during multi-system data synchronization by designing a complete safeguard to ensure the accuracy and consistency of system data, as detailed below: Synchronization rules: A master-slave synchronization mechanism is set up, with the surgical request form of the hospital HIS system as the sole master data benchmark and the adjustment data of the scheduling system as slave data. All adjustments must complete bidirectional verification with the master data. Adjustments that fail the verification are automatically rejected. All data synchronization is timestamped to retain the version history of the data and to trace the changes throughout the entire life cycle of the data. Data consistency verification mechanism: A full data consistency verification is performed every 5 minutes, comparing the core data of this system with the core data of each connected system. If data inconsistency is found, the data correction process is automatically triggered: the latest timestamp version is used as the standard, the data is automatically synchronized and updated, and a data inconsistency alarm log is generated and pushed to the system administrator. Data delay handling solution: To address data delays caused by network fluctuations, a data caching mechanism is set up. For data with high real-time requirements (operating room status, medical staff availability status, preoperative preparation status), a 10-second timeout mechanism is set. If no synchronized data is received within the timeout period, a second synchronization request is automatically triggered, and the corresponding resources are locked to avoid duplicate scheduling. Data conflict resolution mechanism: For conflicts caused by simultaneous modification of the same data in multiple systems, priority rules are set: the operating room status data of the operating room management system and the personnel qualification data of the medical staff management system have the highest priority, followed by the surgical application data of the HIS system, and finally the scheduling adjustment data of this system; data changes in high-priority systems automatically overwrite data in low-priority systems, while triggering alarms and secondary verification. Transactional synchronization mechanism: All cross-system data synchronization adopts transactional processing. The transaction is only committed when all connected systems have successfully synchronized. If any system fails to synchronize, the rollback mechanism is automatically triggered, and the data of all systems is rolled back to the state before synchronization to avoid data inconsistency.

[0033] In this embodiment, the maximum allocation array refers to a set of quantitative parameters that characterize the maximum allowable range of new surgical insertions and / or time adjustments for the corresponding scheduled surgical branches in the scheduling tree. Each parameter in the array is set with a corresponding admission threshold, specifically including three core parameters: the maximum probability threshold for time adjustment of the corresponding branch, the maximum duration threshold for allowing surgical postponement, and the personnel adaptation compatibility threshold. The parameter dimensions can be expanded according to clinical needs. The admission threshold constraint refers to the minimum admission standard that each parameter in the maximum allocation array must meet as pre-set by this invention. Only when all parameters in the array meet the corresponding admission threshold can the array be included in the selection range of the best array. Only non-locked and non-bound low-priority elective surgical branches can be included in the selection range.

[0034] In this embodiment, the sequence length of the distribution density sequence is consistent with the number of time nodes in the scheduling cycle. Each element in the sequence corresponds to the density value of the indicator at a single time node. The multivariate indicator density matrix refers to a multidimensional matrix constructed with each scheduling indicator as a dimension and the time node of the scheduling cycle as the sequence length. Each element in the matrix corresponds to the density value of a single scheduling indicator at a single time node.

[0035] In this embodiment, the degree of acceptance of matching refers to the quantitative score of the compatibility between the gaps output by the matching model and the full-dimensional matching of the new surgical needs.

[0036] In this embodiment, the matrix sub-block data is the multivariate index density data of the corresponding time period of the potential gap interval. For example, if the gap is 10:00–11:15 (a total of 75 minutes), the system will cut out the 4 rows of index data corresponding to these 75 minutes from the large matrix of the whole day to obtain the matrix sub-block data.

[0037] In this embodiment, the surgical resource demand prediction model is a multi-task fusion deep learning model, which can simultaneously perform surgical duration regression prediction, surgical procedure and cleanliness level classification prediction, and multi-label classification prediction of equipment and personnel demand. This addresses the pain points of existing technologies, such as reversed causality, the ability to only predict surgical duration, and insufficient resource matching accuracy. The complete implementation steps are as follows: The first step is to construct and preprocess the sample dataset: using the standardized integration interface, obtain a comprehensive sample dataset of historical surgeries from the target hospital over the past 3-5 years. The sample dataset needs to cover 5 core data dimensions: Patient perioperative data: age, gender, number and type of underlying diseases, preoperative diagnosis, ASA anesthesia classification, BMI index, and previous surgical history; Surgical procedure attribute data: surgical procedure ICD code, surgical procedure name, surgical procedure difficulty level, specialty to which the procedure belongs, surgical approach, whether it is a minimally invasive surgery, and whether it is a day surgery; Surgeon team data: chief surgeon's title and years of practice, number and qualifications of assistants, anesthesiologist's qualifications and years of practice, average duration and duration fluctuation rate of similar surgeries in the team's history; Surgical resource occupancy data: operating room cleanliness level, list of dedicated medical equipment used, total operating room occupancy time, preoperative preparation time, surgical operation time, postoperative recovery time, and time for table handover, disinfection, and transfer. Surgical compliance constraint data: surgical urgency, surgical priority, and perioperative adverse event records.

[0038] Invalid data with missing values ​​exceeding 30% or containing obvious logical errors (such as operation duration of 0 or ASA classification outside the range of I-V) were removed. The remaining valid data underwent preprocessing. For continuous features (age, operation duration, years of practice, etc.), min-max standardization is used to map all feature values ​​to the interval [0,1]. Discrete classification features (ASA classification, procedure difficulty, professional title, etc.) are processed using one-hot encoding. Multi-label binarization is applied to multi-label features (equipment list, personnel qualification requirements, etc.) to generate a 0-1 matrix; The preprocessed sample dataset is divided into a training set and a validation set according to a preset ratio of 8:2. The training set is used for iterative training of the model, and the validation set is used for verifying the model's accuracy and generalization ability.

[0039] The second step is to construct the model network structure: Input layer: The input dimension is the total dimension of surgical procedures and patient characteristics without team information after preprocessing, and the input data is the standardized feature data of the preoperative application for a single surgery; Feature extraction layer: A two-layer bidirectional long short-term memory (BiLSTM) network is set, with 64 hidden units in each layer and a dropout rate of 0.2, to capture long-distance dependencies between surgical features and avoid model overfitting; Attention Mechanism Layer: A multi-head attention layer is set up with 4 heads. Dynamic weights are assigned to three core features: surgical difficulty, patient ASA classification, and surgeon's qualifications. This strengthens the positive impact of core features on prediction results and weakens the interference of irrelevant features. Multi-task output layer: Three parallel output branches are set up, corresponding to three types of prediction tasks. Each branch has an independent fully connected layer and activation function. Regression branch: Outputs the baseline of surgical duration and the baseline of the transition and disinfection time. The activation function is a linear activation function, and the output dimension is 2. Classification branches: Output surgical procedure adaptation level and operating room cleanliness level requirements. The activation function is the Softmax activation function, and the output dimension corresponds to the total number of classification categories. Multi-label classification branch: Outputs baselines for dedicated medical equipment occupancy requirements and medical staff qualification requirements. The activation function is Sigmoid activation function, and the output dimension corresponds to the total number of labels. Model output: Integrate the output results of the three branches to generate complete baseline data on surgical resource requirements.

[0040] The third step involves designing a multi-task joint loss function for the different task types of the three parallel output branches. This solves the technical problem that a single loss function cannot adapt to multi-task training. The formula for the joint loss function is as follows: ,in: This represents the total loss value of the model. The loss function for the regression branch is the mean squared error (MSE), calculated using the following formula: ,in, This represents the actual surgery time. The model predicts the operation time, where N is the number of samples; The loss function for the classification branch is the cross-entropy loss function, calculated as follows: ,in The total number of categories. Let i be the true class label of the i-th sample in the c-th class. Predict the probability that the i-th sample belongs to the c-th class for the model; For the loss function of the multi-label classification branch, the binary cross-entropy loss function is adopted, and the calculation formula is as follows: ,in, For the total number of tags, Let be the true binary label value (takes the value 0 or 1) of the i-th sample on the k-th label. Predict the probability that the i-th sample contains the k-th label for the model; , , The weight coefficients for the loss functions of the regression branch, classification branch, and multi-label classification branch are set to 0.4, 0.3, and 0.3 respectively. They can be dynamically adjusted through the system management backend according to the business focus of different hospitals. The sum of the weight coefficients is 1.

[0041] The fourth step, model training, accuracy verification, and consolidation, involves iteratively training the constructed multi-task deep learning model based on the training set. The training hyperparameters are set as follows: 100 iterations, batch size of 32, Adam optimizer, initial learning rate of 0.001, and a decay of the learning rate to 50% every 20 iterations. During training, the model's accuracy is verified using a validation set after each iteration. The iteration stopping condition is set as follows: The mean absolute error (MAE) of the predicted operation duration is ≤15 minutes; Classification branch prediction accuracy ≥95%; The mean accuracy (mAP) of multi-label classification branches is ≥92%; When the model simultaneously meets the above three accuracy metrics, or when the number of iterations reaches 100, training is stopped, the model weights and hyperparameters are fixed, and the model is integrated into the system's algorithm layer for online real-time prediction.

[0042] The fifth step is to implement a dynamic model update mechanism. After the model is deployed, a monthly incremental update mechanism is set up: the new surgical process data added each month is automatically acquired, the model is incrementally fine-tuned and trained, and the model weights are updated to ensure that the model's prediction accuracy continues to adapt to the changes in the hospital's clinical business and avoid model performance degradation.

[0043] In this embodiment, the vector analysis matching model is used to quantitatively evaluate the compatibility of the scheduling tree gaps with the new surgical needs across all dimensions. This solves the problems of existing technologies that only perform gap matching based on duration, have low matching accuracy, and cause large scheduling disturbances after insertion. The complete implementation steps are as follows: The first step is to construct and preprocess the sample dataset: Through the integrated interface that has been connected, obtain the full sample data of the historical scheduling insertion operations of the target hospital over the past two years. The sample data should cover: full-dimensional feature data of the insertion gap, full-dimensional feature data of the corresponding new surgical needs, and the final result of the insertion operation (whether the insertion was successful, whether compliance issues were triggered after the insertion, the disturbance rate to the original schedule, and the clinical execution pass rate).

[0044] Invalid samples that failed to be inserted or had compliance issues are removed, and valid sample data is retained. In this embodiment, the number of valid samples is no less than 8,000. Feature engineering was performed on the valid sample data to extract 12 core features. All features were then min-max standardized and mapped to the [0,1] interval to generate a standardized training sample set. The 12 core features specifically include: the ratio of the actual duration of the gap after deducting the connection time to the baseline duration of the new surgery; the matching degree between the gap time window and the available time window of the surgeon; the matching degree between the gap time window and the available time window of the anesthesiologist; the matching degree between the operating room resources corresponding to the gap and the resource requirements of the new surgery; the compatibility between the current workload of the medical staff corresponding to the gap and the workload of the new surgery; the absolute value of the priority difference between two adjacent surgeries in the gap; the matching degree of the specialty attributes of two adjacent surgeries in the gap; the proportion of the resource occupancy difference between two adjacent surgeries in the gap; the weight of the time node position of the gap within the daily scheduling cycle; the quantified value of the urgency of the new surgery; the maximum delay time of the original schedule after the gap is inserted; and the risk value of the daily working hours of medical staff exceeding the threshold after the gap is inserted.

[0045] The standardized training sample set is divided into a training set and a test set in a 7:3 ratio. The training set is used for model training, and the test set is used for model classification accuracy verification.

[0046] The second step is model building and training: A binary classification matching model based on Support Vector Machine (SVM) is built to output the degree of matching between gaps and new surgeries. The complete model configuration is as follows: Model input: 12-dimensional normalized insertion vector to be analyzed; Kernel function: The radial basis function (RBF) is used to solve linearly inseparable problems; Hyperparameter settings: Penalty coefficient C is set to 10, kernel function parameter gamma is set to 0.1, and class weights are set to balanced weights to avoid model bias caused by imbalanced samples; Model output: The output sample is the probability value of "acceptable match". The probability value is linearly mapped to the range of 0-100, which is the degree of acceptable match. The model is optimized using 5-fold cross-validation and grid search hyperparameters on the training set, and the classification accuracy is verified on the test set. When the model's classification accuracy is ≥92%, precision is ≥90%, and recall is ≥88%, training is stopped, the model hyperparameters are fixed, and the model is integrated into the system's algorithm layer.

[0047] The third step, the online inference process, involves the following complete steps during real-time online inference: For the target gaps in the locked gap distribution map, the above 12 core features are extracted, standardized, and an insertion vector to be analyzed is generated whose dimensions are completely matched with the model input. Input the insertion vector to be analyzed into a pre-built vector analysis matching model, and the model outputs a score of the acceptance matching degree of the corresponding gap; All locked gaps are sorted in descending order of their acceptance matching scores, and the gap with the highest score is selected as the optimal insertion candidate gap.

[0048] In this embodiment, the hospital HIS system uploads new surgical requirements for surgical needs analysis and resource requirement baseline prediction. The system automatically receives and analyzes the surgical requirements through the integration interface with the HIS system. The specific content includes: patient basic information (male, 52 years old, ASA II, no underlying diseases), preoperative diagnosis of gallstones with chronic cholecystitis, planned laparoscopic cholecystectomy, the surgeon is the associate chief physician of general surgery, the urgency level of the surgery is elective, the planned time is within the next 24 hours, and the surgeon and surgical team have not yet been determined.

[0049] The system extracts full-dimensional feature data of surgical needs, calls the pre-built surgical resource demand prediction model in the algorithm layer, and the model outputs the resource demand baseline data for the surgical needs: the baseline of surgical duration is 120 minutes (including preoperative preparation and postoperative recovery), the baseline of the legally required operating room handover, disinfection and transfer time is 30 minutes, the surgical procedure is adapted to a level III general surgery, the operating room cleanliness level requirement is Class 10,000, the required dedicated medical equipment is a laparoscopic system, and the baseline of medical staff qualification requirements is: the chief surgeon is a general surgery associate chief physician or above, the anesthesiologist has general anesthesia qualifications, and the circulating nurse and scrub nurse are general surgery specialist nurses.

[0050] Based on the initial resource requirements baseline, the system completes the subsequent operating room matching and scheduling process. Once the attending physician and surgical team are determined, the system automatically calls the fine-tuning prediction unit to correct the baseline data and ensure prediction accuracy.

[0051] Operating Room Matching and Booking: Based on the baseline data of resource requirements mentioned above, the system retrieves a list of available operating rooms that meet the access criteria for the next 24 hours from the server backend through its integration interface with the operating room management system. The access criteria are: the operating room has a cleanliness level of 10,000, is equipped with a laparoscopic system, and has a continuous idle time window of more than 120 minutes after deducting the legally mandated 30-minute turnaround time. The retrieved list of available operating rooms is not empty; there are two available operating rooms: Operating Room 3 (general surgery dedicated operating room) and Operating Room 5 (general operating room). The system calculates the matching degree between the two operating rooms and the baseline data of resource requirements. Operating Room 3, a general surgery dedicated operating room, has a matching degree of 98%, while Operating Room 5, a general operating room, has a matching degree of 85%. Operating Room 3, with the highest matching degree, is selected, and a booking tag for this surgical requirement is set for it. The booking result is simultaneously sent back to the hospital's HIS system and the operating room management system to complete the surgical scheduling.

[0052] If the list of available operating rooms is empty (e.g., all Class 10,000 clean operating rooms are full and there is no complete 120-minute continuous idle time window).

[0053] Reserved operating room range: Based on the baseline data of resource demand for surgery, the system locks out the range of reserved operating rooms that meet the basic requirements for surgical procedures, namely, general surgery operating rooms that meet the Class 10,000 cleanliness level and are equipped with laparoscopic systems. A total of 4 reserved operating rooms are locked out, namely operating room 1, operating room 2, operating room 3 and operating room 4. All of the above 4 operating rooms have scheduled surgeries in the next 24 hours and no complete 120-minute continuous idle time window.

[0054] Preprocessing of operating room scheduling data: The system obtains the current scheduling information of 4 operating rooms for the next 24 hours, extracts the scheduling priority, baseline duration of the operation, resource usage attributes, medical staff matching pool and compliance constraints of each scheduled operation, sets a unique identifier for each scheduled operation and stores it in the system's database. The length of each branch corresponds to the sum of the duration of the corresponding operation and the time of the connecting operation.

[0055] At the same time, the system receives real-time preoperative preparation status data from the anesthesia system. For surgical branches that have triggered the preoperative preparation status (the patient has entered the preoperative preparation room), an unadjustable lock tag is automatically set, excluding them from subsequent adjustments and screening.

[0056] The scheduling priority of scheduled surgeries is calculated using a weighted average of pre-set multi-dimensional adjustable weighted factors within the system. The calculation logic is as follows: four basic parameters are obtained: the quantified value of surgical urgency, the quantified value of patient perioperative risk, the surgeon's time window matching degree, and the operating room resource suitability degree. Corresponding weight coefficients are assigned to each of these four basic parameters, with the sum of all weight coefficients being 1. The scheduling priority of the corresponding surgery is obtained by multiplying each basic parameter by its corresponding weight coefficient and then summing the results. In this embodiment, the weight coefficients within the system are configured as follows: surgical urgency weight 0.4, patient perioperative risk weight 0.3, surgeon's time window matching weight 0.2, and operating room resource suitability weight 0.1. Each weight coefficient can be dynamically adjusted through the system management backend. The rules for assigning values ​​to the urgency metrics of surgeries are as follows: 100 for emergency surgeries, 60 for time-limited surgeries, and 30 for elective surgeries. The perioperative risk metric for patients is determined based on their ASA classification: 30 for ASA I-II, 60 for ASA III, and 100 for ASA IV-V. The surgeon's time window matching score ranges from 0 to 100, with 100 for a perfect time window match. The operating room resource suitability score ranges from 0 to 100, with 100 for a perfect resource match. In this embodiment, the scheduling priority for newly added surgeries is 47 points.

[0057] The system extracts the maximum allocation array corresponding to the currently uploaded surgical demand content for each unlocked and unbound independent branch in the scheduling tree. The maximum allocation array includes the maximum probability threshold for the corresponding branch to be replaced, the maximum duration threshold for the surgery to be postponed, and the personnel adaptation and compatibility threshold. In this embodiment, the admission thresholds set in the system are: the maximum probability threshold for replacement is ≥60%, the maximum duration threshold for the surgery to be postponed is ≥30 minutes, and the personnel adaptation and compatibility threshold meets the requirements.

[0058] In this embodiment, the personnel compatibility threshold determination takes into account both the job qualification matching degree of medical staff and the workload balance of the group, which solves the pain points of the existing mismatch between technical personnel and resources and the uneven distribution of workload of medical staff. The complete implementation steps are as follows: The first step is to calculate the job suitability score based on the collaborative matching algorithm: Construct a qualification feature database for medical staff: Based on the integration interface with the medical staff management system, obtain the full-dimensional qualification features of all medical staff in real time, including professional qualifications, specialties, professional titles, experience in assisting similar surgeries, and surgical job suitability levels (chief surgeon / assistant / anesthesiologist / circulating nurse / instrument nurse), and generate a unique standardized qualification feature vector for each medical staff member; Construct a demand feature vector for surgical positions: Extract the baseline of medical staff qualification requirements for newly added surgeries and corresponding scheduled surgeries, and generate a standardized demand feature vector for each surgical position; Bidirectional fit calculation: For each medical staff member in the matching pool of medical staff who have already scheduled surgeries in the corresponding branch, calculate their cosine similarity with the corresponding position for the new surgery and the corresponding position for the previously scheduled surgeries. The calculation formula is as follows: in, For the qualification feature vector of medical staff, This represents the feature vector of surgical job requirements. The cosine similarity score ranges from 0 to 1. Job suitability score generation: Multiply the average cosine similarity between the medical staff and the two surgeries by 100 to obtain the job suitability score of the medical staff. The score range is 0-100. Take the average of the job suitability scores of all medical staff in the matching pool for corresponding positions as the overall job suitability score. Level 1 Judgment: The system presets the compatibility threshold to 80 points. When the overall job compatibility score is ≥80 points, the job qualification matching is deemed to meet the requirements; otherwise, the personnel compatibility threshold is deemed not to meet the requirements.

[0059] The second step involves using the K-Nearest Neighbors (KNN) algorithm to cluster the similarity of workload characteristics among medical staff, and combining this with standard deviation to quantify the workload balance of the group. This solves the problem that standard deviation alone cannot identify local overload. The complete logic is as follows: Load feature vector construction: For all medical staff within the scheduling cycle, extract four core features: daily working hours, cumulative working hours, number of surgeries per day, and percentage of night surgeries within the scheduling cycle, and generate a standardized load feature vector. KNN clustering analysis: The K-nearest neighbor unsupervised clustering algorithm is used to cluster the load feature vectors of all medical staff. The number of clusters K is set to 5, and the medical staff are divided into 5 load level clusters: low load, normal load, critical load, high load, and overload load. Load balancing quantification calculation: Calculate the standard deviation of the ratio of actual working hours to rated working hours within the shift scheduling cycle of medical staff, and use it as the basic load balancing value. The value ranges from 0 to 1. The smaller the value, the more balanced the load distribution of the group. Clustering Correction Coefficient Calculation: Based on the KNN clustering results, the proportion of medical staff in high-load and overload clusters is calculated to generate the clustering correction coefficient. The calculation formula is as follows: ,in, For high-load clusters, The percentage of clusters under overload. The value range is 0-1; Final load balance calculation: Multiply the basic load balance value by the clustering correction coefficient to obtain the final workload balance, with a value range of 0-1; Secondary judgment: The system presets the workload balance threshold to 0.2. When the final workload balance is ≤0.2, the workload balance is judged to meet the requirements; otherwise, the personnel adaptation and compatibility threshold is judged to not meet the requirements.

[0060] The third step is to determine the personnel suitability and compatibility threshold. The personnel suitability and compatibility threshold is only determined to meet the requirements if the job suitability score is greater than or equal to the preset suitability threshold and the workload balance is less than or equal to the preset balance threshold. If either requirement is not met, the personnel suitability and compatibility threshold is determined to not meet the requirements.

[0061] In this embodiment, the weighting coefficients for the weighted calculation scenario, such as priority calculation, branch connection density, allowed insertion probability, comprehensive score, and multi-task loss function, are all determined using a unified scientific method. The complete steps are as follows: Constructing a hierarchical analysis structure: For the corresponding weighted calculation scenario, a three-level hierarchical analysis structure is constructed, consisting of a target layer, a criterion layer, and a scheme layer. The target layer represents the final goal of the weighted calculation (such as surgical scheduling priority), the criterion layer represents the corresponding basic parameters (such as surgical urgency, perioperative risk of patients, etc.), and the scheme layer represents different weight allocation schemes. Constructing a judgment matrix: An expert group consisting of no fewer than 5 experts, including the director of the hospital's operating room, the director of the anesthesiology department, senior head nurses, and medical affairs management personnel, was invited to score the relative importance of each parameter in the criterion layer pairwise using the 1-9 scale method to construct a judgment matrix. Consistency check: Calculate the largest eigenvalue of the judgment matrix and the consistency index CI, calculate the consistency ratio CR. When CR < 0.1, the judgment matrix passes the consistency check; otherwise, the judgment matrix is ​​readjusted. Weight coefficient calculation: The eigenvector corresponding to the largest eigenvalue of the judgment matrix is ​​calculated by the eigenvalue method, and after normalization, the initial weight coefficients corresponding to each basic parameter are obtained. Orthogonal experiment optimization: Design an orthogonal experiment table, with the initial weight coefficients as the center, set up a weight adjustment gradient of 5 levels, and use the resource utilization rate, clinical execution pass rate and compliance rate of the scheduling scheme as the experimental indicators, and determine the optimal combination of weight coefficients through range analysis; Weight coefficient fixing and dynamic adjustment: The optimal weight coefficient is fixed as the system default value. At the same time, an entry point for adjusting the weight coefficient is set in the system management backend, which is only open to the administrators of the hospital's medical affairs management department. It can be dynamically adjusted according to the hospital's business development and changes in clinical needs. After adjustment, consistency verification and trial operation verification are required again.

[0062] In this embodiment of the invention, the default weight coefficients for the core scenarios are set as follows, all determined using the methods described above: Surgical scheduling priority calculation: surgical urgency weight 0.4, patient perioperative risk weight 0.3, surgeon time window matching weight 0.2, operating room resource suitability weight 0.1; Branch connection density calculation: Time dimension weight 0.5, resource dimension weight 0.3, personnel dimension weight 0.2; Insertion probability calculation: Duration dimension weight 0.5, resource dimension weight 0.3, personnel dimension weight 0.2; Optimal array comprehensive score calculation: operating room resource matching degree weight 0.4, personnel suitability degree weight 0.4, surgery delay time weight 0.2; Multi-task loss function: regression branch weight 0.4, classification branch weight 0.3, multi-label classification branch weight 0.3.

[0063] In this embodiment, the branch connection density The calculation formula is: ,in, Adjust the coefficients for the priority dimension, and: The default value for the preset priority difference threshold is 20 points, which can be dynamically adjusted through the hospital management backend. This is the density value of the time dimension; = (the degree of matching between the cleanliness level of the operating room and the specialty suitability attributes of the pre- and post-operative surgeries + the compatibility of the dedicated medical equipment occupied by the pre- and post-operative surgeries) / 2, which is the resource dimension density value; =1 - Coefficient of variation of workload of the core medical team (surgeon, anesthesiologist, circulating / instrument nurse) before and after the two surgeries, where the coefficient of variation = standard deviation of workload / mean workload, i.e., personnel dimensional density value.

[0064] In this embodiment, threshold parameters such as admission threshold, insertion probability threshold, dispersion threshold, fit threshold, and balance threshold are all determined using a method combining historical data statistical distribution, clinical rule constraints, and the optimal critical point of the ROC curve. The complete steps are as follows: Historical data statistics: Obtain historical scheduling and surgical execution data of the target hospital over the past 1-2 years, statistically analyze the probability distribution of the corresponding indicators for the threshold parameters, and determine the reasonable distribution range of the indicators; Clinical hard rules constraints: Determine the hard boundaries of threshold parameters by combining the hospital's internal management system; Determining the optimal critical point of the ROC curve: Using the indicator as the test variable and the compliance and clinical feasibility of the scheduling plan as the state variables, the ROC curve is plotted, the Youden index is calculated, and the indicator value corresponding to the maximum value of the Youden index is taken as the initial value of the threshold parameter. Clinical trial validation: Substitute the initial threshold parameters into the system for a clinical trial run of no less than one month. Based on the trial run results, fine-tune and optimize the threshold parameters to ensure that the threshold parameters take into account resource utilization, clinical compliance, and feasibility. Threshold fixing and dynamic adjustment: The optimized threshold parameters are fixed as the system default values, while dynamic adjustment rules for emergency / elective surgeries are set: the admission threshold and insertion probability threshold for emergency surgeries are automatically reduced by 20% to prioritize the scheduling of emergency surgeries; the default thresholds for elective surgeries are maintained to ensure the stability of the original schedule.

[0065] The default values ​​for the core threshold parameters are set as follows, all determined using the methods described above: Maximum allocation array admission thresholds: maximum probability threshold for allowed replacement ≥ 60%, maximum allowed duration threshold for allowed surgery postponement ≥ 30 minutes, and personnel compatibility thresholds meet the requirements; Allowable insertion probability threshold: 60%; Feature parameter dispersion threshold: 0.3; Job suitability threshold: 80 points; Workload balance threshold: 0.2; The legally mandated time for disinfection and transfer of equipment in a Class 10,000 operating room is 30 minutes, and the legally mandated time for disinfection and transfer of equipment in a Class 100 operating room is 45 minutes.

[0066] The optimal array selection and branch matching process is based on the baseline data of resource requirements for the current surgical needs and a scheduling priority of 47 points. The system selects the optimal array from all maximum allocation arrays for the four available operating rooms, ensuring all parameters meet the admission threshold constraints. After selection, operating rooms 1, 2, and 4 have no arrays meeting the threshold constraints. Operating room 3 has one optimal array; the corresponding branch represents elective surgeries of the same priority, with a maximum allowed replacement probability of 75%, a maximum allowed postponement time of 60 minutes, and satisfactory personnel compatibility. The system counts the number of optimal arrays for all available operating rooms to be 1 (not 0), and initiates the optimal operating room matching and scheduling tree update system to select the unique optimal array from all optimal arrays. If the system marks operating room 3 as the first target operating room, a new branch is created at the branch position corresponding to the optimal array in the scheduling tree for operating room 3. The surgery corresponding to the original branch is postponed by 40 minutes, and the branch for the newly added surgery is inserted within the original time window, completing the scheduling tree update.

[0067] In this embodiment, the optimal array filtering and matching module also includes a progressive threshold relaxation and a fallback mechanism for suboptimal filtering: If the first round of screening does not find the best array that meets the admission threshold constraint, the threshold will be gradually relaxed: the admission threshold will be lowered by 10% in each round, with a maximum reduction of 30%. The array screening will be re-executed after each round of reduction until the best array that meets the conditions is selected; the maximum reduction limit of 30% can be exceeded for emergency surgery. If no array meets the criteria even when the threshold is relaxed to the maximum extent, the suboptimal array screening is performed: the top 3 suboptimal arrays with the fewest number of items that do not meet the threshold and the smallest parameter deviation value are selected and pushed to the system administrator for manual review. The arrays that pass the review are included in the subsequent comprehensive scoring screening range. If the second-best array screening still fails to produce a result that meets the requirements, the system will automatically trigger a multi-operating-room joint adjustment fallback process to ensure that the scheduling process is closed-loop and uninterrupted.

[0068] In this embodiment, if, after screening, none of the available operating rooms meet the threshold constraint in the optimal array, and the number is 0, then the following steps are executed: The first step involves the system determining the branch connection density of the scheduling tree for the four waiting operating rooms based on various scheduling indicators. These indicators include the continuity of surgical time windows, the continuity of operating room resource occupancy, the balance of medical staff workload, and the uniformity of surgical priority distribution. Each scheduling indicator corresponds to one dimension, generating a 4-dimensional multivariate indicator density sequence and constructing a 4×N multivariate indicator density matrix, where N is the number of time nodes within the corresponding scheduling period. In this embodiment, N is 24, corresponding to the hourly nodes in the next 24 hours. In this embodiment, the weighting coefficients within the system are configured as follows: time dimension weight 0.5, resource dimension weight 0.3, and personnel dimension weight 0.2.

[0069] The second step involves the system performing adjacent element connection analysis on the multivariate index density matrix to locate potential gaps between adjacent surgical branches. After deducting the legally mandated 30-minute handover and disinfection transition time, the effective available time of the gap is verified. When the effective available time of the gap is less than the baseline of the new surgical duration, the allowable insertion probability is directly set to 0. The calculation logic is as follows: the ratio of the effective available time of the gap to the baseline of the new surgical duration, the ratio of the matching degree between the operating room resources corresponding to the gap and the new surgical demand to the standard value of the new surgical resource demand, and the ratio of the compatibility between the time window of the gap and the new surgical demand to the standard value of the new surgical personnel demand are obtained respectively. Corresponding weight coefficients are assigned to the three ratios respectively, and the sum of all weight coefficients is 1. The allowable insertion probability is obtained by multiplying the three ratios by the corresponding weight coefficients and summing them. In this embodiment, the weight coefficients in the system are configured as follows: 0.5 for the duration dimension, 0.3 for the resource dimension, and 0.2 for the personnel dimension, with a preset insertion probability threshold of 60%. The system locks gaps with an insertion probability greater than 60%, generating a locked gap distribution map corresponding to each multivariate index density matrix. The effective available time of a potential gap interval = the original time of the potential gap interval - the statutory handover and disinfection transfer time for the corresponding operating room. The potential gap interval refers to the continuous blank time interval between the time windows of two adjacent scheduled surgeries on the scheduling timeline of a single operating room. The statutory handover and disinfection transfer time is predetermined, as shown in Table 1. Table 1. Statutory Disinfection and Transfer Time Table

[0070] The third step involves the system performing an insertion vector transformation on the target gap in each locked gap distribution map, extracting 12-dimensional core features, generating an insertion vector to be analyzed, and inputting the insertion vector to be analyzed into the system's pre-built vector analysis and matching model. The model outputs the acceptance matching degree corresponding to each insertion vector, with a value range of 0 to 100.

[0071] The fourth step involves the system selecting the operating room with the highest acceptance matching degree as the second target operating room, and using the gap with the highest allowable insertion probability in the second target operating room as the branch insertion point. The new branch is established and the scheduling tree is updated in the scheduling tree. No scheduled surgeries need to be adjusted throughout the process, achieving zero-disturbance insertion.

[0072] If, after screening, none of the available operating rooms have a matching locking gap, and the number is 0, then the subsequent minimum disturbance adjustable branch matching process is triggered.

[0073] In this embodiment, the minimum perturbation adjustable branch matching process is as follows: When no suitable zero-disturbance gap is found, the system executes the optimal array filtering and matching process, the complete steps of which are as follows: Based on the baseline data of resource requirements for the current surgical needs and the scheduling priority of 47 points, the system selects the best array from the maximum allocation array of all unlocked and unbound branches of the 4 waiting operating rooms, where all parameters meet the admission threshold constraints.

[0074] After screening, no arrays satisfying the threshold constraints were found in operating rooms 1, 2, and 4. Operating room 3 had one optimal array, and the corresponding branch of the surgery was an elective surgery of the same priority. The maximum probability of replacement was 75%, the maximum allowed delay was 60 minutes, and the personnel adaptability and compatibility met the requirements.

[0075] The system selects the unique best array from all the best arrays. For example, if the system marks operating room 3 as the first target operating room, it executes the scheduling tree branch update process.

[0076] In this embodiment, a comprehensive score is calculated based on a weighted average of operating room resource matching degree, personnel suitability, and surgery delay time: The overall score is calculated as follows: Operating room resource matching weight × resource matching score + personnel suitability weight × personnel suitability score - surgery delay time weight × (Ts / Tmax) × 100. If the actual planned delay time Ts of the scheduled surgery corresponding to the target branch is greater than the preset maximum allowed surgery delay time threshold Tmax in the maximum allocation array of that branch, it will be directly excluded from the optimal array screening range and will not be included in the score. When Ts=0 (no need to postpone scheduled surgeries), Ts / Tmax=0, and there is no negative deduction. When Ts = Tmax, Ts / Tmax = 1, reaching the maximum deduction for this item.

[0077] For example, the system has completed the preliminary processes of the original solution, and the basic boundary conditions are as follows: New surgery to be inserted: Emergency laparoscopic repair of gastrointestinal perforation. The resource requirements baseline output by the surgical resource demand prediction model of this system are as follows: surgery duration baseline 60 minutes, requires a Class 10,000 clean operating room, core equipment is a laparoscopic system, the chief surgeon must be an associate chief physician or above in general surgery, the urgency of the surgery is emergency, and the scheduling priority is 92 points.

[0078] Pre-process results: After system screening, there are no available operating rooms that meet the admission criteria and no optimal array that meets the threshold constraints. The gap insertion module process has been triggered.

[0079] Basic information on the target gap: The system has completed the multivariate index density matrix analysis of the waiting operating room (Class 10,000 clean operating room for general surgery, No. 3), and identified the target gap between two adjacent scheduled surgeries in the scheduling tree of this operating room. The specific information is as follows: The scheduled surgery is an elective laparoscopic cholecystectomy, with a time window of 08:00-10:00 on the same day. The scheduling priority is 47 points. The specialty is general surgery, and the core equipment used is the laparoscopic system. The following surgery has been scheduled: elective tension-free inguinal hernia repair, with a time window of 11:15-12:30 on the same day, a scheduling priority of 42 points, a specialty of general surgery, and using core equipment and basic surgical instruments. Target gap time window: 10:00-11:15 on the same day, actual gap duration 75 minutes, no locked resources occupied, no scheduling conflicts.

[0080] All extracted 12-dimensional core features were uniformly mapped to the [0,1] interval using the min-max normalization method, perfectly matching the input dimension requirements of the vector analysis matching model. The normalization formula is as follows: ,in, The original value of the feature; This is the minimum value of this feature based on 3 years of historical clinical data from the hospital. To correspond to the maximum value of the statistics, ensure that the range of feature values ​​is completely consistent with the model training samples.

[0081] For the 12 core features, extraction, calculation, and standardization were completed dimension by dimension, as shown in Table 2: Table 2 12-dimensional core feature table

[0082] After completing the extraction and standardization of the above 12 core features, the 12 standardized values ​​are sorted by fixed feature numbers and combined in sequence to generate a one-dimensional vector whose dimensions are completely consistent with the input layer of the pre-built vector analysis matching model. The generated vectors are dimension-validated to confirm that they have 12 dimensions and all values ​​are within the range of [0,1]. After the validation is passed, they are marked as vectors to be inserted for analysis.

[0083] The final generated insertion vector to be analyzed is: V=[0.4167,1.0,1.0,1.0,0.92,0.05,1.0,0.5,0.9,1.0,0.0,0.0]; The above-mentioned insertion vector V to be analyzed is input into the vector analysis and matching model pre-built in this system. The model outputs that the acceptance matching degree corresponding to the gap is 94.2 points, which is the highest score among all locked gaps. Based on this, the system marks operating room No. 3 as the second target operating room, uses the gap as the branch insertion point, and completes the establishment of a new branch and the update of the scheduling tree.

[0084] In this embodiment, the scheduling tree branch update mechanism, based on original branch archiving and retention, version ID binding, and transactional atomication, solves the pain points of existing technologies where scheduling adjustments directly modify the original data, causing large disturbances and making rollback impossible. The complete implementation steps are as follows: The first step involves the dual-branch temporary structure construction system locking the time window position and full-dimensional attribute information of the original branch of the scheduling tree corresponding to the final array. The original branch is completely archived and stored in the historical version repository without any modification or destruction. The dual-branch temporary structure construction is then performed: New branch creation: Based on the standardized scheduling sub-segment information of the surgery to be inserted, a new branch is created at the target branch position corresponding to the final array. The new branch is bound to a unique surgery ID corresponding to the surgery to be inserted, and the initial time window of the new branch is the original time window of the original branch. Mirror branch generation: Based on the adjustment rules of the final array, an adjusted mirror branch of the original branch is generated. The mirror branch is bound to a unique version ID associated with the original surgery ID, does not share the same ID with the original branch, inherits the full-dimensional attribute information of the original branch, and only modifies adjustment items such as time window, resource usage, and personnel matching according to the adjustment rules. Dual-branch structure construction: The new branch and the adjusted mirror branch are set to share the same time axis and adjacent non-overlapping time windows, forming a parallel dual-branch temporary structure. The dual-branch structure is completely isolated from other branches of the scheduling tree and is only formally connected to the scheduling tree after pre-synchronization is passed, avoiding data pollution to the original scheduling tree.

[0085] The second step involves the dual-branch pre-synchronization and transaction binding system performing a full-dimensional pre-synchronization check on the dual-branch temporary structure, strongly binding the pre-synchronization with cross-system data synchronization transactions. The core content and execution steps of the pre-synchronization are as follows: Timeline pre-synchronization: Verifying whether the time windows of the new branch and the mirror branch have no overlap or gaps, whether there are no time conflicts with other branches of the scheduling tree, and whether the total length of the timeline matches the time window length of the original branch, ensuring the continuity and compliance of the timeline; Resource occupancy pre-synchronization: Verifying whether the operating room resources and dedicated medical equipment occupancy requirements of the new branch and the mirror branch match the resource capacity of the corresponding operating room, whether there are no conflicts with the resource occupancy of other branches, and synchronously updating the resource occupancy tags and locking status; Personnel matching pre-synchronization: Verifying whether the medical staff matching pools of the new branch and the mirror branch meet the qualification requirements. The process involves several steps: First, checking for overlapping time windows and ensuring the workload of medical staff meets compliance requirements. This includes updating the personnel matching pool and workload statistics. Second, compliance pre-synchronization involves performing a first-level compliance check on the verification unit containing the dual-branch structure to ensure there are no compliance violations and that branch updates do not trigger compliance issues. Third, cross-system transaction pre-synchronization involves synchronizing pre-synchronized data to all connected systems and verifying that each system can receive and execute the data. Pre-synchronization is only considered successful if all systems pass the verification. Fourth, pre-synchronization result determination involves verifying that all pre-synchronization content passes the verification. If any verification fails, pre-synchronization is considered a failure, the dual-branch temporary structure is automatically destroyed, no data in the original scheduling tree is modified, and the process returns to the optimal array filtering process to re-filter the suboptimal array.

[0086] The third step involves the atomic activation and lifecycle management of the dual-branch structure. Atomic activation: After successful pre-synchronization, the system performs an atomic commit operation, completing the following in one go: officially connecting the mirror branch to its corresponding position in the scheduling tree, officially connecting the new branch to its corresponding position in the scheduling tree, updating the sorting of all branches in the scheduling tree, and synchronously updating the locked status of all resources and personnel. During the atomic commit process, if any abnormality occurs, a full rollback is automatically triggered, restoring the original scheduling tree state to prevent data corruption. Lifecycle management: The lifecycle of the dual-branch temporary structure is a maximum of 30 seconds. If pre-synchronization and atomic activation are not completed within 30 seconds, the structure is automatically destroyed, releasing locked resources to prevent long-term resource occupation. Operation log retention: The entire process of branch updates, including original branch archive information, adjustment rules, pre-synchronization results, effective time, and operators, is all retained in the system operation log, which is tamper-proof, traceable, and auditable.

[0087] Fourth, after the abnormal rollback mechanism branch update takes effect, if any violations are found during the subsequent full-cycle compliance verification, or if the system receives objections to the adjustment from clinical departments, the system can perform a rollback operation with one click: destroy the new branch and the mirror branch, retrieve the archived original branch data from the historical version repository to complete the recovery, release all resources locked by the new branch at the same time, and completely restore the scheduling tree state before the branch update, ensuring the stability and traceability of the scheduling plan.

[0088] In this embodiment, a full-process hierarchical compliance verification mechanism based on first-level hard rule compliance judgment, second-level adjustability judgment, closed-loop adjustment, and rollback verification solves the pain point of existing technology compliance verification, which can only report errors and cannot form a closed loop. The complete implementation steps are as follows: First, the scheduling cycle division and first-level adaptability judgment system performs full-scale hierarchical compliance verification on the full-cycle scheduling scheme corresponding to the updated scheduling tree. The following steps are executed first: Verification unit division: The scheduling cycle (default 24 hours) is divided into multiple consecutive verification units. In this embodiment, it is divided into 12 consecutive 2-hour verification units, which can also be adjusted to 1-hour / 4-hour verification units according to the hospital's needs; First-level judgment indicator setting: The first-level adaptability judgment is a hard compliance red line indicator clearly stipulated by national regulations and hospital systems, specifically including: whether the cumulative working hours of medical staff per day do not exceed 8 hours, and when working overtime in the emergency department. The maximum working hours should not exceed 4 hours, and the maximum cumulative working hours per day should not exceed 12 hours; the continuous daily occupation of the operating room should not exceed 14 hours, and the cumulative daily occupation should not exceed 18 hours; the priority ranking of surgeries should comply with the rules of emergency priority and high-risk patient priority, and there should be no situation where low-priority surgeries occupy high-priority surgical resources; the qualifications of medical staff should meet the requirements of surgical positions, and there should be no situation where they practice beyond their qualifications; the cleanliness level and equipment configuration of the operating room should meet the requirements of surgery, and there should be no situation where the hardware is incompatible; the time for disinfection and transfer between operating rooms should comply with the hospital infection control regulations, and whether the locked branches have been adjusted in violation of regulations; Level 1 judgment rule: For the scheduling data in each verification unit, check the above indicators item by item. When all indicators meet the requirements, the verification unit is judged to be qualified in Level 1; when any indicator does not meet the requirements, the verification unit is judged to be unqualified and enters the Level 2 judgment process.

[0089] The second step, the secondary adjustability judgment, is performed on the verification units that failed the primary judgment. The complete steps are as follows: Feature parameter dispersion calculation: Extract the core feature parameters of all scheduling nodes within the unqualified verification unit, including four core parameters: operation duration, operation priority, medical staff working hours, and operating room resource occupancy rate. Calculate the feature parameter dispersion, where dispersion = standard deviation / mean. Secondary judgment rules: The system presets a dispersion threshold of 0.3. When the feature parameter dispersion ≤ the preset dispersion threshold, the unqualified verification unit is judged as an adjustable unit that can meet compliance requirements through local adjustments; when the feature parameter dispersion > the preset dispersion threshold, the unit is judged as an unqualified unit that cannot be adjusted. Judgment result routing: Adjustable units enter a local closed-loop adjustment process; for unqualified units that cannot be adjusted, the system automatically triggers a scheduling scheme rollback mechanism, retrieving the archived original branch data from the historical version repository, rolling the scheduling tree back to its state before the update, and synchronously returning to the scheduling matching process to re-execute the matching until a compliant scheduling scheme is generated.

[0090] The third step involves a local closed-loop adjustment process for adjustable non-conforming verification units. The system automatically performs local closed-loop adjustments, with the following complete steps: Violation Location: The system automatically locates the specific violation point within the non-conforming verification unit, identifying the violation type, the relevant scheduling nodes, and the deviation value of the violation indicator; Adjustment Priority Ranking: Based on the violation type, the adjustment priority is determined, from highest to lowest: Time Window Fine-tuning > Medical Staff Replacement > Surgery Inter-Operating Room Adjustment > Surgery Priority Rearrangement; Minimum Disturbance Adjustment: Local adjustments are performed starting with the least disruptive adjustment method according to the adjustment priority: Time Window Fine-tuning: Without changing the surgical sequence, medical staff, or operating room, the time window of adjacent surgeries is fine-tuned within ±30 minutes to eliminate violations; Medical Staff Replacement: If time window fine-tuning cannot resolve the violation, adjustments are made based on the medical staff's working hours. For violations exceeding the limits, the system will match qualified alternative medical personnel to replace those exceeding the allotted time, thus eliminating the violation. For cross-operating room adjustments: if the above adjustments fail to resolve the violation, low-priority surgeries within the verification unit will be rescheduled to the corresponding time window in other operating rooms of the same suitability level, eliminating the violation. For priority reordering: only for priority conflicts between emergency and elective surgeries, the surgical priority will be reordered to prioritize emergency surgeries and adjust the time window for elective surgeries. For secondary verification after adjustment: after each adjustment, the verification unit will undergo a first-level compliance assessment. If the assessment is satisfactory, the adjustment will stop, and the scheduling tree will be updated. If the assessment still fails, the next level of adjustment will be implemented. For multiple rounds of adjustments as a fallback: if none of the above four adjustment methods can eliminate the violation, the system will automatically mark the unit as unadjustable, triggering a scheduling rollback mechanism and re-executing the scheduling matching process.

[0091] The fourth step is to solidify the full-cycle compliance verification. After all verification units within the scheduling cycle pass the first-level compliance judgment, the system will solidify the updated scheduling tree into the official scheduling data in the database. At the same time, the final scheduling results will be sent back to the hospital's HIS system, operating room management system, and medical staff management system to complete the closed loop of the entire scheduling process. Meanwhile, a scheduling compliance report will be generated and pushed to the hospital's medical affairs management department for archiving.

[0092] In this embodiment, a complete closed-loop processing solution is designed for all extreme abnormal scenarios in the core scheduling process. This ensures that the system can output a compliant and executable scheduling solution under any scenario, solving the pain points of insufficient emergency dispatch capability and inability to close the process in existing technologies. The processing solutions for each extreme scenario are as follows: Extreme Scenario 1: Processing solution for an empty range of waiting operating rooms. When the system locks the range of waiting operating rooms based on the basic adaptation requirements of the surgical procedure, if there is no operating room that meets the basic adaptation requirements of the surgical procedure, i.e., the range of waiting operating rooms is empty, the following closed-loop processing flow is executed: Triggering the hardware adaptation degradation mechanism: The system automatically performs graded degradation on the baseline data of resource requirements for new surgeries. First, the hard constraints of non-core dedicated medical equipment are removed, and only core surgical equipment is retained. The system adheres to strict cleanliness requirements, re-identifying the range of available operating rooms. If the range remains empty after downgrading, the system automatically triggers a cross-specialty operating room adaptation mechanism. Based on the cleanliness level and core equipment requirements, it matches operating rooms from other specialties that meet the hardware requirements, thus re-identifying the range of available operating rooms. If no operating rooms are available after cross-specialty adaptation, the system automatically triggers an emergency priority decision: if the new surgery is an emergency surgery, it is immediately pushed to the hospital's operating room management department and medical affairs department, triggering a manual emergency dispatch channel, and simultaneously outputting a list of all adjustable scheduled elective surgeries and adjustment suggestions; if the new surgery is an elective surgery, the system automatically pushes suggestions for adjusting the surgery time window, recommending the optimal available time window within the next 3 working days, and simultaneously providing feedback to the surgeon requesting the surgery.

[0093] Extreme Scenario 2: Handling Scheme for No Eligible Insertion Gap and No Eligible Optimal Array After system screening, if all waiting operating rooms have no zero-disturbance gaps that meet the requirements and no optimal array that meets the admission threshold constraints, the following closed-loop processing flow is executed: Trigger the dynamic threshold adjustment mechanism: Based on the scheduling priority of new surgeries, gradually lower the admission threshold and insertion probability threshold, lowering them by 10% at each level, with a maximum reduction of no more than 30%; for emergency surgeries, the maximum reduction can be relaxed to 50%. After each level of reduction, the gap locking and optimal array screening process is re-executed; if no match is found after the threshold is lowered... As a result, the system automatically triggers a multi-operating-room joint adjustment mechanism: breaking the scheduling constraints of a single operating room, and based on the resource demand baseline for new surgeries, performing batch fine-tuning of scheduled surgeries in 2-3 waiting operating rooms across operating rooms, generating multiple sets of cross-operating-room adjustment plans, calculating the comprehensive disturbance rate (the change ratio of the original schedule) of each plan; selecting the plan with the lowest comprehensive disturbance rate, and pushing it to the operating room management department for manual review; after approval, executing batch updates of the cross-operating-room scheduling tree to complete the scheduling of new surgeries; if the review fails, triggering the manual emergency dispatch channel, and simultaneously outputting adjustment suggestions.

[0094] Extreme Scenario 3: Priority Conflict Handling Solution for Multiple High-Priority Emergency Surgeries Triggered Simultaneously. When the system receives scheduling requests for two or more high-priority emergency surgeries within the same time period, a multi-emergency priority conflict handling process is triggered: Secondary Priority Quantification and Ranking: Based on the quantified values ​​of surgical urgency and patient perioperative risk, multiple emergency surgeries are prioritized a second time. The ranking rule is: surgical urgency weight 0.6, patient perioperative risk weight 0.4, with higher scores indicating higher priority; Exclusive Resource Locking Mechanism: According to the priority order of the secondary ranking, the scheduling matching process is executed for each emergency surgery in turn, and the matched lock is locked. The system automatically sets exclusive lock tags for designated operating rooms, medical staff, and specialized equipment resources, preventing these locked resources from being used in subsequent low-priority emergency surgery matching processes. A parallel scheduling mechanism is in place: if multiple operating rooms meet the requirements, the system simultaneously performs parallel scheduling matching for multiple emergency surgeries, synchronously locking resources and updating the scheduling tree to ensure that the scheduling response time for emergency surgeries is ≤20 seconds. Conflict fallback handling: if resources cannot simultaneously meet the needs of all emergency surgeries, the system automatically sorts them by priority, pushing alternative operating rooms and time windows to low-priority emergency surgeries, simultaneously pushing these suggestions to the medical affairs department and operating room management department, triggering manual coordination and scheduling. For emergency surgeries, independent threshold flexible adaptation rules are set up and strongly bound to the emergency priority scheduling mechanism: the admission threshold and insertion probability threshold for emergency surgeries are automatically reduced by 20%, and the maximum reduction for life-threatening emergency surgeries can be relaxed to 50%; the personnel suitability and compatibility determination for emergency surgeries follows the rule of "rigid compliance for core positions and flexible exemption for non-core positions": the job suitability of the chief surgeon and anesthesiologist must meet the preset threshold, while the suitability threshold for non-core positions such as circulating nurses and scrub nurses can be reduced by 20%, and the workload balance threshold can be increased to 0.35, without the rigid constraint of simultaneously meeting two thresholds; throughout the entire emergency surgery scheduling matching process, the priority determination weight is automatically adjusted to: surgical urgency weight 0.6, patient perioperative risk weight 0.3, and the other two weights combined 0.1, prioritizing the resource allocation for emergency surgeries.

[0095] Extreme Scenario 4: Conflict Handling Solution for Complete Overlapping Medical Staff Between New and Scheduled Surgeries When the key medical staff, such as the chief surgeon and core anesthesiologist, for a new surgery completely overlap with the matching pool of medical staff for already scheduled surgeries, making personnel matching impossible, the following process is executed: Matching with Replacement Personnel of the Same Qualification: The system automatically filters replacement medical staff from the hospital's medical staff qualification database who perfectly match the original medical staff in terms of qualifications, specialty, and job suitability, generates a replacement personnel list, and calculates the job suitability score between the replacement personnel and the new surgery; if there are replacement personnel with a suitability score ≥ a preset threshold, the system automatically... Replacement personnel are added to the pool of medical staff for newly added surgeries, and the scheduling matching process is re-executed. If no suitable replacement personnel are found, the system automatically triggers the surgery time window staggering adjustment mechanism, calculates the minimum staggering time between the newly added surgery and the already scheduled surgeries, and generates a time adjustment plan for the already scheduled surgeries to ensure that the same medical staff will not have overlapping time windows. The adjustment plan must meet the maximum allowable threshold for the surgery postponement time. If the staggering adjustment still cannot resolve the conflict, the system automatically pushes the conflict details to the surgeon requesting the surgery and the operating room management department, recommends a list of replacement medical staff and alternative surgery time windows, and completes the scheduling adjustment after manual review and confirmation.

[0096] This embodiment successfully integrated and implemented an intelligent operating room scheduling information system. Compared to existing technologies and the hospital's original manual scheduling model, it achieved the following technical effects. All data is based on three consecutive months of clinical trial data from the target hospital, with the original manual scheduling model as the benchmark. The sample size covers 1200 surgeries: System integration and docking effect: Seamless docking with five core business systems of the hospital was achieved, enabling real-time data exchange and eliminating the need for repetitive manual data entry, thus improving the operational efficiency of medical staff by more than 60%; Improved scheduling efficiency: The scheduling matching time for a single new surgery was reduced from 30 minutes manually to less than 20 seconds, improving scheduling matching efficiency by 70%; Improved resource utilization: Operating room resource utilization increased by 29%, and the change rate of the original scheduling plan was controlled within 5%, balancing resource utilization with the stability of clinical work; staff workload balance improved: the workload balance of medical staff improved by 38%, with no compliance issues of medical staff exceeding the daily working hours limit, and the scheduling compliance rate reached 98.7%; emergency dispatch capability improved: the emergency dispatch response time for emergency surgery was shortened to within 20 seconds, with dynamic adjustment capability to meet the hospital's clinical emergency needs; medical safety assurance: the preoperative preparation status was locked, and the hospital infection compliance control for the handover time was achieved. During the trial operation, there were no violations in adjusting the prepared surgery or the handover time not in compliance with hospital infection control regulations, and the medical safety assurance capability was greatly improved.

[0097] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An intelligent scheduling system for operating rooms, characterized in that, include: The baseline prediction module is used to receive the uploaded surgical requirements based on the integrated interface, extract the full-dimensional feature data of the surgical requirements, call the pre-built surgical resource requirement prediction model, and output the resource requirement baseline data corresponding to the surgical requirements. The resource requirement baseline data includes the surgical duration baseline, the legally mandated table-opening disinfection and transfer time baseline, the surgical procedure adaptation level, the operating room cleanliness level requirements, the occupancy requirements of special medical equipment, and the baseline of medical staff qualification requirements. The operating room reservation module is used to retrieve a list of available operating rooms that meet the access criteria from the server backend based on the resource demand baseline data. The access criteria are the cleanliness level of the operating room, equipment configuration, and the requirement that the idle time window after deducting the statutory reception, disinfection and transfer time completely covers the resource demand baseline data. If the list of available operating rooms is not empty, calculate the matching degree between each operating room in the list of available operating rooms and the baseline data of resource demand, filter the operating room with the highest matching degree, and set the reservation tag for the corresponding surgical demand; The operating room range locking module is used to lock the range of available operating rooms that meet the basic adaptation requirements of the surgical procedure based on the resource requirement baseline data of the surgical needs when the list of available operating rooms is empty. The range of available operating rooms includes at least one available operating room, and the basic adaptation requirements of the surgical procedure are the hard requirements of the operating room's cleanliness level and equipment configuration meeting the resource requirement baseline data. The tree construction module is used to obtain the current scheduling information of each waiting operating room within the scope of the waiting operating rooms, construct the tree structure scheduling tree of the corresponding waiting operating rooms, and set an inseparable parent branch for consecutive surgeries by the same surgeon. At the same time, it sets an unadjustable lock tag for the surgical branch that has triggered the preoperative preparation state, and extracts the maximum allocation array of the currently uploaded surgical demand content corresponding to each branch in the scheduling tree. The maximum allocation array is a set of quantitative parameters that characterize the corresponding scheduled surgical branch to be compatible with the insertion and / or adjustment of new surgeries. Each parameter in the maximum allocation array is set with a corresponding admission threshold. The optimal array filtering and matching module is used to filter the optimal array from all the maximum allocation arrays of all available operating rooms based on resource demand baseline data and scheduling priority, so that all parameters meet the admission threshold constraints. The scheduling tree update module is used to count the number of optimal arrays of all available operating rooms. If the number is not 0, it controls the optimal operating room matching module to start working to obtain the first target operating room for scheduling tree update; if the number is 0, it controls the gap insertion module to start working to obtain the second target operating room for scheduling tree update.

2. The intelligent operating room scheduling system according to claim 1, characterized in that, The optimal operating room matching module is used to calculate a comprehensive score from the best array of all available operating rooms based on the operating room resource matching degree, personnel suitability, and surgery delay time, and select the array with the highest comprehensive score as the final array. The available operating rooms corresponding to the final array are marked as the first target operating rooms. In the scheduling tree of the first target operating rooms, a new branch is created at the branch position corresponding to the final array to complete the update of the scheduling tree.

3. The intelligent operating room scheduling system according to claim 1, characterized in that, The gap insertion module is used to determine the scheduling tree for each operating room to be used, and calculate the density value of each scheduling indicator corresponding to each time node within the scheduling cycle based on each scheduling indicator. It generates a multivariate indicator density sequence corresponding to each scheduling indicator. With each scheduling indicator as a row dimension, all multivariate indicator density sequences are stacked row by row to construct an M×N multivariate indicator density matrix, where N is the total number of time nodes within the scheduling cycle; M is the number of scheduling indicators, and each scheduling indicator corresponds to one dimension of the multivariate indicator density matrix. The scheduling indicators include the continuity of surgical time windows, the continuity of operating room resource occupancy, the balance of medical staff workload, and the uniformity of surgical priority distribution. Adjacent element connection analysis is performed on the multivariate index density matrix to locate potential gap intervals between adjacent surgical branches after deducting the statutory disinfection and transfer time. After deducting the statutory disinfection and transfer time preset for the corresponding cleanliness level operating room, the effective usable time of the gap is verified. The matrix sub-block data corresponding to each potential gap interval is extracted, the branch connection density of the corresponding potential gap is calculated, and the allowable insertion probability of the corresponding gap is determined based on the branch connection density. Gaps with allowable insertion probability greater than the preset insertion probability threshold are locked, and a locked gap distribution map corresponding to each multivariate index density matrix is ​​generated. For each target gap in the locked gap distribution map, the insertion vector is transformed to generate the insertion vector to be analyzed. The insertion vector to be analyzed is input into the pre-constructed vector analysis matching model, and the acceptance matching degree corresponding to each insertion vector is output. The operating room with the highest acceptance matching degree is selected as the second target operating room. The gap with the highest allowed insertion probability in the second target operating room is used as the branch insertion point to complete the establishment of new branches and the scheduling tree update.

4. The intelligent operating room scheduling system according to claim 1, characterized in that, The tree construction module includes: The extraction unit is used to extract the scheduling priority, baseline surgery duration, legally mandated handover, disinfection and transfer time, resource occupancy attributes, medical staff matching pool, compliance constraints, consecutive surgery binding identifier, and preoperative preparation status lock identifier for each scheduled surgery in the current scheduling information. It sets a scheduling sub-segment with a corresponding unique identifier for each scheduled surgery, where each scheduling sub-segment corresponds one-to-one with a single scheduled surgery. The tree generation unit is used to automatically generate a tree-structured scheduling tree for each available operating room, based on all scheduling segments corresponding to each available operating room and according to the order of the surgical time windows. The scheduling tree is a multi-branch tree structure with the operating room as the root node, the scheduling segment of a single scheduled surgery as the first-level branch, the parent branch of consecutive surgeries as the second-level root node, the scheduling segment of the consecutive surgeries as the third-level branch, and the order of the surgeries as the branch sorting rule. The length of each branch corresponds to the sum of the duration of the corresponding surgery and the duration of the connecting surgeries, and the attributes of the branches correspond to the full-dimensional information of the scheduling segment.

5. The intelligent operating room scheduling system according to claim 1, characterized in that, Also includes: The first-level judgment module is used to perform hierarchical compliance verification on the full-cycle scheduling scheme corresponding to the updated scheduling tree. It divides the scheduling cycle into multiple consecutive verification units and performs first-level adaptability judgment on the scheduling data in each verification unit. The secondary judgment module is used to determine that the corresponding verification unit is qualified when all the judgment indicators of the primary adaptability judgment meet the requirements; otherwise, the corresponding verification unit is determined to be unqualified. At this time, the unqualified verification unit is subject to secondary judgment. The rule of the secondary judgment is as follows: when the feature parameter dispersion of each scheduling node in the corresponding unqualified verification unit is less than or equal to the preset dispersion threshold, the corresponding unqualified verification unit is determined to meet the compliance requirements through local adjustment; otherwise, the corresponding unqualified verification unit is determined to be an unqualified unit that cannot be adjusted. The feature parameter dispersion is the ratio of the standard deviation to the mean of the core feature parameter of each scheduling node in the verification unit.

6. The intelligent operating room scheduling system according to claim 1, characterized in that, Also includes: The priority calculation module is used to calculate the scheduling priority of scheduled surgeries by weighting them with preset multi-dimensional adjustable weight configuration factors, including: The weight acquisition unit is used to acquire four basic parameters: surgical urgency quantification value, patient perioperative risk quantification value, surgeon time window matching degree, and operating room resource suitability degree. It assigns corresponding weight coefficients to the four basic parameters, and the sum of all weight coefficients is 1. The multiplication unit is used to multiply each basic parameter with its corresponding weight coefficient and then sum them to obtain the score of the corresponding surgery. The score is then matched with the score-priority table to obtain the scheduling priority.

7. The intelligent operating room scheduling system according to claim 1, characterized in that, The tree construction module includes: The scoring determination unit is used to calculate the job suitability score of each medical staff member and the two surgeries based on the matching pool of medical staff members for the corresponding branch and the baseline of medical staff qualification requirements for the current surgeries. The job suitability score ranges from 0 to 100. The similarity calculation unit is used to calculate the workload balance of the medical staff group using the K-nearest neighbor algorithm, with a value ranging from 0 to 1. The judgment unit is required to determine that the personnel suitability and compatibility threshold meets the requirements when the job suitability score is greater than or equal to the preset suitability threshold and the workload balance is less than or equal to the preset balance threshold; otherwise, it is determined that the requirements are not met. The personnel suitability and compatibility threshold is one of the admission thresholds.

8. The intelligent operating room scheduling system according to claim 1, characterized in that, The optimal operating room matching module includes: A new unit is created to lock the time window position and attribute information of the original branch of the scheduling tree corresponding to the final array. Without modifying the original data of the original branch, a new branch to be effective is created at the target branch position based on the scheduling sub-segment information of the surgery to be inserted. Simultaneously, an adjusted mirror branch of the original branch is generated based on the adjustment rules of the final array. The new branch and the adjusted mirror branch share adjacent non-overlapping time windows on the same time axis, forming a parallel dual-branch temporary structure for pre-synchronization. The synchronous update unit is used to bind a new branch to a standardized scheduling sub-segment that uniquely corresponds to the surgery to be inserted. It synchronously updates the operating room resource occupancy tag, medical staff matching pool, dedicated medical equipment lock status, and surgery priority attribute corresponding to the new branch. At the same time, it synchronously updates the time window attribute, resource occupancy attribute, and personnel matching attribute of the adjusted mirror branch.

Images