Dynamic iterative optimization hospital examination resource scheduling system, method and device

Abstract

The invention provides a hospital examination resource scheduling system, method and device capable of dynamically iteratively optimizing, and relates to the technical field of hospital examination resource scheduling. The scheduling system comprises a data base layer, an initial planning layer, a response execution layer and a closed-loop optimization layer. The initial planning layer comprises a schedulable resource calculation module, a schedulable resource allocation module and a static scheduling plan generation module; the response execution layer comprises an emergency identification module and a dynamic scheduling adjustment module; and the closed-loop optimization layer comprises a scheduling effect evaluation module and an iterative optimization module. According to the scheme, multi-source data of patients, equipment, medical staff, consulting rooms and the like are integrated, and based on a resource allocation mechanism of'priority weight + greedy algorithm ', the emergency patients can be effectively ensured to obtain resources preferentially, so that the examination resource allocation is more reasonable, and the utilization rate of the examination resources is higher.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hospital examination resource scheduling, in particular to a hospital examination resource scheduling system, method and device capable of dynamic iteration optimization. BACKGROUND

[0002] In the current hospital medical equipment examination service, resource scheduling faces many technical problems, which are specifically as follows: Most of the existing scheduling systems only collect patient appointment information (such as examination type, appointment time), and do not integrate multi-source data such as patient condition emergency level (such as emergency, general diagnosis), equipment running state (such as failure rate, maintenance cycle), and medical staff scheduling (such as on-duty time, professional direction). In addition, the data has not been standardized pretreated, and there are problems such as non-uniform data format (such as HIS system data in XML format, LIS system data in JSON format), many redundant data (such as repeated patient basic information), and many abnormal data (such as incorrect examination time entry), which leads to inaccurate scheduling data basis. Most of the existing scheduling systems are based on the "first come first served" or "fixed allocation" mode to calculate schedulable resources, without considering the real-time availability of resources (such as temporary equipment failure, temporary medical staff transfer) and patient condition priority. For example, emergency patients often have to wait for a long time because ordinary patients occupy the equipment, and some idle equipment also leads to resource waste because it is not allocated in time. The scheduling plan generated by the existing scheduling system is mostly static fixed plan (such as generating the equipment-patient matching plan for the next day one day in advance), which cannot respond to real-time emergencies such as patient no-show (no-show rate about 15%-20%), equipment sudden failure (monthly failure rate about 5%-8%), and new emergency patients. When emergencies occur, manual adjustment of the scheduling plan is required, which not only consumes time (manual adjustment takes an average of 30-45 minutes), but also easily leads to scheduling confusion (such as repeated allocation of equipment, missed patient allocation). The existing scheduling system does not quantitatively evaluate and feedback optimize the scheduling result, and cannot adjust the scheduling strategy according to the historical scheduling data (such as patient waiting time, equipment utilization rate, and medical staff work load). For example, when the demand for a certain type of examination (such as abdominal medical equipment) increases sharply in a certain period of time, the system cannot predict and adjust the resource allocation ratio in advance through historical data, leading to repeated occurrence of the same problem.

[0003] The above defects lead to low efficiency of hospital examination resource scheduling, serious resource waste, long patient waiting time, and unbalanced medical staff work load, which cannot meet the needs of hospital fine management and efficient medical treatment of patients. SUMMARY

[0004] The application aims to provide a hospital examination resource scheduling system, method and device that can be dynamically and iteratively optimized to solve at least one of the above technical problems in the prior art.

[0005] In a first aspect, to solve the above technical problems, the application provides a hospital examination resource scheduling system that can be dynamically and iteratively optimized, comprising a data base layer, an initial planning layer, a response execution layer and a closed-loop optimization layer. The data base layer comprises a data acquisition module and a preprocessing module. The data acquisition module is configured to acquire patient data, device data, medical staff data and clinic data. The initial planning layer comprises a schedulable resource calculation module, a schedulable resource allocation module and a static scheduling plan generation module. The schedulable resource calculation module is configured to calculate the real-time available amount of each schedulable resource based on the preprocessed multi-source data. The schedulable resource allocation module is configured to allocate the real-time available amount of the schedulable resource according to the allocation priority and the allocation rule by using a greedy algorithm to obtain a resource allocation list. The static scheduling plan generation module generates an initial static scheduling plan based on the resource allocation list and constraint conditions, wherein the constraint conditions include appointment time (of patients), on-duty time (of medical staff) and current examination duration (of devices). The response execution layer comprises a sudden situation identification module and a dynamic scheduling adjustment module. The sudden event identification module identifies sudden events based on the multi-source data of the data acquisition module by using a threshold judgment and an event triggering mechanism. The dynamic scheduling adjustment module dynamically adjusts the static scheduling plan based on the sudden events by using a genetic algorithm and a scheduling strategy to obtain a dynamic scheduling plan. The closed-loop optimization layer comprises a scheduling effect evaluation module and an iterative optimization module. The scheduling effect evaluation module calculates key indicators based on the executed scheduling plan. The iterative optimization module iteratively optimizes the scheduling strategy based on the key indicators by using a reinforcement learning method.

[0006] In a feasible implementation, the patient data comprises: Patient ID, examination type, disease emergency level (e.g., defining emergency patients as priority 1, sub-emergency patients as priority 2 and ordinary patients as priority 3), appointment time and previous examination history, etc., acquired through a hospital HIS system (or LIS system); Patient arrival time and ID card information, etc., acquired through a patient terminal.

[0007] In an embodiment, the device data includes: Device ID, device model, model compatibility, current examination duration, cumulative usage duration, and operation status data of the examination device collected by the device sensor; Current device status, device maintenance period, and fault repair time collected by the device management system.

[0008] In an embodiment, the medical staff data includes: Medical staff ID, professional expertise, on-duty time, and number of patients treated collected by the hospital HIS system; Medical staff scheduling information collected by the scheduling system.

[0009] In an embodiment, the examination room data includes: Examination room ID, location information, and current usage status collected by the hospital management system.

[0010] In an embodiment, the patient data is collected in real time; the device data is collected every 5 minutes; the medical staff data is collected every 30 minutes; and the examination room data is collected every 10 minutes.

[0011] In an embodiment, the preprocessing module is used for data cleaning, data standardization, and data fusion of the collected multi-source data. The data cleaning includes removing abnormal data by an outlier detection algorithm; for example, when the patient arrival time is 25 o'clock, the scheduling system automatically identifies it as abnormal data and removes it. The redundant data is deleted by a duplicate data deletion algorithm; for example, duplicate patient ID information is deleted. The data standardization includes unifying different formats of data into JSON format according to field mapping rules; and normalizing numerical value type data. The data fusion includes merging multi-source data into a unified data table based on key fields; for example, merging the patient's "examination type" and the device's "model compatibility" to generate a patient-device compatibility data table.

[0012] In an embodiment, the specific calculation formula of the real-time available amount includes: Number of schedulable devices = total number of devices - number of faulty devices - number of devices under maintenance - number of allocated devices; Number of schedulable medical staff = total number of on-duty medical staff - number of medical staff who have reached full load after treating patients; The number of schedulable clinics = the total number of clinics - the number of occupied clinics - the number of clinics under cleaning.

[0013] In an embodiment, the allocation priority comprises: setting the allocation weight of the schedulable resource based on the urgency of the illness: The allocation weight corresponding to priority 1 is 0.6; The allocation weight corresponding to priority 2 is 0.3; The allocation weight corresponding to priority 3 is 0.1; In this way, the schedulable resource can be ensured to be allocated to patients with high priority first.

[0014] In an embodiment, the allocation rule comprises: The schedulable equipment, schedulable medical staff and schedulable clinics are allocated to patients with priority 1 first, then to patients with priority 2, and finally to patients with priority 3; According to the patient-equipment matching data table, the matching schedulable equipment is allocated to patients of the corresponding examination type; According to the professional matching degree, the schedulable medical staff with the same professional expertise as the examination type is allocated to the corresponding patient.

[0015] In an embodiment, the standardization formula of the professional matching degree comprises: ; Wherein, represents the standardized value of the professional matching degree, and the value range is , and when the value is 1, it indicates complete matching, and when the value is 0, it indicates complete mismatch; represents the original professional matching degree; represents the maximum value of the professional matching degree; represents the minimum value of the professional matching degree.

[0016] In an embodiment, the resource allocation list comprises patient ID, equipment ID, medical staff ID, clinic ID and patient arrival time; the patient ID is sorted in descending order according to the weight score, and when the weight score is the same, the patient ID with earlier patient arrival time is given priority.

[0017] In an embodiment, the specific calculation formula of the weight score comprises: ; Wherein, represents the weight score; represents the allocation weight of the schedulable resource, which is used to allocate the schedulable resource to patients with high priority first; This indicates the weight of medical staff qualifications, which is used to prioritize highly qualified medical staff to high-priority patients; This indicates the patient-device adaptation weight.

[0018] In one feasible implementation, the static scheduling plan includes a correspondence between schedulable resources and patients arranged according to time slices, as well as the expected average examination duration.

[0019] In one feasible implementation, the scheduling strategy includes allocation weights for schedulable resources, weights for healthcare personnel qualifications, weights for patient-device compatibility, and time slice length.

[0020] In one feasible implementation, the specific formula for calculating the expected average inspection time includes: ; in, Indicates the estimated inspection time for the current inspection type; Indicates the total number of patients; Indicates the number of time slices; Indicates the first The total actual examination time for each patient is calculated using the following formula: ; in, Indicates the duration of the patient's examination; Indicates the patient's preparation time; Indicates the time taken to compile the report.

[0021] In one feasible implementation, the event triggering mechanism includes: When a patient arrives at the hospital later than the scheduled time by more than a preset threshold, a no-show event is identified and generated. When the device's operating status data exceeds a preset threshold, a device fault event is identified and generated. When the number of patients with priority 1 in the urgency level increases, identify and generate new emergency events; When the on-duty hours of medical staff change, identify and generate a job transfer event.

[0022] In one feasible implementation, the genetic algorithm of the dynamic scheduling adjustment module includes: Step a1: Encode the correspondence between schedulable resources and patients in the static scheduling plan into examination task units; Step a2: For each inspection task unit, calculate the fitness using the fitness function and sort them from high to low, retaining the top K inspection task units; Step a3: Perform cross-operations on the retained inspection task units; Step a4, perform mutation operation on the cross-operated examination task units, and iteratively execute step a2 to avoid local optimal solution.

[0023] In an available embodiment, the specific formula of the fitness function comprises: ; wherein, denotes a scheduling plan, i.e. an individual in the genetic algorithm; are weight coefficients, satisfying , and have to embody the principle of "emergency first"; denotes waiting time, used for minimizing patient waiting time, considering that patients who have started examination should not be rescheduled, and only patients who have not started examination are calculated, and the specific expression comprises: ; wherein, denotes a set of all patients who have not started examination; denotes the waiting time of the th patient in (i.e. the length of time from the current time to the examination start time); denotes the priority weight of the th patient, for example, the value of priority 1 is 3, the value of priority 2 is 2, and the value of priority 3 is 1; denotes device utilization, used for maximizing device utilization, and the specific expression comprises: ; wherein, denotes the total number of devices; denotes the utilization of the th device in , and the specific calculation formula is: ; denotes load balance degree, used for balancing the work load of medical staff, and the specific expression comprises: ; wherein, denotes the load (on-duty time or the number of patients who have been treated) standard deviation of all medical staff in , and the specific calculation formula comprises: ; wherein, indicates total number of TCM medical staffs; indicates the load of the medical staffs; indicates the average load of all medical staffs.

[0024] In an implementable embodiment, the reinforcement learning method specifically comprises: Step b1, defining the current multi-source data (i.e. patient data, equipment data, medical staff data and clinic data) as the environment state ; Step b2, defining the adjustment operations such as weight adjustment and time slice length adjustment as actions ; Step b3, constructing a reward value function for giving positive rewards when the key performance indicators (KPIs) reach the targets, and negative rewards when the KPIs do not reach the targets; the specific calculation formula of the reward value includes: R=0.3×(equipment utilization rate-85%)+0.3×(30-patient average waiting time)+0.2×(medical staff load balance-85%)+0.2×(patient satisfaction-90%); Step b4, iteratively updating the action value function by the Q-learning algorithm, and the specific expression includes: ; wherein, denotes the learning rate; denotes the discount factor; denotes the new state after executing the new action ; In each iteration cycle, the action with the maximum value is selected to adjust the current scheduling strategy as the scheduling strategy of the next round.

[0025] In an implementable embodiment, the scheduling system further comprises a multi-end interaction layer; The multi-end interaction layer is used to realize the collaborative interaction among the patient terminal, the medical staff terminal and the administrator terminal.

[0026] In an implementable embodiment, the patient terminal is used to realize the appointment management and real-time progress query.

[0027] In an implementable embodiment, the medical staff terminal is used to realize the task management, collaborative communication and work statistics.

[0028] In a feasible implementation, the administrator terminal is used to realize real-time monitoring, report management and system configuration.

[0029] In a second aspect, based on the same inventive concept, the application further provides a hospital examination resource scheduling method using the above scheduling system, which specifically comprises: Step 1: Collecting multi-source data and performing preprocessing through the data base layer; Step 2: Obtaining a resource allocation list through the schedulable resource allocation module based on the preprocessed multi-source data; Step 3: Generating a static scheduling plan through the static scheduling plan generation module based on the resource allocation list and constraint conditions; Step 4: Obtaining a dynamic scheduling plan through the response execution layer based on the static scheduling plan; Step 5: Iteratively optimizing the scheduling strategy through the closed-loop optimization layer based on the executed scheduling plan.

[0030] In a third aspect, based on the same inventive concept, the application further provides a hospital examination resource scheduling device capable of dynamic iterative optimization, which comprises a processor, a memory and a bus, the memory stores instructions and data readable by the processor, and the processor is used to call the instructions and data in the memory to realize the scheduling system as described above, and the bus is connected between each functional component for transmitting information.

[0031] In a feasible implementation, the scheduling device further comprises a device sensor.

[0032] In a feasible implementation, the scheduling device further comprises a fingerprint recognizer and an NFC card reader for the medical staff to punch in and out.

[0033] In a feasible implementation, the scheduling device further comprises an identity card reader and a code scanning gun for collecting patient information.

[0034] In a feasible implementation, the scheduling device further comprises an LED display screen.

[0035] In a feasible implementation, the model of the processor and the memory comprises: The data base layer: Intel Xeon E5-2690 v4 and 1TB SSD; The initial planning layer: AMD EPYC 7763 and 2TB SSD; The closed-loop optimization layer: Intel Xeon Gold 6248 and 4TB SSD.

[0036] By adopting the above technical solutions, the application has the following beneficial effects: The application provides a dynamically iterative optimization hospital examination resource scheduling system, method and device, multi-source data such as patients, equipment, medical staff and clinics are integrated, data accuracy is improved from 75% of an existing system to 98% through standardized preprocessing, for example, the existing system causes 15% of equipment state misjudgment rate due to non-uniform data format, the application reduces the equipment state misjudgment rate to below 2% through JSON format standardization, and ensures that the scheduling data basis is accurate; The scheme is based on the resource allocation mechanism of "priority weight + greedy algorithm", and can effectively ensure that emergency patients have priority to obtain resources, so that the examination resource allocation is more reasonable, and the examination resource utilization rate is higher. The scheme realizes dynamic adjustment of the scheduling plan through a genetic algorithm for sudden situations such as patient no-show and equipment failure. The scheme realizes a virtuous cycle of "scheduling-evaluation-optimization" through a closed-loop optimization mechanism, so that the scheduling strategy continuously adapts to actual demand. The scheme realizes double improvement of patient and medical staff satisfaction through multi-terminal interaction, and accelerates the examination resource scheduling and communication efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the specific embodiments of the application or the technical solutions in the prior art, the drawings needed in the following description of the specific embodiments or the prior art will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 A dynamically iterative optimization hospital examination resource scheduling system diagram provided by the embodiment of the application; Figure 2 A genetic algorithm flowchart of the dynamic scheduling adjustment module provided by the embodiment of the application; Figure 3 A reinforcement learning method flowchart provided by the embodiment of the application; Figure 4 Another dynamically iterative optimization hospital examination resource scheduling system diagram provided by the embodiment of the application. DETAILED DESCRIPTION

[0039] The technical solutions of the application will be described in detail below with reference to the drawings. Obviously, the described embodiments are some of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0040] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0041] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0042] The present invention will be further explained below with reference to specific embodiments.

[0043] It should also be noted that the specific embodiments or implementation methods described below are a series of optimized settings listed by the present invention to further explain the specific content of the invention, and these settings can be combined or used in conjunction with each other.

[0044] Example 1: like Figure 1 As shown in the figure, this embodiment provides a hospital examination resource scheduling system that can be dynamically iteratively optimized, including a data foundation layer, an initial planning layer, a response execution layer, and a closed-loop optimization layer; The data infrastructure layer includes a data acquisition module and a preprocessing module; The data acquisition module is used to collect patient data, equipment data, medical staff data, and clinic data; The initial planning layer includes a schedulable resource calculation module, a schedulable resource allocation module, and a static scheduling plan generation module; The schedulable resource calculation module is used to calculate the real-time availability of each schedulable resource based on preprocessed multi-source data. The schedulable resource allocation module is used to allocate the real-time availability of schedulable resources according to allocation priority and allocation rules, using a greedy algorithm, to obtain a resource allocation list. The static scheduling plan generation module generates an initial static scheduling plan based on a resource allocation list and constraint conditions, which include appointment time, on-duty time, and current examination duration of equipment, etc. The response execution layer includes a sudden event identification module and a dynamic scheduling adjustment module. The sudden event identification module identifies sudden events based on multi-source data of the data acquisition module through threshold judgment and event triggering mechanism. The dynamic scheduling adjustment module dynamically adjusts the static scheduling plan based on sudden events through genetic algorithm and scheduling strategy to obtain a dynamic scheduling plan. The closed-loop optimization layer includes a scheduling effect evaluation module and an iterative optimization module. The scheduling effect evaluation module calculates key indicators based on the executed scheduling plan. The iterative optimization module iteratively optimizes the scheduling strategy based on the key indicators through reinforcement learning method.

[0045] Further, the patient data includes: Patient ID, examination type (such as using abdominal medical equipment for examination, using heart medical equipment for examination, using gynecological medical equipment for examination), disease emergency degree (such as defining emergency patients as priority 1, defining sub-emergency patients as priority 2, and defining ordinary patients as priority 3), appointment time, and previous examination history, etc. collected through the hospital HIS system (or LIS system); Patient arrival time, ID card information, etc. collected through the patient terminal.

[0046] Further, the equipment data includes: Equipment ID, equipment model, model adaptability, current examination duration, cumulative use duration, and running state data of the examination equipment collected through the equipment sensor; Current equipment state, equipment maintenance cycle (such as maintenance once every 3 months), and fault repair time (such as an average repair time of 2 hours) collected through the equipment management system.

[0047] Further, the medical staff data includes: Medical staff ID, professional expertise, on-duty time, and number of patients already seen, etc. collected through the hospital HIS system; Medical staff scheduling information, etc. collected through the scheduling system.

[0048] Further, the clinic room data includes: Clinic room ID, location information, and current use state, etc. collected through the hospital management system.

[0049] Further, the patient data is collected in real time; the device data is collected once every 5 minutes; the medical staff data is collected once every 30 minutes; and the clinic data is collected once every 10 minutes.

[0050] Further, the preprocessing module is used for data cleaning, data standardization and data fusion of the collected multi-source data. The data cleaning includes removing abnormal data by an outlier detection algorithm (such as the 3σ principle); for example, when the patient arrival time is 25 points, the dispatch system automatically identifies it as abnormal data and removes it. The redundant data is deleted by a deduplication algorithm (such as a hash deduplication method based on patient ID); for example, duplicate patient ID information is deleted. The data standardization includes unifying different formats of data into JSON format according to field mapping rules (for example, “PatientID” corresponds to “patient ID”, and “ExaminationType” corresponds to “examination type”); and normalizing numerical value type data (for example, converting the cumulative use time of the examination device from “hours” to “minutes” for subsequent calculation). The data fusion includes fusing multi-source data into a unified data table based on key fields (such as patient ID, device ID, etc.); for example, the “examination type” of the patient is fused with the “model adaptability” of the device (such as GE Logiq E9 adapting to heart examination) to generate a patient-device adaptation data table.

[0051] Further, the specific calculation formula of the real-time available amount includes: The number of schedulable devices = the total number of devices - the number of faulty devices - the number of devices under maintenance - the number of allocated devices; The number of schedulable medical staff = the total number of on-duty medical staff - the number of medical staff who have reached full load; The number of schedulable clinics = the total number of clinics - the number of occupied clinics - the number of clinics under cleaning.

[0052] Further, the allocation priority includes setting the allocation weight of schedulable resources based on the urgency of the condition: The allocation weight corresponding to priority 1 is 0.6; The allocation weight corresponding to priority 2 is 0.3; The allocation weight corresponding to priority 3 is 0.1; In this way, it can be ensured that high-priority patients have priority in obtaining schedulable resources.

[0053] Further, the allocation rule includes: First, allocate the schedulable devices, schedulable medical staff and schedulable clinics to the patients of priority 1, then to the patients of priority 2, and finally to the patients of priority 3. According to the patient-device matching data table, allocate the matching schedulable devices to the patients of the corresponding examination types. According to the professional matching degree, allocate the schedulable medical staff with professional expertise consistent with the examination type to the corresponding patients.

[0054] Further, the standardization formula of the professional matching degree includes: ; wherein, represents the standardized value of the professional matching degree, and the value range is , and when the value is 1, it indicates complete matching, and when the value is 0, it indicates complete mismatch; represents the original professional matching degree (for example, when the professional expertise is cardiac ultrasound, the matching degree for cardiac ultrasound examination is 100, and the matching degree for abdominal ultrasound examination is 40); represents the maximum value of the professional matching degree; represents the minimum value of the professional matching degree.

[0055] Further, the resource allocation list includes patient ID, device ID, medical staff ID, clinic ID and patient arrival time; the patient ID is sorted in descending order according to the weight score, and when the weight score is the same, the patient ID with earlier patient arrival time is given priority.

[0056] Further, the specific calculation formula of the weight score includes: ; wherein, represents the weight score; represents the allocation weight of the schedulable resource, which is used to allocate the schedulable resource to the patients of high priority first; represents the medical staff qualification weight, which is used to allocate the medical staff of high qualification to the patients of high priority first; represents the patient-device matching weight (for example, if the schedulable device matches the examination type of the patient, the value is 1; otherwise, the value is 0.5).

[0057] Further, the static scheduling plan includes the corresponding relationship between the schedulable resources and the patients according to the time slice (for example, 30 minutes as a time slice) and the expected average examination duration.

[0058] Further, the scheduling strategy includes the allocation weight of the schedulable resource, the medical staff qualification weight, the patient-device matching weight and the time slice length.

[0059] Further, the specific calculation formula of the predicted examination average duration includes: ; Wherein, represents the predicted examination duration of the current examination type; represents the total number of patients; represents the time slice quantity coefficient; represents the actual examination total duration of the i-th patient, and the specific calculation formula includes: ; Wherein, represents the patient examination duration; represents the patient preparation duration; represents the report arrangement duration.

[0060] Further, the event triggering mechanism includes: When the patient arrival time is later than the appointment time by more than a preset threshold (for example, 30 minutes), an appointment cancellation event is identified and generated; When the running state data (for example, vibration value, temperature value) of the equipment is greater than a preset threshold, an equipment failure event is identified and generated; When the number of patients with priority 1 in the severity of illness increases, a new emergency event is identified and generated; When the on-duty time of medical staff changes, a job transfer event is identified and generated.

[0061] Further, as shown in Figure 2 , the genetic algorithm of the dynamic scheduling adjustment module includes: Step a1, encode the correspondence between the schedulable resources in the static scheduling plan and the patients as examination task units; for example, the examination task unit “8:00-8:30_ device 1 _ patient A” indicates that during 8:00-8:30, device 1 examines patient A; Step a2, for the examination task units, calculate the fitness through the fitness function and sort from high to low, and retain the top K examination task units; Step a3, perform a crossover operation on the retained examination task units; for example, for the equipment failure event, perform a crossover operation on the examination task unit “8:30-9:00_ device 2 _ patient B” corresponding to the failed equipment and the examination task unit “8:30-9:00_ device 5 _ patient C” corresponding to the idle equipment, to generate two new examination task units “8:30-9:00_ device 5 _ patient B” and “8:30-9:00_ device 2 _ patient C”; ​Step a4, perform mutation operation on the cross-operated examination task unit, and iteratively execute step a2; for example, adjust the time slice of the examination, and change "9:00-9:30_device 3_patient D" to "9:15-9:45_device 3_patient D" to avoid local optimal solution.

[0062] Further, the specific formula of the fitness function includes: ; Wherein, represents a scheduling plan, that is, an individual in the genetic algorithm; are weight coefficients, satisfying , and to embody the "emergency priority" principle; represents the waiting time, which is used to minimize the patient waiting time, and considering that the patients who have started the examination should not be rescheduled, only the patients who have not started the examination are calculated, and the specific expression includes: ; Wherein, represents a set of all patients who have not started the examination; represents the waiting time of the th patient in (that is, the length of time from the current time to the examination start time); represents the priority weight of the th patient, for example, the value corresponding to the priority 1 is 3, the value corresponding to the priority 2 is 2, and the value corresponding to the priority 3 is 1; represents the device utilization rate, which is used to maximize the device utilization rate, and the specific expression includes: ; Wherein, represents the total number of devices; represents the utilization rate of the th device in , and the specific calculation formula is: ; represents the load balance degree, which is used to balance the work load of medical staff, and the specific expression includes: ; Wherein, represents the load (on-duty time or the number of patients who have been treated) standard deviation of all medical staff in , and the specific calculation formula includes: ; in, express Total number of TCM nurses; express The Middle The workload of individual medical staff; express The average workload of all medical staff in the country.

[0063] Furthermore, such as Figure 3 As shown, the reinforcement learning method specifically includes: Step b1: Define the current multi-source data (i.e., patient data, device data, medical staff data, and clinic data) as the environmental state. ; Step b2: Define the weight adjustment operations (such as the allocation weight of schedulable resources, the weight of medical staff qualifications, or the weight of patient-equipment adaptation) and the time slice length adjustment as actions. ; Step b3: Construct a reward value function to provide a positive reward when a key performance indicator (KPI) reaches its target and a negative reward when it fails to reach its target; reward value The specific calculation formulas include: R = 0.3 × (equipment utilization rate - 85%) + 0.3 × (30 - average patient waiting time) + 0.2 × (medical staff workload balance - 85%) + 0.2 × (patient satisfaction - 90%); Step b4: Iteratively update the action value function using the Q-learning algorithm. The specific expressions include: ; in, This represents the learning rate (which can be set to 0.1). This represents the discount factor (which can be set to 0.9). Indicates the execution of new actions The new state afterwards; Choose for each iteration cycle (e.g., once a day). The action with the highest value adjusts the current scheduling strategy and serves as the scheduling strategy for the next round.

[0064] Furthermore, such as Figure 4 As shown, the scheduling system also includes a multi-terminal interaction layer; The multi-terminal interaction layer is used to enable collaborative interaction between patient terminals, medical staff terminals, and administrator terminals.

[0065] Further, the patient terminal is used to realize appointment management (such as online appointment, appointment rescheduling, appointment cancellation, etc.) and real-time progress (such as the current number of people in line, the estimated waiting time, the location navigation of the examination room, etc.) query.

[0066] Further, the medical staff terminal is used to realize task management (such as the scheduling plan for the day, task confirmation, task adjustment, etc.), collaborative communication (such as cross-examination room resource allocation application, priority reminder, etc.), and work statistics (such as the number of patients seen on the day, examination time, patient satisfaction, personal work report, etc.).

[0067] Further, the administrator terminal is used to realize real-time monitoring (such as device utilization rate, load balancing degree, patient waiting time, emergency, abnormal index, etc.), report management (such as scheduling effect monthly / quarterly report, KPI trend chart, comparison of the effects before and after the optimization of scheduling strategy, emergency statistics, etc.), and system configuration (such as device maintenance period, initial value of each weight, time slice length, permission management, interface management, etc.).

[0068] Embodiment Two: The embodiment provides a hospital examination resource scheduling method using the above scheduling system, which specifically comprises: Step 1: Collecting multi-source data and performing preprocessing through the data base layer; Step 2: Based on the preprocessed multi-source data, obtaining a resource allocation list through the schedulable resource allocation module; Step 3: Based on the resource allocation list and the constraint condition, generating a static scheduling plan through the static scheduling plan generation module; Step 4: Based on the static scheduling plan, obtaining a dynamic scheduling plan through the response execution layer; Step 5: Based on the executed scheduling plan, iteratively optimizing the scheduling strategy through the closed-loop optimization layer.

[0069] Embodiment Three: The embodiment provides a hospital examination resource scheduling device capable of dynamic iterative optimization, comprising a processor, a memory and a bus, the memory stores instructions and data readable by the processor, and the processor is used to call the instructions and data in the memory to realize the scheduling system as described above, and the bus is connected between each functional component for transmitting information.

[0070] Further, the scheduling device further comprises a device sensor.

[0071] Further, the scheduling device further comprises a fingerprint recognizer and an NFC card reader for medical staff clock-in.

[0072] Further, the scheduling device further comprises an ID card reader and a code scanning gun for collecting patient information.

[0073] Further, the scheduling device further comprises an LED display screen.

[0074] Further, the model of the processor and the memory comprises: Data base layer: Intel Xeon E5-2690 v4 and 1TB SSD; Initial planning layer: AMD EPYC 7763 and 2TB SSD; Closed loop optimization layer: Intel Xeon Gold 6248 and 4TB SSD.

[0075] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A hospital examination resource scheduling system capable of dynamic iterative optimization, characterized in that, It includes a data foundation layer, an initial planning layer, a response execution layer, and a closed-loop optimization layer; The data infrastructure layer includes a data acquisition module and a preprocessing module; The data acquisition module is used to collect patient data, equipment data, medical staff data, and clinic data; The initial planning layer includes a schedulable resource calculation module, a schedulable resource allocation module, and a static scheduling plan generation module; The schedulable resource calculation module is used to calculate the real-time availability of each schedulable resource based on preprocessed multi-source data. The schedulable resource allocation module is used to allocate the real-time availability of schedulable resources according to allocation priority and allocation rules, using a greedy algorithm, to obtain a resource allocation list. The static scheduling plan generation module generates an initial static scheduling plan based on the resource allocation list and constraints. The response execution layer includes an emergency identification module and a dynamic scheduling and adjustment module; The emergency event identification module identifies emergencies based on multi-source data from the data acquisition module, through threshold judgment and event triggering mechanisms. The dynamic scheduling adjustment module, based on sudden events, dynamically adjusts the static scheduling plan using a genetic algorithm and scheduling strategy to obtain a dynamic scheduling plan; The closed-loop optimization layer includes a scheduling effect evaluation module and an iterative optimization module; The scheduling effect evaluation module calculates key indicators based on the executed scheduling plan; The iterative optimization module optimizes the scheduling strategy iteratively based on key indicators and using reinforcement learning methods.

2. The scheduling system according to claim 1, characterized in that, The patient data includes: Patient ID, examination type, urgency of condition, appointment time, and previous examination history are collected through the hospital's HIS system; Patient arrival time and ID information collected via patient terminal; The device data includes: The device sensors collect and inspect the device ID, device model, model compatibility, current inspection duration, cumulative usage duration, and operating status data of the device. The equipment management system collects information on the current status of the equipment, the equipment maintenance cycle, and the fault repair time. The data on medical staff includes: The data collected through the hospital's HIS system includes medical staff IDs, professional expertise, on-duty hours, and the number of patients treated. Medical staff scheduling information collected through the scheduling system; The clinic data includes: The clinic ID, location information, and current usage status are collected through the hospital management system.

3. The scheduling system according to claim 2, characterized in that, The degree of urgency of the illness includes: Patients in the emergency department are defined as priority level 1; Patients with sub-emergency symptoms are defined as priority level 2; Patients in the general outpatient clinic are defined as priority level 3.

4. The scheduling system according to claim 3, characterized in that, The specific formula for calculating the real-time availability includes: Number of schedulable devices = Total number of devices - Number of faulty devices - Number of devices under maintenance - Number of devices already assigned; Available medical staff = Total number of on-duty medical staff - Number of medical staff already at full capacity; Number of available clinics = Total number of clinics - Number of clinics in use - Number of clinics under cleaning 5. The scheduling system according to claim 4, characterized in that, The allocation rules include: First, allocate available equipment, available medical staff, and available consultation rooms to patients with priority 1, then to patients with priority 2, and finally to patients with priority 3. Based on the patient-device compatibility data table, the compatible schedulable devices are assigned to patients of the corresponding examination types; Based on professional matching, available medical staff whose professional expertise matches the type of examination are assigned to the corresponding patients.

6. The scheduling system according to claim 5, characterized in that, The standardized formula for the professional matching degree includes: ； in, This represents the standardized value of professional matching, with a range of values ​​of [value range missing]. A value of 1 indicates a complete match, while a value of 0 indicates a complete mismatch. Indicates the original professional matching degree; This indicates the maximum degree of professional matching. This represents the minimum professional matching degree.

7. The scheduling system according to claim 1, characterized in that, The resource allocation list includes patient ID, equipment ID, medical staff ID, consultation room ID, and patient arrival time. The patient IDs are sorted in descending order of weight score, and when the weight scores are the same, the patient ID with the earlier arrival time is ranked higher.

8. The scheduling system according to claim 7, characterized in that, The specific calculation formula for the weighted score includes: ； in, Indicates weighted score; Indicates the allocation weight of schedulable resources; Indicates the weight of medical staff qualifications; This indicates the patient-device adaptation weight.

9. A method for scheduling hospital examination resources using the scheduling system described in any one of claims 1-8, characterized in that, include: Step 1: Collect and preprocess multi-source data through the data foundation layer; Step 2: Based on the preprocessed multi-source data, obtain the resource allocation list through the schedulable resource allocation module; Step 3: Based on the resource allocation list and constraints, generate a static scheduling plan using the static scheduling plan generation module; Step 4: Based on the static scheduling plan, obtain the dynamic scheduling plan through the response execution layer; Step 5: Based on the executed scheduling plan, iteratively optimize the scheduling strategy through a closed-loop optimization layer.

10. A hospital examination resource scheduling device capable of dynamic iterative optimization, characterized in that, It includes a processor, a memory, and a bus. The memory stores instructions and data read by the processor, and the processor is used to call the instructions and data in the memory to implement the scheduling system as described in any one of claims 1-8. The bus connects the functional components for transmitting information.

Images