An intelligent guiding and reminding system for whole process of ophthalmic day surgery

Abstract

The application relates to the field of medical information technology and discloses an intelligent guiding and reminding system for the whole process of ophthalmic day surgery. The system comprises a task modeling module, a state sensing module, a silent continuation engine and an instruction generating module. The task modeling module analyzes preoperative medical orders into task nodes and marks rigid tasks or flexible tasks, and constructs a task dependency graph containing hard dependency edges and soft dependency edges; the state sensing module monitors the task completion state and sets a dynamic tolerance time limit, and generates an execution deviation event when the task is not completed within the time limit; the silent continuation engine resolves the deviation through first-layer elastic buffering, second-layer path reorganization and third-layer fuse help; and the instruction generating module generates a single guiding sentence according to the state of the dependency graph and pushes the guiding sentence to a patient terminal. The application converts the preoperative multi-task time sequence coordination into a continuous guiding instruction flow, actively digests uncertainty when an execution deviation occurs, and helps to improve the preoperative guiding effect.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, and more specifically, to an intelligent triage and reminder system for the entire process of ophthalmic day surgery. Background Technology

[0002] Ophthalmic day surgery is a relatively mature type of day surgery, accounting for over 90% of ophthalmic surgeries such as cataracts and refractive surgeries. Before surgery, patients need to complete multiple preparatory tasks, including various preoperative examinations, preoperative medication, and fasting / abstaining from fluids, with strict time constraints between these tasks. For the main patient group undergoing day surgery for cataracts and other ophthalmic procedures, who are generally older, misunderstandings or omissions in performing these multi-task, time-constrained preoperative preparations are common.

[0003] Chinese patent CN121687466A discloses a client-based perioperative management method and related equipment. This method receives multimodal target user information sent by the client, processes a preset ocular early warning model based on a target training set and a target validation set, generates ocular early warning information and postoperative early warning information for the target user, and sends them to the client. This solution mainly focuses on risk assessment and information push based on the early warning model, with less attention paid to the temporal coordination between multiple preoperative preparation tasks.

[0004] Chinese patent CN119601251A discloses a postoperative disease management system for day surgery based on structured electronic medical records. This system starts with preoperative factors and characteristics, analyzes these characteristics to determine the availability of postoperative health status for users, and constructs an XGBoost model to predict postoperative health status. This solution focuses on risk assessment and prediction of postoperative disease progression and does not involve task guidance in the preoperative stage.

[0005] In existing patient guidance and reminder methods, hospitals generally distribute paper-based lists of precautions to patients or push multiple reminder messages to patients via applications according to a preset schedule. This method is essentially a discrete notification model, with reminders for each task being independent of each other. After receiving multiple independent messages, patients need to understand the content of each task, remember the order of tasks, and schedule them accordingly. When a patient fails to complete a task within the scheduled time, the existing system can only send an overdue alarm, transferring the planning pressure of time conflicts to the patient or their family, increasing the risk of errors in preoperative preparation. Therefore, an intelligent triage and reminder system for the entire process of ophthalmic day surgery is proposed to address the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an intelligent triage and reminder system for the entire process of ophthalmic day surgery. This system aims to address the problem that patients need to coordinate the timing of multiple preoperative tasks themselves under the existing discrete reminder method, and that the system can only respond with passive alarms when deviations occur in task execution, and cannot actively adjust the subsequent guidance path.

[0007] To achieve the above objectives, the present invention provides an intelligent triage and reminder system for the entire process of ophthalmic day surgery, including a task modeling module, a state perception module, a silent calculation engine, and an instruction generation module.

[0008] After the surgical schedule is determined, the task modeling module obtains the patient's preoperative medical orders and parses them into task nodes. Each task node is marked as a rigid task or a flexible task, and a safety window is set for the rigid task. The maximum allowable sliding time for the flexible task is calculated, and a task dependency graph containing hard and soft dependency edges is constructed.

[0009] Furthermore, the rigid task includes at least one of the following: the deadline for fasting and abstaining from drinking before general anesthesia, the deadline for visual recovery after mydriasis, the deadline for delivering corneal topography, and the deadline for delivering the intraocular lens power calculation results.

[0010] If the timing of such tasks deviates from the prescribed duration, it may pose a surgical safety risk or prevent the surgery from being performed on that day. These flexible tasks include at least one of the following: reading preoperative instructions, re-measuring blood pressure, paying fees, and adjusting the timing of antibiotic eye drops administration within the prescribed window.

[0011] These types of tasks have adjustable execution time, and time slippage within their allowable range does not produce adverse clinical consequences. Through the above classification, different types of preoperative tasks are clearly distinguished in the dependency graph, enabling subsequent bias mitigation to employ different handling strategies based on task type.

[0012] Furthermore, the execution order between the two task nodes connected by hard-dependency edges cannot be reversed, reflecting the unshakeable sequential relationship between tasks. Soft-dependency edges point from flexible task nodes to subsequent task nodes affected by their sliding, and are marked with the maximum allowable sliding time, defining the range and magnitude of influence when a flexible task shifts. The temporal constraints and adjustable ranges between preoperative tasks are structurally expressed through the above task dependency graph, providing a basis for adaptive adjustments during subsequent triage.

[0013] Furthermore, when calculating the maximum allowable sliding time for a flexible task, the task modeling module backtracks upstream from the current flexible task node to the first rigid task node or immovable time anchor encountered, and records the end time of that node or anchor as... Search downstream from the current flexible task node until the first rigid task node or immovable time anchor encountered, and record the start time of that node or anchor as... .

[0014] Obtain the estimated duration of the current flexible task. , to The sum of the estimated durations of all rigid task nodes within the interval and the sum of the necessary time intervals between each task. Then the maximum allowed sliding time .

[0015] Calculated The value reflects the total time during which the flexible task can slide freely between upstream and downstream rigid constraints. When the task is delayed, the delay amount is... The portion within this range can be absorbed through the time window of downstream flexible tasks without encroaching on the safety window of rigid tasks.

[0016] The time position of the subsequent task node connected by the soft dependency edge is shifted as a whole within the sliding range of the preceding flexible task. The upper limit of the shift is the maximum allowed sliding time marked on the soft dependency edge.

[0017] The state awareness module obtains the completion status of inspection-type tasks and non-inspection-type tasks.

[0018] Furthermore, for examination-related tasks, this module automatically confirms completion through the report review completion flag in the hospital's laboratory information system or image archiving system, achieving data linkage with the hospital's existing information systems and reducing reliance on patients' proactive reporting. For non-examination-related tasks, this module receives the operation signal from the single confirmation control provided for the current task on the patient's interface. After completing the corresponding action, the patient reports the completion status to the system through a single operation.

[0019] Furthermore, the state-aware module sets a dynamic tolerance time limit for the currently active task, which is set by obtaining the total critical path duration of the remaining incomplete rigid tasks. The remaining time until the start of the surgery. Calculate the ratio ,according to Obtain the dynamic tolerance period. These are preset positive numbers.

[0020] As the surgery approached Decrease As the remaining preparation time decreases, the dynamic tolerance period decreases accordingly, and the system's tolerance for task delays adaptively tightens. In the early preoperative period, the dynamic tolerance period is larger to avoid prematurely triggering deviation alarms; as the operation approaches, the dynamic tolerance period shrinks rapidly to promptly detect and respond to execution deviations. If the completion status is not obtained after the sum of the predetermined completion time and the dynamic tolerance period, an execution deviation event is generated.

[0021] The silent continuation calculation engine includes a first-layer elastic buffer unit, a second-layer path reorganization unit, and a third-layer circuit breaker assistance unit, which resolves deviation events in a three-layer progressive manner.

[0022] When the deviation task is a flexible task and the accumulated delay time does not exceed the maximum allowable sliding time of the flexible task, the first layer of elastic buffer unit shifts the time window of the downstream flexible task along the soft dependency edge by the accumulated delay time, and after each shift, it traces forward along the dependency edge to check whether the downstream rigid task is still within its safety window.

[0023] Furthermore, when the planned start time of the rigid task node is earlier than the earliest start boundary of its safety window, or the planned end time is later than the latest end boundary of its safety window, it is determined that the safety window has been exceeded, the executed translation operation is canceled, and the second-level path reorganization unit is triggered.

[0024] Through a safety window verification mechanism, time window shifting during the flexible buffering process only occurs under conditions that will not jeopardize the execution of rigid tasks. When the flexible buffering is successful, the impact of slight delays on subsequent task scheduling is absorbed and digested internally, and the patient is unaware of the delay in subsequent instructions.

[0025] The second-layer path reorganization unit is activated before the first-layer elastic buffer unit is resolved. It searches the task dependency graph for ready flexible tasks that have all hard dependencies satisfied, whose execution location is in the same or adjacent area as the patient's current location, and which have no mutual exclusion relationship with the suspended tasks. It selects the one with the shortest estimated time and promotes it as the current task, and marks the deviation task as suspended.

[0026] Furthermore, the selection method is as follows: calculate the sum of the standard execution time of each candidate task and the estimated travel time from the patient's current location to the execution location of the candidate task, and select the candidate task with the smallest sum. By comprehensively considering the execution time and travel time, the upgraded alternative task can be completed in a shorter time under the current deviation situation, achieving partial reorganization of the task path without interrupting the triage process.

[0027] If the third-layer circuit breaker emergency response unit fails to resolve the issue in the first two layers, or if calculations indicate that the rigid task safety window will be breached, it will stop automatically adjusting, generate emergency response information, and output it to both the patient and medical staff ends.

[0028] Furthermore, the request for assistance includes the patient's identity, the type of urgent task under threat, the type of failure, and the latest completion deadline. The waiting prompts sent to the patient do not include the reason for the obstruction or alternative solutions, to prevent the patient from making decisions that could affect clinical safety while waiting.

[0029] The help request sent to the medical staff includes the patient's location, the name of the threatened task, and a countdown timer for the latest completion deadline. This allows the medical staff to directly obtain the target, location, and time limit for treatment without needing secondary analysis and judgment to execute priority treatment.

[0030] After the task dependency graph state is updated, the instruction generation module selects task nodes whose predecessor tasks for all incoming hard dependency edges have been completed.

[0031] Furthermore, when multiple tasks exist, the one with the latest completion deadline and the most pressing need is selected as the target for the push notification, generating a single, conversational guidance message that is then sent to the patient. Only one guidance message is sent at a time; no task list or overall schedule is displayed, allowing the patient to focus on only the action they need to perform at any given time.

[0032] Furthermore, after the silent processing engine resolves the content of the single, conversational triage statement, it directly reflects the adjusted highest priority task to be executed. The resolution process is transparent to the patient, and the triage guidance received by the patient is a continuous, uninterrupted stream of single instructions.

[0033] The technical effects and advantages of this invention are as follows: This invention constructs a task dependency graph with flexible annotations to structurally represent the hard temporal constraints and flexible sliding ranges between preoperative tasks. Hard dependency edges indicate that the execution order between connected task nodes cannot be reversed, while soft dependency edges are labeled with the maximum allowable sliding time for flexible tasks, providing a data foundation for subsequent adaptive adjustments.

[0034] Under normal operating conditions, the system uses the dependency graph to filter out tasks that have no unmet hard dependencies in real time and pushes them to the patient one by one with a single, conversational instruction. Patients do not need to face a task list or sort them themselves; they only need to follow the current instruction to complete the corresponding action, which helps to reduce the cognitive burden of multitasking.

[0035] When a delay in the execution of a task is detected, but the delay does not exceed its maximum allowable sliding time, the system automatically shifts the time window of downstream flexible tasks in the background. The patient receives subsequent guidance without being aware of it, helping to reduce manual intervention caused by minor delays. When the delay exceeds the elastic range, the system searches the dependency graph for ready-to-execute alternative tasks, re-determines the current optimal task, and pushes corresponding guidance, adjusting the path without interrupting the guidance process. When automatic adjustment fails to resolve the issue and the safety window of a rigid task is reached, the system issues a task-level assistance command containing the patient's location and the threatened task information to the medical staff, allowing them to directly know the target and time limit for intervention. Thus, various execution deviations are gradually absorbed within the system, ensuring the patient always receives a continuous flow of guidance instructions, which helps improve the orderliness and completeness of preoperative preparation. Attached Figure Description

[0036] Figure 1 This is a system module framework diagram of the present invention.

[0037] Figure 2 This is a flowchart illustrating the overall system workflow of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1 As attached Figures 1 to 2 The system described is an intelligent triage and reminder system for the entire process of ophthalmic day surgery, comprising a task modeling module, a state awareness module, a silent continuation calculation engine, and an instruction generation module. The silent continuation calculation engine includes a first-layer elastic buffer unit, a second-layer path reorganization unit, and a third-layer circuit breaker assistance unit. The following section provides a detailed explanation of the construction and collaborative working process of each module in a specific scenario.

[0040] The generation and classification of task nodes are as follows: Once the surgical schedule is determined, the task modeling module retrieves all preoperative medical orders for the patient from the hospital information system, including examination tasks, medication tasks, and fasting / abstention tasks, and parses each order into an independent task node. Each task node carries the task type, task name, planned execution period, execution location, and estimated duration.

[0041] The task modeling module sets the latest completion time for each task node based on clinical standards for each task type. For rigid task nodes, a safety window is also assigned, defined by the earliest allowed start time boundary and the latest allowed end time boundary. For example, the safety window for the pre-anesthesia fasting and fluid restriction task starts at 22:00 one day before surgery and ends 6 hours before the start of surgery; the safety window for the post-mydriatic vision recovery assessment task is at least 20 minutes of observation after the completion of mydriasis and the recovery of pupillary light reflex.

[0042] Based on clinical safety rules, the task modeling module marks all task nodes as either rigid or flexible tasks.

[0043] If the execution time of a task node deviates from the expected time, it will inevitably cause surgical safety risks or prevent the surgery from continuing on that day. In such cases, it will be marked as a rigid task. If a task node has adjustable execution time and time sliding within its allowable range does not produce adverse clinical consequences, it is marked as a flexible task.

[0044] Rigid tasks include the deadlines for pre-operative fasting and fluid restriction under general anesthesia, the deadline for visual recovery after mydriasis, the deadline for delivering corneal topography maps, and the deadline for delivering intraocular lens power calculation results. Flexible tasks include reading pre-operative instructions, re-measuring blood pressure, paying fees, and adjusting the timing of antibiotic eye drops administration within the medication window.

[0045] The construction of the task dependency graph is as follows: Based on the completed labeling, the task modeling module constructs a task dependency graph. The task dependency graph uses all task nodes as vertices and directed edges to represent temporal dependencies. Directed edges are divided into hard dependency edges and soft dependency edges.

[0046] For hard dependency edges, the task modeling module has a built-in ophthalmic preoperative clinical pathway rule base, which pre-stores the pre-dependent relationship entries between different types of tasks, such as "corneal topography examination is placed before intraocular lens power calculation", "mydriasis is placed before visual recovery assessment", and "fasting blood sampling is placed before preoperative fasting and abstinence status confirmation".

[0047] The task modeling module iterates through the task nodes, querying the rule base pair by pair. If a preceding dependency record is matched, a hard dependency edge is drawn from the preceding task node to the following task node. The execution order of the two task nodes connected by the hard dependency edge cannot be reversed.

[0048] For soft dependency edges, the task modeling module iterates through all subsequent task nodes for each flexible task node. If the interval between the planned start time of a subsequent node and the planned end time of the flexible node is less than a preset association threshold, a soft dependency edge is drawn from that flexible node to that subsequent node. Each soft dependency edge is marked with its maximum allowed sliding time. This represents the upper limit by which the time windows of subsequent tasks can be shifted as a whole when a time shift occurs in the current flexible task. When multiple soft dependency edges originate from the same flexible node, each edge is labeled with the same value. value.

[0049] The calculation of the maximum permissible sliding time is as follows: For each flexible task node, the task modeling module calculates its maximum allowable sliding time as follows: .

[0050] Starting from the current flexible task node, backtrack upstream along the directed edge level by level until the first rigid task node or immovable time anchor point is encountered. The end time of that node or anchor point is recorded as T1. An immovable time anchor point refers to an absolute time reference point that is not subject to sliding adjustment by the system, such as the start time of surgery or the fixed scheduling time of a hospital examination department.

[0051] Search downstream from the current flexible task node along the directed edge step by step until the first rigid task node or immovable time anchor is encountered. The start time of the node or anchor is recorded as T2.

[0052] Get the estimated duration of the current flexible task node Calculate the sum of the estimated durations of all rigid task nodes within the interval from T1 to T2. And the sum of the minimum time intervals required between adjacent tasks within that interval. The maximum allowed sliding time is... .

[0053] In one example, the preoperative blood pressure retest is scheduled for the morning of the day before surgery. Looking upstream, the nearest rigid task is the corneal topography examination and intraocular lens power calculation on the afternoon of the second day before surgery, with a planned deadline of 17:00, i.e., T1=17:00.

[0054] Searching downstream, the most recent mandatory task is the start time of preoperative fasting and fluid restriction at 22:00, i.e., T2 = 22:00. The estimated time for repeat blood pressure measurement itself... The time limit is 10 minutes, and there are no other rigid task nodes between T1 and T2. Necessary intervals between tasks It is also 0.

[0055] therefore The blood pressure measurement can be completed at any time between 17:00 and 21:50 without affecting the subsequent fasting and abstaining from drinking.

[0056] In another example, the prescription for antibiotic eye drops requires four applications daily, with each application at least four hours apart. A planned application time is 2:00 PM. The upstream nearest rigid task is the moment four hours have passed since the last application, and the downstream nearest rigid task is the latest allowed time for the next application. If the window between the two rigid constraints is six hours, and the estimated time for this application is 2 minutes, then... It takes about 1 hour and 58 minutes.

[0057] After the value is calculated, it is marked on each soft dependency edge originating from the flexible task node. The task dependency graph is incrementally updated after the initial construction and after each task is completed. Completed task nodes are marked as completed, and their outgoing edges are no longer involved in subsequent searches.

[0058] The command generation and initial push are as follows: After the task modeling module completes initialization, the instruction generation module starts for the first time. The instruction generation module traverses all nodes in the dependency graph, filtering out completed nodes. For the remaining incomplete nodes, it checks whether all predecessor tasks of their incoming hard dependencies have been completed, excluding nodes whose predecessor dependencies are not satisfied. The remaining nodes constitute the current executable set. If there is only one node in the set, it is directly used as the push target; if there are multiple nodes, they are sorted by their latest completion deadline from nearest to furthest, and the most urgent one is selected.

[0059] The instruction generation module uses the task type, execution location, and recommended execution time period of the node as slot values, fills them into a natural language template, assembles them into a single, conversational triage statement, and pushes it to the patient's end. Only one statement is pushed at a time, without displaying a task list or overall schedule. From the moment of push, the task is considered by the system as a currently active task awaiting execution.

[0060] The monitoring and confirmation of task completion status specifically includes: The system maintains a maximum of one task being active and monitored at any given time. The state awareness module continuously monitors the currently active task and uses different confirmation methods based on the task type.

[0061] For inspection-related tasks, the status awareness module periodically reads the review completion flag of the corresponding inspection report through the hospital information system interface. When the flag changes from incomplete to completed, it automatically confirms that the task node has been completed.

[0062] For non-examination tasks, the patient's end provides a single confirmation button for the currently active task. After the patient completes the action and clicks the button, the confirmation event is uploaded to the status awareness module, which then verifies the confirmation and marks the task as completed.

[0063] The execution of the step loop is as follows: After each task is confirmed, the task dependency graph is incrementally updated. The instruction generation module then automatically triggers, generating the next triage statement according to the same process and pushing it to the patient. The newly pushed task becomes the new active task, and the status awareness module switches the monitoring object to this new task. This step-by-step cycle continues until all preoperative task nodes are marked as completed.

[0064] The setting of the dynamic tolerance period and the generation of deviation events are specifically as follows: During the step-by-step loop, the state-aware module sets a dynamic tolerance period for the currently active task. .

[0065] The setup is as follows: obtain the total critical path duration T3 of the remaining unfinished rigid tasks, and the difference T4 between the current system clock and the start time of the operation. T3 is calculated using the task dependency graph as input and the critical path method. The schedule pressure ratio is then calculated. ,according to Determine the dynamic tolerance period, where The preset normal values ​​are used. The state awareness module recalculates every preset period. .

[0066] When the system clock exceeds the task's scheduled completion time and the latest If the sum of the results has not yet received a completion signal, the state awareness module generates an execution deviation event and submits it to the silent continuation calculation engine. At this time, the task remains incomplete, and the state awareness module suspends its regular monitoring. Once the issue is resolved, the monitoring target will be re-determined based on the new activation task.

[0067] The first layer of elastic buffer is specifically as follows: After receiving a deviation event, the silent calculation engine is first processed by the first layer of elastic buffer unit.

[0068] The first-level elastic buffer unit checks whether the deviation task is a flexible task. If not, it directly triggers the second-level path reassembly unit. If so, it calculates the cumulative delay time. and with Comparison. If Greater than If, it is handed over to the second layer for processing; No more than Execute elastic buffering.

[0069] Before implementing the elastic buffer, the system records the current start time and end time of the plan for each deviation task node and all downstream flexible task nodes reachable along each soft dependency edge, and saves them as a snapshot of the time stamp before the translation.

[0070] The elastic buffer operation is as follows: starting from the biased task node, along each outgoing soft dependency edge, shift the original time window of the first direct successor task node backward as a whole. After translating each node, continue tracing forward along the outgoing dependency edges: If a flexible node is encountered, the translation continues. If a rigid node is encountered, the translation of that branch will stop. If multiple branches converge at a subsequent flexible node, the first branch will be translated normally to the converging node, and subsequent branches will not be translated again. Tracking will continue based on the time window after the translation.

[0071] After each rigid node stops translation, a safety window check is immediately performed: the planned start time is compared with the earliest start boundary, and the planned end time is compared with the latest end boundary. If the start time is not earlier than the earliest boundary and the end time is not later than the latest boundary, the node passes the check.

[0072] If all rigid nodes in all branches pass the check, the elastic buffer is successful, and the deviation event is handled internally. If any branch has a planned start time earlier than the earliest start boundary or a planned end time later than the latest end boundary, the elastic buffer is considered to have failed. The system immediately reads the timestamp snapshot, restores the planned times of all shifted nodes to their original values, and after undoing the changes, hands over the deviation event to the second-level path reorganization unit.

[0073] In one example, the patient planned to read the preoperative instructions at 9:00 AM the day before surgery. It was 2 hours. The actual confirmation time was 9:40. The timeframe is 40 minutes. The system saves snapshots of preoperative instructions, downstream blood pressure re-measurements, and payment. Along the soft dependency edge, the blood pressure re-measurements are shifted 40 minutes after 10:00 to 10:10, and payment is also shifted 40 minutes after that. Forward tracking continues to the start of fasting and fluid restriction, with a planned start time of 22:00. After checking, no boundaries were exceeded, and the flexible buffering was completed.

[0074] The second-level path reorganization is specifically as follows: The second-level path reorganization unit is activated under the following conditions: the deviation task is a rigid task; the deviation task is a flexible task. Greater than The process was terminated when the rigid window was detected to be out of bounds in the first layer.

[0075] The path reorganization unit starts from the current moment and sequentially filters candidate flexible task nodes according to three conditions: All predecessor tasks for incoming hard-dependent edges have been completed; The examination location is in the same or adjacent examination area as the patient's current location; There is no mutual exclusion relationship between this task and any pending tasks. The patient's current location is obtained through the hospital's real-time location subsystem; if location is unavailable, the most recent check-in location will be used instead. Mutual exclusion refers to two tasks that cannot be performed simultaneously or alternately due to occupying the same resource or physiological conflicts, such as fasting blood collection and eating, or antibiotic eye drops and eye irrigation.

[0076] For each task in the candidate set, read the standard execution time. and estimated travel time Calculate the total time occupied. Select The smallest candidate task is promoted to the current optimal task and activated. The instruction generation module generates a triage statement for this task and pushes it to the patient's end, while the status awareness module switches the monitoring object.

[0077] Simultaneously, the original deviation task is marked as suspended, and the reason for suspension and the type of blocking condition are recorded, including missed time window, unmet prerequisites, and resource consumption by mutually exclusive tasks. The state-aware module iterates through the suspended tasks each time a task is confirmed to be complete, checking whether the blocking conditions have been eliminated. Once the conditions are met, the path reorganization unit finds a safe insertion position based on the original dependency relationships and reactivates the task.

[0078] In one example, a patient was unable to have their blood drawn promptly due to a late queue number. Blood drawing is a rigid task, and the first level of processing is skipped after an error event is generated. The system marks the blood drawing as suspended, searches for available blood pressure retests in the same area that are not mutually exclusive, and then... It takes 4 minutes. For 1 minute, The time limit is 5 minutes. The system will prioritize the blood pressure retest and send a notification. Once subsequent blood collection windows open and conditions are met, the system will locate and reactivate the suspended blood collection task.

[0079] In another example, the eye drop task exceeded the latency limit. The task was suspended. The system found another task in the same ready state, the preoperative instructions reading, and promoted it to the current task. The eye drop task will be reactivated once the medication timing is met.

[0080] The third level of circuit breaker emergency call is as follows: The third-level circuit breaker emergency call unit is triggered in the following two situations: the candidate set is empty; or, based on global time calculation, it is found that no matter how the remaining paths are adjusted, the safety window of at least one rigid task will still be breached.

[0081] Global time estimation starts from the current moment and simulates the time accumulation of all incomplete tasks along each dependent branch according to their estimated duration. For flexible tasks... Within the specified range, adjust the relative execution order of the suspended task with adjacent flexible tasks, and attempt to simulate insertion at different positions. If, after traversing all feasible combinations, there are still rigid tasks whose simulated completion time falls outside the safety window, then the task is determined to be unresolvable.

[0082] In one example, the patient arrived at 8:10 AM on the day of surgery, only 50 minutes before the surgery was scheduled to begin. Simulated pupil dilation took 30 minutes, and visual recovery observation took 20 minutes, for a total of 50 minutes. Even after skipping all flexible tasks, the simulated completion time for the visual recovery assessment was after 8:30 AM, exceeding the safety window and triggering a circuit breaker for assistance.

[0083] After the circuit breaker emergency call unit is triggered, all automatic adjustments of the silent continuation engine are frozen. The threatened rigid task nodes are located, and their type names and failure types are extracted. Assuming the latest end boundary of its safety window is T5, and the current clock is T6, the latest completion time limit is determined. Construct a help request information package, including patient identification, threatened task type, failure type, and... .

[0084] Based on the information packet, two outputs are generated: The patient receives a single waiting prompt, simply instructing them to wait at their current location without presenting the reason for the obstruction or alternative solutions; the healthcare provider receives a task-level assistance command, including the patient's location, the name of the threatened task, and... Countdown information. Healthcare workers can use this information to directly proceed with treatment.

[0085] The resolved instruction sequence is as follows: After the silent continuation engine completes resolution at any level, the instruction generation module is retried to generate new guidance statements following the same process. During resolution, the patient's terminal continues to display the last instruction pushed before resolution. Once the new statement is generated, it overwrites the old instruction push, and its content corresponds to the highest priority task to be executed in the adjusted dependency graph. The patient is consistently presented with continuous action guidance, and the background resolution process is completely transparent.

[0086] Through the collaborative work of the aforementioned modules, this system transforms the multi-task time-series coordination of ophthalmic day surgery preoperatively into a continuous, conflict-free flow of guidance instructions. Under normal circumstances, it guides patients to complete tasks step by step, and proactively resolves uncertainties in a three-tiered, progressive manner when execution deviations occur. Patients only need to follow the current, single instruction, effectively reducing the cognitive load of preoperative preparation and the risk of operational omissions.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the technical concept of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fully intelligent triage and reminder system for the entire process of ophthalmic day surgery, characterized in that, It includes a task modeling module, a state awareness module, a silent continuation calculation engine, and an instruction generation module; The task modeling module is used to obtain the patient's preoperative medical orders and parse them into task nodes after the surgical schedule is determined, mark each task node as a rigid task or a flexible task, set a safety window for the rigid task, calculate the maximum allowable sliding time for the flexible task, and construct a task dependency graph containing hard dependency edges and soft dependency edges. The execution order between the two task nodes connected by the hard dependency edge cannot be reversed, and the soft dependency edge points from the flexible task node to the subsequent task node affected by its sliding and is marked with the maximum allowable sliding time. The state awareness module is used to obtain the completion status of inspection-type tasks and non-inspection-type tasks, set a dynamic tolerance time limit for the currently activated task that is adjusted according to the urgency of the remaining rigid tasks, and generate an execution deviation event when the completion status is not obtained after the sum of the predetermined completion time and the dynamic tolerance time limit has expired. The silent continuation calculation engine includes a first-layer elastic buffer unit, a second-layer path reorganization unit, and a third-layer circuit breaker request unit, wherein... The first layer of elastic buffer unit is used to shift the time window of the downstream flexible task along the soft dependency edge by the accumulated delay time when the deviation task is a flexible task and the accumulated delay time does not exceed the maximum allowable sliding time of the flexible task, and to check whether the downstream rigid task is still within its safe window after each shift by forward tracking along the dependency edge. The second-layer path reorganization unit is used to search for ready flexible tasks in the task dependency graph when the first-layer elastic buffer unit has not resolved the hard dependencies. These tasks have all hard dependencies satisfied, their execution location is in the same or adjacent area as the patient's current location, and they have no mutual exclusion relationship with the suspended tasks. The task with the shortest estimated time is selected as the current task, and the deviation task is marked as suspended. The third-layer fuse-based emergency response unit is used to stop automatic adjustment when the first two layers fail to resolve the issue or when it is calculated that the rigid task safety window will be breached. It generates emergency response information, outputs a waiting prompt to the patient, and outputs an emergency response instruction containing location and task information to the medical staff. The instruction generation module is used to, after the task dependency graph state is updated, filter out the task nodes from which all the predecessor tasks of the incoming hard dependency edges have been completed, and determine one of them to generate a single conversational triage statement to push to the patient.

2. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The rigid tasks include at least one of the following: the deadline for fasting and abstaining from drinking before general anesthesia, the deadline for visual recovery after mydriasis, the deadline for delivering corneal topography maps, and the deadline for delivering intraocular lens power calculation results. The flexible tasks include at least one of the following: reading preoperative instructions, re-measuring blood pressure, paying fees, and adjusting the timing of antibiotic eye drops within the medication window.

3. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The task modeling module calculates the maximum allowable sliding time for a flexible task in the following way: Tracing upstream from the current flexible task node to the first rigid task node or immovable time anchor encountered, the end time of that node or anchor is recorded as... ; Search downstream from the current flexible task node until the first rigid task node or immovable time anchor encountered, and record the start time of that node or anchor as _____. ; Obtain the estimated duration of the current flexible task. , to The sum of the estimated durations of all rigid task nodes within the interval and the sum of the necessary time intervals between each task. ; The maximum allowable sliding time .

4. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1 or 3, characterized in that, The soft dependency edge indicates that the time position of the connected subsequent task node is shifted as a whole within the sliding range of the preceding flexible task, and the upper limit of the shift is the maximum allowed sliding time marked on the soft dependency edge.

5. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The state awareness module sets the dynamic tolerance time limit in the following way: Obtain the total critical path duration of the remaining incomplete rigid tasks. The remaining time until the start of the surgery. Calculate the ratio ,according to The dynamic tolerance period is obtained. These are preset positive numbers.

6. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The status awareness module obtains the completion status of examination tasks, including: automatically confirming the completion of examination tasks through the report review completion flag of the hospital's laboratory information system or image archiving system; The status awareness module obtains the completion status of non-examination tasks, including receiving the operation signal of a single confirmation control provided for the current task in the patient's interface.

7. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The first layer of elastic buffer units checks whether the downstream rigid task is still within its safety window by tracing forward along the dependent edge after each translation in the following way: When the planned start time of the rigid task node is earlier than the earliest start boundary of its safety window, or the planned end time is later than the latest end boundary of its safety window, it is determined that the safety window has been exceeded, the executed translation operation is canceled, and the second-level path reorganization unit is triggered.

8. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The second-layer path reorganization unit selects the ready flexible task with the shortest estimated time in the following way: Calculate the sum of the standard execution time of each candidate task and the estimated travel time from the patient's current location to the execution location of the candidate task, and select the candidate task with the smallest sum.

9. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, The help request information generated by the third-layer circuit breaker help request unit includes the patient's identity, the type of the threatened rigid task, the failure type, and the latest completion deadline; The waiting prompts output to the patient do not include the reasons for the blockage or alternative solutions; The help request sent to the medical staff includes the patient's location, the name of the threatened task, and a countdown timer for the latest completion deadline.

10. The intelligent triage and reminder system for the entire process of ophthalmic day surgery according to claim 1, characterized in that, After the instruction generation module filters out the task nodes whose predecessor tasks for all incoming hard dependent edges have been completed, when there are multiple nodes, the one with the latest completion time and the most urgent deadline is used as the push target to generate the single conversational triage statement. Only one triage message is pushed at a time; no task list or overall schedule is pushed.

Images