Bladder irrigation fluid metering method in prostate surgery

Abstract

The invention relates to the technical field of medical information processing and intelligent liquid management, and discloses an intraoperative bladder irrigation liquid metering method for prostate surgery, which comprises the following steps: S1, setting surgical parameters, including inputting the body weight of a patient, preoperative blood sodium concentration and an expected surgical type; s2, the type of flushing fluid used in the operation is selected, and an early warning standard is set based on the sodium concentration of the flushing fluid and a risk threshold value; s3, the input amount of the flushing fluid in the operation is measured in real time through an electromagnetic flow sensor, and the input volume of the fluid is updated; and S4, installing a weighing sensor on the drainage bag, monitoring the mass change of the drainage liquid in real time, and converting to obtain the volume of the drainage liquid. According to the invention, by introducing intraoperative liquid dynamic modeling, blood sodium concentration prediction and feedback calibration mechanisms, the technical effects of intelligent regulation and control, individualized prediction and postoperative data closed-loop return in the whole process of liquid management are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information processing and intelligent fluid management, and specifically provides a method for measuring intraoperative bladder irrigation fluid during prostate surgery. Background Art

[0002] During clinical surgical procedures, especially in scenarios involving a large amount of fluid perfusion or irrigation operations, intraoperative fluid management and electrolyte regulation have always been one of the important factors affecting surgical safety and postoperative recovery quality. Existing technologies manage the intraoperative fluid input and output of patients by relying on the judgment of surgeons' experience, fixed parameter models, or manual recording systems, and supplement them with preoperative static prediction models to estimate electrolyte concentration fluctuations. These methods have been applied to basic surgical management practices for a long time and have certain reference value, but there are still significant limitations in terms of accuracy and dynamics.

[0003] Existing technologies use a static configuration method to construct a fluid prediction model. Its core structure relies on fixed formulas, preoperative average experience data, and manually recorded fluid input and output volumes to judge the fluid balance state. The change in blood sodium concentration is often deduced based on a preset curve, lacking the ability to respond in real time to individual state changes. Such solutions generally have the following problems: First, the true intraoperative fluid absorption amount cannot be sensed in real time, the prediction deviation is significant, and it is difficult to adjust the fluid replacement plan in a timely manner; second, the model itself does not have a feedback mechanism, and postoperative data cannot feed back to the system, resulting in the inability to optimize the prediction path; third, data records are scattered and lack structured archiving, making postoperative traceability and analysis difficult; in addition, existing models are mostly linear or single-variable models, and do not fully consider the dynamic regulation process of multi-variable coupling factors on blood sodium concentration, resulting in insufficient clinical adaptability. The present invention effectively makes up for the above technical shortcomings and significantly improves the intelligence and individualization level of fluid management by introducing real-time prediction modeling, feedback closed-loop mechanism, and visual output methods. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a method for measuring intraoperative bladder irrigation fluid during prostate surgery, which solves the problems in the existing technology that it is difficult to accurately sense the intraoperative fluid absorption amount in real time and the blood sodium concentration prediction lacks a dynamic feedback and individualized regulation mechanism.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for measuring intraoperative bladder irrigation fluid during prostate surgery includes the following steps: S1. Set surgical parameters, including inputting the patient's weight, preoperative blood sodium concentration, and expected surgical type; S2. Select the type of irrigation fluid used during the operation, and set an early warning standard based on the sodium concentration and risk threshold of the irrigation fluid; S3. Measure the input volume of intraoperative irrigation fluid in real time through an electromagnetic flow sensor and update the liquid input volume; S4. Install a weighing sensor on the drainage bag to monitor the mass change of the drainage fluid in real time and convert it to obtain the volume of the drainage fluid; S5. Use an optical sensor to measure the optical absorption rate of the discharged fluid and estimate the blood volume in the discharged fluid; S6. Install a micro flowmeter in the catheter to collect the urine discharge flow rate in real time and calculate the urine output; S7. Calculate the intraoperative fluid absorption amount based on the liquid input amount, output amount, blood volume, and urine output; S8. Predict the change in intraoperative blood sodium concentration based on parameters such as the absorption amount, patient weight, and preoperative blood sodium concentration, and update the blood sodium concentration data in real time; S9. Set a low sodium risk warning threshold. When the predicted blood sodium concentration is lower than the set threshold, trigger a system alarm and give an intervention prompt; S10. Generate a postoperative report. The report content includes intraoperative liquid input and output, absorption amount, blood sodium concentration change, and alarm records, and upload it to the hospital information system.

[0006] Preferably, the flow sensor is an electromagnetic flow sensor, which can measure the liquid flow rate in real time and update the input liquid volume data.

[0007] Preferably, the weighing sensor is used to monitor the mass of the drainage fluid in real time, and the liquid volume is obtained by converting the mass data with the liquid density.

[0008] Preferably, the optical sensor is used to measure the optical absorption rate of the discharged fluid, and the blood volume in the discharged fluid is estimated according to the optical ratio, and the ratio is determined by experimental calibration.

[0009] Preferably, the micro flowmeter can collect the urine flow rate in real time and calculate the urine output, and the calculation of the urine volume is obtained by integrating the flow rate.

[0010] Preferably, the absorption amount calculation step obtains the intraoperative liquid absorption amount by subtracting the blood volume and urine volume from the difference between the liquid input amount and output amount.

[0011] Preferably, the blood sodium concentration prediction step predicts the change in blood sodium concentration based on parameters such as preoperative blood sodium concentration, absorption amount, patient weight, and irrigation fluid type, and updates the intraoperative blood sodium concentration data in real time.

[0012] Preferably, the low sodium risk warning threshold is set according to the patient's preoperative blood sodium concentration and the type of irrigation fluid. If the predicted blood sodium concentration is lower than the set threshold, the system issues an audible and visual alarm.

[0013] Preferably, the report generation step includes intraoperative fluid input and output data, absorption volume, blood sodium concentration prediction and alarm records, and the data is uploaded and managed through the hospital information system.

[0014] The present invention provides a method for measuring bladder washing fluid during prostate surgery. It has the following beneficial effects: 1. The present invention adopts a dynamic modeling method for predicting the real-time intraoperative fluid absorption and blood sodium concentration, achieving the technical effect of instantly grasping body fluid fluctuations and sodium ion levels during surgery. Compared with the existing technology that relies solely on preoperative experience or manual estimation of absorption, it solves the problems of delayed prediction and slow regulation in clinical practice.

[0015] 2. The present invention achieves parameter self-calibration and prediction path optimization by introducing a postoperative feedback closed-loop mechanism and combining it with an error reverse analysis model. Compared with the static model and lack of reverse learning ability in traditional solutions, it effectively makes up for the bottleneck of prediction deviation caused by individual differences between different patients, and enables the system to have adaptive evolution capabilities.

[0016] 3. The present invention constructs a data-driven fluid management system for the entire cycle from preoperative to postoperative, and is supplemented by visual output and report archiving technology, which improves data readability and the standardization of medical records. Compared with the existing technology where data is scattered and lacks structural organization, it solves the practical problems of fragmented postoperative management and difficult information tracing.

[0017] 4. The present invention adopts a blood sodium concentration control strategy based on multivariable function combination modeling, which effectively improves the accuracy of fluid management. Compared with the existing processing methods that only rely on weight conversion or empirical input, it not only fits the physiological curve better in the prediction results, but also solves the risk hidden dangers caused by the delayed response of dependent variables in the fluid control process, and has stronger clinical adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Please see attached Figure 1 The embodiment of the present invention provides a method for measuring bladder washing fluid during prostate surgery, comprising the following steps: S1. Set surgical parameters, including inputting the patient's weight, preoperative serum sodium concentration, and expected surgical type; Before entering the modeling process, the system needs to complete the initialization of the basic data structure and the acquisition of input variables. As the starting link of the entire logical chain, the task of step S1 is not to analyze, but to establish the entry mapping of the parameter path. This step needs to complete the presetting and binding of key variables before and during the operation, which is a necessary prerequisite for realizing dynamic tracking and time series processing. Without complete data, the model cannot be started; without a defined structure, the process cannot be connected.

[0021] In this embodiment, the system first defines a set of master variables , which is used to describe all input types that need to participate in calculations or recordings. It includes but is not limited to the following categories: Basic physiological parameters, such as weight ( ), gender ( ), height ( ; Initial biochemical indexes, such as preoperative serum sodium concentration ( ); Liquid configuration parameters, including the type number of intraoperative perfusion fluid , the sodium concentration of each type of liquid ( ; Configuration items such as the number of drainage channels, the upper and lower limits of the pump speed, and the status of the device interface; Time synchronization reference point .

[0022] Generally, the above variables need to be connected to the hospital information system (HIS) or the anesthesia information system (AIMS) through an interface, or can also be manually entered by the operator in the system control panel, and the system will perform consistency verification on the entered items.

[0023] Specifically, the initial volume estimation of body fluid is constructed using the following empirical expression: ; Where: represents the preoperative basic body fluid volume, in liters; represents the weight (kg), represents the height (cm); is a parameter coefficient preset according to gender and body type.

[0024] In a possible implementation, the system automatically calls different parameter combinations according to the entered gender parameter: Male: , , ; Female: , , 。

[0025] This estimated body fluid volume will serve as the initial assignment basis for the volume variable in the blood sodium prediction formula in step S4 and will be dynamically adjusted according to subsequent data. and will be dynamically adjusted according to subsequent data.

[0026] As an option, the system also supports detecting and calibrating the status of intraoperative device ports. The perfusion device, drainage path, and collection probe need to be registered with a unique code in the system. This code exists as a signal tracking index field in the data path to ensure that each input and output flow has a unique identifier, facilitating traceability of the path during subsequent modeling.

[0027] In some embodiments, to ensure time consistency, the system adopts a centralized time reference and broadcasts it to all modules. Each time step advances synchronously in the form of . The default time resolution is 30 seconds and can be adjusted according to the surgical rhythm.

[0028] In addition, the system performs an integrity scan before the end of step S1 to detect whether there are undefined fields, type errors, or unit inconsistency problems. If a missing field is found, the system returns an interactive prompt and pauses the process startup to avoid invalid variables in the modeling chain.

[0029] In summary, step S1 constructs the definition set and parameter initial space of all variable channels in this system. It does not directly participate in model prediction, but its structural integrity directly determines whether the subsequent calculation path is stable and coherent, which is an essential prerequisite for the system to run.

[0030] S2. Select the type of irrigation fluid used during the operation and set an early warning standard based on the sodium concentration of the irrigation fluid and the risk threshold; After completing the initial variable binding, the system enters the real-time acquisition stage. Step S2 does not construct the model structure nor directly participate in parameter operations. Its core role is to provide a dynamic input channel. The input information types are complex, the data frequencies are inconsistent, and the source devices are heterogeneous. They need to be standardized before entering the subsequent processing path. This step is a key component of the system's perception layer and is the basic interface layer for establishing the model's responsiveness.

[0031] In this embodiment, the system captures various intraoperative input data in real time through an edge acquisition terminal. Including but not limited to: Perfusion fluid flow rate , unit mL / min; drainage fluid flow rate , unit mL / min; sodium concentration of perfusion fluid , unit mmol / L; monitored blood sodium value , unit mmol / L, operating status identifier of the circulation pump, with a Boolean value, timestamp , automatically recorded at the sampling time point.

[0032] Generally, the system default sampling period is 30 seconds. If the device supports high-frequency transmission, it can be switched to 5-second level sampling. After the sampling frequency is unified, the system uses the time window method to align multi-source data.

[0033] Specifically, let the sampling time step be , and the acquisition window be , then: ; ; The above formula converts the non-uniform sampling signal into an equally spaced sequence, adapting to the time structure required for subsequent modeling.

[0034] As an option, if the device cannot provide a continuous signal, the system will use the most recent valid value for short-term interpolation, and the interpolation strategy is only valid within an interval of no more than two time windows.

[0035] In a possible implementation, the system uses the synchronization signal channel to correct the time reference of different devices. Each data packet is attached with a timestamp and the original device time , and corrects the sampling synchronization through the time offset function .

[0036] In some embodiments, the system introduces an intermediate cache module to balance network instability or uneven sampling problems. The cache structure supports a sliding window mechanism, and each window length is time steps to ensure data continuity. The data cleaning strategy will be executed in the next step (S3) and not completed in this step.

[0037] In addition, all acquired variables need to be registered through a unique identifier encoding. The system describes the metadata such as the source, type, unit, update time, etc. of each variable through a registry. For example: Type: Flow rate (continuous type) Unit: mL / min Source: Perfusion controller 1 Update time: 30s This encoding system will be used as the main index in data transmission and storage.

[0038] In summary, step S2 establishes a real-time path for data input, which has a fundamental supporting role in subsequent data cleaning, model input structure, and prediction process, and is an important node that cannot be skipped in the initial stage of system operation.

[0039] S3. Measure the input volume of intraoperative irrigation fluid in real time through an electromagnetic flow sensor and update the liquid input volume; After the intraoperative parameter initialization is completed, the system starts to model the liquid absorption process. This stage marks the transition from static parameter input to dynamic data processing. The core of step S3 lies in estimating the proportion and rate at which the perfusion fluid truly enters the body, providing key inputs for modeling subsequent blood sodium changes. This process requires cross - calculation based on multi - dimensional information such as input - output difference, body fluid dilution reaction, and flow rate time - series fluctuations. Instead of relying on empirical formulas or preset ratios, it introduces time - series modeling logic to construct the actual absorption path during the operation.

[0040] In this embodiment, the absorption estimation module implemented in step S3 uses the time difference between the input liquid volume and the discharged liquid volume as the basic observation item, and combines the current total body fluid volume of the patient and the concentration change trend to establish an absorption rate estimation expression.

[0041] Generally, the intraoperative absorption volume can be obtained by subtracting the discharged volume from the input volume. However, this difference is affected by many factors and does not equal the true absorption amount. The system needs to further introduce a body fluid dilution index to correct this estimate, which is particularly crucial when the sodium concentration changes rapidly.

[0042] Specifically, within each unit time period, the system establishes a liquid absorption rate estimation model according to the following expression: Intraoperative absorption rate = Rate of change of input volume−Rate of change of discharged volume−Physiological retention term Among them, The rate of change of input volume is the sum of all input liquid volumes within a unit time divided by the length of this time period; The rate of change of discharged volume is the sum of drainage, negative pressure aspiration and other liquid discharge volumes within a unit time divided by the length of this time period; The physiological retention term is used to represent the part of the liquid that has not been drained temporarily but has not participated in the body fluid circulation either, and can be automatically fitted by the model according to the blood sodium change trend, with the unit of milliliters per minute.

[0043] In a possible implementation, the system introduces a lag coefficient to quantify the average delay time required for the perfusion liquid to enter the vascular system. This parameter is automatically adjusted according to the intraoperative physiological state of the patient, and the typical value range is between one and five minutes.

[0044] In some embodiments, the system further considers the influence of the sodium concentration of the perfusion fluid on the change of body fluid osmotic pressure, constructs an absorption discrimination boundary. If the absorption estimation value is greater than a certain threshold for multiple consecutive time periods and is accompanied by a stable decrease in sodium concentration, it is determined that the liquid has been truly absorbed; conversely, if the discharged volume is higher than the input volume but the blood sodium does not rise, it is presumed that there is reverse absorption or measurement error.

[0045] As an option, the system constructs the variables in the absorption estimation function as trainable parameter terms, allowing the model to perform multiple rounds of fitting on multiple patient samples and gradually form an absorption pattern template adapted to different surgical procedures.

[0046] In addition, the absorption modeling module in step S3 shares the same timestamp system with the subsequent blood sodium modeling module to ensure the consistency of the data input path and prevent error accumulation caused by variable mismatch.

[0047] In terms of technical implementation, this step is not a single calculation but a sliding window-based iterative update mechanism. The system recalculates the absorption rate every certain period (e.g., 30 seconds to 1 minute), dynamically covering the old values and forming an absorption trend sequence under continuous time.

[0048] In summary, step S3 not only completes the dynamic estimation of the actual absorption process of the perfusion fluid but also constructs an intermediate response mechanism between the liquid flow path and the blood sodium change, which is a crucial link in realizing the closed-loop prediction logic chain.

[0049] S4. Install a weighing sensor on the drainage bag to monitor the mass change of the drained fluid in real time and convert it to obtain the volume of the drained fluid; After the absorption estimation is completed, the system enters the sodium ion concentration modeling stage. The core of this stage is to establish a causal mapping relationship between the sodium flux change and the body fluid concentration change through a time-discrete modeling method. Step S4 predicts the sodium concentration change at each time step based on the absorption liquid volume and component information output by step S3, combined with the body fluid volume and physiological regulation parameters. This process has the characteristics of periodicity, real-time, and recursiveness.

[0050] In this embodiment, the system constructs a basic model for the change of sodium ion concentration based on the principle of mass conservation. The concentration change is based on the ratio of the net increase of sodium ions in a unit time period to the body fluid volume. Its basic expression is as follows: ; Where: represents the change in blood sodium concentration at the th time step, with the unit of mmol / L; represents the total amount of sodium ions absorbed into the body at this time step, with the unit of mmol; represents the amount of sodium ions excreted from the system with the body fluid at this time step, with the unit of mmol; represents the amount of sodium loss caused by the physiological regulation mechanism (such as renal tubular filtration, cell transport) at this time step, with the unit of mmol; represents the estimated effective body fluid volume at this time step, with the unit of L.

[0051] Generally, the sodium input amount It can be obtained by the following calculation: ; Where: is the volume (unit: L) of the th type of liquid absorbed into the body at time step ; is the sodium ion concentration of this type of liquid (unit: mmol / L); is the number of types of liquids participating in the modeling.

[0052] Sodium excretion is similarly defined as: ; Where: represents the volume of the th type of liquid drained or excreted at this time step; is the sodium concentration of this excreted liquid; is the number of types of liquids in the excretion path.

[0053] Specifically, within a time step (e.g., 30 seconds), the system calculates through the above formula, and linearly superimposes this value with the blood sodium concentration at the previous moment to obtain the predicted value at the next moment: ; As an option, when the system calculates , it introduces a physiological adjustment factor to reflect the intraoperative circulatory status or volume responsiveness, and obtains: ; Where: is the basic body fluid volume estimated based on the patient's initial weight, gender, and intraoperative fluid input and output; is the dynamic adjustment coefficient (dimensionless), usually with a value range between 0.85 and 1.15, used to correct short-term volume changes.

[0054] In a possible implementation, the adjustment term can be expressed in the following form: ; Where: is the sodium adjustment coefficient (unit: 1 / min), and its value can be set according to experience in different surgical scenarios, with common values such as 0.002–0.005; is the blood sodium concentration at the current moment.

[0055] In some embodiments, to improve the model's response ability to sudden absorption shocks, the system introduces a second derivative: ; Based on this, the acceleration of blood sodium change is judged, and the time step is dynamically adjusted when necessary. And the high-frequency prediction module is called.

[0056] In addition, this prediction mechanism not only provides numerical output, but also provides a theoretical benchmark for the subsequent error evaluation module, serving as the input reference curve for the closed-loop calibration algorithm.

[0057] In summary, the concentration modeling module in step S4 realizes the systematic mapping from the behavior of body fluid dynamics to the change of electrolyte concentration by constructing a clear sodium flux conservation expression, and has a complete and extensible mathematical structure, ensuring the publicity, accuracy and implementation basis of the modeling process.

[0058] S5. Use an optical sensor to measure the optical absorption rate of the drained fluid and estimate the blood volume in the drained fluid; After completing the recursive modeling of blood sodium concentration, the system does not directly output the prediction result, but introduces an error correction mechanism. This step is used to identify and correct the prediction deviation caused by modeling assumptions, clinical operation deviations or exogenous factors. Step S5 aims to construct a dynamic feedback loop to keep the consistency between the aforementioned predicted concentration trajectory and the actual monitoring value. Its core lies in establishing a numerical response function of the concentration prediction error and adjusting the key parameters of the previous model accordingly to achieve the self-adaptive correction of the model itself.

[0059] In this embodiment, the system first defines the prediction error of blood sodium concentration as the difference between the actual observed value and the model prediction value: ; Where: represents the concentration error at time step , with the unit of mmol / L; is the actually collected blood sodium concentration value; is the model prediction output value.

[0060] Generally, the system will perform a moving window average on this error value to smooth the non-structural offset caused by short-term fluctuations. The expression is as follows: ; Where: represents the window length, which can be configured according to the intraoperative frequency, and the typical value is 4 or 6; is the smoothed error at the current moment.

[0061] As an option, if continuously exceeds the set threshold (for example, 0.8 mmol / L), the system will start the parameter correction module. At this time, the key variables in the original prediction model such as body fluid volume , adjustment term , Absorption Estimation Coefficient will be adjusted.

[0062] Specifically, the correction mechanism constructs the following adjustment expression: ; where: is the parameter to be adjusted participating in the modeling; is the adjustment rate factor, which needs to be preset according to the parameter sensitivity; is the partial derivative of the predicted output with respect to the parameter.

[0063] This correction model realizes the directional adjustment of the core parameter terms affecting the error according to the reverse sensitivity propagation path, ensuring that the model quickly approaches the actual change trajectory without introducing mutation behavior.

[0064] In a possible implementation, the system is provided with a multi-level error response module. If the error is below the threshold, only the body fluid volume parameter is adjusted; if the error exceeds the upper limit, the absorption rate estimation logic and the adjustment term are adjusted simultaneously , and the next time step of the predicted output is frozen to avoid misleading clinical judgment.

[0065] In some embodiments, the system also introduces a non-linear calibrator to map the error to the corrected value of the model output, and acts on the predicted concentration in the form of function transformation instead of directly backpropagating the parameters: ; where: is the error response function obtained by empirical or machine learning modeling, with monotonicity and boundary constraints.

[0066] In addition, the calibration module described in step S5 has memory. The system records each correction operation and its corresponding error to form an error convergence curve, which is used to evaluate the model adaptation efficiency and provide suggestions for subsequent clinical strategy adjustment.

[0067] In summary, step S5 constructs a closed-loop feedback mechanism driven by error. This mechanism is deeply integrated with the prediction model structure, not only immediately corrects the parameters, but also applies secondary optimization to the output results, ensuring dynamic consistency between the prediction and the actual data, and laying a data foundation for the final visual output of the system.

[0068] S6. Install a micro flowmeter in the catheter to collect the urine discharge flow rate in real time and calculate the urine output; After completing the prediction of blood sodium concentration and error correction, the system does not immediately output static values, but enters the dynamic visualization and clinical response prompt phase. This phase aims to transform the output of the complex time series model into an intuitive and structured information carrier, and perform data tagging near specific physiological thresholds for real-time reading and judgment by the intraoperative operator. Step S6 is formally independent of the modeling and calculation module, but logically depends on all the foregoing results, and is the key bridge for realizing the clinical implementation of data.

[0069] In this embodiment, the system first constructs a visualization trajectory with time as the horizontal axis and the predicted value as the vertical axis according to the corrected blood sodium concentration prediction sequence in the previous step.

[0070] Generally, the system uses a combination of line charts, gradient background color bands, and key point labels for display. Among them: The horizontal axis is the time point , and the unit can be configured as seconds or minutes; The vertical axis is the predicted concentration value , and the unit is mmol / L; If there are points in the prediction sequence that exceed the upper limit of the safety threshold or are lower than the lower limit , the system automatically adds warning marks.

[0071] Specifically, the system defines the following by setting a configurable clinical safety range: ; Among them: is the lower limit warning value of blood sodium, and the default is set to 130 mmol / L; is the upper limit warning value of blood sodium, and the default is set to 150 mmol / L; Points outside this range will be given color identification and behavior prompts.

[0072] As an option, the system introduces a visual expression of the trend derivative . That is: ; It is transformed into an arrow direction, color gradient, or numerical dynamic curve to express the current concentration change rate.

[0073] In a possible implementation, the system can overlay the above information on the patient's individual information card in the form of layers, including key parameters such as weight, gender, initial body fluid volume estimate, current absorption rate, adjustment coefficient, etc., and dynamically update them in chronological order.

[0074] In some embodiments, the system also supports displaying the predicted concentration series in a rolling window format in comparison with the real-time measured values, forming a "predicted-actual" dual-channel curve to facilitate clinical intervention judgment.

[0075] In addition, the system provides an abnormal change detection mechanism. When the following conditions occur at multiple consecutive forecast time points, the system will actively pop up a prompt: ,in is a preset change rate threshold (e.g. 1.5mmol / L / min); Or two consecutive prediction results are in the danger zone.

[0076] Prompts may include text warnings, sound signals, console highlights, and suggested actions (such as slowing down the perfusion rate, checking drainage patency, etc.).

[0077] As an extended output, the system also generates a summary of forecast indicators including: The minimum, maximum and average blood sodium values ​​in the next three minutes; Trend stability factor , represents the concentration variance; The prediction model confidence score is constructed based on the residual curve.

[0078] In addition to the graphical interface, the output method also supports structured data export (JSON / XML format) for postoperative analysis or access to remote monitoring platforms.

[0079] In summary, step S6 not only realizes the interactive presentation of model results, but also semantically packages the data in a medically readable way, making the prediction mechanism clinically interpretable and action-oriented, and is the main interaction node between the system and the operator.

[0080] S7. Calculate the amount of fluid absorbed during surgery based on fluid input, output, blood volume, and urine output; After completing the visual output and abnormal prompts, the system needs to establish a data linkage mechanism with external devices to achieve dynamic control or auxiliary decision-making. Step S7 takes over the results of the aforementioned prediction and warning modules, converts the modeling output into control signals or parameter recommendations, and transmits them to surgical equipment, monitoring systems or data platforms. This step connects the model logic chain and the device response channel, and is an important part of the system's construction of a "perception-modeling-response" closed-loop architecture.

[0081] In this embodiment, the system is based on the predicted concentration sequence With dynamic changing trends , construct linkage control rule sets, and generate control factors based on them , used to guide the operation of intraoperative perfusion, drainage or other regulatory devices.

[0082] Under normal circumstances, the control factor can be defined as follows: ; where: is the prediction error; is the concentration change rate; Risk is the current blood sodium risk index; , , are weight coefficients, which are set according to the model sensitivity and surgical procedure requirements.

[0083] As an option, the system adopts a risk stratification mechanism to map the current state to discrete levels. For example: Normal: The control signal is to hold; Borderline: The signal is a warning; Hazardous: The signal is an intervention recommendation.

[0084] Specifically, if the current prediction result satisfies: ; the system will issue an instruction to lower the perfusion rate, where: is the minimum safety concentration threshold; is the change rate boundary value.

[0085] In a possible implementation, the system uses a digital - analog interface to output the control signal through serial or parallel data streams and connect it to the device control bus. This process can be completed by a preset driver protocol without user intervention.

[0086] In some embodiments, the system sets up a response buffer to record each control output and its corresponding input parameters, forming a response linked list. If the device does not feedback the execution status within the set time, the system will repeat sending the signal and mark the current state as "pending confirmation".

[0087] In addition, to support postoperative retrospective analysis, the system archives the logs of all linkage control records. Each record contains fields such as timestamp, concentration prediction value, control parameters, response mark, confirmation status, etc., and can be exported as a structured data file.

[0088] As an extended method, the system can also access an external expert system and call an external decision - making model through an interface protocol. At this time, the control factor can be used as one of the inputs to participate in multi - modal judgment.

[0089] In summary, step S7 constructs a device response module driven by the modeling results, which not only has a control signal generation and execution path, but also realizes operation recording and traceback links, providing data support and technical channels for intraoperative intervention, and finally realizing the functional integration of the prediction modeling system and clinical devices.

[0090] S8. Predict the change of intraoperative blood sodium concentration according to parameters such as absorption volume, patient weight, preoperative blood sodium concentration, etc., and update the blood sodium concentration data in real time; After the generation of the control signal and the linkage response of the peripheral device are completed, the system still needs to archive and evaluate the full-process data. Step S8 no longer focuses on real-time interaction, but turns to posterior analysis and model closed-loop optimization. This step aims to structurally store the intraoperative prediction trajectory, error response, intervention actions and final results, and build a model performance audit system based on this. Its function is not limited to recording, but also provides a technical carrier for subsequent model fine-tuning and experience learning.

[0091] In this embodiment, after each prediction cycle is completed, the system packs the following key data and writes it into an independent data object: Timestamp ; Model prediction value ; Actual acquisition value ; Model error ; Control signal ; External device response status .

[0092] Generally, the system triggers a data archiving process at a fixed cycle (such as every 60 seconds), and appends and stores the above objects to a time series database. This database supports index query, status filtering and time segment aggregation operations, which is convenient for postoperative retrospective analysis.

[0093] Specifically, the system quantifies the performance of the model during the entire surgical process by constructing a performance evaluation function. The evaluation indicators include but are not limited to the following categories: ; ; ; Among them: represents the total number of prediction cycles; represents that there is a linkage control signal output at time point , otherwise it is 0; reflects the activity of the control mechanism.

[0094] In a possible implementation manner, the system can also calculate the standard deviation of the prediction error, which is used to reflect the model stability: ; As an option, the system constructs an adaptive parameter write-back logic based on the above error statistic and response record. If the standard deviation of the error deviates for a long time under a specific surgical procedure, the system will add such samples to the training set queue for re-training or fine-tuning of the model in subsequent versions.

[0095] In some embodiments, the system introduces a model performance label mechanism and constructs a scoring matrix based on multi-dimensional index scoring , each item corresponding to a type of evaluation dimension, such as fitting accuracy, response speed, control timeliness, data integrity rate, etc., and the quantization range can be set from 0 to 1.

[0096] In addition, to support multi-case normalization comparison, the system normalizes all key indicators and outputs a postoperative summary report. The report includes: Prediction - actual curve comparison; High-risk point annotation; Frequency of control behavior triggering; Model performance level; Data quality status.

[0097] The data report can be exported as a PDF or a structured interface file for accessing the in-hospital electronic medical record system, data governance platform, or model development end.

[0098] In summary, step S8 establishes the last link of model life cycle management, constructing a traceable closed loop for modeling, prediction, response, and evaluation, which not only improves the system integrity but also provides data basis and structural support for future model optimization and strategy design.

[0099] S9. Set the low-sodium risk warning threshold. When the predicted blood sodium concentration is lower than the set threshold, trigger the system alarm and give an intervention prompt; After completing the prediction data archiving and model evaluation, the system process enters the final stage, that is, the internal update execution of the model closed-loop tuning strategy. Step S9 is not output to the outside and is not directly open to the user operation interface. Instead, through historical error statistics results, adaptive parameter adjustment logic, dynamic reorganization of training samples, etc., local iteration of model structure parameters or weight redistribution is achieved. This stage does not rely on manual intervention and exists as an internal self-evolution mechanism of the system, marking the transition from a predictive model to an evolutionary model.

[0100] In this embodiment, the system first uses the predicted error residual sequence recorded in step S8 as the input, calculates its central tendency and fluctuation range, and constructs an approximation of the loss function gradient.

[0101] Generally, the system uses a residual mean square loss function in the following form: ; Wherein: ; is the total number of time steps; is the cumulative error loss metric, used to measure the overall prediction deviation level of the current model.

[0102] Specifically, the system constructs a partial derivative vector to perform sensitivity analysis on the core variables of the model. Taking the set of prediction function parameters as the target, calculate the loss contribution of each parameter under the current sample: ; As an option, if the parameter gradient direction remains the same sign in multiple consecutive prediction cycles, the system will perform a round of local iteration with a fixed step size: ; Wherein: is the learning rate, and its value range can be set to 0.001 - 0.01, which is dynamically adjusted according to the convergence rate; The condition for execution is that the gradient direction remains consistent to avoid frequent oscillations.

[0103] In a possible implementation, the model does not perform a complete parameter update, but only fine-tunes a subset of parameters with higher input sensitivity, so as to maintain the stability of the backbone structure.

[0104] In some embodiments, the system introduces a historical sample dynamic scoring mechanism to perform residual weighting on the feature vectors in the training dataset to enhance the modeling weight of low-frequency but highly influential samples.

[0105] The specific weight factor can be defined as: ; Wherein: is the weight of the th sample; is the adjustment factor; is the absolute value of the error corresponding to this sample.

[0106] In addition, the system supports a batch parameter snapshot freezing mechanism. If the overall loss L of the model does not decrease after a round of fine-tuning operations, the system will revoke the parameter update of this round and roll back to the previous snapshot state to prevent the accumulation of negative learning.

[0107] As an extended technical path, the model closed-loop update module can also generate a parameter difference log, record the values of each parameter before and after each round of iteration, and form an auditable internal model change chain.

[0108] In summary, step S9 realizes the adaptive structure adjustment of the model under unsupervised conditions. Its core lies in realizing parameter perturbation through the residual backpropagation path and maintaining the continuous evolution ability of the model structure, which is one of the basic mechanisms to ensure the stable operation of the system in long-term deployment.

[0109] S10. Generate a postoperative report, the content of which includes intraoperative fluid input and output, absorption volume, changes in blood sodium concentration, and alarm records, and upload it to the hospital information system; After the closed-loop update and parameter fine-tuning of the model, it is still necessary to build a stable data interface mechanism. This mechanism is used for data connection and horizontal expansion between different modules of the system, and at the same time provides a unified data protocol standard, providing a technical basis for subsequent external platform docking, cross-model invocation, or clinical system integration. Step S10 does not serve the modeling itself, but ensures the security, consistency, and continuity of data transmission between sub-modules, laying a structural support for the long-term operation of the system in a multi-terminal deployment environment.

[0110] In this embodiment, the system constructs a unified data encapsulation structure and uses a hierarchical coding method to standardize the representation of data such as prediction results, error records, model parameters, control signals, and log summaries.

[0111] Generally, the data encapsulation structure is expanded in the form of key-value pairs. Each module output object contains attributes such as field identifier, data type, time tag, unit description, version number, and checksum.

[0112] Specifically, the system defines the main structure of the data packet as follows: ; Where: represents the complete data object at the th time step; is the field name, such as predicted concentration (C_pred), actual concentration (C_obs), error value (ε); is the field value, supporting multiple types such as numerical, boolean, enumeration, and string; All fields carry a time tag , ensuring consistent timing.

[0113] As an option, the system uses JSONSchema to describe all data structure specifications and also has a data signature field for verifying the integrity of the content.

[0114] In a possible implementation, the system introduces multiplexing transmission logic. That is, multiple data packets within the same time step will be encapsulated as a composite flow unit through a logical channel and then distributed to each target module or external system by the transmission scheduling module.

[0115] In some embodiments, the system supports writing data to redundant paths. For example, when there is an I / O delay in the primary database, the data will be synchronously written to the backup channel, and the switching state will be recorded: ; In addition, the system attaches a protocol version number to each data stream , to support a multi-version compatible structure. If the receiving-end structure is inconsistent with the sending-end version, the system will call a converter for field mapping.

[0116] As an extension, the system opens standardized interfaces (such as RESTful API, WebSocket, or HL7 FHIR) to support the access of third-party platforms, such as anesthesia workstations, intraoperative monitoring systems, hospital data buses, etc.

[0117] The interface call permissions are uniformly managed by the access control policy. Each type of external entity needs to go through the authentication mechanism before it can read or write to the specified structured fields, to avoid cross-domain information leakage.

[0118] In some embodiments, the system can generate a visual flow diagram for data interaction between different modules. Each data transfer event is recorded, including the source module, target module, field summary, delay value, interaction success flag, etc., to form a traceable data operation track.

[0119] In summary, the data encapsulation and interface module described in step S10 not only ensures the efficient connection between the sub-modules within the system, but also establishes a structural bridge for clinical integration and data sharing, and is the basic component to support the deployability and maintainability of the modeling architecture.

[0120] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for measuring bladder washing fluid during prostate surgery, characterized in that: The following steps are involved: S1. Set surgical parameters, including inputting the patient's weight, preoperative blood sodium concentration, and expected type of surgery; S2. Select the type of irrigation fluid used during surgery and set warning criteria based on the sodium concentration of the irrigation fluid and the risk threshold; S3, measuring the input amount of intraoperative flushing fluid in real time through an electromagnetic flow sensor, and updating the liquid input volume; S4, installing a weighing sensor on the drainage bag to monitor the mass change of the drainage fluid in real time and convert it to obtain the volume of the drainage fluid; S5, using an optical sensor to measure the optical absorbance of the discharged fluid, and to estimate the blood volume in the discharged fluid; S6. Install a micro flow meter in the urinary catheter to collect urine discharge flow rate in real time and calculate the urine discharge volume; S7. Calculate the amount of fluid absorbed during surgery based on fluid input, output, blood volume, and urine output; S8. Predict the change of blood sodium concentration during surgery according to the absorption amount, patient weight, and preoperative blood sodium concentration parameters, and update the blood sodium concentration data in real time; S9. Set a low sodium risk warning threshold. When the predicted blood sodium concentration is lower than the set threshold, the system alarm is triggered and an intervention prompt is given; S10. Generate a postoperative report, which includes intraoperative fluid input and output, absorption volume, changes in blood sodium concentration, and alarm records, and upload it to the hospital information system.

2. A method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The flow sensor is an electromagnetic flow sensor, which can measure the flow rate of the liquid in real time and update the input liquid volume data.

3. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The weighing sensor is used to monitor the mass of the drainage fluid in real time, and the mass data is converted with the liquid density to obtain the liquid volume.

4. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The optical sensor is used to measure the optical absorbance of the discharge fluid and to infer the blood volume in the discharge fluid according to the optical ratio, wherein the ratio is determined by experimental calibration.

5. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The micro flow meter can collect urine flow rate in real time and calculate the amount of urine discharged, and the amount of urine is calculated by integrating the flow rate.

6. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The absorption amount calculation step obtains the intraoperative fluid absorption amount by deducting the blood volume and urine volume from the difference between the fluid input volume and the fluid output volume.

7. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The blood sodium concentration prediction step predicts the change of blood sodium concentration by using the parameters of preoperative blood sodium concentration and absorption amount, patient weight, and flushing fluid type, and updates the intraoperative blood sodium concentration data in real time.

8. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The hyponatremia risk warning threshold is set according to the patient's preoperative blood sodium concentration and the type of flushing fluid. If the predicted blood sodium concentration is lower than the set threshold, the system will issue an audible and visual alarm.

9. The method for measuring bladder washing fluid during prostate surgery according to claim 1, characterized in that: The report generation steps include intraoperative fluid input and output data, absorption volume, blood sodium concentration prediction and alarm records, and data uploading and management through the hospital information system.

Images