A method and system for hysteroscopic real-time liquid pressure stabilization control

Abstract

The present invention discloses a hysteroscope real-time liquid pressure stabilization control method and system, which specifically relates to the field of hysteroscopic surgery liquid pressure control, and is used to solve the problems of unclear surgical field and unstable operation caused by pressure fluctuations. The method collects multimodal data in real time and performs preprocessing, and uses a long short-term memory network to build a mathematical model reflecting the operation habits, so as to capture the time correlation and pattern characteristics of the operation; through a dynamic time warping algorithm and multi-fractal spectrum analysis, the matching degree between the operation data and the model and the multi-scale fluctuation characteristics of the liquid pressure are quantified, and the operation model offset sensitivity value is generated to evaluate whether the current operation has the conditions for further control and adjustment; the empirical mode decomposition technology is combined to accurately identify the peak and valley values ​​of abnormal pressure, and a second-order delay control strategy and a proportional integral differential algorithm are used to perform real-time peak clipping and valley filling, so as to finally achieve smooth and stable liquid pressure.

Description

Technical Field

[0001] The present invention relates to the field of hysteroscopic surgery fluid pressure control, and more specifically, to a hysteroscopic real-time fluid pressure stabilization control method and system. Background Art

[0002] As a minimally invasive gynecological diagnosis and treatment method, hysteroscopic surgery is widely used in the diagnosis and treatment of diseases such as endometriosis, uterine fibroids, and intrauterine adhesions. In hysteroscopic surgery, in order to provide a clear surgical field of view and sufficient operating space, the uterine cavity needs to be expanded by continuous injection of liquid media. However, the stability of liquid pressure is directly related to the safety and effectiveness of the operation. Traditional pressure control methods rely on manual adjustment or simple automatic feedback mechanisms, which are difficult to respond to the complex and changing operational requirements and patient physiological conditions during the operation in real time, thereby limiting the accuracy of the operation and the convenience of operation.

[0003] However, in actual applications, the liquid pressure during hysteroscopic surgery often fluctuates slightly, which may be caused by a variety of factors, including unstable equipment performance, inconsistent surgical operations, differences in patients' physiological conditions, and changes in the external environment. These pressure fluctuations may not only lead to unclear surgical fields of view, increase surgical difficulty, and prolong surgical time, but may also impose additional physiological burdens on patients and increase the risk of postoperative complications. The existing technology has shortcomings in real-time monitoring and stable control of liquid pressure, such as slow response speed, imprecise adjustment, and lack of intelligent adaptive capabilities. There is an urgent need for a method and system that can stably control the liquid pressure in the uterine cavity in real time, accurately, and intelligently, so as to improve the safety and efficiency of surgery and reduce potential risks to patients.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a hysteroscope real-time liquid pressure stable control method and system, through the real-time acquisition and preprocessing of multimodal data, a high-precision data foundation is established, which provides complete support for subsequent operation mode modeling and liquid pressure control; combined with the nonlinear time series modeling technology of the long short-term memory network, the dynamic characteristics of the doctor's operation during the operation are fully captured, and a quantifiable operation habit model is formed; through the dynamic time warping algorithm and the multi-fractal spectrum analysis method, a multi-dimensional evaluation mechanism from operation mode matching to pressure fluctuation characteristics is designed, and a deep analysis of operation and control conditions is realized; based on the empirical mode decomposition technology, combined with the second-order delay control and the proportional integral differential algorithm, the abnormal fluctuation of liquid pressure is dynamically adjusted to ensure the smoothness and stability of pressure control. The present invention constructs a set of multi-level adaptive mechanisms from data to control, organically combines the operation habit modeling with the real-time pressure regulation, and forms an efficient, accurate and intelligent liquid pressure control method to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for stabilizing the real-time fluid pressure of a hysteroscope comprises the following steps:

[0008] S1, collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and pre-processes the collected data to ensure the accuracy and consistency of the data;

[0009] S2, using long short-term memory network to perform nonlinear time series modeling on the preprocessed multimodal data and construct a mathematical model reflecting operating habits to capture complex time correlations and operating patterns;

[0010] S3, evaluates the matching degree between real-time operation data and operation habit model through dynamic time warping algorithm, quantifies the degree of mode deviation, and uses multi-fractal spectrum analysis method to extract multi-scale characteristics of pressure fluctuation, and generates operation model deviation sensitivity value to analyze whether the current operation meets the conditions for control adjustment;

[0011] S4, if the conditions for control adjustment are met, the empirical mode decomposition technology is applied to identify abnormal peaks and valleys in the real-time pressure data, and the second-order delay control strategy and proportional integral differential control algorithm are combined to perform peak clipping and valley compensation on these abnormal fluctuations to achieve smooth and stable pressure.

[0012] In a preferred embodiment, step S1 includes the following contents:

[0013] S1.1 deploys a multi-sensor network to collect multi-modal data related to pressure, flow and instrument operation in real time during surgery;

[0014] S1.2, performing noise reduction processing on the collected data to eliminate high-frequency noise and abnormal fluctuations in the signal;

[0015] S1.3, for data missing that may occur during the collection process, data is supplemented based on time correlation and the collaborative relationship between features;

[0016] S1.4, in order to ensure the feature balance and scale consistency of multimodal data, the interval mapping method is used to standardize the data of different physical quantities.

[0017] In a preferred embodiment, step S2 includes the following contents:

[0018] S2.1, select the standardized sequence of pressure, flow and operation information from the preprocessed multimodal data set as modeling input;

[0019] S2.2, adopting long short-term memory network as the modeling framework to handle the temporal correlation and complex nonlinear characteristics in multimodal data;

[0020] S2.3, using the preprocessed multimodal dataset as a training set, the weights and structure of the LSTM model are optimized through supervised learning so that it can predict the operation mode;

[0021] S2.4, after training, the LSTM model can generate high-order feature representations that match the operating habits for the input multimodal time series.

[0022] In a preferred embodiment, step S3 includes the following contents:

[0023] Mode shift degree assessment: mode shift index;

[0024] First, a dynamic time warping algorithm is used to process the real-time operation data sequence. Output sequence with operation habit model Perform timing alignment; in order to quantify the degree of mode deviation, the mode deviation index is used, and its calculation process is as follows:

[0025] Use the dynamic time warping algorithm to align the real-time operation data sequence and the operation habit model output sequence to obtain the best matching path , where each matching point pair satisfy correspond ; Local deviation It is a composite function of the difference in operating data and the difference in time steps between matching point pairs: ;in, Indicates the absolute difference of the operation data, It is used to balance the influence of time step difference on the offset; the local offset of all matching points is accumulated nonlinearly to define the pattern offset index for: .

[0026] In a preferred embodiment, step S3 includes the following contents:

[0027] Assessment of sensitivity to pressure fluctuations: Multiscale pressure sensitivity index;

[0028] The dynamic characteristics of liquid pressure include both low-frequency long-term trends and high-frequency short-term disturbances. In order to measure the sensitivity of the system to pressure fluctuations, a multi-fractal spectrum analysis method is used, using a multi-scale pressure sensitivity index, and the calculation process is as follows:

[0029] The pressure time series Through the multi-fractal decomposition method, it is decomposed into multiple frequency components, each frequency component corresponds to a different time scale, and several subsequences are obtained. ,in Represents different decomposition levels; for each subsequence, calculate its local fractal dimension , combined with the fractal dimensions of each decomposition level, a multi-scale pressure sensitivity index is defined for: ;in, is the number of decomposition levels, the denominator The term is used to reduce the influence of high-level components on the overall sensitivity and emphasize the key role of low-level components.

[0030] In a preferred embodiment, step S3 includes the following contents:

[0031] The mode shift index is integrated with the multi-scale pressure sensitivity index to generate the operation model shift sensitivity value, which is used to evaluate whether the learning effect meets the predetermined standard;

[0032] Firstly, the mode shift index and the multi-scale pressure sensitivity index are nonlinearly mapped, and the two indices are mapped to the standardized range through the hyperbolic tangent function; the operation model shift sensitivity value is obtained by combining the interaction effect of the mode shift index and the multi-scale pressure sensitivity index; the operation model shift sensitivity value is compared with the set deviation threshold to analyze the degree of fit between the current operation and the preset standard, and to evaluate whether the basic conditions for control adjustment are met; if the operation model shift sensitivity value is less than or equal to the deviation threshold, it indicates that the matching degree between the current operation mode and the model and the sensitivity of pressure fluctuations have met the applicable requirements of the control strategy; if it is greater than the deviation threshold, it means that the current state has not met the conditions, and it is necessary to optimize or readjust the association between the operation habit model and the control strategy.

[0033] In a preferred embodiment, step S4 includes the following contents:

[0034] S4.1, in order to effectively identify abnormal fluctuations in the pressure signal, the real-time pressure data is first decomposed by empirical mode decomposition. The decomposition process is as follows:

[0035] Signal decomposition: pressure time series Decompose into multiple eigenmode functions and a residual term ,Right now: ;in, Indicates Layer modal, is the low-frequency residual term representing the long-term trend;

[0036] Decomposition level division: decomposed Corresponding to different frequency components, from high-frequency short-term fluctuations to low-frequency long-term trends;

[0037] Abnormal fluctuation detection: define anomaly detection threshold is the local extreme value range of the eigenmode function, calculate the local amplitude of each mode and compare it with the threshold: , then this point is considered to be an abnormal fluctuation.

[0038] In a preferred embodiment, S4.2, after empirical mode decomposition, dynamically identifying and locating abnormal points detected in each intrinsic mode function level;

[0039] Dynamic threshold adjustment: Dynamically adjust the threshold according to the statistical characteristics of each intrinsic mode function , ensuring that the threshold can adapt to the characteristics of different modes: ;in, and are the mean and standard deviation of the intrinsic mode function, To control the dynamic change rate of the adjustment parameters, is the reference time point;

[0040] Peak and valley detection: Use the adjusted dynamic threshold to confirm the location of abnormal peaks and valleys and their amplitude deviations;

[0041] S4.3, for the detected abnormal peaks and valleys, a second-order delay control strategy combined with an adaptive PID controller is used for dynamic adjustment to achieve smooth pressure adjustment;

[0042] Second-order delay control strategy: Use the second-order delay model to buffer and stabilize pressure changes;

[0043] Combining the second-order delay model and the PID controller, the adjusted liquid flow rate is calculated;

[0044] S4.4, reconstructs the adjusted pressure fluctuation signal with the long-term trend to generate the final smooth pressure signal; verifies whether the adjustment effect meets the expectation by calculating the error between the final smooth pressure signal and the target pressure, ensuring that the pressure fluctuation stability meets the surgical requirements.

[0045] A hysteroscope real-time liquid pressure stabilization control system, comprising: a data acquisition module, an operation modeling module, an evaluation and analysis module and a pressure control module;

[0046] Data acquisition module: collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and pre-processes the collected data to ensure the accuracy and consistency of the data; outputs standardized and pre-processed multimodal data as input for nonlinear modeling in the operation modeling module;

[0047] Operation modeling module: Use long short-term memory network to perform nonlinear time series modeling on preprocessed multimodal data and build a mathematical model that reflects operation habits. The output operation habit model and real-time operation data are transmitted to the evaluation and analysis module to evaluate the matching degree of operation mode and pressure fluctuation sensitivity.

[0048] Evaluation and analysis module: Through the dynamic time warping algorithm and multi-fractal spectrum analysis method, the mode deviation index and multi-scale pressure sensitivity index are calculated to generate the operation model deviation sensitivity value to analyze whether the current operation meets the control adjustment conditions; the output operation model deviation sensitivity value and real-time pressure data are transmitted to the pressure control module to identify abnormal fluctuations and perform peak shaving and valley compensation;

[0049] Pressure control module: Based on the empirical mode decomposition technology, abnormal peaks and valleys in real-time pressure data are identified, and the second-order delay control strategy and proportional integral differential control algorithm are used to cut peaks and fill valleys in pressure fluctuations to achieve smooth and stable pressure output; the output smooth pressure signal is used as the final control result to provide stable liquid pressure support for the surgical process.

[0050] The technical effects and advantages of the hysteroscope real-time liquid pressure stabilization control method and system of the present invention are as follows:

[0051] The present invention collects and preprocesses multimodal data such as liquid pressure, flow rate and instrument movement during surgery through a multi-sensor network, ensuring the high accuracy and consistency of the data, and providing a high-quality foundation for subsequent modeling and control; combines long-term and short-term memory networks to model nonlinear time series of multimodal data, accurately constructs a mathematical model that reflects operating habits, and comprehensively captures complex time correlations and operating modes; through dynamic time warping algorithms and multi-fractal spectrum analysis, quantifies the matching degree between operating data and models and the multi-scale fluctuation characteristics of liquid pressure, generates operating model offset sensitive values, and effectively analyzes whether the current operation has the basis for further adjusting pressure control; based on the empirical mode decomposition technology, accurately identifies the location of abnormal pressure fluctuations, and combines the second-order delay control strategy and proportional integral differential control algorithm to perform real-time peak clipping and valley filling processing on pressure fluctuations, and finally realizes smooth and stable control of liquid pressure. The present invention solves the problems of unclear surgical field of view, increased operating difficulty and patient safety hazards caused by liquid pressure fluctuations during hysteroscopic surgery by constructing a full-process method from data acquisition to model modeling to dynamic evaluation and real-time regulation, and greatly improves the safety, stability and efficiency of surgery. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The present invention is a flowchart of a method for stabilizing the real-time liquid pressure of a hysteroscope.

[0053] Figure 2 The present invention is a structural schematic diagram of a hysteroscope real-time liquid pressure stabilization control system. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments 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.

[0055] Embodiment 1: Figure 1 The present invention provides a hysteroscope real-time liquid pressure stabilization control method, comprising:

[0056] S1, collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and pre-processes the collected data to ensure the accuracy and consistency of the data.

[0057] S2, a long short-term memory network is used to perform nonlinear time series modeling on the preprocessed multimodal data, and a mathematical model reflecting operating habits is constructed to capture complex time correlations and operating patterns.

[0058] S3, evaluates the matching degree between real-time operation data and the operation habit model through the dynamic time warping algorithm, quantifies the degree of mode deviation, and uses the multi-fractal spectrum analysis method to extract the multi-scale characteristics of pressure fluctuations, and generates the operation model deviation sensitivity value to analyze whether the current operation meets the conditions for control adjustment.

[0059] S4, if the conditions for control adjustment are met, the empirical mode decomposition technology is applied to identify abnormal peaks and valleys in the real-time pressure data, and the second-order delay control strategy and proportional integral differential control algorithm are combined to perform peak clipping and valley compensation on these abnormal fluctuations to achieve smooth and stable pressure.

[0060] During hysteroscopic surgery, real-time monitoring and control of intrauterine fluid pressure is crucial to ensure the safety and effectiveness of the surgery. The stability of fluid pressure depends not only on accurate pressure measurement, but also on the synchronous monitoring and analysis of fluid flow rate and motion parameters of surgical instruments. Therefore, building an efficient and accurate data acquisition and preprocessing system is the basis for achieving real-time stable control of fluid pressure.

[0061] Step S1 includes the following contents:

[0062] S1.1 deploys a multi-sensor network to collect multi-modal data related to pressure, flow, and instrument operation in real time during surgery. Liquid pressure data is obtained through high-precision pressure sensors to record real-time pressure changes in the uterine cavity; the liquid injection rate is monitored by flow sensors to capture dynamic changes in liquid; instrument operation information is recorded through position sensors and acceleration sensors to record operation trajectories and frequencies. To ensure data synchronization, all sensors are marked with a unified timestamp to establish the temporal correlation between multi-modal data and form a high-resolution time series data set.

[0063] S1.2, perform noise reduction on the collected data to eliminate high-frequency noise and abnormal fluctuations in the signal. For pressure data, a layer-by-layer decomposition method is used to extract the main components of the signal and eliminate random noise interference; for flow and instrument operation data, a multi-scale filtering method based on segmented analysis is used to isolate the main low-frequency components. Through this layer-by-layer optimization, while ensuring that the data retains key features, interference components that are not related to the actual surgery are eliminated, providing stable input for subsequent calculations.

[0064] S1.3, for data missing that may occur during the acquisition process, data completion is performed based on time correlation and the synergistic relationship between features. First, the time correlation between pressure, flow, and instrument movement is analyzed to infer the reasonable value of the missing interval; second, the high correlation between data is used to supplement the missing parts of flow and movement data through a dynamic extrapolation mechanism, thereby generating continuous and complete time series data in all dimensions. This multi-angle inference ensures that the completed data is consistent with both physical laws and overall trends.

[0065] S1.4, in order to ensure the feature balance and scale consistency of multimodal data, the interval mapping method is used to standardize the data of different physical quantities. The pressure, flow and operation data are converted to a unified scale range while retaining their respective change trends and characteristic distributions. This process not only eliminates the interference between different data dimensions, but also enhances the alignment effect between features, providing a high-quality standardized data set for subsequent time series modeling.

[0066] In hysteroscopic surgery, the doctor's operating habits directly affect the dynamic changes of fluid pressure. In order to achieve stable control of fluid pressure, a mathematical model that can reflect operating habits must be established based on the collected multimodal data. Since operating habits have time dependence and complex nonlinear characteristics, conventional linear models are difficult to accurately describe. Therefore, it is necessary to adopt nonlinear time series modeling technology to capture the complex laws during surgical operations and provide a reliable basis for subsequent control strategies.

[0067] Step S2 includes the following contents:

[0068] S2.1, standardized sequences of pressure, flow and operation information are selected from the preprocessed multimodal data set as modeling input. These data sequences not only contain the immediate characteristics of the operation, but also reflect the long-term trends and dynamic changes during the operation. At the same time, in order to avoid the model being too dependent on a certain feature, feature balancing technology is used to adjust the weights of all input dimensions to a similar level to reduce the possibility of bias.

[0069] S2.2, Long Short-Term Memory Network is used as a modeling framework to handle the temporal correlation and complex nonlinear features in multimodal data. Long Short-Term Memory Network can retain long-term memory while learning short-term dependencies through a built-in gating mechanism, thus being able to fully characterize the doctor's operation mode.

[0070] Time step segmentation and dynamic window design: Divide the time series into several segments of equal length, use a sliding window strategy to construct input data, and dynamically adjust the window length to adapt to the changing characteristics of different operating habits.

[0071] Hidden layer design and memory update: The model gradually extracts high-order features in the sequence by stacking multiple hidden layers. At each time step, the long short-term memory network unit updates its memory state through the input gate, forget gate, and output gate mechanism, thereby effectively capturing the multi-level dynamic features of the time series.

[0072] S2.3, using the preprocessed multimodal dataset as the training set, optimize the weight and structure of the LSTM model through supervised learning, so that it can accurately predict the operation mode.

[0073] Loss function construction: In order to enhance the model's comprehensive understanding of different features, a multi-objective loss function is defined, which includes short-term error (prediction accuracy of the current time step) and long-term trend deviation (overall matching degree of multiple time steps), to ensure that the model pays attention to both local characteristics and global laws.

[0074] Training process and gradient update: The model is trained using an iterative optimization algorithm, and the weights are adjusted through back-propagation and gradient clipping mechanisms to avoid the negative impact of gradient explosion or vanishing on model learning.

[0075] S2.4, after training, the LSTM model can generate high-order feature representations that match the operation habits for the input multimodal time series. These high-order features not only reflect the temporal rules of the operation, but also reveal the dynamic evolution characteristics of the operation mode.

[0076] Construction of a mathematical model of operating habits: The high-order features output by the LSTM model are converted into a mathematical description form to form a complete operating habit model to provide data support for subsequent pressure control strategies.

[0077] Model validation and adjustment: The generalization performance of the model is evaluated through the validation data set, and the parameter configuration of the model is optimized using dynamic error analysis to ensure that it can adapt to different doctors and surgical scenarios.

[0078] Through step S1, the real-time collection, denoising, completion and standardization of multimodal data such as pressure, flow and instrument operation in hysteroscopic surgery are carried out, which eliminates data noise and missing problems, generates a highly consistent and complete time series data set, and provides a high-quality input basis for subsequent modeling; combined with the long short-term memory network used in step S2, by performing nonlinear time series modeling on the preprocessed multimodal data, a mathematical model that can fully reflect the doctor's operating habits is successfully constructed, capturing the complex time correlation and dynamic change laws during the operation process, and laying a foundation for accurate data support and dynamic adjustment for the intelligent control of liquid pressure, thereby improving the accuracy and responsiveness of liquid pressure control during surgery and enhancing the adaptability and robustness of the system to complex scenarios.

[0079] During hysteroscopic surgery, real-time monitoring of the matching degree between operation data and the pre-built operation habit model, as well as the system's sensitivity to liquid pressure fluctuations, are key links in ensuring stable control of liquid pressure. The degree of mode deviation reflects the degree of consistency between the current operation and the doctor's established operation habits, while the pressure fluctuation sensitivity measures the system's ability to respond to changes in liquid pressure. In order to comprehensively evaluate the learning effect, this step designs a multi-dimensional complex evaluation mechanism, which generates a comprehensive decision value through a dynamic time warping algorithm and a multi-fractal spectrum analysis method to determine whether the learning effect meets the predetermined standard, thereby providing a scientific basis for subsequent pressure adjustment.

[0080] Step S3 includes the following contents:

[0081] Assessment of the degree of mode shift: mode shift index.

[0082] First, a dynamic time warping algorithm is used to process the real-time operation data sequence. Output sequence of the operation habit model To more accurately quantify the degree of mode shift, the mode shift index is used, and its calculation process is as follows:

[0083] Use the dynamic time warping algorithm to align the real-time operation data sequence and the operation habit model output sequence to obtain the best matching path , where each matching point pair satisfy correspond .

[0084] Define local offset It is a composite function of the difference in operating data and the difference in time steps between matching point pairs: ;in, Indicates the absolute difference of the operation data, Used to balance the effect of time step differences on the bias.

[0085] The local deviation of all matching points is accumulated nonlinearly to define the pattern deviation index. for: The calculation of the mode shift index emphasizes larger shifts by means of the sum of squares, while introducing a logarithmic term to weaken the cumulative effect of the overall shift, thus achieving a comprehensive quantification of the mode shift.

[0086] The mode shift index is used to quantify the degree of deviation between real-time operation data and the pre-established operation habit model. During hysteroscopic surgery, the doctor's operation behavior has a significant impact on the dynamic changes of liquid pressure. The mode shift index uses a dynamic time warping algorithm to align the real-time operation data with the operation habit model, and accurately measures the difference between the two in time step and operation amplitude. This index reflects the degree of match between the current surgical operation and the established operation mode, and can effectively identify abnormal deviations that may exist during the operation, thereby providing a scientific basis for the system to judge whether the learning effect meets the standard. Through the calculation of the mode shift index, the system can dynamically monitor and evaluate the implementation of operating habits, ensure the accuracy and consistency of liquid pressure control, and improve the overall safety and efficiency of the operation.

[0087] Assessment of sensitivity to pressure fluctuations: A multiscale pressure sensitivity index.

[0088] The dynamic characteristics of liquid pressure include both low-frequency long-term trends and high-frequency short-term disturbances. In order to comprehensively measure the sensitivity of the system to pressure fluctuations, a multi-fractal spectrum analysis method is used, using a multi-scale pressure sensitivity index, the calculation process is as follows:

[0089] The pressure time series Through the multi-fractal decomposition method, it is decomposed into multiple frequency components, each frequency component corresponds to a different time scale, and several subsequences are obtained. ,in Indicates different levels of decomposition.

[0090] For each subsequence, calculate its local fractal dimension , which reflects the complexity and irregularity of pressure fluctuations at this scale.

[0091] Combining the fractal dimensions of each decomposition level, a multi-scale pressure sensitivity index is defined for: ;in, is the number of decomposition levels, the denominator The term is used to reduce the influence of high-level (high frequency) components on the overall sensitivity, emphasize the key role of low-level (low-frequency) components, and adjust the contribution of different fractal dimensions through a logarithmic function.

[0092] The multiscale pressure sensitivity index is used to evaluate the system's responsiveness and sensitivity to fluid pressure fluctuations. Fluid pressure exhibits multiscale dynamic characteristics during hysteroscopic surgery, including low-frequency long-term trends and high-frequency short-term disturbances. The multiscale pressure sensitivity index uses a multi-fractal spectrum analysis method to perform multi-scale decomposition of pressure data, extract the local fractal dimension of each frequency component, and then quantify the complexity and irregularity of pressure fluctuations at different time scales. The index comprehensively considers the amplitude and frequency characteristics of pressure fluctuations, and provides a sensitivity assessment of the system in response to different types of pressure changes. The calculation results of the multiscale pressure sensitivity index can reflect the system's adaptability and response efficiency under multi-level pressure fluctuations, guide the system to optimize the control strategy, ensure the stability of fluid pressure, reduce the risks and discomfort caused by pressure fluctuations during surgery, and improve the overall quality of surgery and patient safety.

[0093] The mode deviation index is integrated with the multi-scale pressure sensitivity index to generate the operation model deviation sensitivity value, which is used to evaluate whether the learning effect reaches the predetermined standard. In order to avoid the limitations of simple weighted summation, an adaptive comprehensive evaluation function is introduced. The specific calculation process is as follows:

[0094] First, the mode shift index and the multi-scale pressure sensitivity index are nonlinearly mapped, and the two indices are mapped to the standardized range through the hyperbolic tangent function: ;in, and They are the empirical scale parameters of the mode shift index and the multi-scale pressure sensitivity index, respectively, which are used to adjust the index value range after mapping.

[0095] Combining the interactive effect of the mode shift index and the multi-scale pressure sensitivity index, the operational model shift sensitivity value is defined, for example, by the following calculation method: ;in, is a minimum value, which is used to avoid the situation where the denominator is zero. This formula initially integrates two standardized exponents in the form of square sum, and introduces an exponential decay term to adjust the interaction between the two, so that the comprehensive decision value can more accurately reflect the overall learning effect of the system.

[0096] Compare the operation model deviation sensitivity value with the set deviation threshold to analyze the degree of fit between the current operation and the preset standard, and evaluate whether the basic conditions for further control adjustment are met. If the operation model deviation sensitivity value is less than or equal to the deviation threshold, it means that the matching degree between the current operation mode and the model and the sensitivity of pressure fluctuation have met the applicable requirements of the control strategy; if it is greater than the deviation threshold, it means that the current state has not met the conditions, and it is necessary to further optimize or readjust the association between the operation habit model and the control strategy.

[0097] After completing the operation mode matching and pressure fluctuation sensitivity analysis, when the operation model offset sensitivity value indicates that the current operation is suitable for further control and adjustment, the real-time liquid pressure data needs to be processed in a refined manner to eliminate abnormal pressure peaks and valleys to ensure the smoothness and stability of the pressure. Liquid pressure signals are essentially complex nonlinear and non-stationary signals. Direct processing can easily lead to response lag or over-adjustment. Therefore, it is necessary to combine signal decomposition technology and advanced control strategies to develop a set of dynamically adaptable pressure smoothing methods to achieve high-precision pressure control during surgery.

[0098] Step S4 includes the following contents:

[0099] S4.1, in order to effectively identify abnormal fluctuations in pressure signals, the real-time pressure data is first decomposed by empirical mode decomposition. Empirical mode decomposition is a data-driven signal decomposition method suitable for processing nonlinear and non-stationary signals. The decomposition process is as follows:

[0100] Signal decomposition: pressure time series Decompose into multiple eigenmode functions and a residual term ,Right now: ;in, Indicates Layer modal, is the low-frequency residual term representing the long-term trend.

[0101] Decomposition level division: decomposed Corresponding to different frequency components, from high-frequency short-term fluctuations to low-frequency long-term trends, the short-term anomalies and long-term change characteristics of the signal are revealed respectively.

[0102] Abnormal fluctuation detection: define anomaly detection threshold is the local extreme value range of the eigenmode function, calculate the local amplitude of each mode and compare it with the threshold: , then this point is considered to be an abnormal fluctuation.

[0103] The process identifies the locations of unusual peaks and valleys and , providing a basis for subsequent peak shaving and valley filling treatment.

[0104] S4.2, after empirical mode decomposition, the abnormal points detected in each intrinsic mode function level are further dynamically identified and located.

[0105] Dynamic threshold adjustment: Dynamically adjust the threshold based on the statistical properties of each intrinsic mode function (such as local mean and rate of change) , ensuring that the threshold can adapt to the characteristics of different modes: ;in, and are the mean and standard deviation of the intrinsic mode function, To control the dynamic change rate of the adjustment parameters, is the reference time point.

[0106] Peak and valley detection: Use the adjusted dynamic threshold to further confirm the location of abnormal peaks and valleys and their amplitude deviations.

[0107] S4.3, for the detected abnormal peaks and valleys, a second-order delay control strategy combined with an adaptive PID controller is used for dynamic adjustment to achieve smooth adjustment of pressure.

[0108] Second-order delay control strategy: The second-order delay model is used to buffer and stabilize pressure changes. The specific form is: ;in, and are the first-order and second-order delay time constants respectively. This model makes the pressure adjustment smoother by delaying the change rate of the signal in sections.

[0109] Adaptive PID Controller: Define PID controller gain parameters , and Dynamic adjustment rules:

[0110] ;

[0111] ;

[0112] ;

[0113] , , : Initial PID gain.

[0114] , , : Dynamic adjustment coefficient.

[0115] : Target pressure value.

[0116] Combine the second-order delay model and PID controller to calculate the adjusted liquid flow rate : ;in, Indicates the real-time pressure error.

[0117] S4.4, reconstruct the regulated pressure fluctuation signal with the long-term trend to generate the final smooth pressure signal .

[0118] Signal reconstruction: ;in, is the adjusted intrinsic mode function.

[0119] Smooth verification: By calculating the error between the final smooth pressure signal and the target pressure, verify whether the adjustment effect meets the expectations and ensure that the pressure fluctuation stability meets the surgical requirements.

[0120] Through the calculation of the mode shift index and multi-scale pressure sensitivity index in step S3, the system can accurately evaluate the matching degree between the real-time operation data and the operation habit model, and comprehensively quantify the multi-scale fluctuation characteristics of the liquid pressure, thereby generating the operation model shift sensitivity value, which is used to analyze whether the current operation has the conditions for further adjusting the pressure control; on this basis, step S4 uses the empirical mode decomposition technology to perform multi-level decomposition of the real-time pressure signal, accurately identify the location of abnormal peaks and valleys, and combine the second-order delay control strategy with the adaptive PID algorithm to dynamically adjust and reconstruct the output of the pressure fluctuation, thus achieving smooth and stable control of the pressure. Through the above process, the system shows higher accuracy and sensitivity in operation mode adaptation and pressure regulation, effectively reducing the interference of surgical field of view and patient physiological risks caused by pressure fluctuations, while improving the safety and efficiency of surgical operations, and providing reliable technical support for intelligent pressure control.

[0121] Embodiment 2: Figure 2 The present invention provides a hysteroscope real-time liquid pressure stabilization control system, comprising: a data acquisition module, an operation modeling module, an evaluation and analysis module and a pressure regulation module.

[0122] Data acquisition module: collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and preprocesses the collected data to ensure the accuracy and consistency of the data; outputs standardized and preprocessed multimodal data as input for nonlinear modeling in the operation modeling module.

[0123] Operation modeling module: Use long short-term memory network to perform nonlinear time series modeling on the preprocessed multimodal data to build a mathematical model that reflects operating habits; the output operating habit model and real-time operating data are transmitted to the evaluation and analysis module together to evaluate the matching degree of operating modes and sensitivity to pressure fluctuations.

[0124] Evaluation and analysis module: Through the dynamic time warping algorithm and multi-fractal spectrum analysis method, the mode deviation index and multi-scale pressure sensitivity index are calculated, and the operation model deviation sensitivity value is generated to analyze whether the current operation meets the conditions for control adjustment; the output operation model deviation sensitivity value and real-time pressure data are transmitted to the pressure control module together for identifying abnormal fluctuations and performing peak shaving and valley filling.

[0125] Pressure control module: Based on the empirical mode decomposition technology, abnormal peaks and valleys in real-time pressure data are identified, and the second-order delay control strategy and proportional integral differential control algorithm are used to cut peaks and fill valleys in pressure fluctuations to achieve smooth and stable pressure output; the output smooth pressure signal is used as the final control result to provide stable liquid pressure support for the surgical process.

[0126] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0127] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0128] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0129] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for hysteroscopic real-time liquid pressure stabilization control, characterized in that: Includes steps: S1, collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and pre-processes the collected data to ensure the accuracy and consistency of the data; S2, using long short-term memory network to perform nonlinear time series modeling on the preprocessed multimodal data and construct a mathematical model reflecting operating habits to capture complex time correlations and operating patterns; S3, evaluates the matching degree between real-time operation data and operation habit model through dynamic time warping algorithm, quantifies the degree of mode deviation, and uses multi-fractal spectrum analysis method to extract multi-scale characteristics of pressure fluctuation, and generates operation model deviation sensitivity value to analyze whether the current operation meets the conditions for control adjustment; S4, if the control adjustment conditions are met, the empirical mode decomposition technology is applied to identify abnormal peaks and valleys in the real-time pressure data, and the second-order delay control strategy and proportional integral differential control algorithm are combined to cut the peaks and fill the valleys of these abnormal fluctuations to achieve smooth and stable pressure; Step S3 includes the following contents: The degree of mode shift is evaluated based on the mode shift index; First, a dynamic time warping algorithm is used to process the real-time operation data sequence. Output sequence of the operation habit model Perform timing alignment; in order to quantify the degree of mode deviation, the mode deviation index is used, and its calculation process is as follows: Use the dynamic time warping algorithm to align the real-time operation data sequence and the operation habit model output sequence to obtain the best matching path , where each matching point pair satisfy correspond ; Local deviation It is a composite function of the difference in operating data and the difference in time steps between matching point pairs: ;in, Indicates the absolute difference of the operation data, It is used to balance the influence of time step difference on the offset; the local offset of all matching points is accumulated nonlinearly to define the pattern offset index for: ; It represents the number of data points after alignment, that is, the number of data points involved in calculating the mode shift index after the dynamic time warping algorithm completes the time series alignment; Evaluation of sensitivity to pressure fluctuations based on a multi-scale pressure sensitivity index; The dynamic characteristics of liquid pressure include both low-frequency long-term trends and high-frequency short-term disturbances. In order to measure the sensitivity of the system to pressure fluctuations, a multi-fractal spectrum analysis method is used, using a multi-scale pressure sensitivity index, and the calculation process is as follows: The pressure time series Through the multi-fractal decomposition method, it is decomposed into multiple frequency components, each frequency component corresponds to a different time scale, and several subsequences are obtained. ,in Represents different decomposition levels; for each subsequence, calculate its local fractal dimension , combined with the fractal dimensions of each decomposition level, a multi-scale pressure sensitivity index is defined for: ;in, is the number of decomposition levels, the denominator The term is used to reduce the impact of high-level components on the overall sensitivity and emphasize the key role of low-level components; The mode shift index is fused with the multi-scale pressure sensitivity index to generate the operational model shift sensitivity value; Firstly, the mode shift index and the multi-scale pressure sensitivity index are nonlinearly mapped, and the two indices are mapped to the standardized range through the hyperbolic tangent function; the operation model shift sensitivity value is obtained by combining the interaction effect of the mode shift index and the multi-scale pressure sensitivity index; the operation model shift sensitivity value is compared with the set deviation threshold to analyze the degree of fit between the current operation and the preset standard, and to evaluate whether the basic conditions for control adjustment are met; if the operation model shift sensitivity value is less than or equal to the deviation threshold, it indicates that the matching degree between the current operation mode and the model and the sensitivity of pressure fluctuations have met the applicable requirements of the control strategy; if it is greater than the deviation threshold, it means that the current state has not met the conditions, and it is necessary to optimize or readjust the association between the operation habit model and the control strategy.

2. A hysteroscopic real-time liquid pressure stabilization control method according to claim 1, characterized in that: Step S1 includes the following contents: S1.1 deploys a multi-sensor network to collect multi-modal data related to pressure, flow and instrument operation in real time during surgery; S1.2, performing noise reduction processing on the collected data to eliminate high-frequency noise and abnormal fluctuations in the signal; S1.3, for data missing that may occur during the collection process, data is supplemented based on time correlation and the collaborative relationship between features; S1.4, in order to ensure the feature balance and scale consistency of multimodal data, the interval mapping method is used to standardize the data of different physical quantities.

3. A hysteroscopic real-time liquid pressure stabilization control method according to claim 2, characterized in that: Step S2 includes the following contents: S2.1, select the standardized sequence of pressure, flow and operation information from the preprocessed multimodal data set as modeling input; S2.2, adopting long short-term memory network as the modeling framework to handle the temporal correlation and complex nonlinear characteristics in multimodal data; S2.3, using the preprocessed multimodal dataset as a training set, the weights and structure of the LSTM model are optimized through supervised learning so that it can predict the operation mode; S2.4, after training, the LSTM model can generate high-order feature representations that match the operating habits for the input multimodal time series.

4. A hysteroscopic real-time liquid pressure stabilization control method according to claim 1, characterized in that: Step S4 includes the following contents: S4.1, in order to effectively identify abnormal fluctuations in the pressure signal, the real-time pressure data is first decomposed by empirical mode decomposition. The decomposition process is as follows: Signal decomposition: pressure time series Decompose into multiple eigenmode functions and a residual term ,Right now: ;in, Indicates Layer modal, is the low-frequency residual term representing the long-term trend; Represents the intrinsic mode function obtained by decomposition during the empirical mode decomposition process the number of Decomposition level division: decomposed Corresponding to different frequency components, from high-frequency short-term fluctuations to low-frequency long-term trends; Abnormal fluctuation detection: define anomaly detection threshold is the local extreme value range of the eigenmode function, calculate the local amplitude of each mode and compare it with the threshold: , then this point is considered to be an abnormal fluctuation.

5. A hysteroscopic real-time liquid pressure stabilization control method according to claim 4, characterized in that: S4.2, after empirical mode decomposition, the abnormal points detected in each intrinsic mode function level are dynamically identified and located; Dynamic threshold adjustment: Dynamically adjust the threshold according to the statistical characteristics of each intrinsic mode function , ensuring that the threshold can adapt to the characteristics of different modes: ;in, and are the mean and standard deviation of the intrinsic mode function, To control the dynamic change rate of the adjustment parameters, is the reference time point, Indicates the current time point; Peak and valley detection: Use the adjusted dynamic threshold to confirm the location of abnormal peaks and valleys and their amplitude deviations; S4.3, for the detected abnormal peaks and valleys, a second-order delay control strategy combined with an adaptive PID controller is used for dynamic adjustment to achieve smooth pressure adjustment; Second-order delay control strategy: Use the second-order delay model to buffer and stabilize pressure changes; Combining the second-order delay model and the PID controller, the adjusted liquid flow rate is calculated; S4.4, reconstructs the adjusted pressure fluctuation signal with the long-term trend to generate the final smooth pressure signal; verifies whether the adjustment effect meets the expectation by calculating the error between the final smooth pressure signal and the target pressure, ensuring that the pressure fluctuation stability meets the surgical requirements.

6. A hysteroscope real-time liquid pressure stabilization control system, used to implement a hysteroscope real-time liquid pressure stabilization control method according to any one of claims 1 to 5, characterized in that: include: Data acquisition module, operation modeling module, evaluation and analysis module and pressure control module; Data acquisition module: collects multimodal data related to fluid pressure in real time during surgery through a multi-sensor network, and pre-processes the collected data to ensure the accuracy and consistency of the data; outputs standardized and pre-processed multimodal data as input for nonlinear modeling in the operation modeling module; Operation modeling module: Use long short-term memory network to perform nonlinear time series modeling on preprocessed multimodal data and build a mathematical model that reflects operation habits. The output operation habit model and real-time operation data are transmitted to the evaluation and analysis module to evaluate the matching degree of operation mode and pressure fluctuation sensitivity. Evaluation and analysis module: Through the dynamic time warping algorithm and multi-fractal spectrum analysis method, the mode deviation index and multi-scale pressure sensitivity index are calculated, and the operation model deviation sensitivity value is generated to analyze whether the current operation meets the conditions for control adjustment; The output operation model offset sensitivity value and real-time pressure data are transmitted to the pressure control module to identify abnormal fluctuations and perform peak shaving and valley compensation. Pressure control module: Based on the empirical mode decomposition technology, abnormal peaks and valleys in real-time pressure data are identified, and the second-order delay control strategy and proportional integral differential control algorithm are used to cut peaks and fill valleys in pressure fluctuations to achieve smooth and stable pressure output; the output smooth pressure signal is used as the final control result to provide stable liquid pressure support for the surgical process.

Images