Endoscope pneumoperitoneum pressure system based on AI image evaluation and intelligent adjusting method

Abstract

The invention discloses an endoscope pneumoperitoneum pressure system based on AI image evaluation and an intelligent adjusting method, and the method comprises the steps: 1, obtaining an original image of the abdominal cavity of a patient through CT equipment, carrying out the standardization processing of the original image, generating a standardized CT image data set, and constructing a three-dimensional dynamic model of the abdominal cavity of the patient based on an AI algorithm; in the second stage, quantitative analysis of subcutaneous fat thickness, visceral organ volume and operable space volume is carried out on the three-dimensional dynamic model by adopting an AI segmentation algorithm, and a personalized pneumoperitoneum pressure target value Ptarget and an adjustment threshold range are generated through a machine learning model in combination with basic illness state data of the patient; in the third stage, working data are collected in real time through a built-in sensor of the endoscope system, and the working data are input into the reinforcement learning model to dynamically calculate a pressure adjusting instruction delta P; through deep fusion of the AI technology and system closed-loop control, the intelligent level of the endoscope system is remarkably optimized, good hardware support is provided, and the function of optimizing the operation environment and effect is achieved.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a laparoscopic pneumoperitoneum pressure system and an intelligent adjustment method based on AI image evaluation. Background Art

[0002] Laparoscopic surgery, as a core technique in minimally invasive surgery, has been widely used in various surgical fields. During surgery, pneumoperitoneum pressure is a key parameter for maintaining a stable operating space in the abdominal cavity. Its setting directly affects surgical field exposure, instrument manipulation freedom, and patient safety. However, individual patient differences (such as subcutaneous fat thickness, organ volume, and cardiopulmonary function) make traditional pneumoperitoneum pressure settings difficult to achieve universal application. Excessively high pressure may cause cardiovascular or respiratory complications, while too low pressure can easily obscure the surgical field and make the procedure difficult. Therefore, an intelligent pressure regulation system that can precisely adapt to the patient's anatomy and physiological state is urgently needed to optimize surgical safety and efficiency.

[0003] In current laparoscopic surgeries, pneumoperitoneum pressure regulation mainly relies on the surgeon's experience or manual adjustment based on fixed thresholds, lacking quantitative analysis of individual patient characteristics. Existing support equipment (such as traditional pneumoperitoneum machines) only perform simple closed-loop control through pressure sensor feedback and are unable to dynamically predict the minimum effective pressure range based on the patient's preoperative imaging data, resulting in blind pressure settings during surgery. For example, for obese patients, high abdominal pressure may increase the risk of visceral damage due to fat compression, but existing technologies cannot identify potential risk areas in advance and adjust pressure thresholds through preoperative modeling. This defect directly leads to an increased incidence of intraoperative complications and insufficient stability of the surgical field.

[0004] In view of this, it is necessary to improve the existing pneumoperitoneum pressure regulation technology to solve the technical problem of lack of individualized regulation design. Summary of the Invention

[0005] The purpose of the present invention is to provide a laparoscopic pneumoperitoneum pressure system and intelligent adjustment method based on AI image evaluation to solve the above technical problems.

[0006] To achieve this object, the present invention adopts the following technical solutions: An intelligent adjustment method for a laparoscopic pneumoperitoneum pressure system based on AI image evaluation, comprising: In the first phase, the original abdominal images of the patient are acquired through CT equipment, standardized and processed to generate a standardized CT image dataset, and a three-dimensional dynamic model of the patient's abdominal cavity is constructed based on an AI algorithm. In the second stage, an AI segmentation algorithm was used to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional dynamic model. Combined with the patient's basic medical data, a machine learning model was used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range. In the third stage, working data is collected in real time through the built-in sensors of the laparoscope system, and the working data is input into the reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP; the working data includes the abdominal pressure P_current, the surgical field clarity score and the patient's vital signs data.

[0007] Optionally, the first stage involves obtaining original abdominal images of the patient using a CT device, standardizing the original images to generate a standardized CT image dataset, and constructing a three-dimensional dynamic model of the patient's abdominal cavity based on an AI algorithm, specifically including: Obtaining original multi-angle tomographic images of the patient's abdominal cavity through a CT device, and performing denoising on the original images using multimodal image fusion technology to generate an original image data set with a high signal-to-noise ratio; The denoised original image dataset is spatially registered, and a feature point-based registration algorithm is used to align the tomographic images at different angles to a unified coordinate system to generate a spatially consistent registered image dataset. The registered image dataset is standardized, and the image contrast and brightness are adjusted using histogram equalization and grayscale normalization algorithms to generate a standardized CT image dataset. Optionally, the registered image dataset is normalized, and image contrast and brightness are adjusted using histogram equalization and grayscale normalization algorithms to generate a standardized CT image dataset, and then the following steps are further included: Based on a standardized CT image dataset, an AI algorithm was used to construct a 3D dynamic model of the patient's abdominal cavity. The anatomical structural features of the subcutaneous fat, organs, and blood vessels in the abdominal cavity were extracted through a convolutional neural network to generate a 3D visualization model. A grid subdivision algorithm is used to smooth the three-dimensional dynamic model, and the continuity of the internal structure of the three-dimensional dynamic model is optimized based on the Laplace operator to generate an optimized three-dimensional dynamic model that can be used for quantitative analysis.

[0008] Optionally, a mesh subdivision algorithm is used at any time to smooth the three-dimensional dynamic model, and the continuity of the internal structure of the three-dimensional dynamic model is optimized based on the Laplace operator to generate an optimized three-dimensional dynamic model that can be used for quantitative analysis, specifically including: Perform initial mesh division on the three-dimensional dynamic model and use triangulation algorithm to generate the initial mesh structure; Based on the initial grid structure, a subdivision algorithm is used to smooth the surface of the three-dimensional dynamic model, and the grid density is increased through iterative subdivision to generate a high-resolution three-dimensional grid model; Performing Laplace smoothing optimization on the high-resolution three-dimensional mesh model, calculating the local curvature of mesh vertices based on the Laplace operator, adjusting the vertex positions to eliminate irregular protrusions and depressions on the model surface, and generating a smooth and continuous three-dimensional mesh model; The internal structure of the smoothed 3D mesh model is optimized, and the position of the internal mesh nodes is adjusted using a volume preservation algorithm based on the Laplace operator. The quality of the optimized 3D mesh model is evaluated and the model is quantitatively analyzed using mesh quality indicators. If the preset threshold is not reached, the mesh is re-subdivided and smoothed until an optimized 3D dynamic model that can be used for quantitative analysis is generated.

[0009] Optionally, in the second stage, an AI segmentation algorithm is used to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional dynamic model. Combined with the patient's basic condition data, a machine learning model is used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range, specifically including: Perform tissue segmentation on the 3D dynamic model, using a deep learning-based semantic segmentation algorithm to segment the subcutaneous fat layer, organ area, and operable space in the model, and generate a segmented 3D tissue distribution map; Quantitative analysis is performed on the segmented three-dimensional tissue distribution map. The voxel counting algorithm is used to calculate the thickness of subcutaneous fat, organ volume, and operable space volume, and a tissue composition data table containing quantitative parameters is generated. Combined with the patient's basic condition data, feature engineering methods are used to extract key physiological indicators, and the key physiological indicators are fused with the tissue composition data table to generate a multidimensional feature data set; Based on a multidimensional feature dataset, a machine learning model was used to train a pneumoperitoneum pressure prediction model. The model parameters were optimized through cross-validation to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range. The generated pneumoperitoneum pressure target value is evaluated for safety, and a rule-based logic engine is used in combination with the patient's basic condition data to dynamically adjust P_target and the adjustment threshold range to keep the pressure setting within a safe range.

[0010] Optionally, the sensor system includes a pressure sensor for monitoring the abdominal pressure P_current, an image recognition module for scoring the clarity of the laparoscopic surgical field, and a monitor for monitoring the patient's vital signs data; Optionally, the third stage collects working data in real time through a built-in sensor system of the laparoscope system, and inputs the working data into a reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP, specifically including: The working data is collected in real time through the sensor system built into the laparoscope system; The real-time monitoring working data set was preprocessed, and the abdominal pressure data P_current was denoised using a sliding window filtering algorithm. The surgical field clarity score was synchronized with the vital sign data using a time series alignment method to generate a standardized real-time data set. The standardized real-time data set is compared with the personalized pneumoperitoneum pressure target value P_target, and the current pressure deviation value ΔP_current is generated using the differential calculation method. The multi-dimensional state feature vector is generated by combining the surgical field clarity score and vital sign data. The multi-dimensional state feature vector is input into the reinforcement learning model, and the pressure adjustment instruction ΔP is generated by the reinforcement learning model; The generated ΔP is verified in real time. A rule-based logic engine is used in combination with vital signs data to determine whether ΔP is within the safe adjustment range. If it exceeds the threshold, the model is triggered to recalculate and generate the corrected pressure adjustment instruction ΔP_1.

[0011] Optionally, the third stage further includes: In the fourth stage, the CO2 injection rate is adjusted through the intelligent pneumoperitoneum machine based on ΔP. If the surgical field clarity score or vital signs data deviate from the preset threshold, the machine learning model is triggered to regenerate P_target, forming a dynamic closed-loop optimization of intraoperative pressure.

[0012] The present invention also provides a laparoscopic pneumoperitoneum pressure system based on AI image evaluation, which adopts the above-mentioned laparoscopic pneumoperitoneum pressure system intelligent adjustment method based on AI image evaluation. The laparoscopic pneumoperitoneum pressure system includes: AI image processing unit, used to standardize raw images, generate standardized CT image datasets, and construct a three-dimensional dynamic model of the patient's abdominal cavity; The analysis unit uses AI segmentation algorithms to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the 3D model and generate a tissue composition data table; A machine learning model unit is used to combine the patient's basic condition data with the tissue composition data table to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range; The sensor system is integrated into the laparoscopic system and is used to collect real-time abdominal pressure P_current, surgical field clarity score and patient vital signs data; Reinforcement learning model unit, used to dynamically calculate the pressure regulation instruction ΔP and optimize the pressure regulation strategy in real time; Intelligent pneumoperitoneum machine, used to automatically adjust the CO2 injection rate according to the pressure regulation instruction ΔP to achieve closed-loop control of pneumoperitoneum pressure; Data storage, used to store preoperative imaging data, intraoperative real-time data and model parameters; Communication module: used to realize data interaction and communication between units.

[0013] Compared with the existing technology, the present invention has the following beneficial effects: first, the original image of the patient's abdominal cavity is obtained by CT equipment, and a standardized CT image data set is generated after standardization processing, and an AI algorithm is used to construct a three-dimensional dynamic model of the patient's abdominal cavity; then, the AI ​​segmentation algorithm is used to quantitatively analyze the subcutaneous fat thickness, organ volume and operable space volume of the model, and combined with the patient's basic condition data, a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range are generated through a machine learning model; finally, the abdominal cavity pressure P_current, surgical field clarity score and patient vital signs data are collected in real time through the built-in sensor of the laparoscopic system, and the reinforcement learning model is used to dynamically calculate the pressure adjustment instruction ΔP, driving the intelligent pneumoperitoneum machine to adjust the CO2 injection rate in real time, forming a closed-loop feedback and dynamic optimization of the intraoperative pressure; this method significantly optimizes the intelligence level of the laparoscopic system through the deep integration of AI technology and system closed-loop control, provides good hardware support, and plays a role in optimizing the surgical environment and effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0016] Figure 1 This is a flow chart of a method for intelligently adjusting the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to the first embodiment of the present invention; Figure 2 This is the second flow chart of the intelligent adjustment method of the laparoscopic pneumoperitoneum pressure system based on AI image evaluation in the first embodiment of this invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below 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 making creative work are within the scope of protection of the present invention.

[0018] In the description of the present invention, it should be understood that the terms "upper," "lower," "top," "bottom," "inner," "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of the present invention and simplify the description. They are not intended to indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. It should be noted that when a component is considered to be "connected" to another component, it may be directly connected to the other component or there may be a centrally located component.

[0019] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0020] Example 1: Combine Figure 1 and Figure 2 As shown, an embodiment of the present invention provides an intelligent adjustment method for a laparoscopic pneumoperitoneum pressure system based on AI image evaluation, comprising: In the first phase, the original abdominal images of the patient are acquired through CT equipment, standardized and processed to generate a standardized CT image dataset, and a three-dimensional dynamic model of the patient's abdominal cavity is constructed based on an AI algorithm. The first stage involves preoperative data acquisition and modeling. CT scans are used to obtain raw abdominal imaging data, including multi-angle tomographic images. To ensure data accuracy and consistency, the raw images undergo standardization, including denoising, registration, and grayscale normalization, to generate a standardized CT image dataset. Subsequently, these standardized images are 3D modeled using AI algorithms (such as 3D U-Net) to construct a 3D dynamic model of the patient's abdominal cavity. This 3D dynamic model, generated by the AI ​​algorithm, dynamically displays the anatomical structures within the abdominal cavity.

[0021] In the second stage, tissue composition quantification and pressure parameter generation, an AI segmentation algorithm is used to quantitatively analyze subcutaneous fat thickness, organ volume, and operable space volume on the three-dimensional dynamic model. Combined with the patient's basic medical data, a machine learning model is used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range. In the second stage, AI segmentation algorithms (such as U-Net or Mask R-CNN) are used to precisely segment the 3D dynamic model, identifying and quantifying subcutaneous fat thickness, organ volume, and operable space volume. Combined with the patient's baseline medical data (such as BMI, cardiopulmonary function score, and age), machine learning models (such as Random Forest or XGBoost) are used to generate a personalized pneumoperitoneum pressure target (P_target) and adjustment threshold range. The output from this stage provides a scientific basis for subsequent intraoperative pressure adjustments, ensuring that the pressure setting meets surgical requirements while ensuring patient safety.

[0022] AI segmentation algorithm: an algorithm used to accurately segment different tissues (such as fat and organs) in a three-dimensional model.

[0023] Quantitative analysis: Quantify the volume, thickness and other parameters of the segmented tissue by calculating the number of voxels.

[0024] P_target: Personalized pneumoperitoneum pressure target value, generated based on the patient's anatomy and physiological status.

[0025] The third stage is the intraoperative real-time monitoring and dynamic response stage. The working data is collected in real time through the built-in sensors of the laparoscope system, and the working data is input into the reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP; the working data includes the abdominal pressure P_current, the surgical field clarity score and the patient's vital signs data.

[0026] In the third phase, the laparoscopic system's built-in multimodal sensors collect real-time operational data, including intraperitoneal pressure (P_current), surgical field clarity scores (calculated using an image recognition algorithm), and patient vital signs (such as heart rate and blood oxygen saturation). After preprocessing (e.g., denoising and time series alignment), this data is input into a reinforcement learning model (e.g., a deep Q-network (DQN)) to dynamically calculate the pressure adjustment command (ΔP). This command is used to adjust the pneumoperitoneum pressure in real time, ensuring that the intraoperative pressure is always adapted to the surgical requirements.

[0027] Multimodal sensor: Integrates multiple sensors (such as pressure sensors, image recognition modules, and vital signs monitoring equipment) for multi-dimensional data collection.

[0028] Reinforcement learning model: An AI model that learns optimal strategies by interacting with the environment and can dynamically optimize pressure regulation strategies.

[0029] ΔP: Pressure regulation instruction, used to adjust the pneumoperitoneum pressure in real time.

[0030] The working principle of the present invention is as follows: first, the original image of the patient's abdominal cavity is obtained through CT equipment, and a standardized CT image data set is generated after standardization processing, and an AI algorithm is used to construct a three-dimensional dynamic model of the patient's abdominal cavity; then, the AI ​​segmentation algorithm is used to quantitatively analyze the subcutaneous fat thickness, organ volume and operable space volume of the model, and combined with the patient's basic disease data, a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range are generated through a machine learning model; finally, the abdominal cavity pressure P_current, surgical field clarity score and patient vital signs data are collected in real time through the built-in sensor of the laparoscopic system, and the reinforcement learning model is used to dynamically calculate the pressure adjustment instruction ΔP, driving the intelligent pneumoperitoneum machine to adjust the CO2 injection rate in real time, forming a closed-loop feedback and dynamic optimization of the intraoperative pressure; this method significantly optimizes the intelligence level of the laparoscopic system through the deep integration of AI technology and system closed-loop control, provides good hardware support, and plays a role in optimizing the surgical environment and effect.

[0031] In this embodiment, it is specifically described that in the first stage, original images of the patient's abdominal cavity are obtained by CT equipment, the original images are standardized to generate a standardized CT image dataset, and a three-dimensional dynamic model of the patient's abdominal cavity is constructed based on the AI ​​algorithm, which specifically includes: S11 uses CT equipment to obtain multi-angle slices of the patient's abdominal cavity. Multimodal image fusion technology is used to denoise these raw images, eliminating motion artifacts and device noise, and generating a high signal-to-noise ratio raw image dataset. Multimodal image fusion technology combines multiple imaging data types (such as CT and MRI) to improve image quality and eliminate noise.

[0032] S12 performs spatial registration on the denoised original image dataset. A feature-point-based registration algorithm is used to align the tomographic images from different angles to a unified coordinate system, generating a spatially consistent registered image dataset. This feature-point-based registration algorithm identifies characteristic points in the images (such as vascular intersections and organ edges) and aligns the images from different angles to a unified coordinate system.

[0033] S13 , normalizing the registered image dataset by adjusting image contrast and brightness using histogram equalization and grayscale normalization algorithms to generate a standardized CT image dataset. Histogram equalization, an image enhancement technique, improves image contrast by adjusting pixel grayscale distribution.

[0034] S14, based on a standardized CT image dataset, uses AI algorithms (such as 3D U-Net) to construct a 3D dynamic model of the patient's abdominal cavity. Using a convolutional neural network, it extracts the anatomical features of the subcutaneous fat, organs, and blood vessels within the abdominal cavity, generating a 3D visualization model. 3D U-Net, a deep learning algorithm for 3D image segmentation, is highly efficient in extracting complex anatomical features.

[0035] S15 uses a mesh subdivision algorithm to smooth the 3D dynamic model and optimizes the continuity of the model's internal structure based on the Laplace operator, generating an optimized 3D dynamic model that can be used for quantitative analysis. The mesh subdivision algorithm improves the surface smoothness of the model by increasing mesh density.

[0036] In this embodiment, it is specifically described that S15 specifically includes: S151 performs initial meshing of the 3D dynamic model, using a triangulation algorithm to generate an initial mesh structure to ensure the geometric integrity of the model surface and internal structure. The triangulation algorithm divides the 3D model surface into a triangular mesh to ensure the integrity of the geometric structure.

[0037] S152, based on the initial mesh structure, a subdivision algorithm is used to smooth the surface of the three-dimensional dynamic model, and the mesh density is increased through iterative subdivision to generate a high-resolution three-dimensional mesh model.

[0038] S153 performs Laplace smoothing optimization on the high-resolution 3D mesh model. This algorithm calculates the local curvature of mesh vertices using the Laplace operator and adjusts vertex positions to eliminate irregular protrusions and depressions on the model surface, generating a smooth and continuous 3D mesh model. Laplace smoothing optimization adjusts mesh vertex positions using the Laplace operator to eliminate surface irregularities and improve model smoothness.

[0039] S154, optimizes the internal structure of the smoothed three-dimensional mesh model, uses a volume preservation algorithm based on the Laplace operator, adjusts the positions of the internal mesh nodes, and ensures the continuity and topological consistency of the internal structure of the three-dimensional mesh model.

[0040] S155, performing a quality assessment on the optimized three-dimensional mesh model, and performing a quantitative analysis on the model using a mesh quality index. If the preset threshold is not reached, the mesh is re-subdivided and smoothed until an optimized three-dimensional dynamic model that can be used for quantitative analysis is generated.

[0041] In this embodiment, specifically, in the second stage, an AI segmentation algorithm is used to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional dynamic model. Combined with the patient's basic condition data, a machine learning model is used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range, specifically including: S21, perform tissue segmentation on the three-dimensional dynamic model, use a semantic segmentation algorithm based on deep learning to segment the subcutaneous fat layer, organ area and operable space in the model, and generate a segmented three-dimensional tissue distribution map.

[0042] S22 performs quantitative analysis on the segmented 3D tissue distribution map, using a voxel counting algorithm to calculate subcutaneous fat thickness, organ volume, and operable space volume, generating a tissue composition data table containing quantitative parameters. The voxel counting algorithm quantifies tissue volume and thickness by counting the number of voxels (volumetric pixels) in the 3D model.

[0043] S23, combining the patient's baseline medical data, uses feature engineering to extract key physiological indicators (such as BMI, cardiopulmonary function score, and age). These indicators are then fused with the tissue composition data table to generate a multidimensional feature dataset. Feature engineering is a technique for extracting useful features from raw data, which is used to enhance the input data for machine learning models.

[0044] S24, based on the multidimensional feature data set, a machine learning model is used to train the pneumoperitoneum pressure prediction model, and the model parameters are optimized through cross-validation to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range.

[0045] S25, conduct a safety assessment on the generated pneumoperitoneum pressure target value, and use a rule-based logic engine combined with the patient's basic condition data to dynamically adjust P_target and the adjustment threshold range to keep the pressure setting within a safe range.

[0046] In this embodiment, it is specifically described that the sensor system includes a pressure sensor for monitoring the abdominal cavity pressure P_current, an image recognition module for scoring the clarity of the laparoscopic surgical field, and a monitor for monitoring the patient's vital signs data.

[0047] In this embodiment, it is specifically described that in the third stage, the working data is collected in real time by the built-in sensor system of the laparoscope system, and the working data is input into the reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP, which specifically includes: S31, the working data is collected in real time through the sensor system built into the laparoscope system; the working data includes abdominal pressure (P_current), surgical field clarity score (calculated based on image recognition algorithm) and patient vital signs data.

[0048] S32 preprocesses the real-time monitoring working data set. A sliding window filtering algorithm is used to denoise the intra-abdominal pressure data P_current. The surgical field clarity score and vital sign data are synchronized using a time series alignment method to generate a standardized real-time data set. The sliding window filtering algorithm is a filtering technique used to remove data noise. It smoothes the data by averaging the data within a sliding window. The time series alignment method synchronizes data collected at different times (such as the surgical field clarity score and vital sign data) to the same time point to ensure data consistency.

[0049] S33, compare the standardized real-time data set with the personalized pneumoperitoneum pressure target value P_target, use the differential calculation method to generate the current pressure deviation value ΔP_current, and combine the surgical field clarity score and vital signs data to generate a multi-dimensional state feature vector.

[0050] S34, the multidimensional state feature vector is input into a reinforcement learning model, such as a deep Q-network (DQN). The reinforcement learning model calculates and generates the pressure adjustment command ΔP. The reinforcement learning model is trained based on historical surgical data and can dynamically optimize the pressure adjustment strategy. The multidimensional state feature vector: A comprehensive feature vector containing pressure deviation values, surgical field clarity scores, and vital sign data, is used as input for the reinforcement learning model. The deep Q-network (DQN), a reinforcement learning model, can learn the optimal strategy through interaction with the environment and dynamically optimize the pressure adjustment command.

[0051] S35 , real-time verification of the generated ΔP is performed using a rule-based logic engine combined with vital sign data to determine whether ΔP is within a safe adjustment range. If it exceeds the threshold, the model is triggered to recalculate and generate a revised pressure adjustment instruction ΔP_1.

[0052] In this embodiment, it is further explained that the third stage also includes: In the fourth stage, the closed-loop feedback and pressure optimization stage, the CO2 injection rate is adjusted through the intelligent pneumoperitoneum machine based on ΔP. If the surgical field clarity score or vital signs data deviate from the preset threshold, the machine learning model is triggered to regenerate P_target, forming a dynamic closed-loop optimization of intraoperative pressure.

[0053] In the fourth stage, based on the pressure regulation command (ΔP), the intelligent pneumoperitoneum device adjusts the CO2 infusion rate and intraperitoneal pressure in real time. If the surgical field clarity score or the patient's vital signs deviate from the preset threshold, the system triggers the machine learning model to regenerate a personalized pneumoperitoneum pressure target (P_target), forming a closed-loop feedback loop and dynamic optimization of intraoperative pressure. This mechanism ensures stable and safe intraoperative pressure and avoids complications caused by excessive or insufficient pressure.

[0054] It should be noted that this solution uses an intelligent pneumoperitoneum machine, which can automatically adjust the CO2 injection rate according to pressure regulation instructions.

[0055] Example 2: The present invention also provides a laparoscopic pneumoperitoneum pressure system based on AI image evaluation, which adopts the laparoscopic pneumoperitoneum pressure system intelligent adjustment method based on AI image evaluation as in Example 1. The laparoscopic pneumoperitoneum pressure system includes: The AI ​​image processing unit is used to standardize the original images, generate standardized CT image data sets, and construct a three-dimensional dynamic model of the patient's abdominal cavity.

[0056] The analysis unit uses AI segmentation algorithms to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional model to generate a tissue composition data table.

[0057] The machine learning model unit is used to combine the patient's basic condition data with the tissue composition data table to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range.

[0058] The sensor system is integrated into the laparoscopic system and is used to collect real-time abdominal pressure P_current, surgical field clarity score and patient vital signs data.

[0059] The reinforcement learning model unit is used to dynamically calculate the pressure regulation instruction ΔP and optimize the pressure regulation strategy in real time.

[0060] The intelligent pneumoperitoneum machine is used to automatically adjust the CO2 injection rate according to the pressure regulation instruction ΔP to achieve closed-loop control of the pneumoperitoneum pressure.

[0061] Data storage is used to store preoperative imaging data, intraoperative real-time data and model parameters.

[0062] Communication module: used to realize data interaction and communication between units.

[0063] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligently adjusting the laparoscopic pneumoperitoneum pressure system based on AI image evaluation, characterized in that: include: In the first phase, the original abdominal images of the patient are acquired through CT equipment, standardized and processed to generate a standardized CT image dataset, and a three-dimensional dynamic model of the patient's abdominal cavity is constructed based on an AI algorithm. In the second stage, an AI segmentation algorithm was used to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional dynamic model. Combined with the patient's basic medical data, a machine learning model was used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range. In the third stage, working data is collected in real time through the built-in sensors of the laparoscope system, and the working data is input into the reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP; the working data includes the abdominal pressure P_current, the surgical field clarity score and the patient's vital signs data.

2. The intelligent adjustment method of the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 1 is characterized in that: In the first phase, the original abdominal images of the patient are acquired through CT equipment, the original images are standardized to generate a standardized CT image dataset, and a three-dimensional dynamic model of the patient's abdominal cavity is constructed based on the AI ​​algorithm. Specifically, the following steps are performed: Obtaining original multi-angle tomographic images of the patient's abdominal cavity through a CT device, and performing denoising on the original images using multimodal image fusion technology to generate an original image data set with a high signal-to-noise ratio; The denoised original image dataset is spatially registered, and a feature point-based registration algorithm is used to align the tomographic images at different angles to a unified coordinate system to generate a spatially consistent registered image dataset. The registered image dataset was standardized, and the image contrast and brightness were adjusted using histogram equalization and grayscale normalization algorithms to generate a standardized CT image dataset.

3. The intelligent adjustment method of the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 2 is characterized in that: The registered image dataset is standardized, and image contrast and brightness are adjusted using histogram equalization and grayscale normalization algorithms to generate a standardized CT image dataset, which then includes: Based on a standardized CT image dataset, an AI algorithm was used to construct a 3D dynamic model of the patient's abdominal cavity. The anatomical structural features of the subcutaneous fat, organs, and blood vessels in the abdominal cavity were extracted through a convolutional neural network to generate a 3D visualization model. A grid subdivision algorithm is used to smooth the three-dimensional dynamic model, and the continuity of the internal structure of the three-dimensional dynamic model is optimized based on the Laplace operator to generate an optimized three-dimensional dynamic model that can be used for quantitative analysis.

4. The intelligent adjustment method for laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 3, characterized in that: The 3D dynamic model is smoothed using a mesh subdivision algorithm at any time, and the continuity of the internal structure of the 3D dynamic model is optimized based on the Laplace operator to generate an optimized 3D dynamic model that can be used for quantitative analysis, specifically including: Perform initial mesh division on the three-dimensional dynamic model and use triangulation algorithm to generate the initial mesh structure; Based on the initial grid structure, a subdivision algorithm is used to smooth the surface of the three-dimensional dynamic model, and the grid density is increased through iterative subdivision to generate a high-resolution three-dimensional grid model; Performing Laplace smoothing optimization on the high-resolution three-dimensional mesh model, calculating the local curvature of mesh vertices based on the Laplace operator, adjusting the vertex positions to eliminate irregular protrusions and depressions on the model surface, and generating a smooth and continuous three-dimensional mesh model; The internal structure of the smoothed 3D mesh model is optimized, and the position of the internal mesh nodes is adjusted using a volume preservation algorithm based on the Laplace operator. The quality of the optimized 3D mesh model is evaluated and the model is quantitatively analyzed using mesh quality indicators. If the preset threshold is not reached, the mesh is re-subdivided and smoothed until an optimized 3D dynamic model that can be used for quantitative analysis is generated.

5. The intelligent adjustment method of laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 1, characterized in that: In the second stage, an AI segmentation algorithm is used to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the three-dimensional dynamic model. Combined with the patient's basic condition data, a machine learning model is used to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range, specifically including: Perform tissue segmentation on the 3D dynamic model, using a deep learning-based semantic segmentation algorithm to segment the subcutaneous fat layer, organ area, and operable space in the model, and generate a segmented 3D tissue distribution map; Quantitative analysis is performed on the segmented three-dimensional tissue distribution map. The voxel counting algorithm is used to calculate the thickness of subcutaneous fat, organ volume, and operable space volume, and a tissue composition data table containing quantitative parameters is generated. Combined with the patient's basic condition data, feature engineering methods are used to extract key physiological indicators, and the key physiological indicators are fused with the tissue composition data table to generate a multidimensional feature data set; Based on a multidimensional feature dataset, a machine learning model was used to train a pneumoperitoneum pressure prediction model. The model parameters were optimized through cross-validation to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range. The generated pneumoperitoneum pressure target value is evaluated for safety, and a rule-based logic engine is used in combination with the patient's basic condition data to dynamically adjust P_target and the adjustment threshold range to keep the pressure setting within a safe range.

6. The method for intelligently adjusting the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 1, characterized in that: The sensor system includes a pressure sensor for monitoring the abdominal cavity pressure P_current, an image recognition module for scoring the clarity of the laparoscopic surgical field, and a monitor for monitoring the patient's vital signs data.

7. The method for intelligently adjusting the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 6, characterized in that: The third stage involves collecting working data in real time through the built-in sensor system of the laparoscope system, and inputting the working data into the reinforcement learning model to dynamically calculate the pressure adjustment instruction ΔP, which specifically includes: The working data is collected in real time through the sensor system built into the laparoscope system; The real-time monitoring working data set was preprocessed, and the abdominal pressure data P_current was denoised using a sliding window filtering algorithm. The surgical field clarity score was synchronized with the vital sign data using a time series alignment method to generate a standardized real-time data set. The standardized real-time data set is compared with the personalized pneumoperitoneum pressure target value P_target, and the current pressure deviation value ΔP_current is generated using the differential calculation method. The multi-dimensional state feature vector is generated by combining the surgical field clarity score and vital sign data. The multi-dimensional state feature vector is input into the reinforcement learning model, and the pressure adjustment instruction ΔP is generated by the reinforcement learning model; The generated ΔP is verified in real time. A rule-based logic engine is used in combination with vital signs data to determine whether ΔP is within the safe adjustment range. If it exceeds the threshold, the model is triggered to recalculate and generate the corrected pressure adjustment instruction ΔP_1.

8. The method for intelligently adjusting the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to claim 1, characterized in that: The third stage also includes: In the fourth stage, the CO2 injection rate is adjusted through the intelligent pneumoperitoneum machine based on ΔP. If the surgical field clarity score or vital signs data deviate from the preset threshold, the machine learning model is triggered to regenerate P_target, forming a dynamic closed-loop optimization of intraoperative pressure.

9. A laparoscopic pneumoperitoneum pressure system based on AI image assessment, characterized in that: The intelligent adjustment method of the laparoscopic pneumoperitoneum pressure system based on AI image evaluation according to any one of claims 1 to 8 is adopted, wherein the laparoscopic pneumoperitoneum pressure system comprises: AI image processing unit, used to standardize raw images, generate standardized CT image datasets, and construct a three-dimensional dynamic model of the patient's abdominal cavity; The analysis unit uses AI segmentation algorithms to quantitatively analyze the subcutaneous fat thickness, organ volume, and operable space volume of the 3D model and generate a tissue composition data table; A machine learning model unit is used to combine the patient's basic condition data with the tissue composition data table to generate a personalized pneumoperitoneum pressure target value P_target and adjustment threshold range; The sensor system is integrated into the laparoscopic system and is used to collect real-time abdominal pressure P_current, surgical field clarity score and patient vital signs data; Reinforcement learning model unit, used to dynamically calculate the pressure regulation instruction ΔP and optimize the pressure regulation strategy in real time; Intelligent pneumoperitoneum machine, used to automatically adjust the CO2 injection rate according to the pressure regulation instruction ΔP to achieve closed-loop control of pneumoperitoneum pressure; Data storage, used to store preoperative imaging data, intraoperative real-time data and model parameters; Communication module: used to realize data interaction and communication between units.

Images