A real-time estimation and early warning system for bleeding volume in laparoscopic hepatectomy

Abstract

The application discloses a real-time bleeding volume estimation and early warning system in laparoscopic hepatectomy, and belongs to the technical field of medical monitoring. The system comprises an image acquisition module, an aspiration flow sensor, a blood concentration sensor, a data processing module and an early warning module; the data processing module is connected with each sensor, and comprises a three-dimensional reconstruction unit, an image segmentation unit and a bleeding volume calculation unit. The three-dimensional reconstruction unit reconstructs a three-dimensional surface model of a surgical area according to an image sequence, the image segmentation unit segments a blood coverage area and maps the blood coverage area to the three-dimensional model to obtain three-dimensional distribution information of blood; the bleeding volume calculation unit calculates the volume of pure blood sucked according to the aspiration flow and the blood concentration, and calculates the total bleeding volume in combination with the residual blood volume; the early warning module issues an early warning when the total bleeding volume or the bleeding rate exceeds a threshold value. The application realizes multi-source data fusion and three-dimensional space analysis of intraoperative bleeding volume, and can accurately estimate the total bleeding volume and dynamically issue an early warning.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, and in particular to a real-time estimation and early warning system for blood loss during laparoscopic liver resection. Background Technology

[0002] Laparoscopic liver resection has become a common procedure in liver surgery due to its advantages of minimal trauma and rapid recovery. However, the liver has a rich blood supply and complex anatomical structure, resulting in a high risk of intraoperative bleeding. Accurate assessment of blood loss is crucial for ensuring surgical safety. Traditional methods for assessing blood loss mainly rely on a rough estimate by subtracting the amount of irrigation fluid from the total volume of fluid in the suction bottle, or on the surgeon's subjective judgment based on visual experience. However, the fluid aspirated during laparoscopic surgery often contains various components such as peritoneal effusion and irrigation fluid, making it difficult to accurately reflect actual blood loss by simply relying on the total aspirated volume. Visual judgment is also affected by factors such as limited field of vision, smoke interference, and irregular blood-covered areas, making quantitative assessment difficult.

[0003] In existing technologies, some systems attempt to monitor changes in bleeding area through image analysis, but two-dimensional images cannot provide information on blood thickness, leading to significant deviations in volume estimation. Other systems use flow sensors to monitor suction flow, but lack effective differentiation of the suction fluid components, failing to address interference from non-blood components such as irrigation fluid. Furthermore, factors such as changes in intra-abdominal pressure and the use of electrocoagulation equipment also affect the dynamic assessment of bleeding status, and existing technologies have not yet established an effective data correction mechanism. Summary of the Invention

[0004] In view of this, the present invention aims to provide a real-time blood loss estimation and early warning system for laparoscopic liver resection to solve or alleviate the technical problems existing in the prior art.

[0005] The technical solution of this invention is implemented as follows: A real-time blood loss estimation and early warning system for laparoscopic liver resection, comprising: The image acquisition module is used to acquire image sequences of the laparoscopic surgical field in real time; A suction flow sensor is used to detect the flow rate of the suction fluid in the suction tube in real time. A blood concentration sensor is used to detect the blood concentration in the liquid in the suction tubing in real time. The data processing module is connected to the image acquisition module, the suction flow sensor, and the blood concentration sensor, and includes a three-dimensional reconstruction unit, an image segmentation unit, and a bleeding volume calculation unit. The three-dimensional reconstruction unit is used to reconstruct a three-dimensional surface model of the surgical area based on the image sequence; The image segmentation unit is used to segment the blood-covered area from the image sequence and map it onto the three-dimensional surface model to obtain the distribution information of blood in three-dimensional space, including the boundary coordinates of the blood-covered area and the local blood thickness. The bleeding volume calculation unit is used to calculate the volume of pure blood aspirated based on the suction flow rate and blood concentration, and to calculate the volume of residual blood in the surgical area based on the distribution of blood in three-dimensional space. Then, it calculates the total bleeding volume, which is the sum of the volume of pure blood aspirated and the volume of residual blood. The early warning module, connected to the data processing module, is used to compare the total bleeding volume or bleeding rate with a preset threshold, and issue an early warning signal when the threshold is exceeded.

[0006] Compared with existing technologies, this invention has the following advantages: Through multi-source data fusion and three-dimensional spatial analysis technology, it significantly improves the accuracy and reliability of blood loss estimation during laparoscopic liver resection. The system combines a suction flow sensor with a blood concentration sensor, enabling precise separation of pure blood components from the suction fluid, effectively overcoming the shortcomings of traditional methods where non-blood substances such as irrigation fluid interfere with blood loss assessment. Simultaneously, based on three-dimensional reconstruction and image segmentation technology, the system can acquire three-dimensional spatial distribution information of blood in the surgical area and calculate local blood thickness, solving the problem that two-dimensional image analysis cannot provide volume information. Furthermore, the system integrates multi-modal correction modules such as pressure monitoring, irrigation flow sensing, and electrocoagulation status recognition, which can dynamically eliminate the interference of intraoperative environmental changes on the estimation results and provide timely warnings of bleeding risk through an early warning mechanism. Therefore, it provides clinical practice with a high-precision, interference-resistant real-time blood loss estimation and dynamic early warning solution.

[0007] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a block diagram of the overall system structure of the present invention; Figure 2 This is a flowchart of the blood loss estimation process of the present invention; Figure 3This is a diagram showing the internal structure of the data processing module of the present invention; Figure 4 This is a flowchart of the Kalman filter data fusion process of the present invention; Figure 5 This is a structural diagram of the early warning module of the present invention. Detailed Implementation

[0010] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0011] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0012] This invention proposes a real-time blood loss estimation and early warning system for laparoscopic liver resection, comprising: The image acquisition module is used to acquire image sequences of the laparoscopic surgical field in real time; A suction flow sensor is used to detect the flow rate of the suction fluid in the suction tube in real time. A blood concentration sensor is used to detect the blood concentration in the liquid in the suction tubing in real time. The data processing module is connected to the image acquisition module, the suction flow sensor, and the blood concentration sensor, and includes a three-dimensional reconstruction unit, an image segmentation unit, and a bleeding volume calculation unit. The 3D reconstruction unit is used to reconstruct a 3D surface model of the surgical area based on an image sequence; The image segmentation unit is used to segment the blood-covered area from the image sequence and map it onto the three-dimensional surface model to obtain the distribution information of blood in three-dimensional space, including the boundary coordinates of the blood-covered area and the local blood thickness. The bleeding volume calculation unit is used to calculate the volume of pure blood aspirated based on the suction flow rate and blood concentration, and to calculate the volume of residual blood in the surgical area based on the distribution of blood in three-dimensional space. Then, it calculates the total bleeding volume, which is the sum of the volume of pure blood aspirated and the volume of residual blood. The early warning module, connected to the data processing module, is used to compare the total bleeding volume or bleeding rate with a preset threshold, and issue an early warning signal when the threshold is exceeded.

[0013] The system is first equipped with an image acquisition module for acquiring real-time image sequences of the laparoscopic surgical field. For example, this module could be a high-resolution endoscopic camera that transmits real-time images of the surgical area to the system via an optical lens. These image sequences provide the foundational data for subsequent visual analysis.

[0014] Simultaneously, the system is equipped with a suction flow sensor to detect the flow rate of the suction liquid in the suction pipe in real time. This sensor can be a turbine flow meter or an ultrasonic flow meter, which determines the flow rate by measuring changes in the physical parameters of the fluid as it passes through the pipe. For example, when the suction device is operating, the sensor can continuously output instantaneous flow rate data of the suction liquid.

[0015] In addition, the system includes a blood concentration sensor for real-time detection of blood concentration in the fluid within the suction tubing. This sensor employs a simple optical transmission or reflection principle, estimating the blood component content by measuring the degree of absorption of light at a specific wavelength by the suction fluid. For example, as the blood content in the suction fluid increases, the sensor's output signal changes accordingly, indicating an increase in blood concentration.

[0016] The image acquisition module, suction flow sensor, and blood concentration sensor are all connected to the data processing module. This data processing module is the central control and computing core of the system, integrating a 3D reconstruction unit, an image segmentation unit, and a blood loss calculation unit. The data processing module is responsible for coordinating the data input from each sensor and performing complex algorithmic processing.

[0017] The 3D reconstruction unit reconstructs a 3D surface model of the surgical area based on the image sequence. For example, by analyzing the displacement or deformation of objects in the image sequence from different viewpoints, this unit uses geometric principles to calculate the depth information of the surgical area, and then constructs a rough 3D model of the area.

[0018] The image segmentation unit segments blood-covered areas from an image sequence and maps them onto a 3D surface model. For example, based on color features or brightness differences in the image, this unit identifies and separates red areas (representing blood) by setting a threshold. Subsequently, through a simple projection transformation, this 2D blood region information is superimposed onto the reconstructed 3D surface model, thereby obtaining preliminary distribution information of blood in 3D space, including the boundary coordinates of the blood-covered areas and the local blood thickness. The local blood thickness can be empirically estimated based on the color intensity or diffusion range of the blood in the 2D image.

[0019] The blood loss calculation unit comprehensively utilizes the suction flow rate, blood concentration, and blood distribution information in three-dimensional space to calculate the total blood loss. Specifically, this unit first calculates the amount of pure blood suctioned out based on data from the suction flow rate sensor and the blood concentration sensor through simple multiplication. For example, if the suction flow rate is... Blood concentration is The amount of pure blood drawn out can be approximated as: Multiply Simultaneously, based on the blood distribution information in three-dimensional space provided by the image segmentation unit, this unit calculates the residual blood volume in the surgical area by simply multiplying and summing the surface area of ​​the blood-covered region and the estimated local blood thickness. Finally, the volume of pure blood aspirated is added to the volume of residual blood to obtain the total bleeding volume.

[0020] Finally, the system is equipped with an early warning module connected to the data processing module. This early warning module is used to compare the calculated total blood loss or bleeding rate with preset thresholds. For example, when the total blood loss exceeds the preset threshold of 100 ml, or the bleeding rate exceeds 10 ml per minute, the early warning module will immediately issue a single-level visual or audible alarm signal to remind medical staff to pay attention to the patient's bleeding situation.

[0021] This invention combines multi-source data such as image acquisition, suction flow detection, and blood concentration detection, and utilizes 3D reconstruction and image segmentation techniques to obtain blood distribution information in three-dimensional space, enabling real-time, quantitative estimation of blood loss during laparoscopic liver resection. The system effectively distinguishes pure blood components in the suction fluid, overcoming the interference problem of non-blood components such as irrigation fluid in traditional methods. Simultaneously, by estimating the volume of residual blood in the surgical area, it compensates for the lack of blood thickness information provided by 2D image analysis, thereby improving the accuracy of blood loss assessment. Furthermore, the introduction of an early warning mechanism helps medical staff promptly grasp the patient's bleeding status, ensuring surgical safety.

[0022] However, in actual surgical settings, the fluid in the suction tubing may contain air bubbles, or the blood composition may be complex, making it difficult for a single detection method to accurately distinguish between them. This could lead to deviations in blood concentration measurements, which in turn could affect the accuracy of total blood loss estimation.

[0023] To address this, the present invention further proposes that the blood concentration sensor includes a multi-wavelength optical detection unit and a bubble detection unit. The multi-wavelength optical detection unit uses at least two light sources of different wavelengths to measure the absorbance of the attracted liquid and calculates the blood concentration based on a preset hemoglobin concentration-absorbance relationship curve. The bubble detection unit monitors bubbles in the aspiration tube in real time; when a bubble is detected, the data processing module discards the concentration data at the corresponding time or uses interpolation compensation. Specifically, the blood concentration... Calculated using the following formula: ,in The absorbance measured by the multi-wavelength optical detection unit. and This is a preset calibration coefficient, which is obtained through experimental calibration and is used to linearly map absorbance to blood concentration.

[0024] The multi-wavelength optical detection unit precisely quantifies the blood component in the attracted liquid using optical methods. This unit typically contains multiple light-emitting diodes (LEDs) or laser diodes capable of emitting light at least two different wavelengths; for example, one wavelength is sensitive to hemoglobin absorption, while the other serves as a reference or is sensitive to other non-blood components. This light passes through the liquid in the attraction tube and is received by a photodetector. By measuring the absorbance (A) of different wavelengths of light in the liquid, the blood concentration can be calculated using a pre-established hemoglobin concentration-absorbance relationship curve. For example, using… The law derives absorbance by measuring the degree of light intensity attenuation at a specific wavelength, and then combines this with a calibration coefficient. and absorbance Linear mapping to blood concentration Calibration coefficient and This is usually achieved through experimental calibration using blood samples of known concentrations under laboratory conditions to ensure the accuracy and reliability of the measurement results.

[0025] The bubble detection unit is a key auxiliary module for improving the accuracy of blood concentration measurement. During laparoscopic surgery, bubbles may form in the suction tubing due to various reasons (such as suction device operation, fluid agitation, etc.). These bubbles can interfere with the optical path of the optical detection unit, leading to abnormal absorbance measurements and introducing errors in blood concentration calculation. The bubble detection unit can be implemented using various technologies, such as additional photoelectric sensors to scatter or interrupt the detection light in the tubing, or using ultrasonic sensors to detect changes in the density of the fluid medium. When the bubble detection unit detects bubbles in the suction tubing in real time, it transmits this information to the data processing module. Upon receiving the bubble signal, the data processing module immediately processes the current blood concentration data. For example, it may directly remove the abnormal data point to avoid its participation in subsequent blood loss calculations; or it may use an interpolation compensation algorithm to estimate a reasonable concentration value at that moment based on the effective concentration data before and after the bubble's appearance, thus ensuring the continuity and integrity of the blood concentration data.

[0026] Through the above technical solutions, the multi-wavelength optical detection unit can utilize the absorbance differences of different wavelength light sources to more accurately identify and quantify blood components in the aspirated liquid, effectively avoiding the influence of non-blood substances or background interference on concentration measurement, thereby improving the accuracy of blood concentration measurement. Simultaneously, the bubble detection unit monitors and processes bubble interference in the aspiration tubing in real time, ensuring the continuity and reliability of blood concentration data by eliminating abnormal data or performing interpolation compensation, avoiding measurement errors caused by bubbles. These improvements enable the blood concentration sensor to provide more stable and accurate real-time blood concentration data, significantly improving the estimation accuracy of the aspirated pure blood volume by the bleeding volume calculation unit, and thus improving the accuracy and reliability of the entire system's real-time estimation of total blood loss during laparoscopic liver resection, providing physicians with more reliable clinical decision-making information.

[0027] However, in actual surgical settings, due to the two-dimensional nature of laparoscopic images and the complexity of the surgical field, it is difficult to obtain a high-precision three-dimensional surface model and accurate information on the blood coverage area by performing three-dimensional reconstruction and blood region segmentation based solely on image sequences. This may affect the accuracy of residual blood volume estimation, and consequently affect the real-time monitoring and early warning effect of total blood loss.

[0028] To address this, the present invention further proposes that the three-dimensional reconstruction unit reconstructs a three-dimensional surface model of the surgical area from an image sequence using stereo vision or structured light projection technology. The image segmentation unit utilizes a deep learning semantic segmentation network to segment the blood region and maps the two-dimensional segmentation results onto the three-dimensional surface model through coordinate transformation, thereby calculating the surface area and corresponding volume of the blood-covered area. The residual blood volume... The calculation is performed by integrating the three-dimensional surface of the blood-covered area: ,in For the three-dimensional surface of the blood-covered area, This refers to the local blood thickness, which is estimated based on the three-dimensional surface curvature and a pre-defined blood spreading model. This represents a small area element.

[0029] Specifically, the 3D reconstruction unit employs stereo vision or structured light projection techniques when reconstructing the 3D surface model of the surgical area. When using stereo vision, the system simultaneously acquires image sequences of the surgical area through at least two image acquisition modules with known relative positions and orientations. By matching feature points in these images and utilizing triangulation principles, the coordinates of each matched feature point in 3D space can be calculated, thereby reconstructing the 3D point cloud data of the surgical area. Subsequently, surface reconstruction algorithms (such as Poisson reconstruction, ...) are used to reconstruct the surface model. Point cloud data (e.g., triangulation) is used to convert point cloud data into a continuous 3D surface model. This method can be implemented using binocular or multi-view endoscopes in existing laparoscopic systems without additional hardware. When using structured light projection technology, the system projects light of a known pattern (e.g., striped light, coded light, or random speckle) onto the surgical area and uses an image acquisition module to capture the light pattern modulated by the object's surface. By analyzing the deformation of the light pattern, the depth information of the object's surface can be accurately calculated, thereby reconstructing a high-precision 3D surface model of the surgical area. Structured light projection technology typically provides higher depth measurement accuracy than pure stereo vision, and is particularly suitable for areas with less texture or strong surface reflection.

[0030] The image segmentation unit utilizes a deep learning semantic segmentation network when segmenting blood regions. Semantic segmentation classifies each pixel in an image into a predefined category. In this application, a deep learning semantic segmentation network (e.g., (e.g., by training on a large dataset of labeled laparoscopic surgical images, the network learns the visual features of blood regions. The trained network can automatically identify and accurately delineate blood-covered areas in image sequences, maintaining high segmentation accuracy and robustness even under conditions of uneven lighting, occlusion, or variable blood morphology.)

[0031] To transform the blood segmentation results in a 2D image into its actual distribution in 3D space, a coordinate transformation is needed to map it onto a 3D surface model. This typically involves using camera intrinsic (focal length, principal point, distortion coefficients) and extrinsic (camera rotation and translation relative to the world coordinate system) matrices to back-project pixels from the 2D image coordinate system onto the 3D surface model. This mapping determines the precise location and shape of the blood-covered area on the 3D model, providing a foundation for subsequent volume calculations. Once the blood-covered area is accurately mapped onto the 3D surface model, its surface area can be calculated by summing the areas of the corresponding triangular facets on the 3D model. The corresponding volume is then estimated based on this surface area, combined with the local blood thickness.

[0032] Residual blood volume The calculation is performed by integrating the three-dimensional surface of the blood-covered area. Divided into countless tiny area elements And the local blood thickness on each micro-element The product is summed. This method fully accounts for the non-uniform distribution of blood on a three-dimensional surface, thus obtaining a more accurate residual blood volume. Among them, local blood thickness The thickness is a key parameter affecting the accuracy of residual blood volume calculation, reflecting the accumulation of blood at different locations. This thickness is estimated based on the three-dimensional surface curvature and a pre-defined blood spreading model. When blood spreads on a surface, its thickness is influenced by the surface geometry (curvature) and the blood's own physical properties (surface tension, viscosity). For example, in concave areas, blood tends to accumulate and become thicker; while in convex or sloping areas, the blood may spread thinner. The pre-defined blood spreading model can be a physical model based on fluid dynamics principles or an empirical model obtained by fitting extensive experimental data, used to describe the spreading behavior of blood on surfaces with different curvatures. By combining the local curvature information of the three-dimensional surface model with the blood spreading model, the local blood thickness on each micro-element can be dynamically and accurately estimated, thereby improving the accuracy of residual blood volume estimation.

[0033] Through the above technical solutions, this system overcomes the limitations of traditional two-dimensional image analysis in acquiring three-dimensional spatial information. Specifically, the three-dimensional reconstruction unit uses stereo vision or structured light projection technology to accurately reconstruct the three-dimensional surface model of the surgical area, providing a precise geometric basis for the distribution of blood in three-dimensional space. Simultaneously, the image segmentation unit utilizes a deep learning semantic segmentation network to achieve pixel-level accurate identification of blood regions in laparoscopic images, significantly improving the accuracy and robustness of segmentation. Based on this, through precise coordinate transformation, the two-dimensional segmentation results are seamlessly mapped onto the three-dimensional surface model, ensuring the accuracy of blood spatial distribution information. Furthermore, by performing integral calculations on the three-dimensional surface of the blood-covered area and combining this with the local blood thickness estimated based on the three-dimensional surface curvature and a preset blood spreading model, this system can precisely calculate the residual blood volume in the surgical area. This method fully considers the uneven spreading characteristics of blood on complex three-dimensional surfaces, greatly improving the accuracy of residual blood volume estimation. This makes the real-time estimation of total blood loss more accurate and reliable, providing clinicians with more precise blood loss monitoring data and more timely early warning information, helping doctors to take timely intervention measures and ensure patient safety.

[0034] However, during laparoscopic surgery, the intra-abdominal pressure is not constant, and its fluctuations may cause deformation of tissues and organs in the surgical area, which in turn affects the actual spread and local thickness of blood on the three-dimensional surface, thus introducing errors in the estimation of residual blood volume and reducing the accuracy of blood loss estimation.

[0035] To address this, the present invention further proposes that the system also includes a pressure monitoring module for real-time monitoring of intra-abdominal pressure; and a data processing module for adjusting the geometric parameters of the three-dimensional surface model based on changes in intra-abdominal pressure, or for correcting the calculated residual blood volume based on pressure changes. Specifically, the corrected residual blood volume... It can be obtained through the following formula: ,in The volume of residual blood before correction. For real-time intra-abdominal pressure, For reference pressure value, This is the pressure correction factor, obtained by fitting experimental data.

[0036] The pressure monitoring module is used to monitor intra-abdominal pressure in real time. This module employs miniature pressure sensors, integrated with laparoscopic cannulas or surgical instruments, to continuously and non-invasively collect intra-abdominal pressure data. These sensors typically possess high sensitivity and good biocompatibility, accurately reflecting subtle changes in intra-abdominal pressure and transmitting the collected pressure signals to the data processing module.

[0037] The data processing module adjusts the geometric parameters of the three-dimensional surface model based on changes in intra-abdominal pressure. Specifically, changes in intra-abdominal pressure cause elastic deformation of intra-abdominal tissues and organs, thus affecting the actual geometry of the surgical area. The data processing module can pre-store or calculate in real-time a mapping model between pressure and deformation, for example, through finite element analysis or machine learning-based methods. When the pressure monitoring module detects intra-abdominal pressure... Deviation from preset reference pressure At that time, the data processing module will dynamically adjust the three-dimensional surface model of the surgical area generated by the three-dimensional reconstruction unit according to the mapping relationship, such as correcting the curvature, surface area or local unevenness of the model, so that the three-dimensional model can more accurately reflect the real geometric state under the current pressure conditions.

[0038] Furthermore, the data processing module corrects the calculated residual blood volume based on pressure changes. Changes in intra-abdominal pressure can affect the spreading characteristics of blood on tissue surfaces (such as spreading area and local thickness). To eliminate this effect, the data processing module employs a correction formula. Obtain the corrected residual blood volume .in, It is the residual blood volume before correction, which is initially estimated by the bleeding volume calculation unit based on the image segmentation results and local blood thickness. It is the intra-abdominal pressure acquired in real time by the pressure monitoring module; It is a preset reference pressure value, usually set as the baseline pressure at the start of surgery or a certain stable pressure; This is a pressure correction factor, obtained through fitting extensive in vitro, animal, or clinical data. It quantifies the impact of unit pressure changes on the estimation of residual blood volume. This correction allows for a more accurate reflection of the actual residual blood volume.

[0039] By introducing a pressure monitoring module to acquire intra-abdominal pressure data in real time, and enabling the data processing module to dynamically adjust the geometric parameters of the three-dimensional surface model based on these pressure changes, or directly correct the calculated results of residual blood volume, this invention can effectively compensate for the impact of intra-abdominal pressure fluctuations on the tissue morphology and blood spread in the surgical area. This significantly improves the accuracy of residual blood volume estimation, thereby making the real-time estimation of total blood loss during laparoscopic liver resection more accurate and reliable. Physicians can obtain information closer to the actual blood loss, helping to promptly assess the patient's physiological condition and take appropriate medical interventions, thus improving surgical safety and patient prognosis.

[0040] In laparoscopic liver resection, surgeons often perform irrigation to maintain a clear surgical field. However, the introduction of irrigation fluid dilutes or washes away the blood in the surgical area. This causes the data processing module to be unable to accurately distinguish between pure bleeding and blood diluted or removed by the irrigation fluid when segmenting blood-covered areas based solely on image sequences. Consequently, this interferes with the estimation of residual blood volume and affects the accuracy of total blood loss estimation.

[0041] To address this, the present invention further proposes that, in order to eliminate the interference of the flushing operation on the estimation of residual blood volume, the system also includes a flushing flow sensor. This flushing flow sensor is used to detect the flow rate of the flushing fluid in the flushing device pipeline in real time. Specifically, the flushing flow sensor employs various technologies, such as an ultrasonic flow meter, which determines the flow rate by measuring the change in the propagation time of ultrasound in the fluid, offering advantages such as non-contact operation and high accuracy; or an electromagnetic induction flow meter, which calculates the flow rate by measuring the electromotive force generated in a magnetic field by a conductive fluid. The flushing flow sensor outputs real-time flushing flow rate data. This provides a crucial input for the subsequent data processing module to correct for residual blood volume.

[0042] After receiving flushing flow data from the flushing flow sensor, the data processing module corrects for changes in the blood-covered area in the image based on this data. Specifically, during flushing, the flushing fluid enters the surgical area, causing some blood to be washed away or diluted, making the blood-covered area identified by the image segmentation unit no longer a complete representation of the actual residual blood volume. By integrating the flushing flow data, the data processing module quantifies the impact of the flushing fluid on the blood-covered area, thereby adjusting the estimation of the residual blood volume.

[0043] To achieve the above corrections, this invention employs a correction formula. .in, This represents the uncorrected residual blood volume, which is the blood volume calculated by the image segmentation unit and the 3D reconstruction unit without considering the effects of flushing. The flushing fluid flow rate is detected in real time by the flushing flow sensor. This indicates the cumulative volume of the flushing liquid, which is determined by the flushing flow rate. The time integral is used to obtain the total amount of irrigation fluid that has entered the surgical area from the start of irrigation to the current moment. The subtraction factor for residual blood volume during flushing is a crucial calibration parameter obtained through experimental calibration. For example, different flushing conditions can be simulated in in vitro experiments or animal models to observe the change in known blood volume under different flushing volumes, thereby determining the η value. This factor reflects the amount of blood that a unit volume of flushing fluid can remove or dilute, and its value may be related to factors such as the properties of the flushing fluid, flushing pressure, blood viscosity, and the surface characteristics of the surgical area. Using this correction formula, the data processing module can calculate... This refers to the corrected residual blood volume, which more accurately reflects the actual amount of residual blood in the surgical area.

[0044] Through the above technical solution, this system introduces a flushing flow sensor, enabling real-time monitoring of the flushing fluid flow rate during surgical flushing. The data processing module utilizes the real-time flushing flow data, combined with a preset correction formula, to quantitatively correct the residual blood volume obtained from image analysis. Specifically, the system can subtract the amount of blood carried away or diluted by the flushing operation from the initially estimated residual blood volume, effectively eliminating the interference of the flushing operation on the estimation of residual blood volume. This significantly improves the accuracy and reliability of total blood loss estimation during surgical flushing, avoiding underestimation or overestimation of blood loss due to flushing. It provides surgeons with more accurate and real-time blood loss information, helping them to promptly assess the patient's physiological condition and make more informed clinical decisions, thereby improving surgical safety and patient prognosis. However, in actual surgical procedures, the use of electrosurgical equipment (such as electrocoagulation devices) can introduce smoke, which may interfere with the image segmentation unit's accurate identification of blood areas, leading to inaccurate estimation of residual blood volume. Furthermore, since electrocoagulation aims to stop bleeding, if the system fails to promptly identify the hemostasis status, bleeding may continue to accumulate, affecting the accuracy of bleeding volume estimation and the timeliness of warnings.

[0045] To address this, the present invention further proposes that the system also includes an electrosurgical equipment monitoring module for acquiring the usage status and power parameters of the electrocoagulation equipment; the data processing module determines whether the bleeding has been stopped based on the usage of the electrocoagulation equipment and adjusts the bleeding volume estimation strategy accordingly, or corrects the image segmentation results during electrocoagulation to avoid smoke interference. Specifically, when the electrocoagulation equipment is activated, the data processing module uses a smoke suppression algorithm to preprocess the image to reduce the impact of smoke on segmentation; when the electrocoagulation equipment is turned off, if the bleeding area does not change significantly, it is determined that the bleeding has been controlled, and the accumulation of bleeding volume is stopped accordingly.

[0046] The electrosurgical equipment monitoring module acquires the real-time usage status (e.g., active or off) and power parameters of the electrocoagulation equipment. This can be achieved in various ways, such as by establishing a wired or wireless communication connection with the electrosurgical equipment to directly read the internal operating status signals and set power values; or by integrating current or voltage sensors into the power line or output line of the electrosurgical equipment to indirectly detect its operating status and power output. The core function of this module is to provide the data processing module with real-time contextual information about the electrocoagulation operation, so that the system can adjust the blood loss estimation logic according to the actual progress of the surgery.

[0047] The data processing module determines whether bleeding has been stopped based on the usage of the electrocoagulation device and adjusts the bleeding volume estimation strategy accordingly. When the electrocoagulation device is activated, the system anticipates that the bleeding volume may decrease or stop. At this time, the data processing module can temporarily reduce the weight of the accumulated bleeding volume or activate specific hemostasis judgment logic. For example, if the area of ​​blood coverage detected by the image segmentation unit continues to decrease or stops increasing within a certain period after electrocoagulation activation, it can be preliminarily determined that the bleeding has been effectively controlled.

[0048] To correct image segmentation results during electrocoagulation to avoid smoke interference, the data processing module preprocesses the images acquired by the image acquisition module using a smoke suppression algorithm when the electrocoagulation equipment is activated. Smoke suppression algorithms can include, but are not limited to: physics-based dehazing algorithms (such as dark channel prior algorithms), which remove the influence of smoke by analyzing the physical properties of smoke in the image; or deep learning-based dehazing networks, which learn smoke removal patterns by training on a large number of smoke-filled and smoke-free images. Furthermore, image enhancement techniques, such as histogram equalization, can be combined. Algorithms, etc., are used to improve image contrast and detail, enabling image segmentation units to more accurately identify blood regions and avoid misidentifying smoke as blood.

[0049] After the electrocoagulation device is turned off, the data processing module continuously monitors the area of ​​blood coverage output by the image segmentation unit. The system sets a time window; within this window, if the rate of change of the blood coverage area is lower than a preset threshold, or if the absolute area value remains below a certain minimum, the system determines that bleeding has been effectively controlled. Once bleeding is determined to be under control, the bleeding calculation unit will correspondingly stop accumulating the total bleeding volume or significantly reduce the accumulation rate to avoid overestimation.

[0050] By introducing an electrosurgical equipment monitoring module, this system can acquire the real-time usage status and power parameters of the electrocoagulation equipment, thus providing crucial surgical context information for the data processing module. When the electrocoagulation equipment is activated, the data processing module uses a smoke suppression algorithm to preprocess the image, effectively reducing the interference of electrocoagulation smoke on the image segmentation unit's identification of blood regions and significantly improving the accuracy of residual blood volume estimation. Furthermore, after the electrocoagulation equipment is turned off, the system can intelligently determine whether bleeding has been effectively controlled based on changes in the bleeding area. If the bleeding area does not change significantly, the system promptly stops accumulating the bleeding volume, avoiding inflated estimates of bleeding volume due to hemostasis procedures. These improvements enable the system to more accurately and in real-time estimate total blood loss and provide more reliable early warnings in laparoscopic surgeries with frequent electrocoagulation operations, thereby assisting surgeons in making better intraoperative decisions and effectively managing the patient's bleeding risk. In some embodiments of the present invention described above, a method is proposed to detect the blood concentration in the fluid of the suction tubing in real time using a blood concentration sensor, and to use this as the basis for calculating the total blood loss. However, in practical applications, due to significant individual differences among patients, such as preoperative hemoglobin levels and blood dilution, the preset conversion curve used by the blood concentration sensor may not accurately reflect the true blood concentration of a specific patient, thus affecting the accuracy of blood loss estimation.

[0051] To address this issue, the present invention further proposes that, in order to solve the above problems, the system also includes a non-invasive hemoglobin monitoring module for real-time acquisition of the patient's current hemoglobin concentration. The data processing module dynamically calibrates the conversion curve of the blood concentration sensor based on this hemoglobin concentration to eliminate the influence of individual differences on blood concentration measurement. The calibrated blood concentration... Calculated using the following formula: ,in The initial concentration measured by the blood concentration sensor. This refers to the hemoglobin concentration acquired in real time by the non-invasive hemoglobin monitoring module. This is the preset reference hemoglobin concentration.

[0052] Specifically, the non-invasive hemoglobin monitoring module is a device that can measure a patient's hemoglobin concentration in real time without invasive procedures. It typically employs the principle of a pulse oximeter, emitting light of a specific wavelength that penetrates tissue and infers hemoglobin concentration based on changes in light absorption. For example, multi-wavelength photoplethysmography or reflectance spectroscopy can be used, with data acquired through sensors on the skin surface. This module provides continuous and non-invasive physiological parameters, avoiding the invasiveness and latency of traditional blood sampling tests, and providing real-time, personalized hemoglobin baseline information for subsequent blood concentration calibration.

[0053] The data processing module dynamically adjusts or corrects the absorbance-concentration conversion relationship of the blood concentration sensor based on the real-time hemoglobin concentration provided by the non-invasive hemoglobin monitoring module. This adjustment is made when the non-invasive hemoglobin monitoring module obtains the patient's current hemoglobin concentration. At that time, the data processing module will compare it with a preset reference hemoglobin concentration. Compare. If and If there is a difference, the initial concentration measured by the blood concentration sensor is considered to be... There may be some discrepancy. The data processing module will adjust the settings based on this difference using a calibration factor (e.g., ...). / ) correction To obtain calibrated blood concentrations that are closer to the true values. This calibration can be linear or a more complex nonlinear adjustment. Its core lies in dynamically adjusting the calibration coefficients or curves to eliminate the influence of individual patient differences on the accuracy of blood concentration measurement and improve the accuracy of blood concentration estimation.

[0054] calibrated blood concentration Through formula Calculations are performed to convert the initial concentration measured by the blood concentration sensor. Combined with the patient's real-time hemoglobin concentration and the preset reference hemoglobin concentration Calculate a more accurate calibrated blood concentration The data processing module receives... , and As input. It could be the average hemoglobin concentration of a standard healthy person, or a baseline value set during system initialization. Data is provided in real time by the non-invasive hemoglobin monitoring module. The data processing module performs multiplication and division operations to obtain... This formula assumes a proportional relationship between the blood concentration sensor reading and the actual hemoglobin concentration. This proportional relationship can be dynamically adjusted based on the patient's own hemoglobin level, thus providing a quantitative and operable calibration method to ensure the accuracy of blood concentration estimation and improve the reliability of total blood loss estimation.

[0055] By introducing a non-invasive hemoglobin monitoring module through the above technical solution, the system can acquire the patient's individualized hemoglobin concentration in real time. The data processing module uses this information to dynamically calibrate the conversion curve of the blood concentration sensor, effectively eliminating the influence of individual patient differences on the accuracy of blood concentration measurement. Specifically, calibration using the aforementioned formula ensures that the initial concentration measured by the blood concentration sensor is accurate. It can adjust according to the patient's current physiological state, thereby obtaining a calibrated blood concentration that is closer to the true value. This significantly improves the accuracy of estimating the amount of pure blood and total blood loss, enabling the early warning module to issue warnings based on more reliable data. This enhances the overall reliability and clinical value of the real-time blood loss estimation and early warning system during laparoscopic liver resection.

[0056] In laparoscopic liver resection, real-time acquired image sequences and data from suction flow sensors and blood concentration sensors can be affected by various factors, such as sensor noise, changes in the surgical environment, and operator uncertainty, leading to fluctuations and errors in the directly calculated blood loss. This uncertainty affects the accuracy of blood loss estimation and the timeliness of early warnings, making it difficult for doctors to accurately grasp the patient's real-time bleeding status, potentially delaying intervention.

[0057] To address this, the present invention further proposes that the data processing module uses a Kalman filter to fuse multi-source data such as suction flow, blood concentration, and residual blood volume, outputting the optimal estimate of total bleeding volume, and generating a predicted bleeding volume for a future period based on a time series prediction algorithm. The state vector of the Kalman filter includes total bleeding volume, bleeding rate, and residual blood volume. The recursive formula of the Kalman filter includes: State prediction: Covariance prediction: Kalman gain: Status Update: Covariance update: ;in This is the system state vector, including total bleeding, etc. This is the state transition matrix; For control matrix; For control input; It is the covariance matrix; For process noise covariance; The observation matrix; To observe the noise covariance; These are observed values, including suction flow rate, blood concentration, and residual blood volume; For Kalman gain.

[0058] Specifically, the Kalman filter is used to estimate the state of a dynamic system from a series of incomplete or noisy measurements. It achieves optimal estimation of the system state using a minimum mean square error criterion by fusing system model predictions and sensor observations. In this invention, the Kalman filter is configured to receive multiple data sources, including suction flow sensors, blood concentration sensors, and residual blood volume calculated by an image segmentation unit. These data may contain noise and uncertainty due to various factors (such as sensor accuracy and surgical environment interference). Through the Kalman filter, the system can effectively fuse these multi-source data, thereby obtaining a more accurate and stable blood loss estimate than from a single data source.

[0059] Time series forecasting algorithms are used to analyze historical data point sequences, identify their inherent patterns, trends, and periodicity, and use this information to predict future data points. In this invention, the algorithm is used to predict the trend of bleeding volume changes over a future period based on historical total bleeding volume data. This can be achieved through various models, such as the autoregressive moving average model (...). Exponential smoothing, or more complex machine learning models, such as recurrent neural networks (RNNs), can be used. Long Short-Term Memory Network (LSTM) )or Structure. By learning from and modeling historical bleeding data, this algorithm can provide forward-looking bleeding volume predictions, offering predictive information to healthcare professionals.

[0060] In a Kalman filter, the state vector It is a set of variables describing the current state of the system. In this invention, the state vector... The bleeding process is defined as including total blood loss, bleeding rate, and residual blood volume. Total blood loss is the total amount of blood accumulated during the operation; bleeding rate reflects how quickly the blood loss changes over time; and residual blood volume represents the amount of blood in the surgical area that has not yet been aspirated. By incorporating these key parameters into the state vector, the Kalman filter can comprehensively and dynamically track and estimate various aspects of the bleeding process.

[0061] The recursive formula of the Kalman filter is the core mathematical framework for its implementation, including five steps: state prediction, covariance prediction, Kalman gain calculation, state update, and covariance update. State prediction ( Based on the optimal state estimate of the previous time step and system state transition matrix Predict the system state at the current moment. Control matrix and control input Used to account for the impact of external controls on the system state. Covariance prediction ( The covariance of the current state estimate. ,in The process noise covariance reflects the uncertainty of the system model. Kalman gain ( Calculate Kalman gain It determines the observed value Weights for state updates. For the observation matrix, To observe the noise covariance, which reflects the noise level measured by the sensor. State update ( Using the observations at the current moment and Kalman gain For the predicted state Make corrections to obtain the optimal state estimate for the current time step. Covariance update ( Update the covariance of the current state estimate. This reflects the uncertainty in the updated state estimate. These formulas together constitute an iterative process, enabling the system to continuously integrate new observation data and constantly optimize the estimates of states such as hemorrhage.

[0062] Through the above technical solution, this invention effectively solves the problem of the impact of multi-source data noise and uncertainty on the accuracy of blood loss estimation. The Kalman filter can intelligently fuse multi-source data such as suction flow, blood concentration, and residual blood volume. Through its iterative prediction and update mechanism, it significantly reduces errors caused by sensor noise and environmental interference, thereby outputting the optimal estimate of total blood loss, enabling doctors to obtain more accurate and stable real-time blood loss data. Based on this, combined with a time series prediction algorithm, the system can generate predicted blood loss values ​​for a future period based on historical blood loss data. This forward-looking predictive capability allows the early warning module to identify potential bleeding risks earlier, such as predicting that blood loss may rapidly exceed a preset threshold within a short period, thus issuing an early warning signal. This provides medical staff with valuable prediction time, enabling them to take more timely intervention measures, effectively avoiding emergencies caused by rapid blood loss accumulation, and greatly improving the precision of bleeding management and patient safety during laparoscopic liver resection.

[0063] In some embodiments of the present invention described above, the system can estimate the total blood loss during surgery in real time and issue an early warning signal when the blood loss or bleeding rate exceeds a preset threshold. However, such threshold-triggered early warning mechanisms are usually passive, providing alerts only when a problem occurs or is about to occur. This may not provide surgeons with sufficient time to make forward-looking decisions or take preventative measures, thereby affecting timely surgical intervention and patient safety.

[0064] To address this, the present invention further proposes that the aforementioned early warning module includes a trend prediction unit and a tiered alarm unit. The trend prediction unit is a key component of the early warning module, its function being to proactively predict the bleeding trend and amount over a future period by analyzing historical data and the current surgical status. This unit elevates traditional passive early warning to proactive prediction, providing the medical team with forward-looking information. Specifically, the trend prediction unit is based on historical bleeding volume sequences. In conjunction with the current surgical stage, deep learning models are used to predict future blood loss. The prediction results are then input into the tiered alarm unit. Historical bleeding volume sequence. This refers to the collection of bleeding data continuously recorded and stored by the system during surgery, arranged in chronological order. This data forms the basis for the predictive model to learn bleeding patterns, reflecting the patient's physiological responses and bleeding patterns at different surgical stages. The current surgical stage refers to information on the specific stage of the surgery, such as laparotomy, liver dissection, vascular ligation, lesion resection, and hemostasis. The bleeding risk and patterns differ significantly at different surgical stages; using this information as input to the predictive model helps the model more accurately understand and predict bleeding. A deep learning model is a machine learning algorithm capable of automatically learning complex features and patterns from large amounts of data. Here, it is used to process time-series data and multimodal inputs (such as historical bleeding volume and surgical stage) to capture the nonlinear relationship and potential patterns of bleeding volume changes over time. The deep learning model may include a Long Short-Term Memory (LSTM) network (…). )or Structure. Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network, particularly suitable for processing and predicting time series data. By introducing a gating mechanism, they effectively solve the gradient vanishing or exploding problems in traditional recurrent neural networks, thus enabling them to learn and remember long-term dependencies. They are very suitable for analyzing historical bleeding sequences. The structure is a deep learning model based on a self-attention mechanism. By processing all elements in a sequence in parallel and utilizing the attention mechanism to capture complex relationships between different time steps, it can efficiently process long sequence data and discover important temporal dependencies. (Future bleeding volume prediction value) The calculation formula given by this deep learning model is as follows: ,in For a well-trained deep learning network, These are network parameters.

[0065] The graded alarm unit is another crucial component of the early warning module. Its function is to issue alarms of different levels and forms based on the severity of bleeding and predictions. This graded mechanism helps avoid "alarm fatigue," ensuring that medical personnel can distinguish between emergencies and general situations and prioritize critical alarms. The graded alarm unit issues different levels of visual, auditory, or vibrational warning signals based on the degree to which total bleeding exceeds a threshold or the predicted outcome. Visual signals can be presented through color changes on the display, flashing icons, text prompts, etc., intuitively displaying the amount of bleeding, bleeding rate, or predicted risk level. Auditory signals can be beeps of different frequencies, volumes, or tones, or voice prompts, used to attract the attention of medical staff in the operating room environment. Vibrational signals can be emitted through wearable devices integrated with surgical instruments or worn by medical personnel, providing a silent but direct form of alertness, particularly suitable for delicate procedures requiring high concentration.

[0066] Through the above technical solution, this invention upgrades the traditional passive early warning mechanism to a mode combining proactive prediction and tiered alarm. The trend prediction unit utilizes a deep learning model, combining historical bleeding sequences and current surgical stage information, to accurately predict future bleeding volume, enabling surgeons to anticipate potential bleeding risks in advance. This forward-looking information allows the medical team ample time to adjust surgical strategies, prepare hemostasis measures, or administer blood transfusions before bleeding reaches a dangerous threshold, thus nipping potential complications in the bud. Simultaneously, the tiered alarm unit issues different levels of visual, auditory, or vibration warning signals based on the real-time total bleeding volume and the severity of the predicted results. This effectively avoids "alarm fatigue" that can result from a single alarm mode, ensuring the effective transmission of critical information and enabling medical staff to prioritize responses based on the urgency of the alarm. This significantly improves the safety and efficiency of the surgical process and reduces perioperative risks for patients.

[0067] However, in practical applications, due to differences between different devices, changes in ambient lighting, and differences in individual physiological characteristics, the images acquired by the image acquisition module may have size deviations, and the results measured by the blood concentration sensor may also have systematic errors. These factors may all affect the accuracy of the blood loss estimation results.

[0068] To address this, the present invention further proposes a real-time blood loss estimation and early warning system for laparoscopic liver resection, which includes a calibration module for calibrating the image acquisition module before the start of surgery. This calibration module uses a known-sized calibration object to determine the ratio between image pixels and actual dimensions. Simultaneously, it calibrates the blood concentration sensor, establishing a personalized absorbance-concentration relationship curve using blood samples of known concentrations. During image calibration, the actual size... With image pixel length The relationship is: ,in The scaling factor is calculated using a calibration material: , The actual length of the calibration object is known. The corresponding pixel length in the image; absorbance in blood concentration calibration. With concentration The relationship is: ,in and These are the coefficients obtained by fitting the calibration samples using the least squares method.

[0069] This calibration module is primarily used to calibrate critical measuring equipment before or periodically before system deployment, ensuring the accuracy and consistency of its measurement data. Its implementation can include a hardware interface and software algorithms. The hardware interface connects to calibration tools or receives calibration data, while the software algorithm executes the calibration process, calculates calibration parameters, and stores them internally for use by subsequent data processing modules. The calibration of the image acquisition module aims to establish a precise correspondence between image pixels and actual physical dimensions. Specifically, before surgery begins, a calibration object of known precise size is placed in the field of view of the image acquisition module. The system uses image processing algorithms to identify the pixel length of this calibration object in the image. And combined with the known actual length of the calibration object Calculate the scaling factor between image pixels and actual size. ,Right now This scaling factor This is then used to calculate the length of any pixel measured in the image. Convert to actual size Their relationship is This calibration process eliminates image size measurement errors caused by optical distortion, focal length variations, or individual differences between devices, ensuring the accuracy of subsequent 3D reconstruction and calculation of the blood coverage area. The calibration of the blood concentration sensor aims to establish a personalized and precise mapping between absorbance and blood concentration. Before the surgery begins, the blood concentration sensor is tested using a series of blood samples with known concentrations. For each blood sample with a known concentration, the sensor measures its absorbance. By collecting multiple sets of absorbance... and corresponding blood concentration For the given data points, regression analysis methods such as the least squares method can be used to fit a personalized absorbance-concentration relationship curve, which is usually expressed as a linear relationship. ,in and These are the calibration coefficients obtained through fitting. These coefficients will be stored in the data processing module and used to monitor the absorbance measured by the sensor in real time during the operation. Accurately converted to blood concentration This improves the accuracy of blood concentration measurement and reduces measurement deviations caused by individual sensor differences or environmental factors.

[0070] Through the above technical solutions, the calibration module precisely calibrates the image acquisition module and blood concentration sensor before the start of surgery, effectively solving the problem of measurement inaccuracies caused by equipment differences, environmental factors, or individual physiological characteristics. The calibration of the image acquisition module ensures a precise correspondence between pixel dimensions in the image and actual physical dimensions, providing a reliable geometric basis for subsequent 3D reconstruction and calculation of the blood-covered area volume. The calibration of the blood concentration sensor establishes a personalized and precise mapping relationship between absorbance and blood concentration, significantly improving the accuracy of blood concentration measurement. These calibration measures guarantee the accuracy of input data from the source, thereby making the calculated volume of aspirated pure blood and the residual blood volume in the surgical area more accurate, ultimately improving the accuracy of total blood loss estimation. More accurate blood loss estimation allows the early warning module to issue warning signals at more appropriate times, providing doctors with more reliable decision-making basis, thus effectively improving surgical safety and patient prognosis.

[0071] The present invention also provides an embodiment of a practical application of the method of the present invention: To verify the effectiveness, accuracy, and clinical applicability of the system and method described in this invention, a set of real-time monitoring data from a laparoscopic left lateral segmentectomy of the liver performed in the hepatobiliary surgery department of a tertiary hospital is provided. This data fully demonstrates the entire workflow of the system, from preoperative calibration, intraoperative real-time monitoring and calculation, multi-source data fusion processing to dynamic early warning.

[0072] I. Application Scenarios and Case Data: This study selected a 52-year-old male patient scheduled for laparoscopic left lateral segmentectomy of the liver at a tertiary-level Class A hospital. The patient weighed 75 kg and was 172 cm tall. Preoperative diagnosis included hepatocellular carcinoma of the left lateral lobe of the liver (approximately 5 cm in diameter), Child-Pugh class A. Routine preoperative examinations showed a hemoglobin concentration of 135 g / L.

[0073] The surgical team used the real-time blood loss estimation and early warning system for laparoscopic liver resection described in this invention to monitor the entire operation. The core parameters of the system are preset as follows: Reference pressure value ( ): 12 mmHg (abdominal pressure after preoperative pneumoperitoneum stabilization); Warning threshold: T0 = 400 mL (total bleeding volume), and bleeding rate trend prediction is enabled simultaneously.

[0074] II. Preoperative calibration and preparation: Before the surgery begins, the circulating nurse assists in completing the system's standardized calibration process.

[0075] 1. Calibration of the image acquisition module: Given the actual length as The checkerboard-patterned marker is placed approximately 5-8 cm in front of the laparoscopic lens to simulate the intraoperative operating distance. The system automatically captures images of the marker and uses an edge detection algorithm to identify the pixel length of the marker in the image. .

[0076] The system calculates the scaling factor according to the formula. : This scaling factor is stored in the system and used to subsequently convert the two-dimensional pixel dimensions in the image into actual physical dimensions.

[0077] 2. Calibration of the blood concentration sensor: A small amount of anticoagulated venous blood was collected from the patient before surgery and diluted with normal saline to known concentrations of 20%, 40%, 60%, and 80%, respectively. The sampling tube of the blood concentration sensor was sequentially immersed in samples of different concentrations, and the corresponding multi-wavelength absorbance values ​​were recorded. The system uses the least squares method to perform linear fitting on the absorbance-concentration data to obtain a personalized calibration curve. In this example, the calibration coefficients obtained from the fitting are: , This calibration relationship will replace the generic curve inside the sensor for real-time blood concentration calculation during this surgery.

[0078] III. Intraoperative Real-time Monitoring and Calculation: The surgery proceeded as planned. During the liver transection phase, transient bleeding occurred. The entire process, from the onset of the bleeding to its control, was fully documented.

[0079] 1. Data Acquisition and Anomaly Triggering: At time point T1, bleeding occurred on the cross-section of the liver parenchyma, and the system's sensors captured the following data in real time: 2. Preliminary processing of single-source data: Blood concentration calculation: According to the calibrated formula, at time T1, the blood concentration is... for: Initial concentration measured by blood concentration sensor The value was 0.85. Meanwhile, the non-invasive hemoglobin monitoring module read the value in real time as follows: =13.5g / dL, system preset reference hemoglobin concentration =15.0 g / dL. Therefore, the actual blood concentration after calibration is: Instantaneous pure blood volume drawn: At time T1, the volume of pure blood drawn per second is: 3. Image analysis and 3D reconstruction: At time T1, the two-dimensional image captured by the image acquisition module is processed by a deep learning semantic segmentation network (such as U-Net) to accurately segment the blood-covered area of ​​the liver wound (figure omitted). At the same time, the three-dimensional reconstruction unit reconstructs the three-dimensional surface model of the liver cross-section in real time using a stereo vision algorithm.

[0080] The image segmentation unit maps the two-dimensional blood-covered region onto a three-dimensional surface model through coordinate transformation, thus obtaining the three-dimensional surface of the blood-covered region. The system estimates the average blood thickness based on the local curvature of the region and a pre-defined blood spreading model. The volume of residual blood was calculated by performing a surface integral over the region. : 4. Multi-source data fusion and correction: Pressure correction: Real-time intra-abdominal pressure at time T1 , and reference pressure Consistent, therefore no correction is needed. Pressure correction factor. In this example, it is set to 0.02.

[0081] Flush Correction: Time T1 Therefore, no correction is needed. Deduction coefficient. Set it to 0.15.

[0082] Electrocoagulation Interference Suppression: At time T3, the electrocoagulation device is activated. Upon receiving the signal, the data processing module immediately initiates a smoke suppression algorithm to preprocess the video stream, effectively removing the interference of smoke generated by electrocoagulation on image segmentation and ensuring the accuracy of residual blood volume estimation.

[0083] 5. Kalman filter fusion and optimal estimation: The system uses the calculated suction flow rate (18 mL / s), blood concentration (49.1%), and residual blood volume (2.15 mL) at time T1 as observed values. Input Kalman filter. Filter state vector. It includes total blood loss, bleeding rate, and residual blood volume.

[0084] After prediction and update iterations (the 30th iteration in this example), the filter outputs the optimal estimate at time T1: Optimal estimate of total blood loss: 125.3 mL (This value is the cumulative amount from the start of the surgery to the present, and is derived by the filter after taking into account the historical aspirated pure blood volume and residual blood volume).

[0085] Optimal estimated bleeding rate: 5.2 mL / min IV. Tiered Early Warning and Dynamic Response: At time T10 (50 seconds after the onset of bleeding), the bleeding was not completely controlled, and the optimal estimate of the total bleeding volume output by the Kalman filter reached 385 mL.

[0086] At this point, the trend prediction unit (based on an LSTM network) calculates the historical bleeding volume sequence over the past 10 minutes. In the current stage of "liver parenchyma transection" surgery, predict the increase in blood loss within the next 30 seconds. This means that without timely intervention, the total blood loss will exceed 420 mL within 30 seconds.

[0087] The tiered alarm unit triggered tiered alarms based on the relationship between the current total blood loss (385 mL) and the preset threshold (400 mL), as well as predictions of future risks. Level 1 Warning (Attention Level, flashing yellow light): Triggered when total blood loss exceeds 300mL, and is continuously displayed in the corner of the screen.

[0088] Level 2 Warning (Action Level, Flashing Orange Light + Intermittent Beeping): At time T10, the current total blood loss (385mL) has not yet exceeded 400mL, but the trend prediction unit predicts that it will soon exceed the threshold. The system determines this as a high-risk trend and immediately upgrades it to Level 2 Warning. The surgeon looks up and sees the orange border on the monitor and hears the intermittent beeping, realizing that the bleeding is at risk of getting out of control.

[0089] Level 3 Warning (Crisis Level, Flashing Red Light + Continuous Alarm): The surgeon will immediately prepare to increase hemostasis efforts. However, if the bleeding continues to rise and exceeds 450mL (the system's second-level hard threshold), the system will issue a continuous, sharp alarm.

[0090] V. Verification of Hemostatic Effect and Case Records: Upon receiving the Level 2 warning, the surgeon quickly adjusted the procedure and precisely applied bipolar electrocoagulation to stop the bleeding at the bleeding point.

[0091] Electrocoagulation status monitoring: The electrocoagulation equipment is reactivated. The data processing module enters "electrocoagulation suppression mode," pausing the bleeding accumulation logic based on image segmentation.

[0092] Hemostasis assessment: After electrocoagulation was stopped, the system continuously monitored five video frames and found no significant increase in the area covered by blood, and the optimal estimated bleeding rate decreased from 5.2 mL / min to 0.8 mL / min. The system determined that the bleeding had been effectively controlled and resumed the accumulation of bleeding volume, but the accumulation rate had significantly decreased at this point.

[0093] Post-surgery, the system automatically generates a standardized case report for this bleeding event and records it in the local database: VI. Conclusion: This clinical application example fully demonstrates the workflow of the system described in this invention in a real surgical scenario. By integrating 3D image reconstruction, multi-source sensor fusion, Kalman filter optimal estimation, and deep learning trend prediction, the system successfully issued a secondary warning approximately 25 seconds before the total blood loss was about to exceed a preset threshold, providing the surgeon with valuable decision-making and reaction time. Ultimately, the patient's total intraoperative blood loss was 412 mL, slightly higher than the initial threshold, but still within the surgeon's control, and the advance warning prevented further accumulation of blood loss. This case fully validates the significant clinical value of this system in improving the accuracy of blood loss estimation, achieving dynamic risk warning, and assisting intraoperative decision-making.

[0094] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A real-time blood loss estimation and early warning system for laparoscopic liver resection, characterized in that, include: The image acquisition module is used to acquire image sequences of the laparoscopic surgical field in real time; A suction flow sensor is used to detect the flow rate of the suction fluid in the suction tube in real time. A blood concentration sensor is used to detect the blood concentration in the liquid in the suction tubing in real time. The data processing module is connected to the image acquisition module, the suction flow sensor, and the blood concentration sensor, and includes a three-dimensional reconstruction unit, an image segmentation unit, and a bleeding volume calculation unit. The three-dimensional reconstruction unit is used to reconstruct a three-dimensional surface model of the surgical area based on the image sequence; The image segmentation unit is used to segment the blood-covered area from the image sequence and map it onto the three-dimensional surface model to obtain the distribution information of blood in three-dimensional space, including the boundary coordinates of the blood-covered area and the local blood thickness. The bleeding volume calculation unit is used to calculate the volume of pure blood aspirated based on the suction flow rate and blood concentration, and to calculate the volume of residual blood in the surgical area based on the distribution of blood in three-dimensional space. Then, it calculates the total bleeding volume, which is the sum of the volume of pure blood aspirated and the volume of residual blood. The early warning module, connected to the data processing module, is used to compare the total bleeding volume or bleeding rate with a preset threshold, and issue an early warning signal when the threshold is exceeded.

2. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, The blood concentration sensor includes a multi-wavelength optical detection unit and a bubble detection unit. The multi-wavelength optical detection unit uses at least two light sources of different wavelengths to measure the absorbance of the attracted liquid and calculates the blood concentration according to a preset hemoglobin concentration-absorbance relationship curve. The bubble detection unit monitors bubbles in the suction tube in real time. When a bubble is detected, the data processing module removes the concentration data at the corresponding time or uses interpolation compensation. Specifically, the blood concentration Calculated using the following formula: ,in The absorbance measured by the multi-wavelength optical detection unit. and This is a preset calibration coefficient, which is obtained through experimental calibration and is used to linearly map absorbance to blood concentration.

3. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, The three-dimensional reconstruction unit reconstructs the three-dimensional surface model of the surgical area from the image sequence using stereo vision or structured light projection technology. The image segmentation unit uses a deep learning semantic segmentation network to segment the blood area and maps the two-dimensional segmentation results onto the three-dimensional surface model through coordinate transformation, thereby calculating the surface area and corresponding volume of the blood-covered area. The residual blood volume The calculation is performed by integrating the three-dimensional surface of the blood-covered area: ,in For the three-dimensional surface of the blood-covered area, This refers to the local blood thickness, which is estimated based on the three-dimensional surface curvature and a pre-defined blood spreading model. This represents a small area element.

4. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, It also includes a pressure monitoring module for real-time monitoring of intra-abdominal pressure; the data processing module adjusts the geometric parameters of the three-dimensional surface model according to changes in intra-abdominal pressure, or corrects the calculation results of residual blood volume according to pressure changes; Specifically, the corrected residual blood volume It can be obtained through the following formula: ,in The volume of residual blood before correction. For real-time intra-abdominal pressure, For reference pressure value, This is the pressure correction factor, obtained by fitting experimental data.

5. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, It also includes a flushing flow sensor for real-time detection of the flushing fluid flow rate in the flushing pipe; the data processing module also receives flushing flow data and corrects changes in the blood coverage area in the image based on the flushing flow data to eliminate interference from the flushing operation on the estimation of residual blood volume. The corrected formula is: ,in For the uncorrected residual blood volume, For flushing flow rate, The subtraction factor for residual blood volume is used for flushing; this factor is calibrated experimentally. This indicates the cumulative volume of the flushing liquid.

6. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, It also includes an electrosurgical equipment monitoring module for acquiring the usage status and power parameters of the electrocoagulation equipment; the data processing module determines whether the bleeding has been stopped based on the usage of the electrocoagulation equipment and adjusts the bleeding volume estimation strategy accordingly, or corrects the image segmentation results during electrocoagulation to avoid smoke interference. Specifically, when the electrocoagulation equipment is activated, the data processing module uses a smoke suppression algorithm to preprocess the image to reduce the impact of smoke on segmentation; when the electrocoagulation equipment is turned off, if the bleeding area does not change significantly, it is determined that the bleeding has been controlled, and the accumulation of bleeding volume is stopped accordingly.

7. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, It also includes a non-invasive hemoglobin monitoring module for real-time acquisition of the patient's current hemoglobin concentration; the data processing module dynamically calibrates the conversion curve of the blood concentration sensor based on the hemoglobin concentration to eliminate the influence of individual differences on blood concentration measurement; calibrated blood concentration Calculated using the following formula: ,in The initial concentration measured by the blood concentration sensor. This refers to the hemoglobin concentration acquired in real time by the non-invasive hemoglobin monitoring module. This is the preset reference hemoglobin concentration.

8. The real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, The data processing module uses a Kalman filter to fuse multi-source data such as suction flow, blood concentration, and residual blood volume, outputting the optimal estimate of total bleeding volume, and generates a predicted bleeding volume for a future period based on a time series prediction algorithm; the state vector of the Kalman filter includes total bleeding volume, bleeding rate, and residual blood volume; The recursive formula for the Kalman filter includes: State prediction: ; Covariance prediction: ; Kalman gain: ; Status Update: ; Covariance update: ; in This is the system state vector, including total bleeding, etc. This is the state transition matrix; For control matrix; For control input; It is the covariance matrix; For process noise covariance; The observation matrix; To observe the noise covariance; These are observed values, including suction flow rate, blood concentration, and residual blood volume; For Kalman gain.

9. A real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, The early warning module includes a trend prediction unit and a graded alarm unit. The trend prediction unit predicts future bleeding volume based on historical bleeding volume sequences and the current surgical stage using a deep learning model, and inputs the prediction results into the graded alarm unit. The graded alarm unit issues different levels of visual, auditory, or vibration warning signals based on the degree to which the total bleeding volume exceeds the threshold or the prediction results. The predicted value of future bleeding Provided by the deep learning model: ,in This is a time series of historical bleeding volumes. For a well-trained deep learning network, These are network parameters, and the network includes a Long Short-Term Memory network or... structure.

10. A real-time blood loss estimation and early warning system for laparoscopic liver resection according to claim 1, characterized in that, It also includes a calibration module, which is used to calibrate the image acquisition module before the operation begins, and to determine the ratio between image pixels and actual size using a calibrator of known size; at the same time, it calibrates the blood concentration sensor and uses blood samples of known concentration to establish a personalized absorbance-concentration relationship curve; In image calibration, actual size With image pixel length The relationship is: ,in The scaling factor is calculated using a calibration material. , The actual length of the calibration object is known. The length of the corresponding pixel in the image; In blood concentration calibration, absorbance With concentration The relationship is: ,in and These are the coefficients obtained by fitting the calibration samples using the least squares method.

Images