Multi-mode intraoperative bleeding amount real-time estimation method and system

Abstract

The invention belongs to the technical field of medical instruments, and discloses a multi-modal intraoperative bleeding amount real-time estimation method and system, the volume proportion of blood in a mixed solution is calculated through multi-band absorbance information, and the interference of irrigation solution dilution on blood loss amount evaluation is effectively solved; the blood amount in the suction bottle is calculated by combining the weight of the mixed liquid and the suction flow, so that the accuracy of liquid amount measurement is improved; the gauze is tracked and recognized through the surgical field video, the blood content is estimated, and automatic evaluation of the blood content of the gauze is achieved; finally, the Kalman filtering algorithm is used for fusing multi-source data, the accumulated blood loss amount and blood loss rate are dynamically calculated, the real-time performance and robustness of estimation are remarkably improved, and second-level blood loss rate information can be provided. In addition, according to the method, graded warning can be carried out according to the blood loss condition, a blood transfusion pump and an anaesthesia machine are controlled in a linkage mode, and auditable logs are recorded; therefore, the accuracy, the real-time performance, the intelligent level and the traceability of intraoperative blood loss volume evaluation are comprehensively improved.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and more specifically, to a multimodal method and system for real-time estimation of intraoperative blood loss. Background Technology

[0002] Accurately assessing intraoperative blood loss during surgical procedures is crucial for timely adjustments to treatment plans and ensuring patient safety. Currently, clinically determining intraoperative blood loss primarily relies on the surgeon's experience, suction bottle volume readings, the weight difference of surgical gauze, or subjective visual judgment. However, these traditional assessment methods have many limitations, leading to significant errors in blood loss estimation.

[0003] Specifically, during surgery, factors such as the introduction of irrigation fluid, the formation of foam or clots in the blood, contamination of the suction bottle's observation window, metallic reflections from surgical instruments, and color casts in the operating room lighting can all severely interfere with the accurate assessment of blood loss. For example, the addition of irrigation fluid dilutes the blood, making the volume of the mixture in the suction bottle inaccurate to reflect the amount of pure blood; the presence of foam and clots affects the accuracy of volume readings; and window contamination and metallic reflections interfere with visual judgment and optical measurements. Furthermore, most existing blood loss assessment methods are single-channel measurements, such as relying solely on weighing or colorimetry, which makes them lack sufficient robustness in the face of complex surgical environments and multiple interfering factors. Single-channel methods cannot comprehensively capture multi-dimensional information during the blood loss process, thus limiting their accuracy and reliability.

[0004] Furthermore, existing technologies struggle to provide blood loss rate information down to the second level, a significant shortcoming for emergencies requiring rapid response. Simultaneously, traditional assessment methods often lack traceable audit records, making it difficult to provide objective evidence in the event of medical disputes. Therefore, there is an urgent need for a solution that overcomes these limitations, enabling real-time, robust, and traceable estimation of blood loss volume and rate under complex surgical procedures and multi-interference conditions, and that can be integrated with anesthesia / transfusion equipment to reduce misjudgment of transfusion timing and perioperative risks.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] The purpose of this application is to provide a multimodal method and system for real-time estimation of intraoperative blood loss, which aims to solve the problems of poor accuracy, insufficient robustness, lack of real-time performance and traceability in existing intraoperative blood loss assessment methods.

[0007] Firstly, this application provides a multimodal real-time estimation method for intraoperative blood loss, the method comprising the following steps: A1. Real-time acquisition of multimodal monitoring data; the multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, weight of the mixture, aspiration flow rate, surgical field video, and patient physiological parameters; A2. Calculate the volume ratio of blood in the mixture based on the multi-band absorbance information; A3. Calculate the amount of blood in the suction bottle based on the weight of the mixture, the suction flow rate, and the volume ratio; A4. Based on the surgical field video tracking and identification of the gauze, and based on the image features of the gauze, estimate the total blood content of the gauze; A5. Based on the amount of blood in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze, calculate the cumulative blood loss and blood loss rate using the Kalman filter algorithm; A6. Based on the cumulative blood loss, the blood loss rate, and the patient's physiological parameters, perform graded alarms, implement coordinated control of the blood transfusion pump and anesthesia machine, and record auditable logs.

[0008] Preferably, the multi-band absorbance information includes absorbance in the 630nm, 660nm, 810nm, 850nm and 940nm bands; The patient's physiological parameters include blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output.

[0009] Preferably, in step A1, the step of obtaining the multi-band absorbance information includes: B1. Acquire the detection light intensity and the corresponding reference light intensity of multiple wavelength bands; the detection light intensity is the light intensity of the incident light after passing through the mixed liquid, and the reference light intensity is the light intensity of the incident light after passing through a preset optical attenuation element. B2. Divide the probe intensity of each band by the corresponding reference intensity to obtain the normalized probe intensity of each band. B3. Calculate the apparent absorbance of each band based on the normalized probe light intensity and the corresponding pre-calibrated reference light intensity. B4. Based on the type of irrigation fluid currently used, look up the table to obtain the baseline absorbance of the irrigation fluid for each band, and subtract the corresponding baseline absorbance of the irrigation fluid from the apparent absorbance of each band to obtain the absorbance after deducting the baseline. B5. Based on the pre-measured hematocrit and hemoglobin concentration of the patient's pure blood, and in conjunction with the type of irrigation fluid currently used, the scattering correction factor for each band is obtained from the table and used to perform scattering correction on the absorbance after baseline subtraction, thus obtaining the final absorbance for each band.

[0010] Preferably, after step B1 and before step B2, the following steps are further included: Obtain the incident light intensity and system temperature for each wavelength band; Compare the rate of change of the reference light intensity of each band relative to its respective pre-calibrated baseline value with a preset rate of change threshold to determine whether significant drift has occurred; calculate the real-time transmittance of the 940nm band based on the probe light intensity of the 940nm band and the corresponding incident light intensity, and compare the rate of decrease of the real-time transmittance relative to the initial transmittance of the 940nm band at the beginning of the operation with a preset rate of decrease threshold to determine whether window contamination has occurred. If a significant drift occurs, the reference light intensity of each band is compared with its respective pre-calibrated baseline value, the drift factor of each band is calculated, and the probe light intensity of each band is divided by the corresponding drift factor to achieve drift compensation of the probe light intensity. If window contamination occurs, the ratio of the real-time transmittance of the 940nm band to the corresponding clean state transmittance is calculated as a contamination factor. The detection light intensity of each band is divided by this contamination factor to achieve contamination compensation of the detection light intensity. Based on the system temperature, the temperature compensation coefficient is determined by referring to a table, which is used to compensate for the temperature of the probe light intensity in each band.

[0011] Preferably, step A2 includes: A201. Based on the absorbance of each wavelength band, and using the Lambert-Beer law, calculate the corresponding hemoglobin concentration for each wavelength band. A202. Weighted summation of hemoglobin concentrations across different bands yields a composite hemoglobin concentration; A203. Based on the overall hemoglobin concentration, the pre-measured hemoglobin concentration of the patient's pure blood, and the hemoglobin concentration of the currently used irrigation fluid, calculate the volume ratio of blood in the mixture based on the dilution principle.

[0012] Preferably, step A3 includes: A301. Obtain the volume of the mixture based on the weight of the mixture and the suction flow rate; A302. Multiply the volume ratio by the volume of the mixture to obtain the amount of blood in the aspiration bottle.

[0013] Preferably, step A301 includes: Based on the pre-measured hematocrit and hemoglobin concentration of the patient's pure blood, and in conjunction with the type of irrigation fluid currently being used, the reference density of the mixed solution is obtained by referring to a table. The mass change rate is calculated based on the weight of the mixture, and the mass flow rate is calculated based on the suction flow rate and the reference density; The mass change rate and the mass flow rate are compared to determine whether the suction flow rate is abnormal. If the suction flow rate is abnormal, the volume of the mixture is calculated based on the weight of the mixture and the reference density; If the suction flow rate is normal, the volume of the mixture is calculated based on the suction flow rate.

[0014] Preferably, step A4 includes: A401. Based on the target recognition algorithm combined with the ReID algorithm, each gauze is tracked and identified from the surgical field video; A402. Extract the image features of each identified gauze, and input them into a pre-trained regression model to estimate the blood content of each gauze piece; the image features include color histogram, mean saturation, and texture features; A403. Calculate the sum of the blood content of each individual piece of the identified gauze to obtain the total blood content of the gauze.

[0015] Preferably, step A5 includes: The cumulative blood loss and blood loss rate form a state vector, the blood volume in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze form an observation vector, and the transfusion rate and fluid resuscitation rate form a control input vector. The cumulative blood loss and blood loss rate are calculated using the Kalman filter algorithm.

[0016] Secondly, this application provides a multimodal real-time intraoperative blood loss estimation system, the system comprising: The multimodal data acquisition module is used to acquire multimodal monitoring data in real time; the multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, weight of the mixture, aspiration flow rate, surgical field video, and patient physiological parameters; The proportion calculation module is used to calculate the volume proportion of blood in the mixture based on the multi-band absorbance information. The blood volume calculation module is used to calculate the blood volume in the suction bottle based on the weight of the mixture, the suction flow rate, and the volume ratio. The gauze total blood content estimation module is used to identify the gauze based on the surgical field video tracking and to estimate the total blood content of the gauze based on the image features of the gauze. The blood loss calculation module is used to calculate the cumulative blood loss and blood loss rate based on the amount of blood in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze, using a Kalman filter algorithm. The decision output module is used to perform graded alarms, coordinate the control of the blood transfusion pump and the anesthesia machine, and record auditable logs based on the cumulative blood loss, the blood loss rate, and the patient's physiological parameters.

[0017] Beneficial Effects: This application provides a multimodal real-time intraoperative blood loss estimation method and system. It accurately calculates the volume ratio of blood in the mixed fluid using multi-band absorbance information, effectively solving the interference of irrigation fluid dilution on blood loss assessment. Combining the weight of the mixed fluid and the suction flow rate to calculate the blood volume in the suction bottle improves the accuracy of fluid volume measurement. By tracking and identifying gauze in the surgical field via video and estimating its blood content, it overcomes the shortcomings of traditional gauze weighing methods and achieves automated assessment. Finally, using a Kalman filter algorithm to fuse multi-source data, it dynamically calculates the cumulative blood loss and blood loss rate, significantly improving the real-time performance and robustness of the estimation, providing blood loss rate information at the second level. Furthermore, this method can also provide graded alarms based on blood loss status, link the transfusion pump and anesthesia machine, and record auditable logs. This effectively addresses the shortcomings of existing technologies, such as difficulty in providing rapid response and lack of traceable audit records, reducing the risk of misjudging transfusion timing and perioperative complications, and comprehensively improving the accuracy, real-time performance, intelligence, and traceability of intraoperative blood loss assessment. Attached Figure Description

[0018] Figure 1 A flowchart of a multimodal intraoperative blood loss real-time estimation method provided in this application.

[0019] Figure 2 This is a schematic diagram of a multimodal intraoperative blood loss real-time estimation system provided in this application.

[0020] Labeling Explanation: 1. Multimodal Data Acquisition Module; 2. Proportion Calculation Module; 3. Blood Volume Calculation Module; 4. Gauze Total Blood Content Estimation Module; 5. Blood Loss Calculation Module; 6. Decision Output Module. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Please refer to Figure 1 This application discloses a multimodal real-time intraoperative blood loss estimation method in some embodiments, the method comprising the following steps: A1. Real-time acquisition of multimodal monitoring data; multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, mixture weight, aspiration flow rate, surgical field video, and patient physiological parameters; A2. Calculate the volume ratio of blood in the mixture based on the multi-band absorbance information; A3. Calculate the amount of blood in the suction bottle based on the weight of the mixture, the suction flow rate, and the volume ratio; A4. Identify the gauze by tracking the surgical field video, and estimate the total blood content of the gauze based on its image features; A5. Based on the blood volume in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze, use the Kalman filter algorithm to calculate the cumulative blood loss and blood loss rate; A6. Based on the cumulative blood loss, blood loss rate, and patient physiological parameters, implement graded alarms, coordinate the control of blood transfusion pumps and anesthesia machines, and record auditable logs.

[0024] The multimodal real-time intraoperative blood loss estimation method proposed in this application aims to overcome the limitations of traditional methods by integrating multiple monitoring data to achieve real-time, accurate, and robust estimation of intraoperative blood loss. Multimodal monitoring data refers to various types of data collected in real time during the operation, including multi-band absorbance information of the mixture in the aspiration bottle, mixture weight, aspiration flow rate, surgical field video, and patient physiological parameters. These data reflect the patient's blood loss and physiological state from different dimensions. Multi-band absorbance information is used to quantify the concentration of blood in the mixture; mixture weight and aspiration flow rate are used to assess the total change in fluid volume in the aspiration bottle; surgical field video is used to identify and analyze blood on the gauze; and patient physiological parameters provide real-time feedback on the patient's overall condition. Through comprehensive analysis of these multimodal data, this application can more comprehensively and accurately assess intraoperative blood loss and, based on this, provide intelligent decision support and coordinated control.

[0025] Specifically, in multimodal intraoperative blood loss real-time estimation methods, the first step is to acquire multimodal monitoring data in real time. For example, multi-band spectral sensors can be installed on the suction bottle to obtain multi-band absorbance information of the mixed fluid; a weighing sensor can monitor the weight of the mixed fluid in the suction bottle in real time; a flow meter can measure the suction flow rate in real time; a high-definition camera can acquire surgical field video; and a patient monitor can be connected to obtain physiological parameters such as blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output. This data is collected in real time and transmitted to the processing unit for subsequent analysis.

[0026] Furthermore, after obtaining multi-band absorbance information, it is necessary to calculate the volume ratio of blood in the mixture based on this information. For example, the Lambert-Beer law can be used to estimate the hemoglobin concentration in the mixture by measuring the absorbance of the mixture at different wavelengths. Then, by combining the hemoglobin concentrations of pure blood and irrigation fluid, the volume ratio of blood in the mixture can be calculated.

[0027] Subsequently, based on the weight of the mixture, the suction flow rate, and the calculated volume ratio, the amount of blood in the suction bottle can be calculated. For example, the total volume of the mixture can be estimated by the changes in the weight of the mixture and the suction flow rate, and then multiplied by the volume ratio of blood in the mixture to obtain the amount of pure blood in the suction bottle.

[0028] Simultaneously, to comprehensively assess blood loss, bleeding on the gauze also needs to be considered. By analyzing the surgical field video, the gauze can be tracked and identified, and its total blood content can be estimated based on its image features, such as color, saturation, and texture. For example, a deep learning model can be trained to recognize gauze in the video and quantify its blood content based on the degree of color change in the gauze.

[0029] After obtaining the blood volume in the suction bottle and the total blood content in the gauze, the cumulative blood loss and blood loss rate can be calculated using the Kalman filter algorithm. The Kalman filter algorithm can fuse multi-source data to provide real-time, dynamic estimation of blood loss, effectively reducing measurement noise and uncertainty. For example, the blood volume in the suction bottle, the weight of the mixed solution, the suction flow rate, and the total blood content in the gauze can be used as observations, while the cumulative blood loss and blood loss rate can be used as state variables for state estimation via Kalman filtering.

[0030] Finally, based on the calculated cumulative blood loss, blood loss rate, and patient physiological parameters, the system can implement tiered alarms, coordinated control of the transfusion pump and anesthesia machine, and record auditable logs. For example, when the cumulative blood loss or blood loss rate exceeds different preset thresholds, the system can issue alarms of different levels to alert medical staff; simultaneously, it can automatically adjust the transfusion rate of the transfusion pump or the depth of anesthesia of the anesthesia machine according to the blood loss situation to maintain the patient's physiological stability; all key data and operation logs will be recorded for subsequent auditing and traceability.

[0031] The multimodal real-time intraoperative blood loss estimation method proposed in this application significantly improves the accuracy and robustness of intraoperative blood loss assessment by integrating multimodal monitoring data. Traditional methods often rely on a single data source, making them susceptible to various interference factors and resulting in large assessment errors. For example, the volume reading of the suction bottle alone cannot distinguish between blood and irrigation fluid, and the weight of gauze alone cannot reflect the blood loss rate in real time. This application constructs a comprehensive data acquisition system by simultaneously acquiring multi-band absorbance information of the mixture in the suction bottle, the weight of the mixture, the suction flow rate, surgical field video, and the patient's physiological parameters.

[0032] Specifically, multi-band absorbance information can accurately calculate the volume ratio of blood in the lavage fluid mixture, effectively solving the interference of lavage fluid dilution on blood loss assessment. The combination of lavage fluid weight and aspiration flow rate provides accurate data on changes in the total volume of fluid in the aspiration bottle. Surgical field video tracking and blood content estimation of gauze compensate for blood loss portions not covered by aspiration bottle data. The introduction of patient physiological parameters closely integrates blood loss assessment with the patient's overall physiological state, providing a more comprehensive basis for clinical decision-making.

[0033] Furthermore, this application employs a Kalman filter algorithm to fuse multi-source data, effectively handling data noise and uncertainty, and achieving real-time, dynamic, and high-precision estimation of cumulative blood loss and blood loss rate. This is a significant advantage compared to traditional methods that only provide lagging or coarse information on blood loss. Real-time acquisition of the blood loss rate enables medical staff to detect and intervene in bleeding more promptly, reducing perioperative risks.

[0034] Furthermore, this application implements tiered alarm systems, integrated control of blood pumps and anesthesia machines, and auditable log recording. The tiered alarm mechanism provides differentiated alerts based on the severity of blood loss, preventing excessive or insufficient intervention. The integrated control with blood pumps and anesthesia machines automates and intelligently manages and treats blood loss, reducing the workload of medical staff and improving treatment efficiency and safety. The auditable log recording provides objective and traceable evidence of the medical process, helping to improve medical quality management and address potential medical disputes.

[0035] In summary, this application overcomes the shortcomings of existing technologies in terms of accuracy, real-time performance, robustness, and traceability by using multimodal data fusion, advanced algorithm application, and intelligent linkage control. It provides a comprehensive, efficient, and reliable solution for real-time estimation of intraoperative blood loss, significantly improving patient surgical safety and medical quality.

[0036] Preferably, the multi-band absorbance information may include absorbance in the 630nm, 660nm, 810nm, 850nm and 940nm bands; Patient physiological parameters may include blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output.

[0037] Specifically, regarding the selection of multi-band absorbance information, the 630nm and 660nm bands are highly sensitive to changes in hemoglobin concentration in blood, effectively reflecting blood concentration information. At the 810nm and 850nm bands, absorbance is less affected by the oxygenation state of hemoglobin, thus providing a more stable reflection of total hemoglobin concentration and reducing measurement errors caused by fluctuations in oxygenation state. The 940nm band exhibits weak hemoglobin absorption and is primarily used to reflect the scattering characteristics of the mixture, and can also be used to detect contamination at the aspiration bottle window, ensuring the accuracy of absorbance measurements.

[0038] The selection of patient physiological parameters aims to provide comprehensive information on the patient's condition. Blood pressure and heart rate are key indicators reflecting the patient's circulatory system status, directly indicating blood volume and cardiovascular compensatory capacity. End-tidal carbon dioxide partial pressure reflects ventilation and perfusion, indirectly indicating tissue oxygenation and metabolic levels. Urine output, as an important indicator of renal perfusion and fluid balance, reflects the patient's overall circulatory function and organ perfusion.

[0039] This application's solution, by clearly defining the specific band selection for multi-band absorbance information, can more accurately capture changes in hemoglobin concentration in the mixed solution. Simultaneously, through the combination of specific bands, it effectively reduces the interference of oxygenation status and scattering effects on the measurement results, thereby improving the accuracy of blood volume ratio calculation. Furthermore, by incorporating patient physiological parameters such as blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output, the system can make a comprehensive judgment based on the patient's real-time physiological responses after calculating cumulative blood loss and blood loss rate. This multi-dimensional data integration makes the tiered alarm system, the coordinated control of the transfusion pump and anesthesia machine, and the recording of auditable logs more scientific and personalized, avoiding the potential bias of making decisions solely based on blood loss.

[0040] Through the above technical solutions, the present application can obtain more representative and accurate multi-band absorbance information, thereby improving the calculation accuracy of the volume ratio of blood in the mixed solution. At the same time, by incorporating key patient physiological parameters such as blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output, the system can more comprehensively evaluate the overall physiological state and compensatory ability of the patient when performing hierarchical alarm and linkage control, thereby providing more timely and accurate clinical decision support, effectively improving the reliability and safety of the intraoperative blood loss real-time estimation method.

[0041] Preferably, in step A1, the step of obtaining multi-band absorbance information may include: B1. Obtain the detection light intensity and the corresponding reference light intensity of multiple bands; the detection light intensity is the light intensity after the incident light passes through the mixed solution, and the reference light intensity is the light intensity after the incident light passes through a preset optical attenuation element; B2. Divide the detection light intensity of each band by the corresponding reference light intensity to obtain the normalized detection light intensity of each band; B3. Calculate the apparent absorbance of each band according to the normalized detection light intensity of each band and the corresponding pre-calibrated reference absorbance; B4. According to the type of the current lavage fluid used, look up the baseline absorbance of the lavage fluid for each band in a table, and subtract the corresponding baseline absorbance of the lavage fluid from the apparent absorbance of each band to obtain the absorbance after baseline subtraction; B5. According to the hematocrit and hemoglobin concentration of the patient's pure blood measured in advance, combined with the type of the current lavage fluid used, look up the scattering correction factor for each band in a table, and use it to perform scattering correction on the absorbance after baseline subtraction to obtain the final absorbance of each band.

[0042] Specifically, in step B1, the detection light intensity refers to the light intensity received by the detector after the incident light of a specific band penetrates the mixed solution in the suction bottle. The reference light intensity refers to the light intensity received by the detector after the incident light of the same band penetrates a preset optical attenuation element (such as a neutral density filter). By simultaneously obtaining the detection light intensity and the reference light intensity, the influence of systematic errors such as light source fluctuations and detector response changes on the measurement results can be effectively eliminated.

[0043] Further, in step B2, dividing the detection light intensity of each band by the corresponding reference light intensity aims to normalize the light intensity signal. This normalization operation can further compensate for the light source intensity drift and detector sensitivity change, thereby obtaining a more stable and accurate normalized detection light intensity.

[0044] In step B3, the apparent absorbance of each wavelength band is calculated based on the normalized probe light intensity and the corresponding pre-calibrated reference light intensity. The reference light intensity can be obtained by pre-normalizing the probe light intensity and the reference light intensity after illuminating an empty, clean absorption bottle with incident light. The formula for calculating apparent absorbance is, for example, as follows: , For the wavelength of the band, for Apparent absorbance in the band, for Normalized probe light intensity in the band for Reference light intensity in the band. This step aims to convert the light intensity measurement into an absorbance value related to the substance concentration.

[0045] Subsequently, in step B4, considering that the mixture contains lavage fluid in addition to blood, and that the lavage fluid itself has a certain absorbance, it is necessary to consult pre-calibrated baseline absorbance data of the lavage fluid according to the type of lavage fluid currently being used (for example, pre-calibrating the baseline absorbance of different types of lavage fluid for different wavelengths, forming a baseline absorbance lookup table, and then consulting the table based on the type of lavage fluid currently being used to obtain the corresponding baseline absorbance). Subtracting the corresponding baseline absorbance of the lavage fluid from the apparent absorbance of each wavelength effectively deducts the contribution of the lavage fluid to the absorbance, thus obtaining a more accurate baseline-deducted absorbance that only reflects the blood components.

[0046] Finally, in step B5, to further improve the accuracy of absorbance measurement, scattering correction is needed for the absorbance after baseline subtraction. This is because particles such as blood cells in the blood scatter light, affecting absorbance measurement. By pre-measuring the hematocrit and hemoglobin concentration of the patient's pure blood, and combining this with the type of irrigation fluid currently used, scattering correction factors for each wavelength band can be obtained from a lookup table (for example, scattering correction factors for each wavelength band can be pre-calibrated for pure blood with different hematocrit and hemoglobin concentrations mixed with different types of irrigation fluid, forming a scattering correction factor lookup table; during the operation, based on the pre-measuring hematocrit and hemoglobin concentration of the patient's pure blood and the type of irrigation fluid currently used, this scattering correction factor lookup table is consulted, and the scattering correction factors for each wavelength band are obtained through interpolation). Using these scattering correction factors to correct the absorbance after baseline subtraction can eliminate errors caused by scattering effects, ultimately obtaining accurate absorbance values ​​for each wavelength band.

[0047] The solution in this application systematically processes the acquisition of multi-band absorbance information through the detailed steps described in B1 to B5 above. First, by simultaneously acquiring and normalizing the probe and reference light intensities, systematic errors such as light source fluctuations and detector drift are effectively eliminated, ensuring the reliability of the raw light intensity data. Second, by calculating the apparent absorbance, the light intensity signal is converted into a physical quantity related to substance concentration. Furthermore, by subtracting the baseline absorbance of the lavage fluid, the interference of the lavage fluid on absorbance is eliminated, making the measurement results more focused on blood components. Finally, through scattering correction, the influence of particulate matter in the blood on light scattering is compensated, thereby obtaining highly accurate final absorbance for each band. It is precisely because of these refined processing steps that the accuracy of subsequent calculations of blood volume proportions based on absorbance is significantly improved.

[0048] The above technical solution overcomes many interference factors present in traditional absorbance measurements, such as unstable light sources, detector drift, background absorption of the irrigation fluid, and blood scattering effects. This allows for the acquisition of high-precision, high-reliability multi-band absorbance information. This accurate absorbance data is a crucial prerequisite for accurately calculating the volume ratio of blood in the mixed fluid, thus providing a solid data foundation for the accuracy and reliability of real-time intraoperative blood loss estimation methods. It significantly improves the accuracy of blood loss estimation and helps physicians make timely clinical decisions.

[0049] In practical applications, optical measurement systems can be affected by various factors, such as long-term drift of optical components, contamination of the suction bottle window, and fluctuations in system temperature. These factors can all lead to errors in the acquired multi-band absorbance information, thus affecting the accuracy of blood loss estimation. If these problems are not addressed, it may result in misjudgment of the patient's blood loss, thereby affecting clinicians' decisions on timely and accurate transfusion or intervention.

[0050] Therefore, in some preferred embodiments, the following steps are included after step B1 and before step B2: Obtain the incident light intensity and system temperature for each wavelength band; Compare the rate of change of the reference light intensity of each band relative to its respective pre-calibrated baseline value with the preset rate of change threshold to determine whether significant drift has occurred; calculate the real-time transmittance of the 940nm band based on the probe light intensity of the 940nm band and the corresponding incident light intensity, and compare the rate of decrease of the real-time transmittance relative to the initial transmittance of the 940nm band at the beginning of the operation with the preset rate of decrease threshold to determine whether window contamination has occurred. If a significant drift occurs, the reference light intensity of each band is compared with its respective pre-calibrated baseline value, the drift factor of each band is calculated, and the probe light intensity of each band is divided by the corresponding drift factor to achieve drift compensation of the probe light intensity. If window contamination occurs, the ratio of the real-time transmittance of the 940nm band to the corresponding clean state transmittance is calculated as a contamination factor. The detection light intensity of each band is divided by this contamination factor to achieve contamination compensation of the detection light intensity. Based on the system temperature, the temperature compensation coefficient is determined by referring to a table, which is used to compensate for the temperature of the probe light intensity in each band.

[0051] Specifically, before acquiring multi-band absorbance information, the system acquires the incident light intensity and system temperature for each band in real time. The incident light intensity is used to calculate transmittance, while the system temperature is used for subsequent temperature compensation. To detect optical system drift, the system continuously compares the rate of change of the reference light intensity for each band relative to its respective pre-calibrated baseline value. If this rate of change exceeds a preset threshold (e.g., 5%, but not limited to this), it is considered a significant drift. Once a significant drift is detected, the system calculates the corresponding drift factor based on the reference light intensity for each band relative to its respective pre-calibrated baseline value. Subsequently, the probe light intensity for each band is divided by this drift factor, thereby achieving probe light intensity drift compensation and correcting for light intensity attenuation or enhancement caused by device aging or environmental changes.

[0052] In addition, to detect whether the suction bottle window is contaminated, the system calculates the real-time transmittance of the 940nm band based on the probe light intensity and the corresponding incident light intensity. This real-time transmittance is compared with the initial transmittance of the 940nm band at the start of the surgery. If the rate of decrease exceeds a preset threshold (e.g., 10%, but not limited to this), window contamination is determined. When window contamination occurs, the system calculates the ratio of the real-time transmittance of the 940nm band to the corresponding clean state transmittance as a contamination factor. The clean state transmittance can be understood as the transmittance of the 940nm band measured at the start of the surgery. Subsequently, the probe light intensity of each band is divided by this contamination factor to achieve contamination compensation of the probe light intensity and eliminate the light signal attenuation caused by window contamination.

[0053] Meanwhile, considering the impact of temperature on optical measurement results, the system determines the corresponding temperature compensation coefficient by looking up a table based on the real-time acquired system temperature (e.g., the temperature of the photodetector and signal conditioning circuit). (For example, the gain drift caused by the probe light intensity in each band under different system temperature conditions is determined in advance through experiments, and the corresponding temperature compensation coefficient is determined based on this gain drift, forming a temperature compensation coefficient lookup table. During the procedure, this table is consulted based on the real-time acquired system temperature, and the corresponding temperature compensation coefficient is determined through interpolation.) This temperature compensation coefficient will be used to compensate for the temperature of the probe light intensity in each band (e.g., multiplying the probe light intensity by the temperature compensation coefficient) to eliminate the impact of temperature fluctuations on measurement accuracy.

[0054] This application's solution effectively addresses common accuracy issues in traditional optical measurements by introducing drift compensation, contamination compensation, and temperature compensation mechanisms. Specifically, by monitoring changes in reference light intensity in real time and performing drift compensation, measurement errors caused by changes in light source or detector performance over time can be corrected. Utilizing the contamination sensitivity of the 940nm band, real-time detection and compensation of contamination at the aspiration bottle window eliminates signal attenuation caused by optical path obstruction, ensuring the authenticity of the optical signal. Furthermore, by acquiring system temperature in real time and performing temperature compensation, the impact of ambient temperature fluctuations on detector response characteristics can be offset, further improving the stability of measurement data. It is precisely due to the introduction of these multiple compensation mechanisms that the acquired multi-band absorbance information becomes more accurate and reliable, thus providing high-quality input data for subsequent blood volume ratio calculations and cumulative blood loss estimation.

[0055] Through the above technical solutions, this application can significantly improve the accuracy and robustness of the real-time estimation method for intraoperative blood loss in multimodal surgery. Specifically, the drift compensation mechanism ensures the stability of the optical measurement system during long-term operation, avoiding measurement deviations caused by device aging. The contamination compensation mechanism effectively addresses the common problem of suction bottle window contamination in the surgical environment, ensuring effective transmission and accurate reception of optical signals. The temperature compensation mechanism eliminates the interference of ambient temperature changes on the measurement results, enabling the system to maintain high accuracy at different operating temperatures. As a result, the acquired multi-band absorbance information is more realistic and reliable, making the estimated results of the blood volume ratio in the mixture, the blood volume in the suction bottle, and the final cumulative blood loss and blood loss rate calculated based on this information more accurate. This provides more reliable data support for clinical decision-making and effectively avoids misjudgments and inappropriate interventions caused by measurement errors.

[0056] In some implementations, step A2 includes: A201. Based on the absorbance of each wavelength band, and using the Lambert-Beer law, calculate the corresponding hemoglobin concentration for each wavelength band. A202. Weighted summation of hemoglobin concentrations across different bands yields a composite hemoglobin concentration; A203. Based on the overall hemoglobin concentration, the pre-measured hemoglobin concentration of the patient's pure blood, and the hemoglobin concentration of the currently used irrigation fluid, calculate the volume ratio of blood in the mixture based on the dilution principle.

[0057] Specifically, in step A201, the Lambert-Beer law is expressed by the following formula: , for Absorbance in the band for The pre-calibrated molar absorptivity of the band. for The corresponding hemoglobin concentration for the band. The optical path length is determined by the structure of the optical system.

[0058] In step A202, the overall hemoglobin concentration can be obtained by calculating the weighted sum of the hemoglobin concentrations in each band. As a preferred embodiment, the weights of the 630nm and 660nm bands can be set higher than those of other bands because these bands are the main absorption peaks of hemoglobin and are more sensitive to changes in hemoglobin concentration. For example, the weights of the 630nm, 660nm, 810nm, 850nm, and 940nm bands can be 25%, 25%, 20%, 20%, and 10%, respectively, but this is not a limitation.

[0059] Specifically, the volume ratio of blood in the lavage fluid can be calculated using the formula α = (C_obs - C_saline) / (C_blood - C_saline). Here, α represents the volume ratio of blood in the lavage fluid, C_obs is the overall hemoglobin concentration, C_blood is the hemoglobin concentration of the patient's pure blood, and C_saline is the hemoglobin concentration of the currently used lavage fluid. Typically, the hemoglobin concentration C_saline of the lavage fluid can be considered zero.

[0060] The proposed solution, through the aforementioned steps, first utilizes the Lambert-Beer law to convert multi-band absorbance information into hemoglobin concentrations for each band, as there is a clear quantitative relationship between absorbance and substance concentration. Subsequently, by weighting and synthesizing these hemoglobin concentrations from different bands, information from various bands can be effectively integrated. Especially when assigning higher weights to bands sensitive to hemoglobin absorption, the accuracy and robustness of concentration estimation are improved, resulting in a more representative composite hemoglobin concentration. Finally, based on the dilution principle, the composite hemoglobin concentration in the mixture is compared with the known hemoglobin concentrations of pure blood and lavage fluid, thereby accurately calculating the blood volume ratio in the mixture. This series of steps ensures that the calculation of the blood volume ratio is more accurate and reliable in complex mixed-fluid environments.

[0061] Through the above technical solution, this application provides a more accurate and reliable method for calculating the volume ratio of blood in a mixed solution. By introducing multi-band absorbance information, the Lambert-Beer law, weighted summation, and the dilution principle, this method effectively overcomes the limitations of single-band measurement and provides a more accurate assessment of hemoglobin concentration in the mixed solution. In particular, by considering and weighting the differences in hemoglobin sensitivity across different bands, the accuracy and anti-interference ability of the calculation results are improved, thus providing a more solid data foundation for subsequent calculations of cumulative blood loss and blood loss rate, thereby enhancing the overall performance and reliability of the real-time intraoperative blood loss estimation system.

[0062] In some implementations, step A3 includes: A301. Obtain the volume of the mixture based on its weight and suction flow rate; A302. Multiply the volume ratio by the volume of the mixture to obtain the amount of blood in the aspiration bottle.

[0063] Specifically, in step A301, the volume of the mixture can be obtained in several ways. For example, the volume of the mixture can be calculated based on the real-time monitored weight of the mixture and a preset mixture density. The density of the mixture can be dynamically adjusted based on the ratio of irrigation fluid to blood, or an average density can be used. Furthermore, the suction flow rate can also be used as supplementary information to calibrate or verify the volume calculated by weight, or in some cases, to directly estimate the volume of the inflowing mixture based on the suction flow rate and time.

[0064] Step A302 aims to separate the blood component from the mixture. By multiplying the blood volume ratio in the mixture calculated in step A2 by the volume of the mixture obtained in step A301, the amount of pure blood in the aspiration bottle can be directly obtained. This process ensures accurate quantification of the actual blood loss.

[0065] The proposed solution breaks down the blood volume calculation process in the aspiration bottle into two sub-steps: obtaining the volume of the mixture and calculating the blood volume based on the volume ratio. This makes the entire calculation process clearer and more controllable. First, by comprehensively analyzing the weight of the mixture and the aspiration flow rate, the total volume of the mixture in the current aspiration bottle can be accurately obtained. Then, combined with the previously calculated blood volume ratio in the mixture, the volume of pure blood can be directly and effectively separated from the total volume of the mixture. This step-by-step calculation method helps improve the accuracy and reliability of blood volume estimation.

[0066] The above technical solution provides a clear breakdown of the blood volume calculation process in the suction bottle, making the logic of the calculation more rigorous and the operation more specific. This refined approach not only improves the transparency of blood volume calculation but also provides more accurate input data for subsequent calculations of cumulative blood loss and blood loss rate, thereby enhancing the overall accuracy and reliability of the real-time intraoperative blood loss estimation method.

[0067] In actual surgical procedures, the measurement of suction flow rate can be affected by various factors, such as air bubbles, blockages, leaks, suction pump malfunctions, or flow sensor drift, leading to abnormal or inaccurate suction flow rate data. Directly relying on abnormal suction flow rate data to calculate the volume of the mixture may introduce significant errors, affecting the accuracy of blood volume calculations in the suction bottle and ultimately resulting in inaccurate estimates of cumulative blood loss.

[0068] Therefore, in some preferred embodiments, step A301 includes: Based on the pre-measured hematocrit and hemoglobin concentration of the patient's pure blood, and in conjunction with the type of irrigation fluid currently being used, the reference density of the mixed solution is obtained by referring to a table. The rate of change of mass is calculated based on the weight of the mixture, and the mass flow rate is calculated based on the suction flow rate and the reference density. Compare the rate of change in mass and the mass flow rate to determine if the attraction flow is abnormal; If the suction flow rate is abnormal, calculate the volume of the mixture based on the weight of the mixture and the reference density; If the suction flow rate is normal, calculate the volume of the mixture based on the suction flow rate.

[0069] Specifically, before obtaining the volume of the lavage fluid, the hematocrit and hemoglobin concentration of the patient's pure blood need to be measured beforehand. These parameters are usually obtained through preoperative blood tests and are used to characterize the basic properties of the patient's blood. At the same time, the type of lavage fluid being used, such as normal saline or Ringer's solution, also needs to be identified. Based on this information, a pre-established database or lookup table can be consulted to obtain the reference density of the current lavage fluid. This reference density is pre-calibrated based on the density characteristics of blood and lavage fluid mixtures in different proportions at different temperatures.

[0070] The mass change rate of the mixture refers to the ratio of the change in the weight of the mixture in the suction bottle to the time interval within a certain time interval, reflecting the real-time accumulation rate of the mixture's mass. The mass flow rate calculated from the suction flow rate and the reference density refers to the product of the flow rate measured by the suction flow sensor and the reference density of the mixture, reflecting the mass flow rate of the mixture entering the suction bottle through the suction system.

[0071] In practical applications, comparing the rate of change in mass with the mass flow rate can help determine if the suction flow is abnormal. For example, when the absolute value of the deviation rate of mass change from the mass flow rate reaches a preset threshold (e.g., 5%), it can be determined that the suction flow is abnormal. This abnormality may indicate problems such as air bubbles, blockages, leaks, or sensor malfunctions in the suction pipeline.

[0072] If the suction flow rate is determined to be abnormal, the suction flow rate data is no longer relied upon directly. Instead, the volume of the mixture is calculated based on the weight and reference density of the mixture. In this case, the volume of the mixture can be obtained by dividing the weight by the reference density. If the suction flow rate is determined to be normal, the volume of the mixture is calculated based on the suction flow rate, i.e., the volume of the mixture is equal to the integral of the suction flow rate over time.

[0073] This application's solution effectively addresses the volume estimation error problem caused by inaccurate suction flow data in traditional methods by introducing a mechanism for judging abnormal suction flow rates. Specifically, by monitoring the mass change rate of the mixture in the suction bottle in real time and comparing it with the mass flow rate calculated based on the suction flow rate and reference density, an inherent verification mechanism can be established. When a significant deviation exists between the two, it indicates that the suction flow rate data may be abnormal. At this time, the system can intelligently switch to a more reliable volume calculation method, namely, calculation using the weight of the mixture and the reference density. This dual verification and adaptive calculation strategy ensures the accuracy and robustness of the mixture volume estimation under different operating conditions.

[0074] Through the above technical solution, this application can effectively identify and address potential abnormalities in intraoperative suction flow, avoiding errors in mixed fluid volume estimation caused by inaccurate suction flow data. This improves the accuracy of blood volume calculation in the suction bottle, thereby enhancing the reliability and precision of the entire multimodal real-time intraoperative blood loss estimation method. By introducing redundancy verification and intelligent switching mechanisms, this solution enhances the system's adaptability and stability in complex and changing surgical environments, providing clinicians with more reliable blood loss monitoring data and facilitating timely and accurate clinical decision-making.

[0075] Preferably, step A4 may include: A401. Based on the target recognition algorithm combined with the ReID algorithm, each gauze is tracked and identified from the surgical field video; A402. Extract the image features of each identified gauze, input them into a pre-trained regression model to estimate the blood content of each gauze piece; the image features include color histogram, mean saturation, and texture features; A403. Calculate the sum of the blood content of each individual gauze piece identified to obtain the total blood content of the gauze.

[0076] Specifically, in step A401, the target recognition algorithm is used to detect and locate gauze in each frame of the surgical field video. To ensure accurate identification and tracking of the same gauze even when it is occluded, moved, or its appearance changes, the ReID (Re-identification) algorithm is used in conjunction. The ReID algorithm learns the unique visual features of the gauze, and can re-identify it as the same gauze even if it temporarily disappears in the video or reappears after overlapping with other gauze, thus achieving continuous tracking of each gauze.

[0077] In step A402, once the gauze is identified and tracked, its image features are extracted. These image features include a color histogram, mean saturation, and texture features. The color histogram reflects the color distribution of blood on the gauze, the mean saturation indicates the blood concentration, and the texture features capture the structural changes on the gauze surface caused by blood absorption. These features are input into a pre-trained regression model. This regression model is trained on a large number of blood-containing gauze samples, learning the mapping relationship between image features and actual blood content, thereby enabling accurate estimation of the blood content of each piece of gauze based on the extracted image features.

[0078] In step A403, the total blood content of the gauze at the current moment can be obtained by summing up the blood content of all the identified and estimated individual gauze pieces.

[0079] This application's solution addresses the inaccuracy of traditional methods in gauze identification and tracking in complex surgical field environments by introducing target recognition and ReID algorithms. This ensures that even if the gauze is obscured, overlapped, or temporarily disappears in the video, it can still be continuously and uniquely identified. Furthermore, by extracting multi-dimensional image features such as color histograms, saturation mean, and texture features, and inputting them into a pre-trained regression model, the estimation of blood content in individual gauze pieces becomes more refined and accurate, overcoming the errors that may exist based solely on simple image analysis. Therefore, this solution can more reliably obtain the total blood content of the gauze, providing more accurate data input for subsequent calculations of cumulative blood loss.

[0080] Through the above technical solutions, this application can significantly improve the accuracy and robustness of intraoperative gauze blood content estimation. Specifically, the application of the ReID algorithm effectively solves the problem of gauze re-identification in complex dynamic surgical fields, avoiding duplicate counting or undercounting, thus ensuring the continuity and accuracy of gauze tracking. Simultaneously, the extraction of multi-dimensional image features combined with the use of regression models makes the estimation of blood content in individual gauze pieces with different blood contents more precise, effectively reducing estimation errors. These improvements collectively ensure the reliability of the total blood content data of the gauze, thereby improving the overall accuracy of the entire multimodal real-time intraoperative blood loss estimation method and providing a more reliable basis for clinical decision-making.

[0081] In some preferred embodiments, the specific implementation is as follows: During the surgery, video frames are first analyzed using target recognition algorithms such as YOLOv5, U-Net, or DeepLab to quickly detect and select all gauze in the frame. Subsequently, using the ReID algorithm, these identified gauze pieces are tracked across frames, assigning a unique ID to each piece. Even if the gauze is obscured by other instruments or temporarily removed from the field of vision, it can be correctly identified upon reappearance. For each tracked gauze piece, the system extracts its image region's color histogram (e.g., in the HSV color space), mean saturation, and texture features based on the gray-level co-occurrence matrix (GLCM), such as contrast, energy, and homogeneity. These features are input into a convolutional neural network (CNN) regression model pre-trained with a large number of gauze images with different blood contents. This model outputs the real-time blood content of each gauze piece (e.g., in milliliters). Finally, the blood content of each gauze pad currently identified and tracked in the surgical field is summed to obtain the total blood content of the gauze pads. This data is then used in the Kalman filter algorithm to more accurately calculate the cumulative blood loss.

[0082] Specifically, step A5 includes: The cumulative blood loss and blood loss rate form a state vector, the blood volume in the suction bottle, the weight of the mixed fluid, the suction flow rate, and the total blood content of the gauze form an observation vector, and the transfusion rate and fluid resuscitation rate form a control input vector. The cumulative blood loss and blood loss rate are calculated using the Kalman filter algorithm.

[0083] Specifically, when using the Kalman filter algorithm to calculate cumulative blood loss and blood loss rate, a suitable system model must first be constructed. The state vector is defined as a combination containing both cumulative blood loss and blood loss rate, allowing the system to dynamically track changes in these two key parameters. Cumulative blood loss refers to the total amount of blood lost by the patient during the surgery, while the blood loss rate reflects the speed at which blood is lost.

[0084] Furthermore, the observation vector was constructed to include the blood volume in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content in the gauze. These observations, acquired in real-time from multimodal monitoring data, provide direct or indirect evidence regarding the patient's blood loss. The blood volume in the suction bottle directly reflects the blood collected via the suction device; the weight of the mixture and the suction flow rate help assess the total volume and rate of aspirated fluid; and the total blood content in the gauze supplements any blood loss that could not be fully measured by suction. Together, these observations provide rich, multi-source information for Kalman filtering, used to correct and optimize the estimation of the state vector.

[0085] Furthermore, the control input vectors are defined as the transfusion rate and the fluid resuscitation rate. The transfusion rate refers to the speed at which blood is transfused into the patient, while the fluid resuscitation rate refers to the speed at which other fluids (such as saline, anesthetic drugs, etc.) are infused into the patient. These control inputs represent external factors that intervene in the patient's physiological state, directly affecting the patient's blood volume and circulatory status. Therefore, incorporating them into the Kalman filter model can more accurately reflect the actual blood loss and compensate for the estimation results. The specific calculation process of the Kalman filter algorithm is existing technology and will not be detailed here.

[0086] The proposed solution explicitly defines the state vector, observation vector, and control input vector of the Kalman filter algorithm, enabling the algorithm to establish a more accurate and comprehensive system dynamic model. By dynamically tracking cumulative blood loss and blood loss rate as state variables, and combining real-time correction with multi-source observation data such as blood volume in the suction bottle, weight of the mixed fluid, suction flow rate, and total blood content in the gauze, the system can effectively integrate information from different modalities, overcoming the limitations of a single data source. Simultaneously, by incorporating the transfusion rate and fluid resuscitation rate as control inputs into the model, the Kalman filter algorithm can effectively compensate for external interventions, thereby more accurately estimating the patient's actual blood loss in complex and variable surgical environments and predicting the blood loss rate in real time, significantly improving the robustness and accuracy of the estimation.

[0087] Through the above technical solutions, this application can significantly improve the accuracy and stability of real-time estimation of intraoperative blood loss. Specifically, the explicit state vector allows the system to focus on the dynamic tracking of core blood loss indicators; the rich observation vector ensures the full utilization of multimodal data, effectively reducing errors that may be caused by a single data source; and the introduction of control input vectors enables the system to effectively model and compensate for medical interventions (such as blood transfusion and fluid replacement), avoiding estimation bias caused by external interventions. Therefore, the solution of this application can provide more reliable cumulative blood loss and blood loss rate, providing clinicians with more accurate decision-making basis, thereby optimizing patient management and improving surgical safety.

[0088] refer to Figure 2 This application provides a multimodal real-time intraoperative blood loss estimation system, which includes: Multimodal data acquisition module 1 is used to acquire multimodal monitoring data in real time. The multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, weight of the mixture, aspiration flow rate, surgical field video, and patient physiological parameters (for details, please refer to step A1 above). The proportion calculation module 2 is used to calculate the volume ratio of blood in the mixture based on multi-band absorbance information (for details, please refer to step A2 above). Blood volume calculation module 3 is used to calculate the blood volume in the aspiration bottle based on the weight of the mixture, the aspiration flow rate, and the volume ratio (for details, please refer to step A3 above). The gauze total blood content estimation module 4 is used to identify the gauze based on the surgical field video tracking and to estimate the total blood content of the gauze based on the image features of the gauze (for details, please refer to step A4 above). The blood loss calculation module 5 is used to calculate the cumulative blood loss and blood loss rate based on the blood volume in the suction bottle, the weight of the mixed solution, the suction flow rate, and the total blood content of the gauze using the Kalman filter algorithm (for details, please refer to step A5 above). The decision output module 6 is used to perform graded alarms, coordinate the control of blood transfusion pumps and anesthesia machines, and record auditable logs based on the cumulative blood loss, blood loss rate, and patient physiological parameters (for details, please refer to step A6 above).

[0089] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A multimodal real-time estimation method for intraoperative blood loss, characterized in that, The method includes the following steps: A1. Real-time acquisition of multimodal monitoring data; the multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, weight of the mixture, aspiration flow rate, surgical field video, and patient physiological parameters; A2. Calculate the volume ratio of blood in the mixture based on the multi-band absorbance information; A3. Calculate the amount of blood in the suction bottle based on the weight of the mixture, the suction flow rate, and the volume ratio; A4. Based on the surgical field video tracking and identification of the gauze, and based on the image features of the gauze, estimate the total blood content of the gauze; A5. Based on the amount of blood in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze, calculate the cumulative blood loss and blood loss rate using the Kalman filter algorithm; A6. Based on the cumulative blood loss, the blood loss rate, and the patient's physiological parameters, perform graded alarms, implement coordinated control of the blood transfusion pump and anesthesia machine, and record auditable logs.

2. The multimodal intraoperative blood loss real-time estimation method according to claim 1, characterized in that, The multi-band absorbance information includes absorbance in the 630nm, 660nm, 810nm, 850nm and 940nm bands; The patient's physiological parameters include blood pressure, heart rate, end-tidal carbon dioxide partial pressure, and urine output.

3. The multimodal intraoperative blood loss real-time estimation method according to claim 2, characterized in that, Step A1, the step of obtaining the multi-band absorbance information includes: B1. Acquire the detection light intensity and the corresponding reference light intensity of multiple wavelength bands; the detection light intensity is the light intensity of the incident light after passing through the mixed liquid, and the reference light intensity is the light intensity of the incident light after passing through a preset optical attenuation element. B2. Divide the probe intensity of each band by the corresponding reference intensity to obtain the normalized probe intensity of each band. B3. Calculate the apparent absorbance of each band based on the normalized probe light intensity and the corresponding pre-calibrated reference light intensity. B4. Based on the type of irrigation fluid currently used, look up the table to obtain the baseline absorbance of the irrigation fluid for each band, and subtract the corresponding baseline absorbance of the irrigation fluid from the apparent absorbance of each band to obtain the absorbance after deducting the baseline. B5. Based on the pre-measured hematocrit and hemoglobin concentration of the patient's pure blood, and in conjunction with the type of irrigation fluid currently used, the scattering correction factor for each band is obtained from the table and used to perform scattering correction on the absorbance after baseline subtraction, thus obtaining the final absorbance for each band.

4. The multimodal intraoperative blood loss real-time estimation method according to claim 3, characterized in that, The steps following step B1 and before step B2 include: Obtain the incident light intensity and system temperature for each wavelength band; Compare the rate of change of the reference light intensity of each band relative to its respective pre-calibrated baseline value with a preset rate of change threshold to determine whether significant drift has occurred; calculate the real-time transmittance of the 940nm band based on the probe light intensity of the 940nm band and the corresponding incident light intensity, and compare the rate of decrease of the real-time transmittance relative to the initial transmittance of the 940nm band at the beginning of the operation with a preset rate of decrease threshold to determine whether window contamination has occurred. If a significant drift occurs, the reference light intensity of each band is compared with its respective pre-calibrated baseline value, the drift factor of each band is calculated, and the probe light intensity of each band is divided by the corresponding drift factor to achieve drift compensation of the probe light intensity. If window contamination occurs, the ratio of the real-time transmittance of the 940nm band to the corresponding clean state transmittance is calculated as a contamination factor. The detection light intensity of each band is divided by this contamination factor to achieve contamination compensation of the detection light intensity. Based on the system temperature, the temperature compensation coefficient is determined by referring to a table, which is used to compensate for the temperature of the probe light intensity in each band.

5. The multimodal intraoperative blood loss real-time estimation method according to claim 2, characterized in that, Step A2 includes: A201. Based on the absorbance of each wavelength band, and using the Lambert-Beer law, calculate the corresponding hemoglobin concentration for each wavelength band. A202. Weighted summation of hemoglobin concentrations across different bands yields a composite hemoglobin concentration; A203. Based on the overall hemoglobin concentration, the pre-measured hemoglobin concentration of the patient's pure blood, and the hemoglobin concentration of the currently used irrigation fluid, calculate the volume ratio of blood in the mixture based on the dilution principle.

6. The multimodal intraoperative blood loss real-time estimation method according to claim 1, characterized in that, Step A3 includes: A301. Obtain the volume of the mixture based on the weight of the mixture and the suction flow rate; A302. Multiply the volume ratio by the volume of the mixture to obtain the amount of blood in the aspiration bottle.

7. The multimodal intraoperative blood loss real-time estimation method according to claim 6, characterized in that, Step A301 includes: Based on the pre-measured hematocrit and hemoglobin concentration of the patient's pure blood, and in conjunction with the type of irrigation fluid currently being used, the reference density of the mixed solution is obtained by referring to a table. The mass change rate is calculated based on the weight of the mixture, and the mass flow rate is calculated based on the suction flow rate and the reference density; The mass change rate and the mass flow rate are compared to determine whether the suction flow rate is abnormal. If the suction flow rate is abnormal, the volume of the mixture is calculated based on the weight of the mixture and the reference density; If the suction flow rate is normal, the volume of the mixture is calculated based on the suction flow rate.

8. The multimodal intraoperative blood loss real-time estimation method according to claim 1, characterized in that, Step A4 includes: A401. Based on the target recognition algorithm combined with the ReID algorithm, each gauze is tracked and identified from the surgical field video; A402. Extract the image features of each identified gauze, and input them into a pre-trained regression model to estimate the blood content of each gauze piece; the image features include color histogram, mean saturation, and texture features; A403. Calculate the sum of the blood content of each individual piece of the identified gauze to obtain the total blood content of the gauze.

9. The multimodal intraoperative blood loss real-time estimation method according to claim 1, characterized in that, Step A5 includes: The cumulative blood loss and blood loss rate form a state vector, the blood volume in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze form an observation vector, and the transfusion rate and fluid resuscitation rate form a control input vector. The cumulative blood loss and blood loss rate are calculated using the Kalman filter algorithm.

10. A multimodal real-time intraoperative blood loss estimation system, characterized in that, The system includes: The multimodal data acquisition module is used to acquire multimodal monitoring data in real time; the multimodal monitoring data includes multi-band absorbance information of the mixture in the aspiration bottle, weight of the mixture, aspiration flow rate, surgical field video, and patient physiological parameters; The proportion calculation module is used to calculate the volume proportion of blood in the mixture based on the multi-band absorbance information. The blood volume calculation module is used to calculate the blood volume in the suction bottle based on the weight of the mixture, the suction flow rate, and the volume ratio. The gauze total blood content estimation module is used to identify the gauze based on the surgical field video tracking and to estimate the total blood content of the gauze based on the image features of the gauze. The blood loss calculation module is used to calculate the cumulative blood loss and blood loss rate based on the amount of blood in the suction bottle, the weight of the mixture, the suction flow rate, and the total blood content of the gauze, using a Kalman filter algorithm. The decision output module is used to perform graded alarms, coordinate the control of the blood transfusion pump and the anesthesia machine, and record auditable logs based on the cumulative blood loss, the blood loss rate, and the patient's physiological parameters.

Images