Multi-source data real-time fusion and decision method and system for surgical field management

Abstract

The application discloses a multi-source data real-time fusion and decision method and system for surgical field management, relates to the technical field of medical data processing, and comprises the following steps: synchronously collecting multi-source sensing data of a surgical field; performing parallel processing on a multi-spectrum image sequence to obtain water turbidity, blood concentration and foreign matter density characteristic indexes; fusing surgical time, instrument motion data and blood concentration trends to identify a current surgical stage; fusing the multi-characteristic indexes into a visual clarity evaluation index according to the surgical stage; and calculating a preliminary operation suggestion value according to the deviation of the index from a target value and generating an intelligent decision data package. The application converts intraoperative evaluation depending on subjective experience into objective and adaptive multi-source data intelligent fusion analysis, solves the problems of inaccurate field state evaluation and lack of quantitative basis for decision-making under multi-factor interference, and provides precise real-time decision support.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a method and system for real-time fusion and decision-making of multi-source data for surgical field of view management. Background Technology

[0002] Spinal endoscopy is a minimally invasive surgical technique that allows direct observation and treatment of spinal lesions through an endoscopic portal. It is primarily used to treat conditions such as lumbar disc herniation, spinal stenosis, and nerve root compression. Its core advantages lie in its minimal trauma, rapid recovery, and minimal damage to normal tissues, making it an important treatment method in the field of spinal surgery. Maintaining a clear and stable surgical field is crucial for ensuring surgical safety and efficiency in minimally invasive surgeries such as spinal endoscopy.

[0003] Currently, the clarity of the surgical field of vision relies primarily on the surgeon's subjective visual judgment, lacking a unified and objective quantitative standard, leading to significant differences in decision-making among surgeons. When the field of vision becomes blurred due to bleeding, tissue debris, or turbid fluid, the procedure must be interrupted. While some existing research attempts to utilize image processing techniques to detect single factors, it has failed to address the challenge of comprehensive assessment under the interference of multiple coupled factors, and lacks the ability to dynamically fuse visual information with multi-source data such as surgical progress to form accurate decision support. Therefore, there is an urgent need for an intelligent method and system capable of real-time fusion of multi-source information and outputting quantitative assessment results and decision-making basis. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a method and system for real-time fusion and decision-making of multi-source data for surgical field management.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A real-time fusion and decision-making method for multi-source data for surgical field management includes the following steps:

[0007] Step S1: Synchronously acquire multi-source sensing data of the surgical field of view. The multi-source sensing data includes at least multispectral image sequences, instrument kinematic data, and physiological parameter data.

[0008] Step S2: Perform parallel processing on the multispectral image sequence, extract and quantify the water turbidity feature index reflecting the change in liquid transparency through the first calculation path, extract and quantify the blood color concentration feature index reflecting blood content through the second calculation path, and extract and quantify the foreign matter density feature index reflecting the distribution state of solid foreign matter based on the deep learning segmentation model through the third calculation path.

[0009] Step S3: Based on the proportion of time the surgery has been performed, the kinematic data of the instruments, and the recent trend of the blood color concentration characteristic index, identify the surgical stage in real time;

[0010] Step S4: Based on the identified surgical stage, the water turbidity characteristic index, blood color concentration characteristic index, and foreign matter density characteristic index are fused into a visual clarity assessment index.

[0011] Step S5: Calculate the operation suggestion parameters based on the deviation between the visual clarity evaluation index and the preset target value, and generate an intelligent decision data package containing the operation suggestion parameters.

[0012] Furthermore, in step S2, the first calculation path specifically includes the following steps:

[0013] Based on the white light channel image in the multispectral image sequence, calculate the texture component based on the local binary mode texture entropy and the contrast of the gray-level co-occurrence matrix;

[0014] Analyze the diffusion area of ​​a preset reference spot in the image to obtain the light scattering component;

[0015] The mean square error of the current image and the dynamically updated reference image in the a channel and b channel of the Lab color space and the saturation S channel of the HSV color space are calculated respectively, and the mean square error of the a, b and S channels is synthesized to obtain the color shift component.

[0016] The texture component, light scattering component, and color offset component are weighted and summed according to preset weights to obtain the water turbidity characteristic index.

[0017] Furthermore, in step S2, the second calculation path specifically includes the following steps:

[0018] The average reflectance of the first wavelength channel image and the second wavelength channel image in the multispectral image sequence were calculated respectively, and the spectral absorbance ratio characterizing the specific absorption of hemoglobin was calculated.

[0019] Convert the white light channel image to the HSV color space and count the proportion of pixels that meet the preset blood-red hue, saturation, and brightness threshold ranges.

[0020] The blood color concentration characteristic index is obtained by linearly combining the spectral absorption ratio and the pixel ratio according to a preset coefficient.

[0021] Furthermore, in step S2, the third calculation path specifically includes the following steps:

[0022] The white light channel image in the multispectral image sequence is input into a pre-trained U-Net convolutional neural network to obtain the initial mask;

[0023] The initial mask is subjected to morphological opening and closing operations in sequence to obtain a segmentation mask;

[0024] The area ratio of foreign object pixels in the segmentation mask is calculated as the basic foreign object density;

[0025] The white light channel image is divided into multiple grids and the statistical uniformity of foreign object density in each grid is calculated;

[0026] The foreign matter density characteristic index is calculated based on the basic foreign matter density and the statistical uniformity.

[0027] Furthermore, step S3 specifically includes the following steps:

[0028] Step S3.1: Calculate the proportion of the surgery already performed to the total estimated time;

[0029] Step S3.2: Using the current time as a reference, analyze the movement pattern and operation frequency of the surgical instruments within the predetermined time period;

[0030] Step S3.3: Calculate the average value of the hemochromatin concentration index within the same predetermined time period as in step S3.2;

[0031] Step S3.4: Based on the feature vector composed of the average values ​​of the ratio, movement pattern, operation frequency and blood color concentration index, identify the current surgical stage through preset rules;

[0032] The surgical phase includes at least the initial entry phase, decompression phase, hemostasis phase, and termination phase.

[0033] Furthermore, step S4 specifically includes the following steps:

[0034] Step S4.1: Based on the identified surgical stage, retrieve the basic weight coefficients corresponding to the turbidity, blood color concentration and foreign body density indicators from the preset mapping table;

[0035] Step S4.2: Adjust the weights according to the patient's specific pathological characteristics to obtain the final weight coefficients used for fusion calculation;

[0036] Step S4.3: Using the final weighting coefficients, calculate the visual clarity index using a nonlinear fusion formula.

[0037] Furthermore, step S5 specifically includes the following steps:

[0038] Step S5.1: Calculate the deviation between the current visual acuity index and the preset target value;

[0039] Step S5.2: Based on the deviation and real-time physiological parameter data, a set of optimized operation suggestion parameters are calculated and generated through a decision model. The operation suggestion parameters include: flow rate suggestion value and flushing mode suggestion. The flushing modes include: standard continuous flushing, pulsed perturbation flushing and continuous gentle flushing.

[0040] Step S5.3: Perform compliance verification and formatted encapsulation on the operation suggestion parameters, and output them as the intelligent decision data package.

[0041] A real-time multi-source data fusion and decision-making system for surgical field management, used to implement any of the aforementioned real-time multi-source data fusion and decision-making methods for surgical field management, including:

[0042] The multi-source synchronous data acquisition module is used to acquire multispectral visual data, instrument kinematic data, and physiological parameter data with high precision and synchronously.

[0043] The parallel feature calculation engine module has parallel processing capabilities and is used to perform real-time extraction and quantification calculations of the water turbidity feature index, blood color concentration feature index and foreign matter density feature index.

[0044] The context awareness and analysis module is used to process temporal and motion data to achieve dynamic recognition of surgical stages;

[0045] An adaptive data fusion and evaluation module is used to generate a visual acuity evaluation index based on weight adjustment and multi-feature nonlinear fusion performed according to the surgical stage.

[0046] The intelligent decision parameter generation module is used to generate and output intelligent decision data packages based on the evaluation index and physiological parameter feedback.

[0047] The human-computer interaction and data interface module is used to visualize all intermediate data, evaluation results and decision parameters, and provides data output interface and manual intervention interface.

[0048] Furthermore, the multi-source synchronous data acquisition module includes a multispectral imaging unit. The multispectral imaging unit uses a miniature camera with a diameter of 3.5 mm and integrates three LED light sources of three bands: first wavelength, second wavelength, and white light. It adopts a time-division multiplexing light source control strategy to ensure that the excitation of different band light sources and the camera exposure are synchronized within microsecond-level errors.

[0049] Furthermore, the human-computer interaction and data interface module provides a touch screen interface, allowing the operator to view, confirm, or modify the suggested parameters in the intelligent decision data package.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0051] 1. This invention transforms the traditional fuzzy judgments that rely on subjective experience into objective and quantitative characteristic indices of water turbidity, blood color concentration, and foreign matter density through multispectral imaging and parallel analysis models, providing accurate and reliable multi-source data for subsequent intelligent analysis and decision-making.

[0052] 2. This invention can monitor and fuse multi-source data in real time to quickly assess the state of the field of vision. When clarity decreases, the system can respond quickly, automatically generate and output optimized decision suggestions, reduce operational delays, and provide surgeons with continuous and efficient decision support.

[0053] 3. This invention dynamically identifies the surgical stage and adaptively adjusts the weighting strategy of multi-feature fusion accordingly, so that the final evaluation results and decision recommendations can be highly matched with the actual needs and risk focus of the current surgical stage.

[0054] 4. This invention incorporates real-time physiological parameters, such as intraspinal pressure, as key constraints into the decision-making model. When a safety risk is detected, the system will proactively embed strong safety constraints into the generated decision suggestions, prioritizing the output of suggested solutions that ensure the patient's physiological safety. This avoids secondary risks that may arise from solely pursuing a clear field of vision at the source of decision-making. Attached Figure Description

[0055] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0056] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0057] Figure 2 This is a system schematic diagram according to an embodiment of the present invention;

[0058] Figure 3 A logic diagram for generating decision parameters in embodiments of the present invention;

[0059] Figure 4 This is a logic diagram of image processing and intelligent analysis according to an embodiment of the present invention;

[0060] Figure 5 This is a logic diagram of the rinsing mode according to an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] like Figure 1 As shown, the multi-source data real-time fusion and decision-making method for surgical field management includes the following steps:

[0063] Step S1: Synchronously acquire multi-source sensing data of the surgical field of view. The multi-source sensing data includes at least multispectral image sequences, instrument kinematic data, and physiological parameter data.

[0064] Step S2: Perform parallel processing on the multispectral image sequence, extract and quantify the water turbidity feature index reflecting the change in liquid transparency through the first calculation path, extract and quantify the blood color concentration feature index reflecting blood content through the second calculation path, and extract and quantify the foreign matter density feature index reflecting the distribution state of solid foreign matter based on the deep learning segmentation model through the third calculation path.

[0065] Step S3: Based on the proportion of time the surgery has been performed, the kinematic data of the instruments, and the recent trend of the blood color concentration characteristic index, identify the surgical stage in real time;

[0066] Step S4: Based on the identified surgical stage, the water turbidity characteristic index, blood color concentration characteristic index, and foreign matter density characteristic index are fused into a visual clarity assessment index.

[0067] Step S5: Calculate the operation suggestion parameters based on the deviation between the visual clarity evaluation index and the preset target value, and generate an intelligent decision data package containing the operation suggestion parameters.

[0068] In step S2, the first calculation path specifically includes the following steps:

[0069] Based on the white light channel image in a multispectral image sequence, the texture component based on the local binary mode texture entropy and gray-level co-occurrence matrix contrast is calculated using the following formula:

[0070]

[0071] in, Represents texture components, This represents the entropy value of the local binary pattern histogram. Indicates the number of LBP sampling points. Represents the contrast eigenvalues ​​of the gray-level co-occurrence matrix;

[0072] Analyze the diffusion area of ​​a preset reference spot in the image to obtain the light scattering component. The calculation formula is as follows:

[0073]

[0074] in, Represents the light scattering component. This indicates the number of reference spots selected in the image. This represents the area of ​​the i-th light spot in the actual image. This represents the area of ​​the corresponding light spot in the reference clear image;

[0075] Calculate the mean square error of the current image and the dynamically updated reference image in the Lab color space (a channel, b channel) and the HSV color space (S channel of saturation), respectively. Then, synthesize the mean square errors of the a, b, and S channels to obtain the color shift component. The calculation formula is as follows:

[0076]

[0077] in, Indicates the color offset component. This represents the mean square error of the α channel in the Lab color space. This represents the mean square error of the b channel in the Lab color space. This represents the mean square error of the S-channel saturation in the HSV color space.

[0078] The texture component, light scattering component, and color shift component are weighted and summed according to preset weights to obtain the water turbidity characteristic index. The calculation formula is as follows:

[0079]

[0080] in, Indicates the turbidity index of water bodies. , and These represent the weights of the texture component, light scattering component, and color offset component, respectively.

[0081] The reference clear image is determined in the following way: during system initialization, the first frame image captured when the endoscope first enters the body and the field of vision is clean is used as the initial reference image; during the operation, when the visual acuity index (VCI) of 5 consecutive frames is higher than the set threshold, the system determines that the current field of vision is clear, and updates the reference image using the moving average method based on the images within this time period to adapt to the slow changes in tissue color or light.

[0082] In spinal endoscopic surgery, the irrigation fluid is usually clear saline. When the field of view becomes cloudy, the color shift mainly manifests as: 1. Yellow / brown shift: caused by the dissolution of tissue debris, proteins, etc.; 2. Red shift: caused by blood contamination; 3. Grayish-white shift: caused by fat particles, bone marrow components, etc. Therefore, the following three color channel combinations are selected for monitoring: 1. Lab color space a channel: mainly detects red-green color-opposite changes, sensitive to blood; 2. Lab color space b channel: mainly detects yellow-blue color-opposite changes, sensitive to tissue fluid and pus; 3. HSV color space saturation S channel: detects changes in color purity, with saturation decreasing in cloudy fluids.

[0083] In step S2, the second calculation path specifically includes the following steps:

[0084] The average reflectance of the first wavelength channel image (415 nm) and the second wavelength channel image (540 nm) in the multispectral image sequence were calculated separately, and the spectral absorbance ratio, which characterizes the specific absorption of hemoglobin, was calculated using the following formula:

[0085]

[0086] in, Indicates the spectral absorbance ratio. This indicates the reflectivity at 415nm, which is the blue light band. This indicates the reflectivity at 540nm, which is the green light band. This represents a constant used to prevent division by zero errors;

[0087] Convert the image to the HSV color space and count the percentage of pixels whose hue is in the range of 0-10 and 160-180, and whose saturation and brightness are both higher than the threshold.

[0088] The blood color concentration characteristic index is obtained by linearly combining the spectral absorbance ratio and the pixel ratio according to a preset coefficient. The calculation formula is as follows:

[0089]

[0090] in, Indicates blood color concentration index. This indicates the number of pixels that satisfy the HSV condition. This represents the total number of pixels in the image.

[0091] The saturation and brightness thresholds are both set to 50.

[0092] In step S2, the third calculation path specifically includes the following steps:

[0093] The white light channel image in the multispectral image sequence is input into a pre-trained U-Net convolutional neural network to obtain the initial mask;

[0094] The initial mask is subjected to morphological opening and closing operations in sequence to eliminate noise and fill holes to obtain the foreign object region. The morphological opening and closing operations use elliptical structural elements, wherein the structural element size of the opening operation is smaller than that of the closing operation. The opening operation uses a 3x3 pixel structural element to remove noise points, and the closing operation uses a 5x5 pixel structural element to fill small holes to obtain the segmentation mask.

[0095] The area ratio of foreign object pixels in the segmentation mask is calculated as the basic foreign object density;

[0096] The white light channel image is divided into multiple grids, and the statistical uniformity of foreign object density in each grid is calculated using the following formula:

[0097]

[0098] in, Indicates the statistical uniformity of foreign matter density. Indicates the standard deviation of foreign matter density in the mesh. This represents the average density of foreign objects in the grid. This represents a constant used to prevent division by zero errors;

[0099] The foreign matter density characteristic index is calculated based on the basic foreign matter density and the statistical uniformity. The calculation formula is as follows:

[0100]

[0101] in, Indicates the density characteristic index of foreign matter. Indicates the basic foreign matter density. This represents the distribution influence coefficient. The more uneven the distribution of foreign objects, the greater the impact on the field of vision. It is usually taken as 0.3.

[0102] The pre-trained U-Net convolutional neural network is specifically designed to identify foreign objects such as bone fragments, nucleus pulposus tissue, and ligament fragments. The input is a surgical field image, and the output is an initial mask.

[0103] Step S3 specifically includes the following steps:

[0104] Step S3.1: Calculate the proportion of the surgery already performed to the total estimated time;

[0105] Step S3.2: Using the current time as a reference, analyze the movement pattern and operation frequency of the surgical instruments within the predetermined time period;

[0106] Step S3.3: Calculate the average value of the hemochromatin concentration index within the same predetermined time period as in step S3.2;

[0107] Step S3.4: Based on the feature vector composed of the average values ​​of the ratio, movement pattern, operation frequency and blood color concentration index, identify the current surgical stage through preset rules;

[0108] The predetermined duration is typically set to 30 seconds;

[0109] The surgical phase includes at least the initial entry phase, decompression phase, hemostasis phase, and termination phase, specifically: 1. Initial entry phase: 0-20% of the time, instrument movement is rapid, with minimal bleeding; 2. Decompression phase: 20-80% of the time, instrument movement is delicate, with more tissue debris; 3. Hemostasis phase: triggered when the blood color concentration characteristic index is continuously higher than 0.3 for more than 10 seconds; 4. Termination phase: 80-100% of the time, instrument movement is reduced, focusing mainly on residual tissue removal.

[0110] The motion mode and operating frequency of the surgical instrument are obtained through any one or a combination of the following methods: 1. Real-time identification and tracking of the position and motion trajectory of the instrument tip in the image using an endoscopic image visual tracking algorithm; 2. Acquiring motion data through an inertial measurement unit sensor integrated on the surgical instrument handle or power unit.

[0111] Surgical stage identification is based on the following feature vectors:

[0112]

[0113] in, This represents the feature vector from step S3.4. This indicates the percentage of the total estimated time that the surgery has already taken place. This indicates the average blood color concentration over the past 30 seconds. The coding indicates the movement pattern of the equipment, specifically including 0-stationary, 1-delicate operation, and 2-rapid movement. This indicates the operator's operating frequency, that is, the number of times the instrument moves per unit of time.

[0114] Step S4 specifically includes the following steps:

[0115] Step S4.1: Based on the identified surgical stage, retrieve the basic weight coefficients corresponding to the turbidity, blood color concentration and foreign body density indicators from the preset mapping table;

[0116] Step S4.2: Adjust the weights according to the patient's specific pathological characteristics to obtain the final weight coefficients used for fusion calculation;

[0117] Step S4.3: Using the final weighting coefficients, calculate the visual clarity index using a nonlinear fusion formula. The calculation formula is as follows:

[0118]

[0119] in, Indicates visual acuity index. , and Let represent the nonlinear response functions corresponding to water turbidity, blood color concentration, and foreign matter density index, respectively. , and These represent the weighting coefficients corresponding to water turbidity, blood color concentration, and foreign matter density index, respectively. This represents the superposition effect function when both the turbidity index and the blood color concentration index of the water body exist simultaneously.

[0120] The calculation formulas for the nonlinear response function are as follows:

[0121]

[0122]

[0123]

[0124] in, This represents the size penalty factor, which is 1.2 when Y is greater than 0.1, and 1.0 otherwise.

[0125] The specific calculation method of the superposition effect function is as follows: when both the water turbidity index and the blood color concentration index are greater than 0.3, the value is taken as -0.15 multiplied by the product of the water turbidity index and the blood color concentration index; otherwise, the value is taken as 0.

[0126] The basic weight coefficients corresponding to the preset mapping table are shown in Table 1:

[0127] Table 1:

[0128]

[0129] The adjustment of weights based on the patient's specific pathological characteristics includes: for patients with coagulation disorders, increasing the weight of the hemochromatogram index by 0.15; for patients with spinal stenosis, increasing the weight of the foreign body density index by 0.1; and focusing on bony structures. After the weights are adjusted, they are normalized so that the sum of the three weight values ​​is 1.

[0130] Step S5 specifically includes the following steps:

[0131] Step S5.1: Calculate the deviation between the current visual acuity index and the preset target value;

[0132] Step S5.2: Based on the deviation and real-time physiological parameter data, a set of optimized operation suggestion parameters are calculated and generated through a decision model. The operation suggestion parameters include: flow rate suggestion value and flushing mode suggestion. The flushing modes include: standard continuous flushing, pulsed perturbation flushing and continuous gentle flushing.

[0133] Step S5.3: Perform compliance verification and formatted encapsulation on the operation suggestion parameters, and output them as the intelligent decision data package.

[0134] Specifically, step S5.2 includes the following steps:

[0135] Step S5.2.1: Based on the aforementioned deviation, calculate the flow adjustment amount based on visual assessment to improve visual clarity. The calculation formula is as follows:

[0136]

[0137] in, This represents the flow adjustment based on visual assessment at time t. This represents the deviation between the current visual acuity index at time t and the preset target value. The preset target acuity index ranges from [0.7, 0.9], with a default value of 0.8. This represents the strength coefficient, with a value ranging from 50 to 100 mL / min. This represents the saturation coefficient, with a value range of [0.1, 0.3]. This indicates a sign function; a positive number within the parentheses represents 1, otherwise it represents 0.

[0138] The intensity coefficient value is positively correlated with the expected maximum flow requirement of the surgical channel. For cervical spine surgery with a narrow surgical field and delicate operation, a lower limit value such as 50-60 mL / min is used; for lumbar spine surgery with a larger surgical field and potential for more bleeding, an upper limit value such as 80-100 mL / min is used. A preset surgical type template can be selected during system initialization.

[0139] The smaller the saturation coefficient value, the more sensitive the system is to small deviations and the faster the response, but it may be more prone to fluctuations; the larger the saturation coefficient value, the smoother the system response. It is initially set to 0.2 and can be fine-tuned by ±0.1 in the human-computer interaction interface according to the operator's preference for response speed.

[0140] Step S5.2.2: Calculate the flow correction based on safety constraints using real-time physiological parameter data and preset safety pressure thresholds and maximum permissible pressures. The calculation formula is as follows:

[0141]

[0142] in, This represents the flow correction based on safety constraints at time t. This represents the maximum negative compensation flow, typically set to 30% to 40% of the system's maximum safe flow. This represents the intraspinal pressure monitored at time t. This indicates the safe pressure threshold, which is 30-40 mmHg. This indicates the maximum permissible pressure, typically 60-80 mmHg. The nonlinear compensation function is represented by the following formula:

[0143]

[0144] Step S5.2.3: The traffic adjustment amount based on visual evaluation is superimposed with the traffic correction amount based on security constraints, and then summed with the preset base traffic to obtain a preliminary traffic suggestion value. The calculation formula is as follows:

[0145]

[0146] in, This represents the initial suggested flow rate at time t. This indicates the preset baseline flow rate, such as 30 mL / min, to maintain a minimum level of field of view cleaning;

[0147] Step S5.2.4: Based on the real-time identified surgical stage, visual acuity index, and real-time physiological parameter data, construct a decision tree and provide a flushing mode suggestion from the decision tree output.

[0148] like Figure 5 As shown, mode A is standard continuous flushing, which provides stable and balanced cleaning when there is general visual interference. It is the default mode and is enabled when the triggering conditions of mode B or C are not met.

[0149] Mode B is pulsed perturbation flushing, which utilizes fluid shear force to actively disturb, loosen, and remove attached or aggregated solid foreign objects. When the surgical phase is the "decompression phase," the foreign object density index is greater than 0.35 and the pressure... < Triggered at time;

[0150] Mode C is a continuous, gentle flushing method that prioritizes safety during periods of active bleeding or critical pressure, preventing further disturbance from the water flow or a sudden increase in pressure. It is triggered when any of the following conditions are met: the surgical phase is identified as the "hemostasis phase"; or the real-time pressure... > .

[0151] Step S5.3 specifically includes: to ensure that the suggested parameters are within a reasonable range, the initial traffic suggestion value is subjected to compliance constraint processing based on preset upper and lower limits to obtain the final traffic suggestion parameters. The calculation formula is as follows:

[0152]

[0153] in, This represents the final suggested flow rate parameter at time t. This indicates the preset maximum safe flow rate, typically 120-150 mL / min. This indicates the preset minimum allowable flow rate, typically 10-20 mL / min;

[0154] The constrained traffic recommendation parameters and their metadata are encapsulated into a structured intelligent decision data package.

[0155] like Figure 2 As shown, a multi-source data real-time fusion and decision-making system for surgical field management is used to implement any of the multi-source data real-time fusion and decision-making methods for surgical field management, including:

[0156] The multi-source synchronous data acquisition module is used to acquire multispectral visual data, instrument kinematic data, and physiological parameter data with high precision and synchronously.

[0157] The parallel feature calculation engine module has parallel processing capabilities and is used to perform real-time extraction and quantification calculations of the water turbidity feature index, blood color concentration feature index and foreign matter density feature index.

[0158] The context awareness and analysis module is used to process temporal and motion data to achieve dynamic recognition of surgical stages;

[0159] An adaptive data fusion and evaluation module is used to generate a visual acuity evaluation index based on weight adjustment and multi-feature nonlinear fusion performed according to the surgical stage.

[0160] The intelligent decision parameter generation module is used to generate and output intelligent decision data packages based on the evaluation index and physiological parameter feedback.

[0161] The human-computer interaction and data interface module is used to visualize all intermediate data, evaluation results and decision parameters, and provides data output interface and manual intervention interface.

[0162] The multi-source synchronous data acquisition module includes a multispectral imaging unit. The multispectral imaging unit uses a miniature camera with a diameter of 3.5 mm and integrates three LED light sources with three wavelengths: 415 nm, 540 nm, and white light. It adopts a time-division multiplexing light source control strategy to ensure that the excitation of different wavelength light sources and the camera exposure are synchronized within microsecond-level errors.

[0163] The human-computer interaction and data interface module provides a touch screen interface, allowing surgeons to view, confirm, or modify the suggested parameters in the intelligent decision data package.

[0164] like Figure 3As shown, the intelligent decision generation module receives visual acuity index and real-time physiological parameters, such as spinal canal pressure, from the upstream module in each processing cycle. The module first calculates the deviation from the target value and, based on a nonlinear decision model, calculates a suggested flow adjustment amount to improve visual acuity. Simultaneously, the module performs real-time safety checks on the physiological parameters: if the parameters are normal, the correction amount is zero; if the parameters are too high, a negative flow correction amount is calculated based on a preset nonlinear constraint function, and these are then added together to obtain a preliminary suggested value. This value undergoes compliance verification to ensure it is within preset reasonable parameter limits. Finally, all relevant parameters and metadata are formatted and encapsulated into a structured intelligent decision data package and output to the human-computer interaction and data interface module.

[0165] like Figure 4 As shown, the system first receives raw multispectral image sequences and instrument movement time-series data. The image data is then fed into a parallel feature computation engine. This engine comprises three independent processing paths: a turbidity analysis path calculates texture, light scattering, and color shift components to output a water turbidity feature index; a blood color concentration analysis path analyzes specific spectral reflectance and the HSV color space to output a blood color concentration feature index; and a foreign matter density analysis path segments the object using a pre-trained U-Net network and combines morphological and spatial distribution analysis to output a foreign matter density feature index. Simultaneously, a context awareness and analysis module processes the instrument movement data, dynamically constructs feature vectors, and identifies the current surgical stage in real time through built-in logic. Finally, the three quantified visual feature indices and the identified surgical stage information are output together, providing core input for subsequent data fusion and decision-making.

[0166] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A multi-source data real-time fusion and decision method for surgical field management, characterized in that, The method comprises the following steps: Step S1, synchronously collecting multi-source perception data of a surgical field, the multi-source perception data comprising at least a multispectral image sequence, instrument kinematics data and physiological parameter data; Step S2, performing parallel processing on the multispectral image sequence, extracting and quantifying a water turbidity characteristic index reflecting liquid transparency changes through a first calculation path, extracting and quantifying a blood color concentration characteristic index reflecting blood content through a second calculation path, and extracting and quantifying a foreign body density characteristic index reflecting foreign body distribution states through a third calculation path based on a deep learning segmentation model; Step S3, identifying a surgical stage in real time based on a proportion of a time of surgery performed, the instrument kinematics data and a recent change trend of the blood color concentration characteristic index; Step S4, fusing the water turbidity characteristic index, the blood color concentration characteristic index and the foreign body density characteristic index into a visual clarity evaluation index according to the identified surgical stage; Step S5, calculating an operation recommendation parameter according to a deviation of the visual clarity evaluation index from a preset target value, and generating an intelligent decision data packet comprising the operation recommendation parameter; In the step S2, the first calculation path specifically comprises the following steps: calculating a texture component based on a local binary pattern texture entropy and a gray level co-occurrence matrix contrast based on a white light channel image in the multispectral image sequence; analyzing a diffusion area of a preset reference light spot in the image to obtain a light scattering component; calculating a mean square error of the current image and a dynamically updated reference image in a Lab color space a, b channel and an HSV color space saturation channel to obtain a color shift component; weighting and summing the texture component, the light scattering component and the color shift component according to preset weights to obtain the water turbidity characteristic index; In the step S2, the second calculation path specifically comprises the following steps: calculating average reflectivity of 415 nm channel images and 540 nm channel images in the multispectral image sequence respectively, and calculating a spectral absorption ratio representing hemoglobin specific absorption; converting the white light channel image to the HSV color space, and counting a pixel proportion satisfying preset blood color hue, saturation and brightness threshold ranges; linearly combining the spectral absorption ratio and the pixel proportion according to a preset coefficient to obtain the blood color concentration characteristic index; In the step S2, the third calculation path specifically comprises the following steps: inputting the white light channel image in the multispectral image sequence into a pre-trained U-Net convolutional neural network to obtain an initial mask; performing morphological opening operation and closing operation on the initial mask in sequence to obtain a segmentation mask; calculating an area proportion of foreign body pixels in the segmentation mask as a basic foreign body density; dividing the white light channel image into multiple grids and calculating statistical uniformity of foreign body densities of the grids; calculating the foreign body density characteristic index according to the basic foreign body density and the statistical uniformity.

2. The method of claim 1, wherein, The step S3 specifically comprises the following steps: Step S3.1, calculating a proportion of a time of surgery performed to a total predicted time; Step S3.2, analyzing a motion mode and an operation frequency of a surgical instrument in a recent time window; Step S3.3, calculating the average of the blood color concentration index in the latest time period; Step S3.4, identifying the current surgical stage through a preset rule based on the feature vector composed of the ratio, motion pattern, operation frequency, and the average of the blood color concentration index; The surgical stage at least includes an initial access stage, a decompression stage, a hemostasis stage, and a finishing stage.

3. The method of claim 2, wherein, The step S4 specifically includes the following steps: Step S4.1, querying the basic weight coefficients corresponding to the turbidity, blood color concentration, and foreign matter density indicators from a preset mapping table according to the identified surgical stage; Step S4.2, adjusting the weight according to the specific pathological characteristics of the patient to obtain the final weight coefficient for fusion calculation; Step S4.3, calculating the visual clarity index by a nonlinear fusion formula using the final weight coefficient.

4. The method of claim 3, wherein, The step S5 specifically includes the following steps: Step S5.1, calculating the deviation of the current visual clarity index from a preset target value; Step S5.2, calculating and generating a set of optimized operation recommendation parameters based on the deviation and real-time physiological parameter data through a decision model, wherein the operation recommendation parameters include: a flow rate recommendation value and a flushing mode recommendation, and the flushing mode includes: standard continuous flushing, pulse disturbance flushing, and continuous gentle flushing; Step S5.3, performing compliance verification and format packaging on the operation recommendation parameters, and outputting as the intelligent decision data package.

5. The multi-source data real-time fusion and decision system for surgical field management, which is implemented based on the multi-source data real-time fusion and decision method for surgical field management according to any one of claims 1-4, characterized in that, It includes: A multi-source synchronous data acquisition module for high-precision synchronous acquisition of multispectral visual data, instrument kinematics data, and physiological parameter data; A parallel feature calculation engine module with parallel processing capability for performing real-time extraction and quantitative calculation of the water turbidity feature index, blood color concentration feature index, and foreign matter density feature index; A context awareness and analysis module for processing time series and motion data to achieve dynamic identification of the surgical stage; An adaptive data fusion and evaluation module for performing weight adjustment and multi-feature nonlinear fusion according to the surgical stage to generate a visual clarity evaluation index; An intelligent decision parameter generation module for generating and outputting an intelligent decision data package based on the evaluation index and physiological parameter feedback; A human-computer interaction and data interface module for visualizing all intermediate data, evaluation results, and decision parameters, and providing data output interfaces and manual intervention interfaces.

6. The system of claim 5, wherein, The multi-source synchronous data acquisition module includes a multispectral imaging unit that uses a miniature camera with a diameter of 3.5mm and integrates 415nm, 540nm, and white light three-band LED light sources, adopts a time division multiplexing light source control strategy to ensure that the excitation of different waveband light sources and camera exposure are synchronized within a microsecond error.

7. The system of claim 6, wherein, The human-computer interaction and data interface module provides a touch screen interface that supports the operator to view, confirm, or modify the recommended parameters in the intelligent decision data package.

Images