Interventional procedure intelligent platform management system

Abstract

The present application relates to medical information and intelligent operation management technical field, specifically to an interventional operation intelligent platform management system;Including multi-source data sensing step: collecting medical image data of the object to be operated, real-time running state of interventional equipment and consumable inventory flow data;Anatomy modeling and probability generation step: construct anatomical feature model and generate resource consumption probability vector;Physical coupling prediction step: combine running state and probability vector, predict operation time and equipment thermal load curve, and generate equipment availability evaluation result;Risk quantification and intervention step: calculate the operation continuity risk index, and generate graded intervention instructions when the index exceeds the safety threshold;Interactive response step: display instructions or trigger resource allocation operation;The present application effectively eliminates the unexpected interruption of operation caused by space-time dislocation, and improves the safety and efficiency of operation.

Description

Technical Field

[0001] This invention relates to the field of medical information technology and intelligent surgical management technology, specifically to an intelligent platform management system for interventional surgery. Background Technology

[0002] Existing interventional surgery platform management systems mainly involve monitoring the inventory of surgical consumables and maintaining the operational status of interventional equipment. In related technologies, in order to ensure the supply of surgical resources, static inventory management systems are usually used to count and inventory consumables; or, doctors rely on their subjective experience based on two-dimensional images to estimate the operation time and required instruments, and use the fault detection logic built into the equipment to determine whether the equipment is available.

[0003] However, when faced with cases with complex anatomical structures, the former approach is difficult to accurately match the specific consumable requirements caused by vascular variations, while the latter approach ignores the thermodynamic and physical limits of the equipment under specific surgical loads. Therefore, the management methods in related technologies are difficult to balance the dynamic precision of resource allocation with the continuity of equipment operation, which can easily lead to unexpected interruptions during surgery. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent platform management system for interventional surgery. Specifically, the objective of this invention can be achieved through the following technical solutions:

[0005] Multi-source data acquisition terminal, central processing server, and interactive response terminal;

[0006] The multi-source data acquisition terminal is used to acquire medical image data of the patient to be operated on, real-time operating status data of the interventional device, and consumable inventory flow data, and sends the medical image data, the real-time operating status data, and the consumable inventory flow data to the central processing server.

[0007] The central processing server is used to construct an anatomical feature model based on the medical image data, and to generate a resource consumption probability vector representing the possibility of demand for consumables of different specifications based on the anatomical feature model.

[0008] The central processing server is also used to combine the real-time operating status data with the resource consumption probability vector to predict the operation time and equipment heat load curve, so as to generate equipment availability assessment results.

[0009] The central processing server is also used to calculate a surgical continuity risk index that characterizes the possibility of surgical interruption based on the resource consumption probability vector, the equipment availability assessment results, and the consumable inventory flow data, and to generate a graded intervention instruction and send it to the interactive response terminal when the surgical continuity risk index exceeds a preset safety threshold.

[0010] The interactive response terminal is used to display the tiered intervention instructions or to perform resource allocation operations based on the tiered intervention instructions.

[0011] As a further aspect of the present invention: the central processing server is also used for:

[0012] The medical image data is subjected to three-dimensional reconstruction and visual feature extraction to obtain parameters such as vascular tortuosity, calcification degree, and bifurcation angle.

[0013] The vascular tortuosity parameter, the calcification degree parameter, and the bifurcation angle parameter are mapped to a preset instrument requirement database, wherein the instrument requirement database stores the correspondence between anatomical geometric parameters and consumable specifications established based on historical surgical data.

[0014] Based on the mapping results, the probability distribution of different specifications of interventional consumables used in the current surgery is calculated, which serves as the resource consumption probability vector.

[0015] As a further aspect of the present invention: the central processing server is also used for:

[0016] The surgical complexity coefficient is determined based on the geometric topology of the anatomical feature model;

[0017] Using the surgical complexity coefficient and historical surgical data, the expected duration and expected fluoroscopic dose of the current surgery are predicted;

[0018] The heat accumulation and dissipation process of the X-ray tube of the interventional device during the expected duration is simulated to generate the heat load curve of the device.

[0019] In response to the equipment heat load curve reaching a preset equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be in a state of insufficient cooling, and a scheduling blockage warning is generated;

[0020] If the equipment heat load curve does not reach the equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be equipment ready.

[0021] As a further aspect of the present invention: the central processing server is also used for:

[0022] Real-time monitoring of surgical progress and identification of the current surgical stage;

[0023] Based on the current surgical stage and the resource consumption probability vector, determine the target consumables required for the next stage;

[0024] Retrieve the physical location information of the target consumable from the consumable inventory flow data;

[0025] Calculate the logistics transport time for the target consumable to reach the sterile surgical area based on the physical location information;

[0026] In response to the logistics transmission time being greater than the preset supply buffer time, a logistics allocation instruction is generated as the tiered intervention instruction;

[0027] A supply readiness confirmation signal is generated in response to the logistics transmission time being less than or equal to the supply buffer time.

[0028] As a further aspect of the present invention: the central processing server is also used for:

[0029] Acquire real-time video stream data from the operating room;

[0030] The real-time video stream data is analyzed using computer vision algorithms to identify the doctor's current movement characteristics and the status of equipment use;

[0031] The current surgical stage is updated based on the action characteristics and the instrument usage status.

[0032] As a further aspect of the present invention: the central processing server is also used for:

[0033] Real-time calculation of the temporal overlap probability between the risk of resource supply delay and the risk of physiological variation in demand, as a spatiotemporal misalignment probability;

[0034] The surgical continuity risk index is generated based on the spatiotemporal misalignment probability quantification.

[0035] The surgical continuity risk index is compared with a preset warning threshold and a lockout threshold, wherein the lockout threshold is greater than the warning threshold;

[0036] In response to the surgical continuity risk index being less than or equal to the warning threshold, a normal monitoring signal is generated;

[0037] In response to the surgical continuity risk index being greater than the warning threshold and less than or equal to the lockout threshold, it is determined that the procedure is in a low-risk range, and a warning message is generated.

[0038] If the surgical continuity risk index is greater than the locking threshold, it is determined to be in a high-risk range, and a mandatory locking command is generated to suspend the surgical schedule or trigger emergency resource allocation.

[0039] As a further aspect of the present invention: the central processing server is also used for:

[0040] After the surgery, obtain data on actual consumable consumption and actual surgery duration;

[0041] Calculate the deviation between the actual consumable consumption data and the resource consumption probability vector;

[0042] The algorithm model parameters used to generate the resource consumption probability vector are corrected based on the deviation value.

[0043] As a further aspect of the present invention: the multi-source data acquisition terminal includes:

[0044] Radio frequency identification (RFID) reading module, used to collect the consumable inventory flow data in real time;

[0045] The device's underlying telemetry interface is used to collect the X-ray tube heat capacity data and rack position data of the interventional device;

[0046] An image transmission interface is used to acquire the medical image data.

[0047] Compared with existing technologies, this system achieves precise mapping from patient medical images to consumable needs by constructing an anatomical feature model and combining it with deep learning algorithms; it achieves the effect of scientifically predicting the probability of using specific specifications of consumables based on objective anatomical parameters such as vascular tortuosity, degree of calcification, and bifurcation angle; compared with existing technologies that rely on doctors' subjective experience to estimate or static inventory counting, this invention can identify special instrument needs caused by variations in anatomical structure, and solves the problem of unexpected surgical interruptions caused by experience bias or mismatch between inventory and patient condition. Attached Figure Description

[0048] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 This is a structural diagram of the intelligent platform management system for interventional surgery of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0051] Please see Figure 1 The present invention provides an intelligent platform management system for interventional surgery, comprising: a multi-source data acquisition terminal, a central processing server, and an interactive response terminal;

[0052] Among them, the multi-source data acquisition terminal is used to acquire medical image data of the patient to be operated on, real-time operating status data of the interventional device, and consumable inventory flow data, and send the medical image data, real-time operating status data, and consumable inventory flow data to the central processing server.

[0053] The central processing server is used to build anatomical feature models based on medical image data, and to generate resource consumption probability vectors that characterize the likelihood of demand for consumables of different specifications based on the anatomical feature models.

[0054] The central processing server is also used to combine real-time operating status data with resource consumption probability vectors to predict operation duration and equipment thermal load curves in order to generate equipment availability assessment results.

[0055] The central processing server is also used to calculate the surgical continuity risk index, which characterizes the possibility of surgical interruption, based on the resource consumption probability vector, equipment availability assessment results, and consumable inventory flow data. When the surgical continuity risk index exceeds the preset safety threshold, it generates graded intervention instructions and sends them to the interactive response terminal.

[0056] The interactive response terminal is used to display tiered intervention instructions or to perform resource allocation operations based on tiered intervention instructions.

[0057] This embodiment details the overall architecture of the intelligent interventional platform management system for interventional surgery. The multi-source data acquisition terminal, serving as the system's data perception layer, aims to break down the barriers between the physical and digital worlds, achieving digital mapping of all surgical elements. This terminal accesses the hospital's internal network and the underlying equipment bus in real time via a hardware interface layer to acquire medical imaging data of the patient, including but not limited to raw CTA, MRA, or DSA sequences in DICOM format. Simultaneously, it collects real-time operating status data of the interventional equipment, covering the current heat capacity of the X-ray tube, the physical coordinates and rotation angle of the gantry, and consumable inventory data, i.e., the real-time location, batch number, and expiration date of consumables obtained based on RFID or visual recognition technology. This terminal transmits the aforementioned data to the central processing server via fiber optic cable or a high-bandwidth local area network.

[0058] The central processing server, serving as the core of computing and decision-making, constructs an anatomical feature model based on medical imaging data. This model is a digital geometric topology generated after three-dimensional reconstruction of the surgical site, such as the coronary artery or intracranial vessels, including the centerline coordinates of the vessels, the rate of change of diameter, and spatial curvature information. Simultaneously, the anatomical feature model also includes a surface contour mesh of the surgical object extracted from the medical imaging data. This grid is a set of triangular facets generated by segmenting DICOM data based on skin thresholds and using the moving cube algorithm, used to define the external physical boundary of the surgical area;

[0059] Based on this, the server generates a resource consumption probability vector, which is a multi-dimensional vector. Each dimension corresponds to a specific interventional consumable specification, and the value represents the statistical probability of using the consumable under that anatomical structure. Furthermore, the server introduces physical and physiological coupled computational logic, combining real-time operating status data with the resource consumption probability vector to predict the operation time and equipment thermal load curve, and generate equipment availability assessment results. This assessment does not simply consider the equipment failure status, but calculates whether the current thermal capacity of the equipment can support the expected fluoroscopy time caused by the specific patient's anatomical structure.

[0060] The server calculates the surgical continuity risk index based on resource consumption probability vectors, equipment availability assessment results, and consumable inventory flow data. This indicator has a range of values ​​within The dimensionless scalar is used to quantify the risk of forced shutdown during critical surgical steps due to resource shortages or equipment overheating. In response to the index exceeding the preset safety threshold, the server generates a graded intervention instruction and sends it to the interactive response terminal. This terminal is deployed in the operating room control room and logistics management center to display graded intervention instructions such as pop-up warnings or audible and visual alarms, or to automatically trigger resource allocation operations such as AGV logistics robots based on the instructions.

[0061] This embodiment achieves a leap from static resource management to dynamic surgical continuity assurance by constructing a probabilistic mapping between anatomical features and resource consumption and taking into account the physical limits of the equipment. The system can identify in advance the hidden risks that may lead to equipment overheating or shortage of certain consumables due to the complexity of the patient's anatomy causing the operation time to be prolonged, even if the inventory is in stock and the equipment is fault-free. This effectively eliminates unexpected surgical interruptions caused by spatiotemporal misalignment and significantly improves the safety and efficiency of interventional surgery in complex clinical environments.

[0062] In a preferred embodiment of the present invention, the central processing server is further configured to:

[0063] Three-dimensional reconstruction and visual feature extraction are performed on medical imaging data to obtain parameters such as vascular tortuosity, calcification degree, and bifurcation angle.

[0064] The parameters of vascular tortuosity, calcification degree, and bifurcation angle are mapped to a preset instrument requirement database, which stores the correspondence between anatomical geometric parameters and consumable specifications established based on historical surgical data.

[0065] Based on the mapping results, the probability distribution of different specifications of interventional consumables used in the current surgery is calculated as a resource consumption probability vector.

[0066] This embodiment further specifies the process by which the central processing server generates the resource consumption probability vector. The central processing server performs three-dimensional reconstruction and visual feature extraction on medical image data, and uses a deep learning-based blood vessel segmentation algorithm to extract the blood vessel tree structure. Specifically, the algorithm adopts a V-Net three-dimensional fully convolutional neural network architecture. The input is the original DICOM volume data. Multi-scale features are fused through an encoder-decoder structure and skip connections. The Dice coefficient loss function is used for training to solve the sample imbalance problem where blood vessel pixels account for a very small proportion in the volume data, thereby accurately obtaining the blood vessel centerline and vessel wall boundary.

[0067] Based on the segmentation results, the system calculates key geometric parameters: vascular tortuosity parameter, which is the ratio of the actual length of the vessel centerline to the Euclidean distance; calcification degree parameter, which is the proportion of voxels whose image grayscale value (HU) exceeds the preset calcification threshold; and bifurcation angle parameter, which is the angle between the tangent vectors of the centerlines of the main vessel and the branch vessels; the server maps the above parameters to the preset device demand database.

[0068] To ensure the readability of the computer program and the closed-loop operation logic, this embodiment clearly defines the underlying data structure and construction method of the database: the database is not a simple relational table lookup, but a parameter repository using a key-value pair structure; where the Key is the Global Trade Item Code (GTIN) for the intervention consumables, and the Value is a set of pre-trained regression model parameter tuples. , including feature weight vector and bias scalar These parameters were obtained by training a logistic regression dataset on at least 1000 historical surgical cases, with L2 regularization employed during training. The penalty coefficient... To prevent overfitting, binary cross-entropy is used as the loss function. Gradient descent optimization was performed to ensure that the stored parameters accurately reflected the statistical relationship between anatomical features and consumable usage. This represents the total number of historical surgical data samples. For the first The predicted probability of each sample;

[0069] For training strategies targeting multiple consumable specifications, this embodiment explicitly adopts an independent binary classification training mechanism of one device, one model; specifically, for each GTIN in the database, the system independently constructs a logistic regression model; when training the model corresponding to that GTIN, if the first... If the GTIN is included in the historical surgical consumption list, then the sample will be tagged. Defined as 1, otherwise defined as 0; this processing method effectively solves the multi-label classification problem, enabling the model to output independent probabilities of multiple potential applicable consumables for the same anatomical feature;

[0070] Before performing probability calculations, to eliminate the interference of differences in the physical dimensions of various parameters on the algorithm weights, the system performs Min-Max standardization on all geometric parameters. To ensure the consistency and reproducibility of the model input, this embodiment uses a preset global statistical boundary as the normalization benchmark, rather than relying solely on the maximum or minimum value of a single sample. Specifically, the boundary range of the vascular tortuosity parameter is set as follows: based on physiological limit statistics. The range of calcification degree parameters is: The range of bifurcation angle parameters is: Degree; the normalization formula is defined as

[0071]

[0072] in, These are the currently measured anatomical parameter values. and These are the lower and upper limits of the corresponding parameters in the aforementioned statistical boundaries, for example, for vascular tortuosity. ;

[0073] The `clip` function is used to truncate outliers that exceed statistical boundaries to the endpoints of an interval; specifically, its mathematical definition is: when the input value... When, return ;when When, return ;when When, return itself;

[0074] Based on the mapping and normalization results, that is, using the data retrieved from the database... With real-time computing Calculate the probability distribution of different specifications of interventional consumables used in the current surgery, as the resource consumption probability vector; in this embodiment, the first... The probability of using certain types of consumables as follows:

[0075]

[0076] in, The probability of using a specific type of consumable is derived from the model output, and its value range is... ; : The normalized anatomical feature vector, which includes the processed parameters of vascular tortuosity, calcification degree and bifurcation angle; : Feature weight vector, derived from the Value field in the database, physically represents the sensitivity of the consumable to each anatomical feature; : Bias item, derived from the Value field in the database, physically represents the basic usage rate of consumables of this specification; Sigmoid activation function;

[0077] This embodiment directly maps objective anatomical geometric features to the probability of consumable demand, and strictly defines the physical boundaries of the features, the normalization method, and the parameter storage structure of the database, effectively eliminating the bias of estimation based solely on the doctor's experience; especially for complex lesions such as severe calcification or large-angle bifurcation, the system can accurately predict the demand for special instruments such as rotational burrs or double-lumen catheters, ensuring that instruments are found and avoiding critical steps caused by temporarily searching for special consumables during the operation, greatly improving the accuracy of surgical preparation;

[0078] The central processing server is also used for:

[0079] The surgical complexity coefficient is determined based on the geometric topology of the anatomical feature model;

[0080] Using surgical complexity coefficients and historical surgical data, predict the expected duration and fluoroscopic dose of the current surgery;

[0081] The heat accumulation and dissipation process of the X-ray tube of the interventional device is simulated during the expected duration to generate the device's heat load curve;

[0082] In response to the equipment heat load curve reaching the preset equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be in a state of insufficient cooling, and a scheduling blockage warning is generated;

[0083] If the equipment heat load curve does not reach the equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be equipment ready.

[0084] This embodiment further specifies the implementation logic of equipment availability assessment, particularly the accurate generation process of the heat load curve; the central processing server determines the surgical complexity coefficient based on the geometric topology of the anatomical feature model; the specific calculation logic involves constructing a weighted evaluation model. ,in, This represents the total number of feature factors participating in the evaluation, in this embodiment. , Represents the normalized version of the first... Each of the following topological feature factors corresponds specifically to the length of the diseased vessel segment, the number of vessel bifurcations, the maximum curvature value, and the length of the CTO occluded segment. To ensure the accuracy of feature alignment during matrix operations by the computer program, this embodiment explicitly defines the index mapping relationship of the feature factors: Mapped to the length of the diseased vessel segment. This is mapped to the number of blood vessel bifurcations. Mapped to the maximum curvature value, The mapping is to the length of the CTO occlusion segment; the selection of the above four characteristic factors is based on the lesion complexity assessment criteria published by the American Society of Cardiology and Interventional Angiography, and these characteristics have been shown to have a significant statistical correlation with the amount of equipment consumed in interventional procedures.

[0085] Regarding weight To avoid uncertainty caused by subjective assignment, this embodiment uses the entropy weight method to objectively calculate based on historical data.

[0086] The specific steps are as follows: Construct a standardized evaluation matrix: Select... For example, that is Historical surgical data were used to construct the original matrix. For the positive index in this embodiment, i.e., the larger the value, the more complex it is, the range transformation method is used for standardization: This eliminates the influence of dimensions; calculates the characteristic weight: calculates the first The sample at the th Contribution under each indicator ,in, For the first The first sample The original value of the indicator, For the first The set of values ​​for the item index across all samples. For the first The sample at the th The standardized value under each indicator The corresponding feature weights; during this calculation process, the system has built-in logic to prevent division by zero: if the denominator If the indicator remains unchanged across all samples, then a mandatory order is issued. This is to ensure the numerical stability of the normalization process;

[0087] Calculate information entropy:

[0088]

[0089] Among them, when At that time, it was stipulated ; Calculate the weights: In the specific configuration of this embodiment, the weight vector calculated by the above algorithm is: lesion length weight. Fork count weight Maximum curvature weight CTO length weight ;

[0090] Using this coefficient and historical surgical data, the estimated duration and fluoroscopic dose of the current surgery are predicted. Specifically, this prediction process is based on a gradient boosting decision tree (GBDT) model, with inputs including a surgical complexity coefficient and the physician's historical efficiency index. Based on the doctor's past For example, that is The normalized index is calculated from similar surgical data, and the formula is:

[0091]

[0092] in, The standard average duration for this type of surgery is defined in this embodiment as the arithmetic mean of the durations of all historical surgeries of the same type stored in the database, and is updated quarterly. For the first The actual surgical time is used to correct for the model's personalized prediction bias regarding the operating speed of a specific surgeon. The specific prediction process is divided into two stages: baseline prediction and personalized correction. The surgical complexity coefficient is then used to adjust the model's prediction. The input feature vectors are fed into the pre-trained GBDT model, which outputs the baseline estimated duration of this type of surgery. Using the doctor's historical efficiency index The baseline duration is weighted and adjusted to obtain the final estimated operation time. The calculation formula is:

[0093]

[0094] Similarly, the expected total fluoroscopic dose The same logic was used for the correction;

[0095] Based on this, in order to generate the expected fluoroscopic dose rate curve for the entire process The system employs a complexity-based time-domain warping mapping technique: utilizing the K-Means algorithm, where, By combining the dynamic time warping (DTW) distance metric, historical surgical dose curves were clustered into five standard templates. The system retrieves the value from the database and matches the current complexity coefficient. Historical standard surgical dose curve templates for the same cluster ;

[0096] The specific determination rule is as follows: calculate the current complexity coefficient. With pre-stored The Euclidean distance between the cluster centers is used to select the template corresponding to the cluster center with the smallest distance as the target. ;in, For the normalized time axis; the system establishes a normalized time axis Timeline to actual forecast linear mapping relationship Based on this, the system uses the WarpingPath generated by the DTW algorithm to modify historical templates. The typical feature points in the image are aligned to ensure that nonlinear features are in place. Logical consistency over time;

[0097] Considering that the time integral of the dose rate should equal the total dose, and that the prediction time axis has undergone scaling, this embodiment has made a dimensional correction to the mapping formula:

[0098]

[0099] Among them, the integral term The normalized total area under the standard template curve is used as a normalization factor to ensure total dose conservation; the denominator is introduced... This is to offset the integral area gain caused by time axis stretching, ensuring This processing method can accurately reproduce the nonlinear temporal characteristics of intermittent fluoroscopy-continuous angiography-high-power exposure during surgery, rather than a simple mean distribution, thus providing realistic input for thermal simulation.

[0100] The heat accumulation and dissipation process of the X-ray tube of the interventional device is simulated during the expected duration to generate a heat load curve for the device. This embodiment uses the following thermodynamic model to calculate the timing. X-ray tube heat capacity :

[0101]

[0102] in, The amount of heat accumulated in the X-ray tube at time t, expressed in HU. The initial heat capacity before surgery begins is usually set to 0 or the residual heat after the previous surgery is read on the device. : Input power generates heat, i.e. Considering the nonlinear variation of X-ray heat generation efficiency with tube voltage, This is the dose-power conversion factor. Since we are in the preoperative prediction stage and cannot obtain real-time equipment tube voltage, this embodiment uses physical simulation calculation based on anatomical features: the system calculates the equivalent tissue thickness (in cm) of the X-ray beam passing through the target blood vessel region based on the constructed anatomical feature model. The specific calculation logic is as follows: obtain the physical coordinates of the current interventional equipment rack to determine the ray projection vector. Calculate vectors using ray casting algorithm With body surface contour grid The two intersection points, i.e., the ray incident points. With the point of ray emission The equivalent tissue thickness is defined as the Euclidean distance between two points:

[0103]

[0104] And based on the automatic exposure control AEC characteristic curve of the equipment. Predicting the transistor voltage; to ensure the executability and numerical uniqueness of this mapping relationship in the computer program, this embodiment uses the characteristic curve. The mathematical expression for is explicitly defined as the following piecewise function:

[0105]

[0106] in, The unit is cm. The unit is kV; the function is clearly defined across all its domains: when the equivalent tissue thickness is less than or equal to 10 cm, the function value is strictly locked at a constant of 60 kV; when it is greater than 30 cm, it is locked at a constant of 110 kV; in the linear transition region from 10 cm to 30 cm, the function guarantees numerical continuity at the boundary points, i.e., at 10 cm and 30 cm, for example... This avoids abrupt changes during the calculation process;

[0107] Then using physical formulas

[0108]

[0109] Dynamically generate conversion factors, where, HU / mGy is the reference voltage. The baseline value is used to ensure the physical authenticity of the heat input; Heat dissipation power, using a nonlinear function Expression; coefficient This involves parsing the electronic data sheet of the access device and matching the specific thermal constants of the model from a pre-set physical parameter library of the device; taking the experimental model involved in this embodiment as an example, The value is 0.045. , Values ;

[0110] In order to transform the above continuous-time integral equation into computer-executable discrete instructions and ensure that those skilled in the art can reproduce the generation process of the curve, this embodiment uses the fourth-order Runge-Kutta method (RK4) for numerical iterative solution; compared with the Euler method, RK4 can more effectively suppress the cumulative error over a long time span.

[0111] Set the calculation time step Define the net heat power function Then from arrive The logic for updating the hot capacity state at any given moment is as follows:

[0112] Calculate the initial slope:

[0113] Calculate the midpoint prediction slope A:

[0114] Calculate the midpoint prediction slope B:

[0115] Calculate the final slope:

[0116] Update heat capacity value:

[0117]

[0118] Through the above iterative algorithm, the system can accurately track the dynamic changes of heat load in the X-ray tube under nonlinear heat dissipation conditions;

[0119] If the equipment's heat load curve reaches a preset equipment heat capacity limit threshold within the expected duration, the system determines the equipment availability assessment result as insufficient cooling and generates a scheduling interruption warning; the preset equipment heat capacity limit threshold is... It is not set arbitrarily, but rather is the physical rating value obtained by the system through the device's underlying interface from the AnodeHeatContentRating field in its DICOM conformance declaration file. For example, this value is typically found in an angiography machine. HU; The system typically reserves a 5% safety margin when making judgments, meaning the actual judgment threshold is... .

[0120] In a preferred embodiment of the present invention, the central processing server is further configured to:

[0121] Real-time monitoring of surgical progress and identification of the current surgical stage;

[0122] Based on the current surgical stage and the probability vector of resource consumption, determine the target consumables required for the next stage;

[0123] Retrieve the physical location information of the target consumable from the consumable inventory flow data;

[0124] Calculate the logistics transportation time of the target consumables to the sterile surgical area based on physical location information;

[0125] In response to logistics transmission time exceeding the preset supply buffer time, a logistics allocation instruction is generated as a tiered intervention instruction;

[0126] A supply readiness confirmation signal is generated in response to a logistics transmission time being less than or equal to the supply buffer time.

[0127] The central processing server is also used for:

[0128] Acquire real-time video stream data from the operating room;

[0129] Using computer vision algorithms to analyze real-time video stream data, we can identify the doctor's current movement characteristics and the status of equipment use;

[0130] Update the current surgical stage based on motion characteristics and instrument usage status.

[0131] This embodiment relates to intelligent navigation of dynamic logistics during surgery and visual recognition during the surgical phase; the central processing server acquires real-time video stream data from the operating room and analyzes the data using computer vision algorithms; in specific implementation, the system adopts a Two-Stream CNN architecture, in which the spatial stream network is based on the YOLO algorithm to detect the bounding boxes and positions of instruments such as catheters, guidewires, and syringes in real time, and the temporal stream network is based on 3D-CNN, such as the I3D model, to extract the spatiotemporal features of the doctor's hand movements. The system inputs the detected instrument states and movement features into a Hidden Markov Model (HMM) for temporal decoding;

[0132] To ensure the algorithm's feasibility, this embodiment explicitly defines the state space and observation space of the HMM: defining the set of hidden states. : Puncture establishment, The angiography catheter is in place. Road map imaging, Microguidewire passage, balloon or stent delivery, Release and Post-expansion The surgical stage corresponding to the embodiment; define the observation vector. This involves concatenating the device category confidence vector output by the spatial flow CNN with the action classification probability vector output by the temporal flow CNN; to construct complete HMM parameters, this embodiment explicitly defines the emission probability. The computational logic is as follows: The inverse Bayesian transform is used to transform the posterior probability vector output by the CNN. Convert to likelihood, i.e. ,in, To determine the state at the current moment for the neural network confidence level This represents the prior distribution probability of this state in historical data; coupled with a forcibly set upper triangular state transition matrix. To ensure that the surgical procedure is irreversible, this embodiment specifically defines a matrix. elements The acquisition method is as follows: from the historical data of the period Jump to stage frequency ,for Reverse jump command To comply with the one-way surgical procedure;

[0133] Calculate the transition probability Among them, the summation term in the denominator Explicitly includes In the case of a situation where the frequency of state self-maintenance is included; when the system is initialized without historical data, the default setting is used: the diagonal element represents the self-maintenance probability. The second diagonal element represents the probability of sequential flow. All other elements are 0; the model calculates the maximum a posteriori probability path using the Viterbi algorithm, thereby determining the current path in real time. This allows for precise identification of the doctor's current actions, such as stepping on a foot switch, injecting contrast agent, and the status of instruments used, and updates the current surgical stage accordingly, such as puncture, angiography, guidewire passage, or stent release.

[0134] Based on this, the target consumables required for the next stage are determined according to the probability vector of the current surgical stage and resource consumption. To address the logical problem of how the probability vector is mapped to a specific surgical stage, this embodiment pre-configures a surgical stage-consumable category mapping matrix in the server. To improve retrieval efficiency, this The matrix is ​​stored in memory using a hash mapping data structure, with the key being the surgical stage status code and the value being a list of the corresponding consumable categories;

[0135] For example, when the current stage is identified as That is, when the microguidewire passes through, the next stage is... That is, balloon / stent delivery, the system queries The target consumable category is coronary stents;

[0136] The system iterates through all components belonging to the coronary stent category in the previously generated resource consumption probability vector and selects the specification with the highest probability value; it retrieves the physical location information of the consumable in the consumable inventory flow data and calculates the logistics transmission time for the consumable to reach the sterile surgical area; this embodiment adopts a tiered setting strategy for the preset supply buffer time, and the system stores a mapping table between surgical stages and buffer times, for example: 15 minutes for the low-risk angiography stage; 5 minutes for the high-risk stent release stage; the system executes the following logical judgment:

[0137]

[0138] in, Logistics transmission time is calculated based on physical location and AGV real-time speed. Based on the current stage of surgery The system retrieves the preset supply buffer time from the table. If the logistics transmission time is longer than the preset supply buffer time, it indicates that the current logistics speed is lagging behind the urgency requirements of the current surgical stage. The system generates a logistics allocation instruction as a graded intervention instruction to trigger logistics delivery in advance. Conversely, it generates a supply readiness confirmation signal.

[0139] This embodiment constructs a zero-latency supply chain system; by visually recognizing the progress of surgery in real time and combining the time difference with the physical location of inventory, the system can predict potential supply delays and automatically trigger allocation before the doctor realizes that a certain consumable is needed. This ensures that when the doctor reaches out for the instrument, the instrument is already next to the sterile table, achieving precise synchronization between the flow of materials and the flow of surgery, and greatly reducing the waiting time during surgery.

[0140] In a preferred embodiment of the present invention, the central processing server is further configured to:

[0141] Real-time calculation of the temporal overlap probability between the risk of resource supply delay and the risk of physiological variation in demand, as a spatiotemporal misalignment probability;

[0142] A surgical continuity risk index is generated based on the probability quantification of spatiotemporal misalignment.

[0143] The surgical continuity risk index is compared with preset warning thresholds and lockout thresholds, where the lockout threshold is greater than the warning threshold.

[0144] A normal monitoring signal is generated in response to the surgical continuity risk index being less than or equal to the warning threshold;

[0145] If the surgical continuity risk index is greater than the warning threshold but less than or equal to the lockout threshold, the procedure is determined to be in a low-risk range, and a warning message is generated.

[0146] If the surgical continuity risk index exceeds the lockout threshold, indicating a high-risk zone, a mandatory lockout command is generated to suspend the surgical schedule or trigger emergency resource allocation.

[0147] This embodiment details the calculation and tiered intervention mechanism of the surgical continuity risk index, clarifying the mathematical definitions of each probability indicator and the source of key thresholds. The central processing server calculates in real time the temporal overlap probability of resource supply delay risk and physiological variation demand risk as the spatiotemporal misalignment probability. To satisfy this logic, this embodiment constructs the following hierarchical probability model: defining the physiological variation demand risk as the resource consumption probability. , which corresponds to the j-th component in the aforementioned resource consumption probability vector; where, The logistics transmission time is calculated using the following formula: ,in, This represents the path planning and travel time of the AGV robot. This represents the estimated time for manual pickup and handover.

[0148] Define the risk of resource supply delay Logistics transit time in minutes The safety buffer time is measured in minutes. Sigmoid probability mapping:

[0149]

[0150] in, For the unit The delay sensitivity coefficient is explicitly set to in this embodiment. The selection of this parameter is based on clinical risk gradient testing: when the actual transmission time Exceeding the buffer time At 5 minutes, substituting into the formula yields the following result. In other words, the system classifies this level of delay as a near-deterministic risk event, thereby driving the model to respond quickly to timeout behavior; the spatiotemporal misalignment probability of the above two factors is calculated. That is, the probability of time overlap, based on the joint probability formula for independent events: ;

[0151] Surgical continuity risk index generated based on spatiotemporal misalignment probability quantification. To fully cover the aforementioned dual risk considerations based on resource consumption and equipment availability, this embodiment logically completes the risk index model by introducing an equipment heat load risk term; peak heat capacity is extracted from the heat load curve. And combined with the heat capacity limit Calculate the probability of thermal failure of equipment

[0152]

[0153] in, The steepness coefficient is obtained by fitting the thermal characteristic curve of the experimental model. This is a dimensionless safety threshold ratio, and its physical meaning represents the proportion of the peak heat capacity of the X-ray tube. The critical inflection point, i.e., the center offset of the Sigmoid function, is such that when the ratio exceeds this threshold, the probability of thermal failure increases exponentially. An exponential model is constructed using probabilistic union logic to characterize the combined risk of system interruption due to consumable shortages or equipment overheating.

[0154]

[0155] in, The total number of key consumables involved in this surgical procedure; the selection criteria for key consumables are clearly defined as: only those with a high probability of resource consumption. Consumables are included in the calculation scope to prevent low-probability noise interference;

[0156] : represents the Index subscripts for key consumables;

[0157] item Characterizes the probability that the equipment is in a thermally safe state;

[0158] item : Represents the probability that all key consumables have not experienced spatiotemporal misalignment;

[0159] Constant 100: used to... The probability values ​​of the interval are mapped as follows Integer scale;

[0160] The system compares the calculated R with preset warning and lockout thresholds. Regarding the acquisition of these two key thresholds, this embodiment does not set them arbitrarily, but rather uses statistical boundaries derived from receiver operating characteristic (ROC) curve analysis of historical surgical interruption events. To ensure the reproducibility of the threshold settings, this embodiment defines in detail the ROC curve construction process: a sample set is constructed by extracting no fewer than 2000 historical surgical records from the hospital's historical database, and a ground truth label, GroundTruth, is defined for unexpected surgical interruption events: if the surgical record clearly states that the surgical pause due to insufficient consumables or C-arm tube overheating resulted in a waiting time exceeding 10 minutes, it is marked as a positive sample. Otherwise, it is marked as a negative sample, i.e. ;

[0161] Using the risk index calculation logic described above in this manual, a retrospective calculation is performed on each surgery in the sample set to obtain the corresponding risk index value. During this backtracking calculation process, the system strictly adheres to the principle of predictive model consistency: [The following is a separate, unrelated sentence:] used for calculation... Input parameters, such as estimated operation duration, estimated fluoroscopy dose, and related parameters. All predictions are based on the preoperative medical imaging data of the historical case and are regenerated by calling the prediction algorithm, rather than directly using the posterior data of the actual occurrence of the case, such as the actual operation time. That is, during the retrospective calculation, the system will forcibly shield the actual end timestamp and actual radiation dose record of the historical case to ensure that the data input into the model is limited to the information available before the operation, and to prevent overfitting caused by future information leakage.

[0162] This process ensures that the threshold determined by the ROC curve can truly reflect the system's ability to identify risks in the preoperative stage, rather than overfitting to historical results.

[0163] Using a step size of 1, iterate through the truncation thresholds within the range of 0 to 100. For each Values, statistical sample sets and True positive count TP and and The number of false positives (FP) is calculated, and then the true positive rate (TPR) and false positive rate (FPR) are calculated and an ROC curve is plotted. Based on this curve, a warning threshold is determined. Set as the Youden Index on the ROC curve, i.e. The risk score corresponding to the maximum point aims to maximize the balance between sensitivity and specificity. In this embodiment, it is calculated... Lock threshold Specificity is set as the specificity on the ROC curve, i.e. The risk score corresponding to 99% is designed to ensure that mandatory locking is triggered only when the risk is almost certain. In this embodiment, it is calculated... ;

[0164] When R is less than or equal to the warning threshold, a normal monitoring signal is generated; when R is between the warning threshold and the lock threshold, it is determined to be a low-risk zone and a warning message is generated; when R is greater than the lock threshold, it is determined to be a high-risk zone and a mandatory lock command is generated.

[0165] The central processing server is also used for:

[0166] After the surgery, obtain data on actual consumable consumption and actual surgery duration;

[0167] Calculate the deviation between actual material consumption data and the resource consumption probability vector;

[0168] The algorithm model parameters used to generate the resource consumption probability vector are corrected based on the deviation value.

[0169] This embodiment involves the system's adaptive evolution capability; after the surgery, the server obtains the actual consumable consumption data, denoted as a vector. And actual surgical time data; calculate the probability vector of actual consumable consumption data and preoperative predicted resource consumption, denoted as... The deviation value between them is calculated, for example, using the cross-entropy loss function; based on this deviation value, the algorithm model parameters used to generate the resource consumption probability vector are corrected using the backpropagation algorithm, i.e., the weight w in the aforementioned embodiment;

[0170] To ensure the stability and mathematical rigor of model parameter updates, this embodiment specifically employs the Online Stochastic Gradient Descent (OnlineSGD) algorithm for correction: For the logistic regression model, based on the gradient derivation of the loss function with L2 regularization, the weight vector... The update formula is defined as:

[0171]

[0172] in, The preset learning rate, The L2 regularization coefficient is the term. Weight decay was implemented to prevent overfitting. The residual scalar between the predicted probability and the actual tag used. This is the normalized anatomical feature vector for the patient; similarly, the bias term is updated to...

[0173]

[0174] in, For the updated bias term, The original bias term before the update; this specific mathematical feedback mechanism ensures that the system can fine-tune the model boundary using each real surgical data, thereby achieving gradual convergence of prediction accuracy;

[0175] This embodiment uses a closed-loop feedback mechanism to enable the system to continuously learn the operating habits and preferences of the hospital's doctors as the number of surgeries accumulates. This means that the longer the system is used, the more accurate its prediction of which doctor will choose instruments under specific anatomical structures, thereby continuously reducing prediction bias and significantly improving the matching efficiency and intelligence level of the supply chain.

[0176] In a preferred embodiment of the present invention, the multi-source data acquisition terminal includes:

[0177] Radio frequency identification (RFID) reading module, used to collect consumable inventory flow data in real time;

[0178] The device's underlying telemetry interface is used to collect data on the X-ray tube heat capacity and rack position of the interventional equipment;

[0179] Image transmission interface, used for acquiring medical image data.

[0180] This embodiment details the hardware configuration of the multi-source data acquisition terminal; the RFID reading module is deployed in the intelligent consumable cabinet and the operating room ceiling, using the UHF band to collect consumable inventory flow data in real time, enabling contactless batch inventory and location tracking; the device's underlying telemetry interface connects to the DSA server room via CAN bus or DICOM-RT protocol, collecting X-ray tube heat capacity data (anode heat capacity percentage) and rack position data (C-arm angle) of the interventional equipment in read-only mode, ensuring data real-time performance at the millisecond level; the image transmission interface uses the DICMC-MOVE instruction set to collect patients' preoperative medical image data through the image archiving and communication system PACS gateway.

[0181] This embodiment ensures the accuracy and timeliness of the data through specific hardware selection; in particular, the introduction of the underlying telemetry interface enables the system to obtain physical parameters of the equipment that are inaccessible to traditional IT systems, such as heat capacity data, providing indispensable physical boundary conditions for the system's thermodynamic simulation and ensuring the physical authenticity of the risk assessment model.

[0182] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent platform management system for interventional surgery, characterized in that, include: Multi-source data acquisition terminal, central processing server, and interactive response terminal; The multi-source data acquisition terminal is used to acquire medical image data of the patient to be operated on, real-time operating status data of the interventional device, and consumable inventory data, and sends the medical image data, the real-time operating status data, and the consumable inventory data to the central processing server. The central processing server is used to construct an anatomical feature model based on the medical image data, and to generate a resource consumption probability vector representing the possibility of demand for consumables of different specifications based on the anatomical feature model. The central processing server is also used to combine the real-time operating status data with the resource consumption probability vector to predict the operation time and equipment heat load curve, so as to generate equipment availability assessment results. The central processing server is also used to calculate a surgical continuity risk index that characterizes the possibility of surgical interruption based on the resource consumption probability vector, the equipment availability assessment results, and the consumable inventory data, and to generate a graded intervention instruction and send it to the interactive response terminal when the surgical continuity risk index exceeds a preset safety threshold. The interactive response terminal is used to display the tiered intervention instructions or to perform resource allocation operations based on the tiered intervention instructions. The central processing server is also used for: The medical image data is subjected to three-dimensional reconstruction and visual feature extraction to obtain parameters such as vascular tortuosity, calcification degree, and bifurcation angle. The vascular tortuosity parameter, the calcification degree parameter, and the bifurcation angle parameter are mapped to a preset instrument requirement database, wherein the instrument requirement database stores the correspondence between anatomical geometric parameters and consumable specifications established based on historical surgical data. Based on the mapping results, the probability distribution of different specifications of interventional consumables used in the current surgery is calculated, which serves as the resource consumption probability vector.

2. The intelligent platform management system for interventional surgery according to claim 1, characterized in that, The central processing server is also used for: The surgical complexity coefficient is determined based on the geometric topology of the anatomical feature model; Using the surgical complexity coefficient and historical surgical data, the expected duration and expected fluoroscopic dose of the current surgery are predicted. The heat accumulation and dissipation process of the X-ray tube of the interventional device during the expected duration is simulated to generate the heat load curve of the device. In response to the equipment heat load curve reaching a preset equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be in a state of insufficient cooling, and a scheduling blockage warning is generated; If the equipment heat load curve does not reach the equipment heat capacity limit threshold within the expected duration, the equipment availability assessment result is determined to be equipment ready.

3. The intelligent platform management system for interventional surgery according to claim 1, characterized in that, The central processing server is also used for: Real-time monitoring of surgical progress and identification of the current surgical stage; Based on the current surgical stage and the resource consumption probability vector, determine the target consumables required for the next stage; Retrieve the physical location information of the target consumable from the consumable inventory data; Calculate the logistics transport time for the target consumable to reach the sterile surgical area based on the physical location information; In response to the logistics transmission time being greater than the preset supply buffer time, a logistics allocation instruction is generated as the tiered intervention instruction; A supply readiness confirmation signal is generated in response to the logistics transmission time being less than or equal to the supply buffer time.

4. The intelligent platform management system for interventional surgery according to claim 3, characterized in that, The central processing server is also used for: Acquire real-time video data from the operating room; The real-time video data is analyzed using computer vision algorithms to identify the doctor's current movement characteristics and the status of equipment use. The current surgical stage is updated based on the action characteristics and the instrument usage status.

5. The intelligent platform management system for interventional surgery according to claim 1, characterized in that, The central processing server is also used for: Real-time calculation of the temporal overlap probability between the risk of resource supply delay and the risk of physiological variation in demand, as a spatiotemporal misalignment probability; The surgical continuity risk index is generated based on the spatiotemporal misalignment probability quantification. The surgical continuity risk index is compared with a preset warning threshold and a lockout threshold, wherein the lockout threshold is greater than the warning threshold; In response to the surgical continuity risk index being less than or equal to the warning threshold, a normal monitoring signal is generated; In response to the surgical continuity risk index being greater than the warning threshold and less than or equal to the lockout threshold, it is determined that the procedure is in a low-risk range, and a warning message is generated. In response to the surgical continuity risk index being greater than the locking threshold, it is determined to be in a high-risk range, and a mandatory locking command is generated to suspend the surgical schedule or trigger emergency resource allocation. The probability of temporal overlap between the risk of delay in the supply of real-time computing resources to the central processing server and the risk of demand due to physiological variations is used as the probability of spatiotemporal misalignment. Define physiological variation demand risk as the probability of resource consumption. , which corresponds to the j-th component in the aforementioned resource consumption probability vector; where, The logistics transmission time is calculated using the following formula: ,in, This represents the path planning and travel time of the AGV robot. This represents the estimated time for manual pickup and handover. Define the risk of resource supply delay Logistics transit time in minutes The safety buffer time is measured in minutes. Sigmoid probability mapping: ； in, For the unit The delay sensitivity coefficient is explicitly set to... The selection of this parameter is based on clinical risk gradient testing: when the actual transmission time Exceeding the buffer time At 5 minutes, substituting into the formula yields the following result. In other words, the system classifies this level of delay as a near-deterministic risk event, thereby driving the model to respond quickly to timeout behavior; the spatiotemporal misalignment probability of the two is calculated. That is, the probability of time overlap, based on the joint probability formula for independent events: ; Surgical continuity risk index generated based on spatiotemporal misalignment probability quantification. To fully cover the dual risks based on resource consumption and equipment availability, the risk index model was logically supplemented by introducing an equipment heat load risk term; peak heat capacity was extracted from the heat load curve. And combined with the heat capacity limit Calculate the probability of thermal failure of equipment ； in, The steepness coefficient is obtained by fitting the thermal characteristic curve of the experimental model. This is a dimensionless safety threshold ratio, and its physical meaning represents the proportion of the peak heat capacity of the X-ray tube. The critical inflection point, i.e., the center offset of the Sigmoid function, is such that when the ratio exceeds this threshold, the probability of thermal failure increases exponentially. An exponential model is constructed using probabilistic union logic to characterize the combined risk of system interruption due to consumable shortages or equipment overheating. ； in, The total number of key consumables involved in this surgical procedure; the selection criteria for key consumables are clearly defined as: only those with a high probability of resource consumption. Consumables are included in the calculation scope to prevent low-probability noise interference; : represents the Index subscripts for key consumables; item Characterizes the probability that the equipment is in a thermally safe state; item : Represents the probability that all key consumables have not experienced spatiotemporal misalignment; Constant 100: used to... The probability values ​​of the interval are mapped as follows Integer scale.

6. The intelligent platform management system for interventional surgery according to claim 1, characterized in that, The central processing server is also used for: After the surgery, obtain data on actual consumable consumption and actual surgery duration; Calculate the deviation between the actual consumable consumption data and the resource consumption probability vector; The algorithm model parameters used to generate the resource consumption probability vector are corrected based on the deviation value.

7. The intelligent platform management system for interventional surgery according to any one of claims 1-6, characterized in that, The multi-source data acquisition terminal includes: Radio frequency identification (RFID) reading module, used to collect the consumable inventory data in real time; The device's underlying telemetry interface is used to collect the X-ray tube heat capacity data and rack position data of the interventional device; An image transmission interface is used to acquire the medical image data.

Images