CN-121983214-A - Urinary surgery data management system based on cloud platform

CN121983214ACN 121983214 ACN121983214 ACN 121983214ACN-121983214-A

Abstract

The invention relates to the technical field of operation data management, in particular to a cloud platform-based urinary operation data management system. According to the invention, through calculating pixel differences of adjacent frames in an original surgical video, a representative reference frame sequence is automatically screened and established, and then an average tissue reference texture map is constructed based on the sequence, when an intra-operative picture has tissue abnormality, the intra-operative picture can be intelligently classified and corresponding surgical event metadata is generated, the identified and filtered key events are associated with the accurate index of the original video frame, and are bound into a unified surgical event ID and archived for storage, so that a retrievable structured data asset is formed, the unstructured state of the surgical video is fundamentally changed, and the analysis and utilization depth and retrieval efficiency of surgical image data are remarkably improved.

Inventors

YU ZHOU
ZHANG TIANGE
HUANG XIANGJIANG
Yuan Yeqing
GUO JINAN

Assignees

深圳爱递医药科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260225

Claims (10)

1. Urinary surgery data management system based on cloud platform, characterized in that it comprises: the intraoperative image structuring module acquires an original video frame sequence of urinary surgery, calculates pixel differences among frames, screens video frames with the differences lower than a threshold value to be a reference frame sequence, and transmits the reference frame sequence to the physiological characteristic referencing module; the physiological characteristic referencing module divides a video frame into grid matrixes based on the reference frame sequence to calculate tissue texture characteristic values, calculates the average value of the tissue texture characteristic values at the same position of each grid matrix to generate a tissue reference texture map, and transmits the tissue reference texture map to the operation field event intelligent recognition module; the intelligent recognition module of the surgical field event generates a current texture map of a current frame, calls a support vector machine model to use the organization reference texture map as a reference class, generates surgical event metadata if a classification result is an organization abnormality, and transmits the surgical event metadata to the data asset archiving management module; and the data asset archiving management module filters the surgical event metadata with the quantity lower than a frequency threshold, establishes index association with the original video frame sequence of the urinary surgery for the unfiltered surgical event metadata, binds the surgical event metadata, the index association and the original video frame sequence of the urinary surgery to a surgical event ID, and stores the surgical event metadata, the index association and the original video frame sequence of the urinary surgery as urinary surgery association data items.
2. The cloud platform-based urological data management system of claim 1, wherein the reference frame sequence comprises a stable frame image, a frame sequence number, a frame time stamp, the tissue reference texture map is specifically grid cell coordinates, average texture feature values, texture feature standard deviations, the surgical event metadata is specifically referred to as event type identification, abnormal region coordinate set, texture deviation metrics, and the urological related data items comprise a patient unique identification, an original video file, an event index file, a cleaning metadata record.
3. The cloud platform-based urological data management system of claim 2, wherein the intra-operative image structuring module comprises: The method comprises the steps that an inter-frame difference quantization sub-module obtains an original video frame sequence of urinary surgery, the original video frame sequence of urinary surgery is traversed sequentially by taking a frame as a unit, pixel level processing is carried out on every two adjacent video frames in time, the sum of absolute difference values of pixel gray values of coordinate points corresponding to the two video frames is calculated, and an inter-frame change quantization value is generated; The low-difference frame screening submodule calls the inter-frame change quantized values, loads preset static pixel difference thresholds, judges the numerical value magnitude relation between each inter-frame change quantized value in the original video frame sequence of the urinary surgery and the static pixel difference thresholds, screens target video frames with quantized values smaller than the static pixel difference thresholds, and establishes a low-difference frame index set; And the reference sequence construction sub-module performs index retrieval in the original video frame sequence of the urinary surgery according to the low-difference frame index set, positions and extracts all video frames pointed by the internal identifiers of the low-difference frame index set, and performs queue reorganization operation according to the original time stamp sequence of the video frames to establish a reference frame sequence.
4. The cloud platform based urological data management system of claim 3, wherein the physiological characteristic benchmarking module comprises: The frame gridding processing sub-module traverses all video frames in the reference frame sequence, cuts the two-dimensional pixel space of each video frame into a plurality of rectangular units with equal size according to preset transverse and longitudinal segmentation quantity parameters, distributes coordinate identifiers for each rectangular unit, and establishes a grid matrix in the frame; the texture quantization submodule calls the intra-frame lattice matrix, extracts a gray level co-occurrence matrix of an internal pixel point aiming at each rectangular unit marked by the coordinates in the intra-frame lattice matrix, calculates the contrast and the angular second moment of the gray level co-occurrence matrix as texture descriptors, and carries out weighted fusion operation on the texture descriptors to obtain texture characteristic values; And traversing all intra-frame lattice matrixes in the reference frame sequence according to the tissue texture characteristic values by the reference texture averaging sub-module, grouping the tissue texture characteristic values corresponding to the rectangular units with the same coordinate identification, executing accumulation summation on the tissue texture characteristic values in each group, and calculating arithmetic mean to generate the tissue reference texture map.
5. The cloud platform based urological data management system of claim 4, wherein the surgical event intelligent identification module comprises: The method comprises the steps that a real-time texture map generation submodule obtains a current frame, space segmentation is carried out on the current frame based on preset grid matrix parameters, a pixel gray level co-occurrence matrix is extracted for each segmented matrix unit, contrast and angular second moment are calculated, and weighting fusion operation is carried out to generate an instantaneous tissue texture matrix; The texture state classification and discrimination submodule calls the instantaneous tissue texture matrix and the tissue reference texture map, performs difference calculation on a plurality of matrix unit values of the instantaneous tissue texture matrix and reference values of corresponding positions of the tissue reference texture map one by one, and performs square sum operation on all the difference values to obtain tissue abnormality deviation; And the surgical event metadata encoding submodule loads a preset event triggering threshold according to the tissue abnormal deviation degree, judges whether the tissue abnormal deviation degree exceeds the event triggering threshold, records the timestamp of the current frame when the judgment result is yes, integrates the numerical value of the tissue abnormal deviation degree and generates surgical event metadata.
6. The cloud platform based urological data management system of claim 5, wherein the data asset archiving management module comprises: The metadata low-frequency filtering sub-module performs accumulation counting on data items in the surgical event metadata set to obtain metadata instance number, carries out numerical value size judgment on the metadata instance number and an event frequency threshold, filters the surgical event metadata set when the metadata instance number is lower than the event frequency threshold, reserves the surgical event metadata higher than the event frequency threshold, and establishes a core event metadata cluster; The event video index association sub-module invokes the core event metadata cluster, acquires the original video frame sequence of the urinary surgery, searches video frames with matched time stamps in the original video frame sequence of the urinary surgery based on the time stamp attribute of a plurality of pieces of metadata in the core event metadata cluster, and establishes index association pointing to the physical storage position of the corresponding video frame in the sequence for each piece of metadata; And the data asset binding and grading sub-module calls an identifier generation service to create a surgery event ID according to the index association, encapsulates the core event metadata cluster, the index association and the complete original video frame sequence of the urological operation as an integral data unit, binds the surgery event ID as a main key of the data unit, and stores the surgery event ID as a urological operation association data item.
7. The cloud platform-based urological data management system of claim 6, wherein in the tissue texture quantification sub-module, the weighted fusion operation is specifically: firstly, respectively acquiring the contrast ratio and the angular second moment of the gray level co-occurrence matrix for each rectangular unit; and then calling a preset contrast weight coefficient and an angle second moment weight coefficient, and passing through the formula: ; Calculating the texture characteristic value of the tissue; Wherein, the Representing the values of the texture features of the tissue, Representing the contrast of the gray level co-occurrence matrix, An angular second moment representing the gray level co-occurrence matrix, As a result of the contrast weighting coefficients, Is the angular second moment weight coefficient, and And (3) with The sum of (2) is 1.
8. The cloud platform-based urological data management system of claim 7, wherein said surgical event metadata encoding submodule further comprises, prior to loading a preset event trigger threshold: Firstly, obtaining texture feature standard deviations of all grid cells recorded in the organization reference texture map, calculating arithmetic average values of the texture feature standard deviations of all grid cells, and generating reference texture fluctuation degrees; Then loading a dynamic adjustment factor associated with the reference texture fluctuation degree, and carrying out product operation on the preset event trigger threshold value and the dynamic adjustment factor to obtain an adaptive event trigger threshold value; finally, judging the abnormal deviation degree of the tissue by using the self-adaptive event trigger threshold; the dynamic adjustment factor is determined by the following formula based on the reference texture fluctuation degree: Calculating to obtain; Wherein, the Representing the dynamic adjustment factor in question, Representing the degree of fluctuation of the texture of the reference, Representing a preset lower limit value of the dynamic adjustment factor, Representing a preset upper limit value of the dynamic adjustment factor, Representing the baseline texture volatility reference midpoint for calibration, A slope coefficient representing the response sensitivity of the control factor, Is the base of natural logarithms.
9. The cloud platform-based urological data management system of claim 8, wherein the texture state classification discrimination sub-module calculates the degree of tissue abnormality deviation, in particular: Calling the instantaneous organization texture matrix and the organization reference texture map, and based on the standard deviation of the texture characteristics of each grid unit recorded in the organization reference texture map, passing through the formula: ; calculating the abnormal deviation degree of the tissue; Wherein, the For the degree of abnormal deviation of the tissue, For the total number of cells of the grid matrix, For the first of the instantaneous texture matrices The number of the individual units is determined, Reference texture map for the tissue The average texture feature value of the individual locations, Reference texture map for the tissue The standard deviation of the texture features for each location, Is a smooth constant for circumventing the denominator zero.
10. The cloud platform-based urological data management system of claim 9, wherein the metadata low frequency filtering submodule comprises: Firstly, according to the timestamp attribute of the surgical event metadata, carrying out time sequence sequencing on the surgical event metadata set, and establishing an event time sequence; secondly, setting a time window threshold, traversing the event time sequence, and identifying and dividing a plurality of continuous operation event metadata with time stamp difference values smaller than the time window threshold into a temporary event cluster; thirdly, counting the number of the surgical event metadata included in each temporary event cluster to generate a cluster event count; and finally, comparing the intra-cluster event count with the event frequency threshold, only reserving the temporary event cluster with the count value not smaller than the event frequency threshold, merging all reserved surgical event metadata in the temporary event cluster, and establishing the core event metadata cluster.

Description

Urinary surgery data management system based on cloud platform Technical Field The invention relates to the technical field of operation data management, in particular to a urinary operation data management system based on a cloud platform. Background The technical field of operation data management relates to the digital acquisition, storage and organization of various information generated in an operation process, including the archiving of medical record information and inspection images of a patient before operation, the recording of operation steps and instrument use conditions in operation and the management of follow-up visit data after operation. The traditional urology operation data management system is an application system for classifying and managing various data generated by urology operation, mainly aims at the arrangement of preoperative urinary tract images and inspection data of patients involved in urology operation, records operation steps and equipment in operation, and archives of post-operation recovery related data, and the traditional urology operation data management generally adopts a database-based storage mode to combine manual recording and tabular management to realize data collection and maintenance. The traditional urinary surgery data management relies on manual entry and tabular maintenance, the key information and the image data in the surgery process are mutually split in this way, a doctor is difficult to quickly position a recorded text event, such as instrument replacement or tissue abnormality, to a specific time point in a few hours of surgery video when reviewing after the surgery, the difficulty of data checking and analysis is increased, information incompleteness is possibly caused by omission or deviation of subjective recording, massive surgery video data cannot be effectively utilized in a structured mode, potential teaching and scientific research values are greatly limited, and a data chain is broken in key links. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a urological operation data management system based on a cloud platform. In order to achieve the purpose, the invention adopts the following technical scheme that the urology operation data management system based on the cloud platform comprises: the intraoperative image structuring module acquires an original video frame sequence of urinary surgery, calculates pixel differences among frames, screens video frames with the differences lower than a threshold value to be a reference frame sequence, and transmits the reference frame sequence to the physiological characteristic referencing module; the physiological characteristic referencing module divides a video frame into grid matrixes based on the reference frame sequence to calculate tissue texture characteristic values, calculates the average value of the tissue texture characteristic values at the same position of each grid matrix to generate a tissue reference texture map, and transmits the tissue reference texture map to the operation field event intelligent recognition module; the intelligent recognition module of the surgical field event generates a current texture map of a current frame, calls a support vector machine model to use the organization reference texture map as a reference class, generates surgical event metadata if a classification result is an organization abnormality, and transmits the surgical event metadata to the data asset archiving management module; and the data asset archiving management module filters the surgical event metadata with the quantity lower than a frequency threshold, establishes index association with the original video frame sequence of the urinary surgery for the unfiltered surgical event metadata, binds the surgical event metadata, the index association and the original video frame sequence of the urinary surgery to a surgical event ID, and stores the surgical event metadata, the index association and the original video frame sequence of the urinary surgery as urinary surgery association data items. As a further scheme of the invention, the reference frame sequence comprises a stable frame image, a frame sequence number and a frame time stamp, the tissue reference texture map is specifically grid unit coordinates, average texture characteristic values and texture characteristic standard deviations, the operation event metadata is specifically referred to as event type identification, abnormal region coordinate sets and texture deviation measurement, and the urology operation associated data items comprise a patient unique identification, an original video file, an event index file and a cleaning metadata record. As a further aspect of the present invention, the intra-operative image structuring module includes: The method comprises the steps that an inter-frame difference quantization sub-module obtains an original video frame sequence of urinary surgery, the