CN-121983287-A - Building method, auxiliary judging method, device and system of auxiliary judging model of bladder outlet obstruction

CN121983287ACN 121983287 ACN121983287 ACN 121983287ACN-121983287-A

Abstract

The invention belongs to the technical field of computer-aided diagnosis, and discloses a method for establishing an auxiliary judging model of bladder outlet obstruction, an auxiliary judging method, an auxiliary judging device and an auxiliary judging system, wherein the establishing method comprises the steps of obtaining a plurality of urodynamic sample data to establish a data set; the sample data comprises multichannel time sequence data, a space-time model is built and trained to obtain an auxiliary judgment model, the space-time model is specifically characterized in that the multichannel time sequence data is input into a coding module for coding, the output of the coding module is subjected to multichannel and single-channel convolution through a variable channel convolution module to obtain a convolution stacking feature vector with a plurality of output channels, and then the convolution stacking feature vector is fused by a self-attention module so as to be used for a classification network to recognize and output a judgment result. The method can effectively capture the coupling relation between the multiple channels and the time relation in the channels by constructing the space-time model, and is beneficial to improving the judgment accuracy of the intelligent judgment model according to the multiple time sequence data.

Inventors

HU YOUMIN
LI GUANG
ZHAO YU
SHEN AO
LIU QICHANG
MA QICHENG

Assignees

华中科技大学

Dates

Publication Date: 20260505
Application Date: 20260128

Claims (10)

1. The method for establishing the auxiliary judging model of the bladder outlet obstruction is characterized by comprising the following steps of: Acquiring a plurality of urodynamic sample data, preprocessing the sample data, and establishing a data set, wherein the sample data comprises multichannel time series data; The method comprises the steps of constructing a space-time model which simultaneously considers the coupling relation between time sequences in channels and the channels, training and testing to obtain an auxiliary judgment model, wherein the space-time model comprises a coding module, a variable channel convolution module, a self-attention module and a classification network, multi-channel time sequence data are input into the coding module for coding, the output of the coding module carries out multi-channel dimension and single-channel dimension convolution through the variable channel convolution module to obtain a convolution stacking feature vector with a plurality of output channels, and then the self-attention module fuses the convolution stacking feature vector so as to be used for the classification network to carry out identification and output judgment results.
2. The method for establishing an auxiliary judgment model of bladder outlet obstruction according to claim 1, wherein the sample data further comprises a static index; The space-time model also comprises a gating unit arranged between the self-attention module and the classification network, and the gating unit is used for fusing the output of the self-attention module and the static index to obtain a multivariate relation; The space-time model also comprises a graph module and a dynamic graph neural network module, the time series data of the multiple channels are divided into multiple segments according to time slices, one segment generates a full-connection graph through the graph module, the dynamic graph neural network module is used for integrating the multiple full-connection graphs and the multivariate relation output by the gating unit, and the output of the dynamic graph neural network module is used for the classification network to recognize and output a judgment result.
3. The method for building the bladder outlet obstruction auxiliary judgment model according to claim 2, wherein the gating unit is expressed as: ; Wherein, the For the output tensor of the gating cell, As an input tensor to the gating cell, The feature vector of the static index is represented, The multiplication gating cell weight matrix and the bias vector, In order to activate the function, And The cell weight matrix and bias vector are gated for addition.
4. The method for establishing the auxiliary judging model of the bladder outlet obstruction according to claim 2, wherein an average pooling layer, a full-connection layer and a sigmoid classification function layer are utilized to form the classification network, and finally, the characteristics are converted into probability values between 0 and 1; And a time graph pooling module is further arranged between the dynamic graph neural network module and the classification network and is used for performing time graph pooling on the output of the dynamic graph neural network module to obtain a new adjacency matrix and a new output tensor after graph node aggregation, wherein the adjacency matrix is an expression form of a full-connection graph.
5. The method for building a bladder outlet obstruction aid determination model according to any one of claims 1-4, wherein the multi-channel time series data comprises a plurality of urine flow rate, urine volume, bladder pressure, detrusor pressure, abdominal pressure, and lower urinary tract electromyographic signals, wherein the urine flow rate is based on the urine volume data in a central differential manner: a calculation is performed, wherein, For the current moment in time the urine flow rate, For the urine volume at the previous moment, For the urine volume at the later moment, Is the reciprocal of the sampling value; Detrusor pressure is calculated from data of bladder pressure and abdominal pressure over time by: Wherein, the method comprises the steps of, Is the detrusor pressure value at the current moment, As the value of the bladder pressure at the current moment, Is the abdominal pressure value at the current moment.
6. The method of building a bladder outlet obstruction aiding determination model according to claim 5, wherein the static indicators comprise a plurality of the age, maximum flow rate, flow time, start of urination to maximum flow time, volume of urination, flow rate at 2 seconds start of urination, flow rate, maximum detrusor pressure, maximum detrusor flow under detrusor pressure; Wherein the urine flow velocity is according to And (3) calculating, wherein, Indicating the velocity of the urine flow, For the maximum urine flow rate, In order to initiate the urine flow rate, Indicating the time taken to begin urination to the maximum urine flow rate.
7. The method for establishing the bladder outlet obstruction auxiliary judgment model according to claim 5, wherein the preprocessing of the sample data specifically comprises: Abnormal data caused by identification precision in the sample data is eliminated by utilizing a sliding window and a threshold detection method; detecting mutation data of adjacent points, setting a difference threshold value of the adjacent points, judging as an abnormal point if detecting that the difference value of the adjacent points is larger than the difference threshold value, and processing the abnormal point as an average value of previous and last sampling; preprocessing the sample data further comprises filtering the urine volume, bladder pressure and abdominal pressure data by using a Kalman filtering algorithm.
8. An auxiliary judging method for bladder outlet obstruction is characterized by comprising the following steps: Acquiring urine dynamic data to be detected, and preprocessing the data to be detected; inputting the preprocessed data to be detected into the auxiliary judgment model in any one of claims 1-7 to carry out auxiliary judgment on bladder outlet obstruction.
9. An auxiliary judging device for bladder outlet obstruction, which is characterized by comprising: The data processing module is used for acquiring the urodynamic data to be detected and preprocessing the data to be detected; the auxiliary judging module is used for inputting the preprocessed data to be detected into the auxiliary judging model in any one of claims 1-7 to carry out auxiliary judgment on bladder outlet obstruction.
10. A system for building an auxiliary judging model of bladder outlet obstruction, characterized in that the system comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the method for building the auxiliary judging model of bladder outlet obstruction according to any one of claims 1-7 when executing the computer program.

Description

Building method, auxiliary judging method, device and system of auxiliary judging model of bladder outlet obstruction Technical Field The invention belongs to the technical field of computer-aided diagnosis, and particularly relates to a method, a device and a system for establishing an auxiliary judging model of bladder outlet obstruction. Background Bladder outlet obstruction (Bladder Outlet Obstruction, BOO) is defined by the International incontinence Association as a generic term for obstruction during urination, accompanied by reduced urinary flow rate and/or increased post-urination residue and detrusor pressure, and urination dysfunction is prone to increased pressure in the bladder, causing reflux of urine, damaging the kidneys. The diagnosis of bladder outlet obstruction is derived from the result of clinical urodynamic examination, and at present, the diagnosis of bladder outlet obstruction still depends on the experience of doctors or is carried out according to BOOI (BOO Index) values, however, due to the difference of the depths of the experience of doctors and the low accuracy of BOOI, the phenomenon of false positive/false negative is easy to occur. Urodynamic examination the physiological signal data of urine flow rate, bladder pressure, abdominal pressure, detrusor pressure, lower urinary tract electromyographic signals and the like which change with time are obtained through a weighing sensor, a pressure sensor and a myoelectric sensor. Most of the existing researches extract indexes such as maximum urine flow rate, average urine flow rate, maximum detrusor pressure and the like from physiological signal data changing along with time through a data statistics analysis method to evaluate urination dysfunction, and further develop related intelligent algorithms such as algorithms of logistic regression, BP neural network and the like to perform intelligent diagnosis. However, high accuracy discrimination is difficult to realize only by indexes, comprehensive discrimination is often needed by referring to the detection data of urodynamic time series, and urodynamic detection results have the problems of multiple time scales, multiple-channel relation coupling, channel delay and the like, so that the method becomes a key problem in the development of intelligent diagnosis algorithms. In view of the above problems, currently, no preferred technical solution is proposed. Disclosure of Invention Aiming at the above defects or improvement demands of the prior art, the invention provides a method for establishing an auxiliary judging model of bladder outlet obstruction, an auxiliary judging method, an auxiliary judging device and an auxiliary judging system, which are used for solving the problem that the development of an intelligent diagnosis algorithm model according to multi-time sequence data is difficult due to multi-time scale and multi-channel relation coupling in the current multi-time sequence data. In order to achieve the above object, according to a first aspect of the present invention, there is provided a method for establishing an auxiliary judgment model of bladder outlet obstruction, comprising: Acquiring a plurality of urodynamic sample data, preprocessing the sample data, and establishing a data set, wherein the sample data comprises multichannel time series data; The method comprises the steps of constructing a space-time model which simultaneously considers the coupling relation between time sequences in channels and the channels, training and testing to obtain an auxiliary judgment model, wherein the space-time model comprises a coding module, a variable channel convolution module, a self-attention module and a classification network, multi-channel time sequence data are input into the coding module for coding, the output of the coding module carries out multi-channel dimension and single-channel dimension convolution through the variable channel convolution module to obtain a convolution stacking feature vector with a plurality of output channels, and then the self-attention module fuses the convolution stacking feature vector so as to be used for the classification network to carry out identification and output judgment results. According to the method for establishing the auxiliary judging model of bladder outlet obstruction, provided by the invention, the sample data also comprise static indexes; The space-time model also comprises a gating unit arranged between the self-attention module and the classification network, and the gating unit is used for fusing the output of the self-attention module and the static index to obtain a multivariate relation; The space-time model also comprises a graph module and a dynamic graph neural network module, the time series data of the multiple channels are divided into multiple segments according to time slices, one segment generates a full-connection graph through the graph module, the dynamic graph neural network module is used for i