CN115188467A - Urodynamics examination and diagnosis method and device - Google Patents

Urodynamics examination and diagnosis method and device Download PDF

Info

Publication number
CN115188467A
CN115188467A CN202210872036.3A CN202210872036A CN115188467A CN 115188467 A CN115188467 A CN 115188467A CN 202210872036 A CN202210872036 A CN 202210872036A CN 115188467 A CN115188467 A CN 115188467A
Authority
CN
China
Prior art keywords
urinary tract
lower urinary
diagnosis
layer
detrusor
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210872036.3A
Other languages
Chinese (zh)
Inventor
胡忠民
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Wendao Renaissance Intelligent Technology Co ltd
Original Assignee
Wuhan Wendao Renaissance Intelligent Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuhan Wendao Renaissance Intelligent Technology Co ltd filed Critical Wuhan Wendao Renaissance Intelligent Technology Co ltd
Priority to CN202210872036.3A priority Critical patent/CN115188467A/en
Publication of CN115188467A publication Critical patent/CN115188467A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/20—Measuring for diagnostic purposes; Identification of persons for measuring urological functions restricted to the evaluation of the urinary system
    • A61B5/202—Assessing bladder functions, e.g. incontinence assessment
    • A61B5/204—Determining bladder volume
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/20—Measuring for diagnostic purposes; Identification of persons for measuring urological functions restricted to the evaluation of the urinary system
    • A61B5/202—Assessing bladder functions, e.g. incontinence assessment
    • A61B5/205—Determining bladder or urethral pressure
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/20—Measuring for diagnostic purposes; Identification of persons for measuring urological functions restricted to the evaluation of the urinary system
    • A61B5/207—Sensing devices adapted to collect urine
    • A61B5/208—Sensing devices adapted to collect urine adapted to determine urine quantity, e.g. flow, volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Pathology (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Primary Health Care (AREA)
  • Urology & Nephrology (AREA)
  • Physiology (AREA)
  • Veterinary Medicine (AREA)
  • Epidemiology (AREA)
  • Surgery (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

The invention relates to the technical field of computers, and provides a urodynamics examination and diagnosis method and device. The method comprises the steps of collecting urination physiological parameters and detrusor pressure curves of a plurality of patients; performing second machine learning according to the urination physiological parameters of the patient with detrusor overactivity and lower urinary tract obstruction; performing third machine learning according to the urination physiological parameters of the patient with the lower urinary tract obstruction but without detrusor over activity; diagnosing over activity of detrusor, and if the patient has over activity of detrusor through diagnosis, diagnosing the obstruction of the lower urinary tract according to the result of learning by a second machine; otherwise, the lower urinary tract obstruction is diagnosed according to the result of the third machine learning. The invention diagnoses the excessive activity of the detrusor muscle firstly, and then selects and uses the corresponding machine learning result to diagnose the lower urinary tract obstruction so as to improve the correct rate of diagnosis.

Description

Urodynamics examination and diagnosis method and device
Technical Field
The invention relates to the technical field of computers, in particular to a urodynamics examination and diagnosis method and device.
Background
Lower urinary tract obstruction refers to the symptom of abnormally increased resistance when urine flows through the urethra due to prostatic enlargement (male), urethral overactivity or abnormal urethral stricture during urination of a patient, and is one of important indexes for evaluating the urination function of a patient. At present, the lower urinary tract obstruction of a male patient is diagnosed mainly by drawing a pressure-urine flow curve of the urination period of the patient and by means of a Shaefer nomogram (also called a linear passive urethral resistance diagram, linPURR diagram); in female patients, however, the diagnosis is currently performed mainly by the doctor's experience in cooperation with the observation of the original images of the detrusor pressure curve and the uroflow curve.
At present, it is difficult to clearly demarcate whether or not obstruction occurs in the lower urinary tract because of differences among patients. Therefore, the current diagnosis of lower urinary tract obstruction mainly relies on the abundant professional knowledge and diagnosis experience of physicians, and the diagnosis result is given after the urodynamic examination and diagnosis curve of the patients is carefully analyzed. The diagnosis effect greatly depends on the professional level of professional doctors, the cost of manual image recognition is high, but because the levels of doctors are different, and the number of excellent doctors cannot meet the daily increase of the number of patients, the development of a diagnosis model based on an artificial intelligent algorithm to reduce the manual diagnosis burden of doctors is of great significance.
Although the conventional urodynamic examination equipment and the matched software in the hospital can extract some characteristic parameters related to the diagnosis of the lower urinary tract obstruction, the conventional data acquisition method mainly depends on artificial classification and cannot realize intelligent classification, and the conventional intelligent treatment process has the problem of data overfitting, so that the data treatment result is not accurate enough.
Detrusor Overactivity is also commonly associated with lower urinary tract obstruction patients, and Detrusor Overactivity (DO) is an important association with urinary incontinence and bladder activity, and is one of the important indicators for urodynamic disease diagnosis. Detrusor over activity DO is primarily diagnosed by observing involuntary contraction of the detrusor during the bladder filling phase. In general, DO may be associated with urinary incontinence and bladder activity, which provides an important indicator for the diagnosis of urodynamic disease. However, there is currently a clinical lack of international standard definition for DO. Therefore, identification of DO mainly relies on the synergistic observation of the bladder pressure curve and the abdominal pressure curve during urodynamic examination, which requires a certain professional knowledge and expert experience.
Lower urinary tract obstruction and detrusor overactivity are important indexes for urodynamic examination and diagnosis, particularly for diagnosis of benign prostatic hyperplasia, but in the prior art, the examination and diagnosis are carried out by depending on expert experience. In addition, some patients with lower urinary tract obstruction have compensatory detrusor overactivity, and in this case, the urination physiological parameters such as the urine flow rate curve of the patient may be changed, so that the physiological parameters are different from those of the patient with simple lower urinary tract obstruction, but the patient with detrusor overactivity does not necessarily have lower urinary tract obstruction, and if the diagnosis of the lower urinary tract obstruction is uniformly carried out according to the two types of patients without discrimination, the diagnosis result may be inaccurate.
In view of this, overcoming the drawbacks of the prior art is a problem to be solved urgently in the art.
Disclosure of Invention
The invention aims to solve the technical problem that the prior art does not distinguish the patients with lower urinary tract obstruction and detrusor over activity from the patients with lower urinary tract obstruction but without detrusor over activity, so that the diagnosis result is not accurate.
The invention adopts the following technical scheme:
in a first aspect, the present invention provides a urodynamic examination diagnostic method comprising:
collecting urination physiological parameters and detrusor pressure curves of a plurality of patients;
forming a detrusor overactivity diagnosis data set according to a detrusor pressure curve of a patient, and performing first machine learning by using a detrusor overactivity diagnosis model and the detrusor overactivity diagnosis data set;
forming a first lower urinary tract obstruction diagnosis data set according to urination physiological parameters of a patient with detrusor overactivity and lower urinary tract obstruction, and performing second machine learning by using a lower urinary tract obstruction diagnosis model and the first lower urinary tract obstruction diagnosis data set;
forming a second lower urinary tract obstruction diagnostic data set according to urination physiological parameters of patients with lower urinary tract obstruction but without detrusor overactivity, and performing third machine learning by using a lower urinary tract obstruction diagnostic model and the second lower urinary tract obstruction diagnostic data set;
when urodynamics examination and diagnosis are carried out, firstly, detrusor overactivity diagnosis is carried out according to the result of the first machine learning, and if the fact that the detrusor overactivity of the patient exists is obtained through diagnosis, lower urinary tract obstruction diagnosis is carried out according to the result of the second machine learning; otherwise, the lower urinary tract obstruction is diagnosed according to the result of the third machine learning.
Preferably, the urination physiological parameter specifically includes one or more of a real-time urine flow rate, a total urination time, a urine flow time, a peak reaching time, a total urine volume, a maximum urine flow rate, an average urine flow rate, a maximum detrusor pressure, a urethra opening pressure, a detrusor pressure at a maximum urine flow rate, a maximum bladder capacity, and a urine flow rate at a maximum detrusor pressure.
Preferably, the lower urinary tract obstruction diagnosis model is composed of an input layer, a hidden layer and an output layer, and specifically comprises:
obtaining nodes of the input layer according to the corresponding data set, and obtaining nodes of the input layer according to the nodes of the input layer and a formula h j ＝ReLU(∑(X i W ij +b ij ) Computing to obtain a hidden layer node; wherein, X i The ith node, h, representing the input layer j J-th node, W, representing a hidden layer ij Representing the weight between the ith input layer node and the jth hidden layer node, b ij Representing the offset between the ith input layer node and the jth hidden layer node, reLU () representing the activation function, which is
Figure BDA0003753080730000031
According to the hidden layer node and formula
Figure BDA0003753080730000032
Calculating to obtain an output layer node; wherein, W ij Represents the weight between the ith hidden layer node and the jth output layer node, B ij Represents the offset between the ith hidden layer node and the jth output layer node, k is the value of the number of input layer nodes minus 1, and o represents a one-dimensional matrix composed of the output layer nodes.
Preferably, the forming a diagnostic data set of detrusor overactivity according to a detrusor pressure curve of the patient specifically comprises:
sampling health samples, and performing overlapped sampling on the detrusor pressure curve by using a sliding sampling window to obtain a plurality of health sample data;
sampling a non-healthy sample, taking a peak of a detrusor pressure curve of the non-healthy sample as a sampling midpoint, and extracting data of first preset time before the sampling midpoint and second preset time after the sampling midpoint in the detrusor pressure curve as non-healthy sample data;
the detrusor overactivity diagnostic dataset consists of a plurality of health sample data and a plurality of non-health sample data.
Preferably, before the forming of the detrusor overactivity diagnostic dataset according to the detrusor pressure curve of the patient, denoising the detrusor pressure curves of all patients by using a wavelet soft threshold method specifically comprises:
using the formula
Figure BDA0003753080730000041
The noise reduction is carried out, and the noise reduction is carried out,
Figure BDA0003753080730000042
representing a threshold value, W, for calculating the preceding and following wavelet coefficients x And the threshold value of the wavelet coefficient after calculation is represented, and lambda is a critical threshold value.
Preferably, the lower urinary tract obstruction diagnostic model consists of two Convolutional Neural Network (CNN) layers and a Back Propagation Neural Network (BPNN), wherein each CNN layer consists of a plurality of convolutional kernels, convolutional layers and pooling layers;
inputting the detrusor overactivity diagnostic dataset from a first CNN layer, performing convolution and pooling operations in the first CNN layer, the convolution and pooling operations comprising, according to a formula
Figure BDA0003753080730000043
Performing convolution operation according to formula
Figure BDA0003753080730000044
Taking out the maximum output value of the convolution layer;
wherein the content of the first and second substances,
Figure BDA0003753080730000045
representing the output value of the sample training set X after convolution operation with the ith convolution kernel in the first CNN layer, wherein X is an input sample, i represents the ith convolution kernel in the first CNN layer, i =1,2, \8230;, c1, c1 is the number of convolution kernels in the first CNN layer, conv1D () represents one-dimensional convolution operation,
Figure BDA0003753080730000046
is the weight of the ith convolution kernel in the first CNN layer,
Figure BDA0003753080730000047
is the deviation of the ith convolution kernel in the first CNN layer, maxpoling () represents the largest pool operation, the largest pooling window is used to perform step length movement of a preset length on the sample training set X, the largest value in the pooling window is selected as the output data of each movement, and ReLU () is a rectifying linear unit function;
performing convolution and pooling operation in the second CNN layer according to the pooling result obtained by the first CNN layer, inputting the pooling result obtained by the second CNN layer into the flat layer F5 of the BPNN, transmitting the output result of the flat layer F5 to the full connection layer FC6 for data dimension reduction, and outputting the output result of the full connection layer FC6 according to a formula
Figure BDA0003753080730000051
Classifying to obtain the output layer, M, of the BPNN j Is the weight matrix of the output layer, n j Is the offset vector of the output layer, j is the ith network node of the output layer, k is the total number of network nodes of the output layer, and O is the one-dimensional vector of the output.
Preferably, the forming a first lower urinary tract obstruction diagnosis data set according to the urination physiological parameters of the patient with both detrusor overactivity and lower urinary tract obstruction includes:
and respectively aiming at males and females, calculating the correlation between each urination physiological parameter and the lower urinary tract obstruction diagnosis result, and selecting the corresponding physiological parameter according to the correlation to generate a corresponding lower urinary tract obstruction diagnosis data set.
Preferably, the calculating the correlation between each urination physiological parameter and the diagnosis result of the lower urinary tract obstruction specifically comprises:
according to the formula
Figure BDA0003753080730000052
Calculating the correlation; wherein X represents corresponding physiological parameters, Y represents the diagnosis result of the lower urinary tract obstruction, cov (X, Y) represents the covariance of X and Y, D (X) represents the variance of X, and D (Y) represents the variance of Y.
Preferably, for males, according to the correlation from large to small, selecting a first preset number of physiological parameters to generate a corresponding first lower urinary tract obstruction diagnosis data set;
and selecting a second preset number of physiological parameters with the correlation larger than a preset numerical value from large to small according to the correlation of women to generate a corresponding first lower urinary tract obstruction diagnosis data set.
In a second aspect, the present invention also provides a urodynamic examination diagnosis apparatus for implementing the urodynamic examination diagnosis method of the first aspect, the apparatus including:
at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor for performing the urodynamic examination diagnostic method of the first aspect.
In a third aspect, the present invention also provides a non-transitory computer storage medium having stored thereon computer-executable instructions for execution by one or more processors for performing the urodynamic examination diagnostic method of the first aspect.
According to the invention, two training data sets are respectively formed according to the patient with both detrusor over activity and lower urinary tract obstruction and the patient with only lower urinary tract obstruction, so that the urination physiological parameters of the patient are not influenced in a cross way, and meanwhile, a diagnosis model of detrusor over activity is also generated, so that when diagnosis is carried out, the detrusor over activity is diagnosed firstly, and then the diagnosis of the lower urinary tract obstruction is carried out by selecting and using a corresponding machine learning result, so that the diagnosis accuracy is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly described below. It is obvious that the drawings described below are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
FIG. 1 is a schematic flow chart of a urodynamic examination diagnosis method provided by an embodiment of the invention;
FIG. 2 is a schematic diagram of a lower urinary tract obstruction diagnostic model provided by an embodiment of the invention;
FIG. 3 is a schematic flow chart of a urodynamic examination diagnosis method provided by an embodiment of the present invention;
FIG. 4 is a schematic flow chart of a urodynamic examination diagnosis method provided by an embodiment of the invention;
FIG. 5 is a schematic flow chart of a urodynamic examination diagnosis method provided by an embodiment of the present invention;
FIG. 6 is a schematic diagram illustrating the correlation between physiological parameters of urination and diagnostic results of a female according to an embodiment of the present invention;
FIG. 7 is a schematic diagram illustrating correlation between physiological parameters of urination and diagnosis results of a male according to an embodiment of the present invention;
FIG. 8 is a schematic diagram illustrating the comparison between the training effect of machine learning by selecting different urination physiological parameters for men and women and the training effect of machine learning by selecting the same urination physiological parameter for men and women according to the embodiment of the present invention;
FIG. 9 is a usage configuration chart of a detrusor over activity diagnostic model provided by an embodiment of the invention;
FIG. 10 is a table comparing training accuracy results of a detrusor over activity diagnostic model provided by embodiments of the present invention with other existing models;
FIG. 11 is a graph of urodynamic examination of male urination time provided by an embodiment of the present invention;
FIG. 12 is a graph illustrating a urodynamic test of urination in a female patient according to an embodiment of the present invention;
fig. 13 is a schematic structural diagram of a urodynamic examination and diagnosis device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In the description of the present invention, the terms "inner", "outer", "longitudinal", "lateral", "upper", "lower", "top", "bottom", and the like indicate orientations or positional relationships based on those shown in the drawings, and are for convenience only to describe the present invention without requiring the present invention to be necessarily constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.
In addition, the technical features involved in the respective embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
Example 1:
an embodiment of the present invention provides a urodynamic examination diagnosis method, as shown in fig. 1, the method including:
in step 201, the physiological parameters of urination and detrusor pressure curves for a plurality of patients are collected.
In step 202, a detrusor overactivity diagnostic dataset is formed from a detrusor pressure curve of the patient, and a first machine learning is performed using a detrusor overactivity diagnostic model and the detrusor overactivity diagnostic dataset.
In step 203, a first lower urinary tract obstruction diagnostic data set is formed according to urination physiological parameters of a patient with both detrusor overactivity and lower urinary tract obstruction, and a second machine learning is performed using a lower urinary tract obstruction diagnostic model and the first lower urinary tract obstruction diagnostic data set.
In step 204, a second lower urinary tract obstruction diagnostic data set is formed based on the urination physiological parameters of the patient with lower urinary tract obstruction but without detrusor overactivity, and a third machine learning is performed using the lower urinary tract obstruction diagnostic model and the second lower urinary tract obstruction diagnostic data set.
In step 205, when performing urodynamics examination and diagnosis, firstly, diagnosing detrusor overactivity according to the result of the first machine learning, and if the patient has detrusor overactivity through diagnosis, diagnosing lower urinary tract obstruction according to the result of the second machine learning; otherwise, the lower urinary tract obstruction is diagnosed according to the result of the third machine learning.
The detrusor pressure curve and the urination physiological parameters are acquired through the existing medical detection equipment. The detrusor pressure curve is collected from the filling phase of the patient and the voiding physiological parameter is collected from the voiding phase of the patient. In urodynamic examination diagnosis, since the pressure of the detrusor cannot be directly measured, the detrusor pressure is usually measured clinically and indirectly by subtracting the abdominal pressure from the intravesical pressure. The method comprises the steps of collecting the urination physiological parameters and the detrusor pressure curves of patients with confirmed diseases, and collecting the urination physiological parameters and the detrusor pressure curves of patients without corresponding symptoms, wherein in a subsequent text, the patients without corresponding symptoms are described in the form of healthy people, the corresponding urination physiological parameters or the detrusor pressure curves of the patients with confirmed diseases are described in the form of healthy samples, and the urination physiological parameters or the detrusor pressure curves of the patients with confirmed diseases are described in the form of unhealthy samples.
In the embodiment, two training data sets are respectively formed according to the patients with detrusor overactivity and lower urinary tract obstruction, so that urination physiological parameters of the patients are not influenced by each other, and meanwhile, a diagnosis model of detrusor overactivity is also generated, so that when diagnosis is carried out, detrusor overactivity is diagnosed firstly, and then corresponding machine learning results are selected to diagnose the lower urinary tract obstruction, so that the diagnosis accuracy is improved.
The urination physiological parameter specifically includes one or more of real-time urine flow rate, total urination time, urine flow time, peak arrival time, total urine volume, maximum urine flow rate, average urine flow rate, maximum detrusor pressure, urethra opening pressure, detrusor pressure at maximum urine flow rate, maximum bladder capacity, and urine flow rate at maximum detrusor pressure.
As shown in fig. 2, the lower urinary tract obstruction diagnosis model is composed of an input layer, a hidden layer and an output layer, and specifically includes:
obtaining nodes of the input layer according to the corresponding data set, and obtaining nodes of the input layer according to the nodes of the input layer and a formula h j ＝ReLU(∑(X i W ij +b ij ) Computing to obtain a hidden layer node; wherein, X i The ith node, h, representing the input layer j J-th node, W, representing a hidden layer ij Representing the weight between the ith input layer node and the jth hidden layer node, b ij Representing the offset between the ith input layer node and the jth hidden layer node, and ReLU () representing the activation function, which is
Figure BDA0003753080730000091
According to the hidden layer node and formula
Figure BDA0003753080730000092
Calculating to obtain an output layer node; wherein, W ij Represents the weight between the ith hidden layer node and the jth output layer node, B ij Represents the offset between the ith hidden layer node and the jth output layer node, k is the value of the number of input layer nodes minus 1, and o represents a one-dimensional matrix composed of the output layer nodes.
Labels of 0,1 and 2 are respectively set for no lower urinary tract obstruction, suspected lower urinary tract obstruction and lower urinary tract obstruction in the classification results. O represents the output result, the value of which is a one-dimensional vector, the position index (0, 1, 2) corresponding to the maximum element in the vector corresponds to the final classification label, if O is represented as [ O ] 1 ，O 2 ，O 3 ]Wherein O is 1 The corresponding position index is 0, and when the value is maximum, no lower urinary tract obstruction is diagnosed; o is 2 The corresponding position index is 0, and when the value is maximum, the suspected lower urinary tract obstruction is diagnosed; o is 3 The corresponding position index is 0, and when the value is the maximum, the presence of lower urinary tract obstruction is diagnosed.
When the second machine learning is carried out, obtaining nodes of an input layer by using the first lower urinary tract obstruction diagnosis data set; when performing the third machine learning, the node of the input layer is obtained using the second lower urinary tract obstruction diagnostic dataset.
It should be noted that the meaning of the parameters in all formulas in this embodiment does not only depend on the expression form, but also depends on the model where the parameters are located, for example, when the same parameter appears in different models, different meanings may exist, and details are not repeated later.
The forming of the detrusor overactivity diagnostic dataset according to the detrusor pressure curve of the patient, as shown in fig. 3, specifically includes:
in step 301, a plurality of health sample data is obtained by sampling a health sample and performing overlapping sampling on the detrusor pressure curve using a sliding sampling window.
In step 302, a non-healthy sample is sampled, a peak of a detrusor pressure curve of the non-healthy sample is taken as a sampling midpoint, and data of first preset time before the sampling midpoint and second preset time after the sampling midpoint in the detrusor pressure curve are extracted as the non-healthy sample data.
In step 303, the detrusor over activity diagnostic data set consists of a plurality of health sample data and a plurality of non-health sample data.
Wherein, the sampling size of the sliding sampling window, the first preset time and the second preset time are obtained by analyzing according to the change rule of the detrusor pressure curve by a person skilled in the art.
Typically, the first predetermined time is the same as the second predetermined time as used by one skilled in the art.
Before the detrusor overactivity diagnostic dataset is derived from the detrusor pressure curve, as shown in FIG. 4, the method further comprises:
in step 300, the detrusor pressure curves for all patients are denoised using wavelet soft thresholding. The method specifically comprises the following steps: using the formula
Figure BDA0003753080730000101
The noise reduction is carried out, and the noise reduction is carried out,
Figure BDA0003753080730000102
representing a threshold value, W, for calculating the preceding and following wavelet coefficients x A threshold value representing the wavelet coefficient after calculation, wherein lambda is a critical threshold value and the noise amplitude is less than
Figure BDA0003753080730000103
And selecting the absolute standard deviation of the high-frequency coefficient of the first layer of wavelet decomposition as a sigma value, wherein the number of layers of the wavelet decomposition is 6.
Where detrusor overactivity can be represented by DO events, the primary purpose of the training model is to determine whether the input samples are DO events. For the most recent test patient signals, it is desirable to extract samples that may contain DO events after wavelet soft threshold denoising to improve diagnostic efficiency. Generally, DO events occur with increasing magnitude in the detrusor pressure curve, pdet curve. Therefore, the signal range over which the DO event occurs must contain at least one local peak. The wavelet soft threshold denoised curve generally meets the smoothing requirement, but sometimes a glitch occurs, which may result in a false identification of a local peak. Typically, this artifact appears on the pressure curve as a rapid rise in a short time and then return to the pre-rise level.
Aiming at the problems, the section sets a threshold strategy and rapidly screens the suspicious DO sample. In this strategy, three important parameters, respectively, prominence, width and relative height, need to be set in advance.
The projection degree: this parameter is used to show the saliency of the local maxima and then to filter out small perturbations and spikes on the curve. The parameter is calculated by finding a trough on both sides of the maximum value which does not exceed the higher peak value, and then calculating the difference between the maximum value and the trough, wherein the smaller value is the significant value.
Width: this parameter is used to calculate the peak width of the signal.
Relative height: this parameter is associated with a parameter width that is used to select the relative height of the measured peak width, expressed as a percentage of the peak height. For example, the lowest profile width of a peak is calculated for a relative height of 1, and half the height of the peak is calculated for 0.5.
By combining the Pdet curve characteristics of the patient, on the basis of multiple experiments, the three parameters of the protrusion degree, the width and the relative height are finally determined to be 10, 8 and 0.4 respectively.
Based on the threshold strategy described above, most of the perturbations and artifacts can be filtered out and the peak points most likely to contain the DO event are selected. Then, the found peak point is taken as the midpoint, and the signal with the length of 30s is taken as the final suspected DO sample.
The lower urinary tract obstruction diagnosis model consists of two Convolutional Neural Network (CNN) layers and a Back Propagation Neural Network (BPNN), wherein each CNN layer consists of a plurality of convolutional kernels, convolutional layers and pooling layers.
Inputting the detrusor overactivity diagnostic dataset from a first CNN layer, performing convolution and pooling operations in the first CNN layer, the convolution and poolingThe operation comprises, according to the formula
Figure BDA0003753080730000111
Performing convolution operation according to formula
Figure BDA0003753080730000112
The maximum output value of the convolution layer is extracted.
Wherein the content of the first and second substances,
Figure BDA0003753080730000113
represents the output value of the sample training set X after convolution operation with the ith convolution kernel in the first CNN layer, wherein X is the input sample, i represents the ith convolution kernel in the first CNN layer, i =1,2, \ 8230;, c1, c1 is the number of convolution kernels in the first CNN layer, conv1D () represents one-dimensional convolution operation,
Figure BDA0003753080730000114
is the weight of the ith convolution kernel in the first CNN layer,
Figure BDA0003753080730000115
is the deviation of the ith convolution kernel in the first CNN layer, maxporoling () represents the maximum pool operation, the maximum pooling window is used to perform step length movement of a preset length on the sample training set X, the maximum value in the pooling window is selected as the output data of each movement, and ReLU () is a rectification linear unit function.
Performing convolution and pooling operation in the second CNN layer according to the pooling result obtained by the first CNN layer, inputting the pooling result obtained by the second CNN layer into the flat layer F5 of the BPNN, transmitting the output result of the flat layer F5 to the full connection layer FC6 for data dimension reduction, and outputting the output result of the full connection layer FC6 according to a formula
Figure BDA0003753080730000121
Classifying to obtain the output layer, M, of the BPNN j Is the weight matrix of the output layer, n j Is the offset vector of the output layer, j is the ith network node of the output layer, k is the total network node of the output layerNumber, O is the one-dimensional vector of the output.
The convolution and pooling operation of the second CNN layer is consistent with the convolution and pooling operation of the first CNN layer in terms of implementation concept, and is not described herein again. The preset length is obtained by analyzing the change rule of the detrusor pressure curve by a person skilled in the art. In the subsequent text or figures, the detrusor overactivity diagnostic model can be described in terms of a CNN model. And the output of the O is similar to that of the lower urinary tract obstruction diagnosis model, and represents the diagnosis result by the position index of the maximum value, wherein when the position index is 0, the detrusor overactivity does not exist, when the position index is 1, the suspected detrusor overactivity is represented, and when the position index is 0, the detrusor overactivity is represented.
The first lower urinary tract obstruction diagnostic data set is formed according to the urination physiological parameters of the patient with both detrusor overactivity and lower urinary tract obstruction, as shown in fig. 5, and specifically includes:
in step 401, the correlation between each urination physiological parameter and the diagnosis result of the lower urinary tract obstruction is calculated for each of the male and female.
In step 402, corresponding physiological parameters are selected according to the correlation, and a corresponding lower urinary tract obstruction diagnostic data set is generated.
Wherein said generating a corresponding lower urinary tract obstruction diagnostic dataset comprises generating a first lower urinary tract obstruction diagnostic dataset for a male and a first lower urinary tract obstruction diagnostic dataset for a female.
The method for calculating the correlation between the physiological parameters of urination and the diagnosis result of the lower urinary tract obstruction specifically comprises the following steps:
according to the formula
Figure BDA0003753080730000122
Calculating the correlation; wherein X represents corresponding physiological parameters, and Y represents lower urinary tract obstruction diagnosis nodeIf yes, cov (X, Y) represents the covariance of X and Y, D (X) represents the variance of X, and D (Y) represents the variance of Y.
And selecting a first preset number of physiological parameters to generate a corresponding first lower urinary tract obstruction diagnosis data set according to the correlation from large to small for male.
And selecting a second preset number of physiological parameters with the correlation larger than a preset numerical value from large to small according to the correlation of women to generate a corresponding first lower urinary tract obstruction diagnosis data set.
Wherein, the first preset number is obtained by analyzing the physiological parameters of the male and the correlation of the diagnosis result and the accuracy requirement of the male lower urinary tract obstruction diagnosis together by the technicians in the field.
The preset numerical value and the second preset quantity are obtained by analyzing the physiological parameters of the male and the correlation of the diagnosis result and the accuracy requirement of the male lower urinary tract obstruction.
For example, if the correlation between the calculated physiological parameter of urination of women and the diagnosis result is shown in fig. 6, and the correlation between the calculated physiological parameter of urination of men and the diagnosis result is shown in fig. 7, the following options are available, including:
for male, the first six characteristic indexes with the maximum correlation coefficient are selected, namely one or more physiological parameters of urination time, maximum urine flow rate, average urine flow rate, maximum detrusor pressure, detrusor pressure at the maximum urine flow rate and initial urine flow pressure are selected to generate a corresponding lower urinary tract obstruction diagnosis data set.
For women, 7 characteristic indexes with the relation number larger than 0.2 in the 11 characteristic indexes are selected, namely one or more physiological parameters of total urine volume, maximum urine flow rate, average urine flow rate, maximum detrusor pressure, detrusor pressure at the maximum urine flow rate, urine flow initial pressure and maximum bladder capacity are selected to generate a corresponding lower urinary tract obstruction diagnosis data set.
Aiming at the analysis results, six characteristic indexes with the maximum correlation coefficient are respectively selected for men and women to establish a diagnosis model, and are compared with results of simultaneously using the established model with eleven indexes, and the comparison results are shown in fig. 8. As can be seen from the model training results, although the precision of the model screened based on the features is reduced to some extent in the training set, the precision of the test set is improved to a great extent as a whole, which shows that through the feature screening, a part of redundant information is eliminated, and the overfitting phenomenon of the model is relieved to a great extent.
In the embodiment, the correlation between each urination physiological parameter and the diagnosis result is calculated for the male and the female respectively, so that the optimal physiological parameter is selected to form a data set for machine learning, and the accuracy of the lower urinary tract obstruction diagnosis result of the male and the female is improved.
The terms "first," "second," and "third" in the present embodiment have no special limiting meanings, and are used for descriptive purposes only for convenience of describing different individuals among the objects, and should not be interpreted as having special limiting meanings in order or otherwise.
Example 2:
based on the method described in embodiment 1, the invention combines with a specific application scenario and uses technical expressions in a related scenario to describe an implementation process in a characteristic scenario.
Machine learning was performed as with 44 cases of data, where 18 patients diagnosed with symptoms of detrusor overactivity, 8 of 18 with simultaneous symptoms of lower urinary tract obstruction; another 14 patients had confirmed the presence of lower urinary tract obstruction symptoms only.
The embodiment specifically includes: samples of 44 patient data for model training were divided, with 18 patients presenting with DO symptoms. By analyzing the detrusor pressure Pdet curves for these 18 patients, 146 DO events were collected. To ensure that a DO event is recorded completely for each DO sample collected, each sample has a duration of 30s and the peak of the DO curve is taken as the sampling midpoint. If the peak of the DO event is within the first 15 seconds or the last 15 seconds, the first 30 seconds and the last 30 seconds of data are taken as DO samples, respectively.
And for the health sample data, overlapping sampling is adopted, and the number of samples of non-DO events is increased so as to facilitate the subsequent training of the diagnosis model. This is done by sliding the sampling window over the Pdet signal. In this study, the width of the sampling window was set to 300 and the step size was set to 50. A total of 1863 samples of non-DO events were collected by sampling data from patients without DO. The machine learning is performed by substituting the sampled samples into the detrusor overactivity diagnosis model, which is explained in detail in embodiment 1 and will not be described herein again.
To ensure the effectiveness of the diagnostic model, a number of experiments were performed during the model building process and compared with several classical models, as shown in fig. 13, including:
k-nearest neighbor algorithm (KNN), where 6 is chosen as the number of neighbors.
Support Vector Machines (SVM) in which a linear kernel is selected as the kernel when studying binary classification.
The Back Propagation Neural Network (BPNN) is characterized in that the adopted structure is 300-200-100-2. Random gradient descent is adopted as a network parameter optimization strategy, the learning rate is 0.01, and the batch size is 20.
An Automatic Encoder (AE) is adopted, wherein the structure is 300-128-642-64-128-300. Adam was chosen as the optimizer, learning rate 0.001, batch size 20.
In making the comparison, the detrusor overactivity diagnostic model used was constructed as shown in FIG. 9, wherein a convolution kernel size of 3 × 1 × 60 represents a convolution kernel size of 3, step size of 1, and kernel number of 60. The parameters of the optimizer used are the same as those of AE.
Through multiple 5-fold cross validation experiments, the superiority of the detrusor overactivity diagnosis model is further proved. The result shows that the CNN provided by the application has the highest classification precision and smaller standard deviation. And the convergence speed of the model is found to be very high after the training process of the model is further analyzed, so that the machine learning efficiency is improved.
The urination physiological parameters of 8 patients with lower urinary tract obstruction symptoms simultaneously in 18 patients with detrusor overactivity are used for establishing a first lower urinary tract obstruction diagnosis data set, the urination physiological parameters of the other 14 patients with lower urinary tract obstruction only are used for a second lower urinary tract obstruction diagnosis data set, and the urination physiological parameters of other patients are used as data of healthy people to participate in machine learning, specifically:
preliminarily selecting 11 parameters of a urine volume curve of a urination period and water filling flow to extract total urination time, urine flow time, peak reaching time, total urine volume, maximum urine flow rate, average urine flow rate, maximum detrusor pressure (Pmax), urethra opening pressure, detrusor pressure (Pdet.Qmax) at the maximum urine flow rate, maximum bladder capacity and urine flow rate (Q.Pmax) at the maximum detrusor pressure. Wherein, the urine flow time, the peak reaching time, the total urine volume, the maximum urine flow rate, the average urine flow rate and the like are obtained by the urine flow rate curve. For male and female, the correlation coefficient method is used to calculate the correlation degree between the 11 characteristic parameters and the final label. The correlation formula used in the calculation is given in example 1, and will not be described herein.
And respectively counting correlation analysis data between the female and male urination stage characteristic indexes and the labels based on the existing patient historical data. The following conclusion can be obtained through analysis, namely, the correlation between the maximum detrusor pressure Pdet.Qmax and the initial urine flow pressure of male and female at the maximum urine flow rate and the label is maximum; secondly, the distribution of the secondary important urine flow indexes of the male and the female has certain difference, wherein the female is sensitive to the maximum bladder capacity and the total urine volume, and the male is sensitive to the urination time.
The correlation between each calculated physiological parameter and the diagnosis result is shown in fig. 6 and fig. 7, and for the difference between the male and female sensitivities to different urine flow indicators, there is a certain difference between the selection of the male and female core urine flow indicators.
For women, 7 characteristic indexes with the relation number larger than 0.2 in 11 characteristic indexes are selected, namely total urine volume, maximum urine flow rate, average urine flow rate, pmax, pdet.Qmax, initial urine flow pressure and maximum bladder capacity; for male, the first 6 characteristic indexes with the maximum correlation coefficient are selected and respectively comprise urination time, maximum urine flow rate, average urine flow rate, pmax, pdetAnd (6) pressing. Forming data sets special for men and women respectively, and forming data sets aiming at samples with detrusor over activity and lower urinary tract obstruction symptoms simultaneously and samples with lower urinary tract obstruction symptoms only respectively, so that 4 data sets for training are formed totally, and the 4 data sets are substituted into a lower urinary tract obstruction diagnosis model respectively for machine learning, wherein in the model substituted by the two data sets for women, the number of nodes of an input layer is 7, and the used formula for converting a hidden layer into an output layer is
Figure BDA0003753080730000161
In the model with two male data sets substituted, the number of nodes in the input layer is 6, and the formula for switching from the hidden layer to the output layer is used
Figure BDA0003753080730000162
The meaning of each parameter and other formulas in the model are specifically described in embodiment 1, and are not described herein again.
In order to verify the training results of the diagnosis model for lower urinary tract obstruction differentiated for men and women, the training results are compared with the training results obtained by using 11 characteristic indexes to participate in training at the same time, and although the accuracy of the model screened based on the characteristics is reduced in the training set, the accuracy of the test set is improved to a greater extent on the whole, which shows that a part of redundant information is eliminated through the characteristic screening, the overfitting phenomenon of the model is reduced to a greater extent, and the training results after the differentiation of the men and women are more accurate along with the increase of samples.
The following will explain the application effect of the present embodiment by an actual diagnostic process, which specifically includes:
taking the example of the young 62 year old male patient, who was diagnosed in 3 months in 2021, first, the data related to the urination period of the patient was collected by the conventional urodynamic examination diagnosis, as shown in fig. 13, in which the first row from top to bottom is a urine flow rate curve, the second row is a urine volume curve, the third row is a detrusor pressure curve, and the last row is a water injection volume curve, and these data were statistically analyzed to calculate specific values of six core indicators, including a urination time 449.1 sec, a maximum urine flow rate 5.1 ml/sec, an average urine flow rate 3.7 ml/sec, pmax81.4 cm water column, pdet.qmax75.1 cm water column, and a urine flow initial pressure 73.7 cm water column, respectively. Inputting the detrusor pressure curve into a detrusor overactivity diagnosis model, diagnosing to obtain that no detrusor overactivity exists, substituting the six indexes into a second machine for learning in sequence to obtain an output result of [0, 1], wherein the position index corresponding to the maximum value is 2, and diagnosing to obtain that the patient has lower urinary tract obstruction.
We also exemplified the method for female patients, and also exemplified 58 year-old female patients who were diagnosed in 2021 and 2 months, and also obtained data of urine flow rate, urine volume, detrusor pressure, and perfusion water volume according to the curves shown in fig. 12 collected by urodynamic examination diagnosis, and calculated seven core features for female, including total urine volume 579 ml, maximum urine flow rate 24.6 ml/s, average urine flow rate 14.3 ml/s, pmax30.6 cm water column, pdet.qmax18 cm water column, initial urine flow pressure 23.9 cm water column, and maximum bladder volume 508 cc. Inputting the detrusor pressure curve into a detrusor overactivity diagnosis model, diagnosing to obtain that no detrusor overactivity exists, and inputting the six indexes into second machine learning to obtain a diagnosis result which is that no lower urinary tract obstruction exists and is consistent with the diagnosis result of a doctor.
Example 3:
fig. 13 is a schematic diagram showing the configuration of the urodynamic examination and diagnosis device according to the embodiment of the present invention. The urodynamic examination diagnostic apparatus of the present embodiment includes one or more processors 21 and a memory 22. In fig. 13, one processor 21 is taken as an example.
The processor 21 and the memory 22 may be connected by a bus or other means, and the bus connection is exemplified in fig. 13.
The memory 22, which is a non-volatile computer-readable storage medium, may be used to store non-volatile software programs and non-volatile computer-executable programs, such as the urodynamic test diagnostic method of example 1. The processor 21 executes the urodynamic examination diagnostic method by executing non-volatile software programs and instructions stored in the memory 22.
The memory 22 may include high speed random access memory and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, the memory 22 may optionally include memory located remotely from the processor 21, and these remote memories may be connected to the processor 21 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The program instructions/modules are stored in the memory 22 and, when executed by the one or more processors 21, perform the urodynamic examination diagnostic methods of examples 1 and 2 described above, e.g., perform the various steps shown in fig. 1, 3-5 described above.
It should be noted that, for the information interaction, execution process and other contents between the modules and units in the apparatus and system, the specific contents may refer to the description in the embodiment of the method of the present invention because the same concept is used as the embodiment of the processing method of the present invention, and are not described herein again.
Those of ordinary skill in the art will appreciate that all or part of the steps of the various methods of the embodiments may be performed by associated hardware as instructed by a program, which may be stored on a computer-readable storage medium, which may include: a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic or optical disk, and the like.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims (10)

1. A method of urodynamic examination diagnosis comprising:
collecting urination physiological parameters and detrusor pressure curves of a plurality of patients;
forming a detrusor overactivity diagnosis dataset according to a detrusor pressure curve of a patient, and performing first machine learning by using a detrusor overactivity diagnosis model and the detrusor overactivity diagnosis dataset;
forming a first lower urinary tract obstruction diagnosis data set according to urination physiological parameters of a patient with detrusor overactivity and lower urinary tract obstruction, and performing second machine learning by using a lower urinary tract obstruction diagnosis model and the first lower urinary tract obstruction diagnosis data set;
forming a second lower urinary tract obstruction diagnostic data set according to urination physiological parameters of patients with lower urinary tract obstruction but without detrusor overactivity, and performing third machine learning by using a lower urinary tract obstruction diagnostic model and the second lower urinary tract obstruction diagnostic data set;
when urodynamics examination and diagnosis are carried out, firstly, the diagnosis of over activity of detrusor is carried out according to the result of the first machine learning, and if the fact that the patient has over activity of detrusor is obtained through diagnosis, the diagnosis of lower urinary tract obstruction is carried out according to the result of the second machine learning; otherwise, the lower urinary tract obstruction is diagnosed according to the result of the third machine learning.
2. The urodynamic examination diagnostic method of claim 1, wherein the physiological parameters of urination specifically include one or more of real-time urine flow rate, total urination time, urine flow time, time to peak, total urine volume, maximum urine flow rate, average urine flow rate, maximum detrusor pressure, urethra opening pressure, detrusor pressure at maximum urine flow rate, maximum bladder capacity, and urine flow rate at maximum detrusor pressure.
3. The urodynamic examination and diagnosis method according to claim 1, wherein the lower urinary tract obstruction diagnosis model is composed of an input layer, a hidden layer and an output layer, and specifically comprises:
obtaining nodes of the input layer according to the corresponding data set, and obtaining nodes of the input layer according to the nodes of the input layer and a formula h j ＝ReLU(∑(X i W ij +b ij ) ) calculate to get hiddenA layer node; wherein, X i I-th node, h, representing the input layer j J-th node, W, representing a hidden layer ij Representing the weight between the ith input layer node and the jth hidden layer node, b ij Representing the offset between the ith input layer node and the jth hidden layer node, reLU () representing the activation function, which is
Figure FDA0003753080720000021
According to the hidden layer node and formula
Figure FDA0003753080720000022
Calculating to obtain an output layer node; wherein, W ij Represents the weight between the ith hidden layer node and the jth output layer node, B ij Represents the offset between the ith hidden layer node and the jth output layer node, k is the value of the number of input layer nodes minus 1, and o represents a one-dimensional matrix composed of the output layer nodes.
4. The urodynamic examination diagnostic method of claim 1, wherein the forming of the detrusor overactivity diagnostic dataset from the detrusor pressure curve of the patient comprises:
sampling health samples, and performing overlapped sampling on the detrusor pressure curve by using a sliding sampling window to obtain a plurality of health sample data;
sampling a non-healthy sample, taking a peak of a detrusor pressure curve of the non-healthy sample as a sampling midpoint, and extracting data of first preset time before the sampling midpoint and second preset time after the sampling midpoint in the detrusor pressure curve as the non-healthy sample data;
the detrusor overactivity diagnostic dataset consists of a plurality of health sample data and a plurality of non-health sample data.
5. The urodynamic examination diagnostic method of claim 1, wherein before the forming of the diagnostic data set of detrusor overactivity according to the detrusor pressure curve of the patient, denoising the detrusor pressure curves of all patients by a wavelet soft threshold method further comprises:
using the formula
Figure FDA0003753080720000023
The noise reduction is carried out, and the noise reduction is carried out,
Figure FDA0003753080720000024
threshold, W, representing the calculated front and rear wavelet coefficients x And the threshold value of the wavelet coefficient after calculation is represented, and lambda is a critical threshold value.
6. The urodynamic examination diagnostic method of claim 1, wherein the diagnostic model of lower urinary tract obstruction is composed of two Convolutional Neural Network (CNN) layers and a Back Propagation Neural Network (BPNN), each CNN layer is composed of a plurality of convolutional kernels, convolutional layers and pooling layers;
inputting the detrusor overactivity diagnostic dataset from a first CNN layer group, performing convolution and pooling operations in the first CNN layer group, the convolution and pooling operations comprising, according to a formula
Figure FDA0003753080720000031
Figure FDA0003753080720000032
Performing convolution operation according to formula
Figure FDA0003753080720000033
Taking out the maximum output value of the convolution layer;
wherein the content of the first and second substances,
Figure FDA0003753080720000034
representing the output value of the convolution operation between the sample training set X and the ith convolution kernel in the first CNN layer, wherein X is the input sample, i represents the ith convolution kernel in the first CNN layer, i =1,2, \ 8230;, c1, c1 is the number of convolution kernels in the first CNN layer, conv1D () represents a one-dimensional convolution operation, W i 1 Is the weight of the ith convolution kernel in the first CNN layer,
Figure FDA0003753080720000036
is the deviation of the ith convolution kernel in the first CNN layer, maxpoling () represents the largest pool operation, the largest pooling window is used to perform step length movement of a preset length on the sample training set X, the largest value in the pooling window is selected as the output data of each movement, and ReLU () is a rectifying linear unit function;
performing convolution and pooling operation in the second CNN layer according to the pooling result obtained by the first CNN layer, inputting the pooling result obtained by the second CNN layer into the flat layer F5 of the BPNN, transmitting the output result of the flat layer F5 to the full connection layer FC6 for data dimension reduction, and outputting the output result of the full connection layer FC6 according to a formula
Figure FDA0003753080720000035
Classifying to obtain the output layer, M, of the BPNN j Is the weight matrix of the output layer, n j Is the offset vector of the output layer, j is the ith network node of the output layer, k is the total number of network nodes of the output layer, and O is the one-dimensional vector of the output.
7. The urodynamic examination diagnostic method of claim 1, wherein the forming a first diagnostic data set of lower urinary tract obstruction based on the physiological parameters of urination of a patient with both detrusor overactivity and lower urinary tract obstruction comprises:
and respectively aiming at males and females, calculating the correlation between each urination physiological parameter and the lower urinary tract obstruction diagnosis result, selecting the corresponding physiological parameter according to the correlation, generating a corresponding first lower urinary tract obstruction diagnosis data set, and respectively performing machine learning by using the corresponding first lower urinary tract obstruction diagnosis data set.
8. The urodynamic examination and diagnosis method according to claim 7, wherein the calculating the correlation between each physiological parameter of urination and the diagnosis result of the lower urinary tract obstruction specifically comprises:
according to the formula
Figure FDA0003753080720000041
Calculating the correlation; wherein X represents corresponding physiological parameters, Y represents the diagnosis result of the lower urinary tract obstruction, cov (X, Y represents the covariance of X and Y, D (X) represents the variance of X, and D (Y) represents the variance of Y.
9. The urodynamic examination diagnostic method of claim 8, wherein for a male, a first preset number of physiological parameters are selected to generate a corresponding first lower urinary tract obstruction diagnostic dataset according to the correlation from large to small;
and selecting a second preset number of physiological parameters with the correlation larger than a preset numerical value from large to small according to the correlation of women to generate a corresponding first lower urinary tract obstruction diagnosis data set.
10. A urodynamic examination diagnostic device, comprising:
at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor for performing the urodynamic examination diagnostic method of any one of claims 1-9.
CN202210872036.3A 2022-07-19 2022-07-19 Urodynamics examination and diagnosis method and device Pending CN115188467A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210872036.3A CN115188467A (en) 2022-07-19 2022-07-19 Urodynamics examination and diagnosis method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210872036.3A CN115188467A (en) 2022-07-19 2022-07-19 Urodynamics examination and diagnosis method and device

Publications (1)

Publication Number Publication Date
CN115188467A true CN115188467A (en) 2022-10-14

Family

ID=83521743

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210872036.3A Pending CN115188467A (en) 2022-07-19 2022-07-19 Urodynamics examination and diagnosis method and device

Country Status (1)

Country Link
CN (1) CN115188467A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116432564A (en) * 2023-06-15 2023-07-14 天津市第五中心医院 Urodynamic state analysis method and analysis system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116432564A (en) * 2023-06-15 2023-07-14 天津市第五中心医院 Urodynamic state analysis method and analysis system
CN116432564B (en) * 2023-06-15 2023-08-18 天津市第五中心医院 Urodynamic state analysis method and analysis system

Similar Documents

Publication Publication Date Title
Mousavi et al. Deep convolutional neural network for classification of sleep stages from single-channel EEG signals
Jin et al. Multi-domain modeling of atrial fibrillation detection with twin attentional convolutional long short-term memory neural networks
CN113040715B (en) Human brain function network classification method based on convolutional neural network
Acharya et al. Automated diagnosis of epileptic EEG using entropies
Gil et al. Application of artificial neural networks in the diagnosis of urological dysfunctions
CN109117890B (en) Image classification method and device and storage medium
KR102354674B1 (en) Method and apparatus for diagnosing dysuria by simultaneously measuring urine flow rate and residual urine
CN104636580A (en) Health monitoring mobile phone based on human face
EP3863022A1 (en) Method and system for automatically characterizing liver tissue of a patient, computer program and electronically readable storage medium
CN114464322B (en) Female pelvic floor dysfunction disease risk early warning model and construction method and system thereof
CN112950614A (en) Breast cancer detection method based on multi-scale cavity convolution
CN115188467A (en) Urodynamics examination and diagnosis method and device
CN117116477A (en) Construction method and system of prostate cancer disease risk prediction model based on random forest and XGBoost
Ravishankar et al. A deep learning approach for ovarian cysts detection and classification (OCD-FCNN) using fuzzy convolutional neural network
Zhou et al. A pilot study: detrusor overactivity diagnosis method based on deep learning
CN115064257A (en) Method, device, medium and equipment for establishing detrusor overactivity diagnosis model
CN115295149A (en) Method for constructing data model of intestinal constipation
CN115206521A (en) Method and device for diagnosing lower urinary tract obstruction
CN115116606A (en) Urodynamics examination and diagnosis device
US20230309931A1 (en) Deep learning-based simple urine flow test result learning method and lower urinary tract symptom diagnosis method
Zaylaa et al. Automatic Monodimensional EHG Contractions' Segmentation
Kim et al. Automatic segmentation of liver/kidney area with double-layered fuzzy C-means and the utility of hepatorenal index for fatty liver severity classification
Enabling et al. OH, USA
CN117898682B (en) Epileptic prediction system based on gastrointestinal electric signals and construction method
Abbaraju et al. Machine learning for automated bladder event classification from single-channel vesical pressure recordings

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination