CN117373674B - Aortic valve stenosis persistence risk prediction method, system, equipment and medium - Google Patents
Aortic valve stenosis persistence risk prediction method, system, equipment and medium Download PDFInfo
- Publication number
- CN117373674B CN117373674B CN202311659808.6A CN202311659808A CN117373674B CN 117373674 B CN117373674 B CN 117373674B CN 202311659808 A CN202311659808 A CN 202311659808A CN 117373674 B CN117373674 B CN 117373674B
- Authority
- CN
- China
- Prior art keywords
- aortic valve
- valve stenosis
- data
- sample data
- risk
- 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.)
- Active
Links
- 206010002906 aortic stenosis Diseases 0.000 title claims abstract description 127
- 208000003017 Aortic Valve Stenosis Diseases 0.000 title claims abstract description 115
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000002688 persistence Effects 0.000 title claims description 26
- 238000013058 risk prediction model Methods 0.000 claims abstract description 59
- 238000000605 extraction Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000002776 aggregation Effects 0.000 claims abstract description 25
- 238000004220 aggregation Methods 0.000 claims abstract description 25
- 230000002085 persistent effect Effects 0.000 claims abstract description 15
- 230000004927 fusion Effects 0.000 claims description 50
- 230000007211 cardiovascular event Effects 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 18
- 230000007246 mechanism Effects 0.000 claims description 16
- 238000011176 pooling Methods 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 9
- 230000036772 blood pressure Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 6
- 238000013434 data augmentation Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000012217 deletion Methods 0.000 claims description 4
- 230000037430 deletion Effects 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 4
- 238000003780 insertion Methods 0.000 claims description 4
- 230000037431 insertion Effects 0.000 claims description 4
- 238000006467 substitution reaction Methods 0.000 claims description 4
- 238000010876 biochemical test Methods 0.000 claims description 2
- 230000000747 cardiac effect Effects 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 4
- 238000013528 artificial neural network Methods 0.000 description 9
- 238000001514 detection method Methods 0.000 description 8
- 238000003062 neural network model Methods 0.000 description 7
- 230000011218 segmentation Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 6
- 206010019280 Heart failures Diseases 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 4
- 230000001537 neural effect Effects 0.000 description 4
- 230000002980 postoperative effect Effects 0.000 description 4
- 208000006011 Stroke Diseases 0.000 description 3
- 238000009530 blood pressure measurement Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 239000003814 drug Substances 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000003902 lesion Effects 0.000 description 3
- 238000012502 risk assessment Methods 0.000 description 3
- 206010049993 Cardiac death Diseases 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 2
- 206010011906 Death Diseases 0.000 description 2
- 208000035965 Postoperative Complications Diseases 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 208000024891 symptom Diseases 0.000 description 2
- 238000004260 weight control Methods 0.000 description 2
- 208000009304 Acute Kidney Injury Diseases 0.000 description 1
- 206010050559 Aortic valve calcification Diseases 0.000 description 1
- 208000004434 Calcinosis Diseases 0.000 description 1
- 206010008190 Cerebrovascular accident Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 208000004880 Polyuria Diseases 0.000 description 1
- 206010058046 Post procedural complication Diseases 0.000 description 1
- 208000033626 Renal failure acute Diseases 0.000 description 1
- 108090000190 Thrombin Proteins 0.000 description 1
- 208000007536 Thrombosis Diseases 0.000 description 1
- 208000003959 Ventricular Outflow Obstruction Diseases 0.000 description 1
- 201000011040 acute kidney failure Diseases 0.000 description 1
- 208000012998 acute renal failure Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000003146 anticoagulant agent Substances 0.000 description 1
- 229940127219 anticoagulant drug Drugs 0.000 description 1
- 210000001765 aortic valve Anatomy 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000002308 calcification Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000001684 chronic effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 230000035619 diuresis Effects 0.000 description 1
- 210000003021 heart conduction system Anatomy 0.000 description 1
- 230000002489 hematologic effect Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000003907 kidney function Effects 0.000 description 1
- 210000004185 liver Anatomy 0.000 description 1
- 230000003908 liver function Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000012495 positive regulation of renal sodium excretion Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
- 229960004072 thrombin Drugs 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a method, a system, equipment and a medium for predicting the persistent risk of aortic valve stenosis, belongs to the technical field of artificial intelligence, and aims to solve the technical problem that the persistent risk of aortic valve stenosis cannot be predicted in the prior art. It comprises the following steps: acquiring clinical sample data and labels of patients with aortic valve stenosis, and constructing an aortic valve stenosis continuous risk prediction model, wherein the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module; training the aortic valve stenosis continuous risk prediction model by using the acquired sample data and the label to obtain a mature aortic valve stenosis continuous risk prediction model; and acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, relates to risk prediction of aortic valve stenosis, and in particular relates to a method, a system, equipment and a medium for predicting persistent risk of aortic valve stenosis.
Background
Aortic stenosis is a common senile cardiovascular disease with high mortality after symptoms, but there is currently no effective medical intervention to delay or prevent its progression. The mortality rate of patients who are not treated surgically for 2 years after symptoms are present is over 50%. At the same time, more than 30% of patients with aortic stenosis may develop stroke and heart failure due to the compliance of aortic valve calcifications and left ventricular outflow obstruction. Minimally invasive interventional therapy and postoperative regular medical intervention are the most effective methods for reducing cardiovascular events of patients with high-risk aortic valve stenosis, so accurate risk assessment and effective technical means for screening high-risk patients are of great importance to such people. Patients with aortic valve stenosis require continuous risk prediction and assessment of cardiovascular events occurring in patients with aortic valve stenosis, either during treatment or before and after treatment.
Artificial intelligence has been widely used in clinical medicine in a number of fields such as lesion segmentation, semantic recognition, and risk assessment. The neural network is a nonlinear model simulating a brain information processing mechanism, and is an important way for realizing artificial intelligence. The description and analysis of dynamic systems is an important research content in the development and practical application of neural network theory. Neural network approaches such as attention mechanisms, memory mechanisms, etc. have proven to have good processing power for continuous data changes and incremental updates. Neural network models have been successfully used for dynamic medical data analysis. 2021, nenad et al published a neural network model based on the electronic medical record system for predicting the persistent risk of acute renal failure patients with an accuracy of over 90%. In addition, the neural ordinary differential equation network is used as a newly-proposed neural network model, breaks through the defect of insufficient dynamic property of the traditional neural network model, and is particularly suitable for dynamic medical data analysis, such as medical dynamic images or indexes of continuously-changing severe patients. These works show that the development of the modern advanced neural network method is expected to break through the technical bottleneck restricting the sustainable risk prediction and early warning of aortic valve stenosis.
The invention patent with application number 202111538987.9 discloses a TAVR postoperative complication risk value prediction method based on a polymeric neural network, wherein a constructed polymeric neural network model comprises a pre-operation module, an intra-operation module and a polymeric neural network model of a post-operation module, wherein the pre-operation module, the intra-operation module and the post-operation module comprise an input layer, a hidden layer and an output layer, and each hidden layer comprises a full connection layer, a batch normalization layer and a nonlinear activation function layer; the hidden layer of the preoperative module is connected with the hidden layer of the intraoperative module, and the hidden layer of the intraoperative module is connected with the hidden layer of the postoperative module; through the neural network model, probability values of postoperative complications can be given in stages, and prediction accuracy of postoperative risks is improved.
As in the prior art described above, most of existing risk prediction of aortic valve stenosis is to obtain data of a certain time (pre-operation, intra-operation and/or post-operation) as sample data, and the sample data is not persistent, so that the disease development of the patient and the dynamic nature of the data change cannot be overcome, and the accuracy of dividing the calcification of the aortic valve is low, so that the sustainable risk prediction cannot be realized. Thus, there is a need for innovations and improvements in the network architecture to enable continuous risk prediction for aortic stenosis.
Disclosure of Invention
The invention aims at: in order to solve the technical problem that the persistent risk prediction of aortic valve stenosis cannot be performed in the prior art, the invention provides a method, a system, equipment and a medium for predicting the persistent risk of aortic valve stenosis.
The invention adopts the following technical scheme for realizing the purposes:
a method of predicting persistent risk of aortic stenosis, comprising the steps of:
step S1, acquiring sample data and labels;
acquiring clinical sample data and labels of patients with aortic valve stenosis, wherein the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data under different time, and the labels comprise labels of cardiovascular events and labels of cardiovascular events;
s2, constructing an aortic valve stenosis persistence risk prediction model;
constructing an aortic valve stenosis continuous risk prediction model, wherein the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; the fused characteristics are input into a prediction module for prediction, and a prediction result is output;
the feature extraction module comprises three convolution layers, three maximum pooling layers, three ReLU activation functions, two full-connection layers, a fusion layer and a normal micro-layering, wherein image data in clinical sample data are sequentially arranged in a crossing mode and then input into the fusion layer after passing through the three convolution layers, the three maximum pooling layers and the three ReLU activation functions, clinical data in the clinical sample data are input into the fusion layer after passing through the first full-connection layer, output of the fusion layer is used as input of a second full-connection layer, and output of the second full-connection layer is used as input of the normal micro-layering;
step S3, training an aortic valve stenosis persistence risk prediction model;
training the aortic valve stenosis continuous risk prediction model constructed in the step S2 by adopting the sample data and the label obtained in the step S1 to obtain a mature aortic valve stenosis continuous risk prediction model;
s4, predicting in real time;
and acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model.
Further, in step S1, the substantially unchanged risk factor sample data includes cardiac image data, demographic data, and medical record data;
the variable risk factor sample data at different times includes electrocardiographic data, home blood pressure monitoring data, BMI self-test data, hematology test data, and biochemical test data.
Further, preprocessing is carried out on the heart image data in the acquired sample data, wherein the preprocessing comprises normalization processing and data augmentation processing;
firstly cleaning medical record data in sample data, then carrying out data enhancement, and finally converting text into numerical values by using text codes; methods of data enhancement include synonym substitution, random deletion, inverting sentences, random insertion.
Further, in the usual micro-hierarchy, a mapping from the input data space to the memory space is established, expressed as:
wherein,representing the status of the network,/->；/>Representing an external input variable,/->；/>Weights representing the weights available for ordinary differential learning, +.>Representing the offset.
Further, in step S2, the dynamic aggregation network module dynamically aggregates the current-stage features and the previous-stage features output by the feature extraction module by using a feature fusion algorithm; the formula of the feature fusion algorithm is expressed as:
wherein,the output of the dynamic weight mechanism is represented and realized through a GRU gating mechanism; />And->Representing the current stage feature and the last stage feature, respectively.
Further, in step S3, when the aortic valve stenosis persistence risk prediction model is trained, the learning rate is set to 0.001 by the network, and the learning rate decays ten times after every 30 learning iterations; the convolution weight is initialized by using Gaussian distribution, one training batch is set to be 16, and the learning iteration number is 500; the network training uses a BP feedback propagation algorithm to calculate gradients and update weights.
An aortic valve stenosis persistence risk prediction system comprising:
the system comprises a sample data and label acquisition module, a data processing module and a data processing module, wherein the sample data and label acquisition module is used for acquiring clinical sample data and labels of patients with aortic valve stenosis, the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data under different time, and the labels comprise labels of cardiovascular events and labels of cardiovascular events;
the model construction module is used for constructing an aortic valve stenosis continuous risk prediction model, and the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; the fused characteristics are input into a prediction module for prediction, and a prediction result is output;
the feature extraction module comprises three convolution layers, three maximum pooling layers, three ReLU activation functions, two full-connection layers, a fusion layer and a normal micro-layering, wherein image data in clinical sample data are sequentially arranged in a crossing mode and then input into the fusion layer after passing through the three convolution layers, the three maximum pooling layers and the three ReLU activation functions, clinical data in the clinical sample data are input into the fusion layer after passing through the first full-connection layer, output of the fusion layer is used as input of a second full-connection layer, and output of the second full-connection layer is used as input of the normal micro-layering;
the model training module is used for training the aortic valve stenosis continuous risk prediction model constructed by the model construction module by adopting the sample data and the label acquired by the label acquisition module to acquire a mature aortic valve stenosis continuous risk prediction model;
the real-time prediction module is used for acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model.
A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method described above.
A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method described above.
The beneficial effects of the invention are as follows:
1. in the invention, the sample data are basically invariable risk factor sample data obtained continuously and variable risk factor sample data under different time, and are used as training samples for learning an aortic valve stenosis persistence prediction model, and the model can learn aortic valve stenosis from the persistence sample data and generate the dynamic property of data change of cardiovascular events, so that the model has the persistence risk prediction capability of the aortic valve stenosis, realizes persistence risk prediction of the aortic valve stenosis, and effectively improves persistence risk prediction precision of the aortic valve stenosis.
2. According to the invention, irregular time series data generated in diagnosis and treatment of patients with aortic valve stenosis are processed through a neural frequent differential equation network, so that the problem that a traditional prediction model cannot incorporate dynamic variable risk factors is solved, and the prediction accuracy and efficiency of the prediction model are remarkably improved.
3. According to the invention, feature cascading is realized based on a self-adaptive attention mechanism, a progressive prediction network is established, and the dynamic evolution characteristics of aortic valve stenosis are combined, so that the accurate risk prediction and assessment of the whole period of a patient are realized, and the singleness and one-sided performance of the traditional prediction model and system are broken through.
4. According to the invention, the trained model can be rapidly and timely predicted for continuous risk through automatic aortic valve stenosis, so that manpower and material resources for aortic valve stenosis patient management are saved, and a beneficial help is provided for follow-up assistance in accurate diagnosis and evaluation.
Drawings
FIG. 1 is a schematic diagram of the structure of the present invention;
fig. 2 is a schematic structural diagram of a feature extraction module in the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention.
Thus, all other embodiments, which can be made by one of ordinary skill in the art without undue burden from the invention, are intended to be within the scope of the invention.
Example 1
The embodiment provides a method for predicting persistent risk of aortic stenosis, as shown in fig. 1, specifically comprising the following steps:
step S1, acquiring sample data and labels;
clinical sample data of patients with aortic valve stenosis are obtained, wherein the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data at different times, and the label comprises cardiovascular events generated by patients with aortic valve stenosis at any time. Cardiovascular events are a collection of events that occur in patients with total cause death, cardiac death, stroke, and readmission of heart failure. The label is numerical data after being coded by text, for example, the label is 0, which indicates that the patient does not generate any cardiovascular event; a label of 1 indicates that the patient has all-cause death; the label is 2, which indicates that the patient has cardiac death; the label is 3, which indicates that the patient has cerebral apoplexy; the label is 4, which indicates that the patient is admitted to the hospital after heart failure.
The substantially unchanged risk factor sample data means that the sample cannot easily change with time, and mainly comprises heart image data, demographic data and medical history data. The heart image data, including "heart multi-layer spiral CT" and "heart ultrasound", are the main methods and data types for analyzing the structure and lesion degree of the valve itself. Demographic data, "gender" will affect the type of valve decay in the patient, in addition to the "age" directly affecting the risk of mortality competition in the patient. The medical history data of patients, including 'diabetes history', 'hypertension history', 'chronic renal history', are important risk factors of patients suffering from cardiovascular diseases such as valvular disease, and have direct effects on the progress of valvular disease and the improvement of the risks of all cardiovascular events of the patients.
Variable risk factor sample data at different times means that samples can easily change over time, and this part of data is ignored in past clinical practice and is difficult to process by common methods. The system mainly comprises electrocardiographic data, home blood pressure monitoring data, BMI self-test data, hematology detection data and biochemical detection data. The central electric data is mainly 'electrocardiogram signal' data continuously acquired by a patient, and the deterioration condition of a heart conduction system can be analyzed by taking the electrocardiographic data into consideration because lesions of a valve and surgical treatment have direct influence on heart electric activity. The home blood pressure data are continuous 'blood pressure measurement values' obtained by a patient through home equipment, and besides the home blood pressure measurement values are used for analyzing the obstruction degree of a valve on an outflow channel, the home 'blood pressure measurement values' are helpful for analyzing the influence of post-operation heart rehabilitation drugs on blood pressure reduction, so that the optimal dosage is obtained. Continuous dynamic monitoring of BMI helps to analyze patient dynamic weight changes, and increases and decreases thereof may help to analyze whether or not aortic valve stenosis patients exhibit natriuresis or excessive diuresis, thereby avoiding the occurrence of heart failure. In addition to monitoring the effect of long-term post-operative medication on liver and kidney functions, hematological and biochemical monitoring can also be used for evaluating the use pointer of patients for anticoagulants for preventing valve thrombosis by monitoring the international thrombin source time. The above data are critical for risk assessment of aortic stenosis, and previous lack of methods for effectively processing such data in series
The label comprises a label with a cardiovascular event and a label without a cardiovascular event, wherein the label with the cardiovascular event refers to various cardiovascular events generated by patients with aortic valve stenosis at any time.
The method comprises the steps of preprocessing image data in acquired sample data, wherein the preprocessing comprises normalization processing and data augmentation processing, and the data augmentation processing method comprises random rotation, random inversion and random contrast enhancement.
The medical record data in the sample data is firstly subjected to data cleaning, then subjected to data enhancement, and finally converted into numerical values by utilizing text codes; methods of data enhancement include synonym substitution, random deletion, inverting sentences, random insertion.
After the data of the two types are processed, the data is input into an aortic valve stenosis continuous risk prediction model for continuous learning of the aortic valve stenosis continuous risk prediction model.
S2, constructing an aortic valve stenosis persistence risk prediction model;
constructing an aortic valve stenosis continuous risk prediction model, wherein the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; and the fused characteristics are input into a prediction module for prediction, and a prediction result is output.
The feature extraction module is mainly used for learning the medical history data (namely sample data) of the patient in different time phases to obtain features with more discernment; the dynamic aggregation network module comprises a selectable fusion of current stage features (including current stage dynamically variable patient history data features and substantially constant patient background data features) and previous stage features; the prediction module predicts the risk (cardiovascular event) of the aortic valve stenosis according to the fusion characteristics.
As shown in fig. 2, the feature extraction module includes three convolutional layers, three max pooling layers, three ReLU activation functions, two fully connected layers, a fusion layer, and a normally micro-hierarchical layer. The image data (including heart image data and electrocardio data) in the clinical sample data are input into the fusion layer after passing through three convolution layers, three maximum pooling layers and three ReLU activation functions which are sequentially arranged in a crossing mode, the clinical data (including demographic data, medical history data, family blood pressure monitoring data, BMI self-test data, hematology detection data and biochemistry detection data) in the clinical sample data are input into the fusion layer after passing through the first full-connection layer, the output of the fusion layer is used as the input of the second full-connection layer, and the output of the second full-connection layer is used as the input of the normal micro-layering.
Three convolution layers and three maximum pooling layers are used, and the characteristics of the image are gradually extracted; each convolution layer is followed by a ReLU activation function to increase nonlinearity; the maximum pooling layer is beneficial to reducing the space dimension of the feature map and keeping key information; clinical data is input into a fully connected layer (i.e., the first fully connected layer) that has 512 neurons for learning complex features of the clinical data. In order to effectively combine the image features and the clinical features, they are connected together using a fusion layer and then processed through a fully connected layer (i.e., the second fully connected layer), the output of which contains information from both branches to better perform the aortic stenosis persistence risk prediction task.
The normal micro-layer is used for re-polymerizing the extracted features to obtain features with more discriminant ability, so that the accuracy of predicting the continuous risk of aortic valve stenosis is improved. Unlike the ordinary neural ordinary differential equation, ordinary micro-layering separates learning neurons from memory neurons, and thus has a clearer dynamic feature capturing capability. And the ordinary differential layer is given an external input, mapping the input to its global attractor. I.e., a normally micro-hierarchy, which establishes a mapping from input data space to memory space, expressed as:
wherein,representing the status of the network,/->；/>Representing an external input variable,/->；/>Weights representing the weights available for ordinary differential learning, +.>Representing the offset. Given an initial value y (0), the trajectory of the equation starting from y (0) can be expressed by y (t, y (0)) or simply y (t), the ordinary differential equation having a global attractor for each external input, and this attractor being the only solution to the above-mentioned hidden mapping equation.
As shown in fig. 1, the basic unchanged risk factor sample data and the variable risk factor sample data under different time are respectively and independently input into a feature extraction module, the feature extraction module performs feature extraction, outputs corresponding basic unchanged risk factor features and variable risk factor features, and then inputs the two basic unchanged risk factor features and the variable risk factor features into a dynamic aggregation network module together for feature fusion.
Furthermore, the dynamic aggregation network module aims at characterizing the current phase of the current time phaseFeatures of the previous phase from the previous time phase +.>Fused together to generate a composite feature +.>This integrated feature will be used for risk prediction. Therefore, in this embodiment, a dynamic weight control mechanism is introduced into the dynamic aggregation network module, which allows features to be selectively fused according to the current situation. This weight is typically generated by a neural network module (which may be an existing neural network module) and may be dynamically adjusted based on the input data and the context information. This weighting mechanism can be expressed as a function W (∈) that accepts input features and context information and outputs weights for feature fusion. The method comprises the following steps:
the dynamic aggregation network module adopts a feature fusion algorithm to dynamically aggregate the current-stage features and the previous-stage features output by the feature extraction module; the formula of the feature fusion algorithm is expressed as:
wherein,the output of the dynamic weight mechanism is represented and realized through a GRU gating mechanism; />And->Representing the current stage feature and the last stage feature, respectively.
On this basis, the present embodiment uses the GRU (existing method) to generate a weight between 0 and 1, which represents the degree of contribution of the current-stage feature and the previous-stage feature in feature fusion. This weight generation formula is as follows:
wherein,and->Respectively representing the current stage characteristic and the last stage characteristic; />Representing an update gate, wherein the update gate mainly controls the update degree of the current time stage characteristics, allows a model to determine whether to integrate the current characteristics, and has a calculation formula of +.>；/>Indicating whether the reset gate, the main control, retains the information of the last time period, allows the model to determine whether to rely on the characteristics of the last time period, and the calculation formula is as followsWherein->、/>、/>And->Trainable parameters are predicted for the persistent risk of aortic stenosis.
In this embodiment, the prediction module is not innovative, and the prediction module in the prior art can be directly adopted, so that creative labor is not required.
Step S3, training an aortic valve stenosis persistence risk prediction model;
training the aortic valve stenosis continuous risk prediction model constructed in the step S2 by adopting the sample data and the label obtained in the step S1 to obtain a mature aortic valve stenosis continuous risk prediction model.
And inputting the sample data subjected to data processing into an aortic valve stenosis continuous risk prediction model, and training the aortic valve stenosis continuous risk prediction model. When the aortic valve stenosis persistence risk prediction model is trained, the learning rate is set to be 0.001 by the network, and the learning rate decays ten times after every 30 learning iterations; the convolution weight is initialized by using Gaussian distribution, one training batch is set to be 16, and the learning iteration number is 500; the network training uses a BP feedback propagation algorithm to calculate gradients and update weights. The network learning updates a parameter for each batch, after each iteration learning, the segmentation model judges the segmentation evaluation result, if the current error is smaller than the error of the previous iteration, the current segmentation model is saved, and then training is continued until the maximum iteration times are reached.
In addition, the loss function adopted in training is the cross entropy loss function in the prior art, and creative labor is not required.
S4, predicting in real time;
and acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model.
Example 2
The present embodiment provides an aortic valve stenosis duration risk prediction system, specifically including:
the system comprises a sample data and a label acquisition module, wherein the sample data and the label acquisition module are used for acquiring clinical sample data and labels of patients with aortic valve stenosis, the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data under different time, and the labels comprise cardiovascular events generated by the patients with aortic valve stenosis at any time.
The risk factor sample data which is basically unchanged means that the sample cannot easily change with time, and mainly comprises heart image data, demographic data and medical record data. The variable risk factor sample data at different times means that the sample can easily change along with time, and mainly comprises electrocardio data, household blood pressure monitoring data, BMI self-test data, hematology detection data and biochemical detection data.
The label comprises a label with a cardiovascular event and a label without a cardiovascular event, wherein the label with the cardiovascular event refers to various cardiovascular events generated by patients with aortic valve stenosis at any time.
The method comprises the steps of preprocessing image data in acquired sample data, wherein the preprocessing comprises normalization processing and data augmentation processing, and the data augmentation processing method comprises random rotation, random inversion and random contrast enhancement.
The medical record data in the sample data is firstly subjected to data cleaning, then subjected to data enhancement, and finally converted into numerical values by utilizing text codes; methods of data enhancement include synonym substitution, random deletion, inverting sentences, random insertion.
After the data of the two types are processed, the data is input into an aortic valve stenosis continuous risk prediction model for continuous learning of the aortic valve stenosis continuous risk prediction model.
The model construction module is used for constructing an aortic valve stenosis continuous risk prediction model, and the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; and the fused characteristics are input into a prediction module for prediction, and a prediction result is output.
The feature extraction module is mainly used for learning the medical history data (namely sample data) of the patient in different time phases to obtain features with more discernment; the dynamic aggregation network module comprises a selectable fusion of current stage features (including current stage dynamically variable patient history data features and substantially constant patient background data features) and previous stage features; the prediction module predicts the continuous risk of aortic stenosis according to the fusion characteristics.
As shown in fig. 2, the feature extraction module includes three convolutional layers, three max pooling layers, three ReLU activation functions, two fully connected layers, a fusion layer, and a normally micro-hierarchical layer. The image data (including heart image data and electrocardio data) in the clinical sample data are input into the fusion layer after passing through three convolution layers, three maximum pooling layers and three ReLU activation functions which are sequentially arranged in a crossing mode, the clinical data (including demographic data, medical history data, family blood pressure monitoring data, BMI self-test data, hematology detection data and biochemistry detection data) in the clinical sample data are input into the fusion layer after passing through the first full-connection layer, the output of the fusion layer is used as the input of the second full-connection layer, and the output of the second full-connection layer is used as the input of the normal micro-layering.
Three convolution layers and three maximum pooling layers are used, and the characteristics of the image are gradually extracted; each convolution layer is followed by a ReLU activation function to increase nonlinearity; the maximum pooling layer is beneficial to reducing the space dimension of the feature map and keeping key information; clinical data is input into a fully connected layer (i.e., the first fully connected layer) that has 512 neurons for learning complex features of the clinical data. In order to effectively combine the image features and the clinical features, they are connected together using a fusion layer and then processed through a fully connected layer (i.e., the second fully connected layer), the output of which contains information from both branches to better perform the aortic stenosis persistence risk prediction task.
The normal micro-layer is used for re-polymerizing the extracted features to obtain features with more discriminant ability, so that the accuracy of predicting the continuous risk of aortic valve stenosis is improved. Unlike the ordinary neural ordinary differential equation, ordinary micro-layering separates learning neurons from memory neurons, and thus has a clearer dynamic feature capturing capability. And the ordinary differential layer is given an external input, mapping the input to its global attractor. I.e., a normally micro-hierarchy, which establishes a mapping from input data space to memory space, expressed as:
wherein,representing the status of the network,/->；/>Representing an external input variable,/->；/>Weights representing the weights available for ordinary differential learning, +.>Representing the offset. Given an initial value y (0), the trajectory of the equation starting from y (0) can be expressed by y (t, y (0)) or simply y (t), the ordinary differential equation having a global attractor for each external input, and this attractor being the only solution to the above-mentioned hidden mapping equation.
As shown in fig. 1, the basic unchanged risk factor sample data and the variable risk factor sample data under different time are respectively and independently input into a feature extraction module, the feature extraction module performs feature extraction, outputs corresponding basic unchanged risk factor features and variable risk factor features, and then inputs the two basic unchanged risk factor features and the variable risk factor features into a dynamic aggregation network module together for feature fusion.
Furthermore, the dynamic aggregation network module aims at characterizing the current phase of the current time phaseFeatures of the previous phase from the previous time phase +.>Fused together to generateGeneral features->This integrated feature will be used for risk prediction. Therefore, in this embodiment, a dynamic weight control mechanism is introduced into the dynamic aggregation network module, which allows features to be selectively fused according to the current situation. This weight is typically generated by a neural network module (which may be an existing neural network module) and may be dynamically adjusted based on the input data and the context information. This weighting mechanism can be expressed as a function W (∈) that accepts input features and context information and outputs weights for feature fusion. The method comprises the following steps:
the dynamic aggregation network module adopts a feature fusion algorithm to dynamically aggregate the current-stage features and the previous-stage features output by the feature extraction module; the formula of the feature fusion algorithm is expressed as:
wherein,the output of the dynamic weight mechanism is represented and realized through a GRU gating mechanism; />And->Representing the current stage feature and the last stage feature, respectively.
On this basis, the present embodiment uses the GRU (existing method) to generate a weight between 0 and 1, which represents the degree of contribution of the current-stage feature and the previous-stage feature in feature fusion. This weight generation formula is as follows:
wherein,and->Respectively representing the current stage characteristic and the last stage characteristic; />Representing an update gate, wherein the update gate mainly controls the update degree of the current time stage characteristics, allows a model to determine whether to integrate the current characteristics, and has a calculation formula of +.>；/>Indicating whether the reset gate, the main control, retains the information of the last time period, allows the model to determine whether to rely on the characteristics of the last time period, and the calculation formula is as followsWherein->、/>、/>And->Trainable parameters are predicted for the persistent risk of aortic stenosis.
In this embodiment, the prediction module is not innovative, and the prediction module in the prior art can be directly adopted, so that creative labor is not required.
The model training module is used for training the aortic valve stenosis continuous risk prediction model constructed by the model construction module by adopting the sample data and the label acquired by the label acquisition module to obtain a mature aortic valve stenosis continuous risk prediction model.
And inputting the sample data subjected to data processing into an aortic valve stenosis continuous risk prediction model, and training the aortic valve stenosis continuous risk prediction model. When the aortic valve stenosis persistence risk prediction model is trained, the learning rate is set to be 0.001 by the network, and the learning rate decays ten times after every 30 learning iterations; the convolution weight is initialized by using Gaussian distribution, one training batch is set to be 16, and the learning iteration number is 500; the network training uses a BP feedback propagation algorithm to calculate gradients and update weights. The network learning updates a parameter for each batch, after each iteration learning, the segmentation model judges the segmentation evaluation result, if the current error is smaller than the error of the previous iteration, the current segmentation model is saved, and then training is continued until the maximum iteration times are reached.
In addition, the loss function adopted in training is the cross entropy loss function in the prior art, and creative labor is not required.
The real-time prediction module is used for acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model.
Example 3
A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of an aortic valve stenosis persistence risk prediction method.
The computer equipment can be computing equipment such as a desktop computer, a notebook computer, a palm computer, a cloud server and the like. The computer equipment can perform man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch pad or voice control equipment and the like.
The memory includes at least one type of readable storage medium including flash memory, hard disk, multimedia card, card memory (e.g., SD or D interface display memory, etc.), random Access Memory (RAM), static Random Access Memory (SRAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), programmable Read Only Memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk provided on the computer device, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), or the like. Of course, the memory may also include both internal storage units of the computer device and external storage devices. In this embodiment, the memory is often used to store an operating system and various application software installed on the computer device, such as program codes of the aortic valve stenosis persistence risk prediction method. In addition, the memory may be used to temporarily store various types of data that have been output or are to be output.
The processor may be a central processing unit (CentralProcessingUnit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor is typically used to control the overall operation of the computer device. In this embodiment, the processor is configured to execute the program code stored in the memory or process data, such as the program code of the aortic valve stenosis persistence risk prediction method.
Example 4
A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of an aortic valve stenosis persistence risk prediction method.
Wherein the computer-readable storage medium stores an interface display program executable by at least one processor to cause the at least one processor to perform the steps of the aortic valve stenosis persistence risk prediction method as described above.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk), comprising several instructions for causing a terminal device (which may be a mobile phone, a computer, a server or a network device, etc.) to perform the aortic valve stenosis persistence risk prediction method according to the embodiments of the present application.
Claims (8)
1. A method for predicting persistent risk of aortic stenosis, comprising the steps of:
step S1, acquiring sample data and labels;
acquiring clinical sample data and labels of patients with aortic valve stenosis, wherein the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data under different time, and the labels comprise labels of cardiovascular events and labels of cardiovascular events;
s2, constructing an aortic valve stenosis persistence risk prediction model;
constructing an aortic valve stenosis continuous risk prediction model, wherein the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; the fused characteristics are input into a prediction module for prediction, and a prediction result is output;
the feature extraction module comprises three convolution layers, three maximum pooling layers, three ReLU activation functions, two full connection layers, a fusion layer and a normal micro layering; the method comprises the steps that image data in clinical sample data are input into a fusion layer after passing through three convolution layers, three maximum pooling layers and three ReLU activation functions which are sequentially arranged in a crossing mode, the clinical data in the clinical sample data are input into the fusion layer after passing through a first full-connection layer, output of the fusion layer serves as input of a second full-connection layer, and output of the second full-connection layer serves as input of a normal micro-layering;
step S3, training an aortic valve stenosis persistence risk prediction model;
training the aortic valve stenosis continuous risk prediction model constructed in the step S2 by adopting the sample data and the label obtained in the step S1 to obtain a mature aortic valve stenosis continuous risk prediction model;
s4, predicting in real time;
acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model;
in the usual micro-hierarchy, a mapping from the input data space to the memory space is established, expressed as:
wherein,representing the status of the network,/->；/>Representing an external input variable,/->；/>Weights representing the weights available for ordinary differential learning, +.>Representing the offset.
2. The method of claim 1, wherein in step S1, the substantially unchanged risk factor sample data includes cardiac image data, demographic data, and medical record data;
the variable risk factor sample data at different times includes electrocardiographic data, home blood pressure monitoring data, BMI self-test data, hematology test data, and biochemical test data.
3. A method for predicting persistent risk of aortic valve stenosis as set forth in claim 2, wherein,
preprocessing the heart image data in the acquired sample data, wherein the preprocessing comprises normalization processing and data augmentation processing;
firstly cleaning medical record data in sample data, then carrying out data enhancement, and finally converting text into numerical values by using text codes; methods of data enhancement include synonym substitution, random deletion, inverting sentences, random insertion.
4. The method for predicting the persistent risk of aortic valve stenosis as set forth in claim 1, wherein in step S2, the dynamic aggregation network module dynamically aggregates the current-stage features and the previous-stage features outputted by the feature extraction module by using a feature fusion algorithm; the formula of the feature fusion algorithm is expressed as:
wherein,the output of the dynamic weight mechanism is represented and realized through a GRU gating mechanism; />And->Representing the current stage feature and the last stage feature, respectively.
5. The method for predicting the persistent risk of aortic valve stenosis according to claim 1, wherein in step S3, the learning rate is set to 0.001 by the network, and the learning rate decays ten times after every 30 learning iterations; the convolution weight is initialized by using Gaussian distribution, one training batch is set to be 16, and the learning iteration number is 500; the network training uses a BP feedback propagation algorithm to calculate gradients and update weights.
6. An aortic valve stenosis persistence risk prediction system, comprising:
the system comprises a sample data and label acquisition module, a data processing module and a data processing module, wherein the sample data and label acquisition module is used for acquiring clinical sample data and labels of patients with aortic valve stenosis, the clinical sample data comprises basically unchanged risk factor sample data and variable risk factor sample data under different time, and the labels comprise labels of cardiovascular events and labels of cardiovascular events;
the model construction module is used for constructing an aortic valve stenosis continuous risk prediction model, and the aortic valve stenosis continuous risk prediction model comprises a feature extraction module, a dynamic aggregation network module and a prediction module;
the feature extraction module is used for extracting the basic risk factor feature and the variable risk factor feature of the basic unchanged risk factor sample data and the variable risk factor sample data under different time respectively; basically, the characteristics of the risk factors are not changed, the characteristics of the risk factors are input into a dynamic aggregation network module to be subjected to characteristic fusion, and the fused characteristics are output; the fused characteristics are input into a prediction module for prediction, and a prediction result is output;
the feature extraction module comprises three convolution layers, three maximum pooling layers, three ReLU activation functions, two full connection layers, a fusion layer and a normal micro layering; the method comprises the steps that image data in clinical sample data are input into a fusion layer after passing through three convolution layers, three maximum pooling layers and three ReLU activation functions which are sequentially arranged in a crossing mode, the clinical data in the clinical sample data are input into the fusion layer after passing through a first full-connection layer, output of the fusion layer serves as input of a second full-connection layer, and output of the second full-connection layer serves as input of a normal micro-layering;
the model training module is used for training the aortic valve stenosis continuous risk prediction model constructed by the model construction module by adopting the sample data and the label acquired by the label acquisition module to acquire a mature aortic valve stenosis continuous risk prediction model;
the real-time prediction module is used for acquiring real-time clinical sample data, inputting an aortic valve stenosis continuous risk prediction model, and outputting a risk prediction result of the aortic valve stenosis by the aortic valve stenosis continuous risk prediction model;
in the usual micro-hierarchy, a mapping from the input data space to the memory space is established, expressed as:
wherein,representing the status of the network,/->；/>Representing an external input variable,/->；/>Weights representing the weights available for ordinary differential learning, +.>Representing the offset.
7. A computer device, characterized by: comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized by: a computer program is stored which, when executed by a processor, causes the processor to perform the steps of the method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311659808.6A CN117373674B (en) | 2023-12-06 | 2023-12-06 | Aortic valve stenosis persistence risk prediction method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311659808.6A CN117373674B (en) | 2023-12-06 | 2023-12-06 | Aortic valve stenosis persistence risk prediction method, system, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117373674A CN117373674A (en) | 2024-01-09 |
| CN117373674B true CN117373674B (en) | 2024-02-02 |
Family
ID=89402614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311659808.6A Active CN117373674B (en) | 2023-12-06 | 2023-12-06 | Aortic valve stenosis persistence risk prediction method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117373674B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363294A (en) * | 2018-03-26 | 2019-10-22 | 辉达公司 | Neural network is indicated using the path in network to improve the performance of neural network |
| CN113449636A (en) * | 2021-06-28 | 2021-09-28 | 苏州美糯爱医疗科技有限公司 | Artificial intelligence-based automatic classification method for severity of aortic stenosis |
| CN114242234A (en) * | 2021-12-15 | 2022-03-25 | 四川大学 | TAVR postoperative complication risk value prediction method based on aggregation neural network |
| CN114333769A (en) * | 2021-09-29 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Speech recognition method, computer program product, computer device and storage medium |
| CN115186701A (en) * | 2022-06-15 | 2022-10-14 | 西门子（中国）有限公司 | Bearing life prediction method, device, electronic device and storage medium |
| CN116229564A (en) * | 2022-12-27 | 2023-06-06 | 北京理工大学 | Gait recognition method and system based on dynamic aggregation network |
| CN116596145A (en) * | 2023-05-23 | 2023-08-15 | 北京理工大学 | Traffic flow space-time data prediction method, system, equipment and medium |
| CN116935892A (en) * | 2023-08-15 | 2023-10-24 | 浙江大学 | Industrial valve anomaly detection method based on audio key feature dynamic aggregation |
| CN117081941A (en) * | 2023-06-25 | 2023-11-17 | 中国工商银行股份有限公司 | Flow prediction method and device based on attention mechanism and electronic equipment |
| CN117153377A (en) * | 2023-10-11 | 2023-12-01 | 中山大学附属第一医院 | Model for predicting death risk of adult patient with moderately severe aortic valve stenosis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230015122A1 (en) * | 2019-12-17 | 2023-01-19 | Koninklijke Philips N.V. | Aortic stenosis classification |
-
2023
- 2023-12-06 CN CN202311659808.6A patent/CN117373674B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363294A (en) * | 2018-03-26 | 2019-10-22 | 辉达公司 | Neural network is indicated using the path in network to improve the performance of neural network |
| CN113449636A (en) * | 2021-06-28 | 2021-09-28 | 苏州美糯爱医疗科技有限公司 | Artificial intelligence-based automatic classification method for severity of aortic stenosis |
| CN114333769A (en) * | 2021-09-29 | 2022-04-12 | 腾讯科技（深圳）有限公司 | Speech recognition method, computer program product, computer device and storage medium |
| CN114242234A (en) * | 2021-12-15 | 2022-03-25 | 四川大学 | TAVR postoperative complication risk value prediction method based on aggregation neural network |
| CN115186701A (en) * | 2022-06-15 | 2022-10-14 | 西门子（中国）有限公司 | Bearing life prediction method, device, electronic device and storage medium |
| CN116229564A (en) * | 2022-12-27 | 2023-06-06 | 北京理工大学 | Gait recognition method and system based on dynamic aggregation network |
| CN116596145A (en) * | 2023-05-23 | 2023-08-15 | 北京理工大学 | Traffic flow space-time data prediction method, system, equipment and medium |
| CN117081941A (en) * | 2023-06-25 | 2023-11-17 | 中国工商银行股份有限公司 | Flow prediction method and device based on attention mechanism and electronic equipment |
| CN116935892A (en) * | 2023-08-15 | 2023-10-24 | 浙江大学 | Industrial valve anomaly detection method based on audio key feature dynamic aggregation |
| CN117153377A (en) * | 2023-10-11 | 2023-12-01 | 中山大学附属第一医院 | Model for predicting death risk of adult patient with moderately severe aortic valve stenosis |
Non-Patent Citations (5)
| Title |
|---|
| A risk prediction model in asymptomatic patients with severe aortic stenosis: CURRENT-AS risk score;Eri Minamino-Muta 等;《European Heart Journal - Quality of Care and Clinical Outcomes》;第6卷(第2期);第166-174页 * |
| Deep Learning Based Centerline-Aggregated Aortic Hemodynamics: An Efficient Alternative to Numerical Modeling of Hemodynamics;Pavlo Yevtushenko 等;《IEEE Journal of Biomedical and Health Informatics》;第26卷(第4期);第1815- 1825页 * |
| 刘维永 等.加强现代临床医学研究,规范术前风险评估.《中国胸心血管外科临床杂志》.2011,第18卷(第1期),第1页. * |
| 基于双注意力机制和多尺度特征的点云场景分割;于丽丽 等;《激光与光电子学进展》;第58卷(第24期);第1-9页 * |
| 无症状严重主动脉瓣狭窄的风险评估与管理;姚建民;《中华老年多器官疾病杂志》;第10卷(第1期);第92-96页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117373674A (en) | 2024-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tripoliti et al. | Heart failure: diagnosis, severity estimation and prediction of adverse events through machine learning techniques | |
| Cramer et al. | Predicting the incidence of pressure ulcers in the intensive care unit using machine learning | |
| Liu et al. | Predicting heart failure readmission from clinical notes using deep learning | |
| Fang et al. | Feature Selection Method Based on Class Discriminative Degree for Intelligent Medical Diagnosis. | |
| CN114512228A (en) | Traditional Chinese medicine disease auxiliary diagnosis system, equipment and storage medium | |
| Swapna et al. | Diabetes detection using ecg signals: An overview | |
| CN113270192A (en) | Hemodialysis risk prediction system based on combination of dynamic and static data and depth self-encoder | |
| De Falco et al. | A genetic programming-based regression for extrapolating a blood glucose-dynamics model from interstitial glucose measurements and their first derivatives | |
| Gong et al. | Prognosis analysis of heart failure based on recurrent attention model | |
| Sheikhalishahi et al. | Benchmarking machine learning models on eICU critical care dataset | |
| Teo et al. | Current trends in readmission prediction: an overview of approaches | |
| CN109157211A (en) | A kind of portable cardiac on-line intelligence monitoring diagnosis system design method | |
| Yuan et al. | Efficient symptom inquiring and diagnosis via adaptive alignment of reinforcement learning and classification | |
| Liang et al. | Heart failure disease prediction and stratification with temporal electronic health records data using patient representation | |
| CN117373674B (en) | Aortic valve stenosis persistence risk prediction method, system, equipment and medium | |
| CN115547502B (en) | Hemodialysis patient risk prediction device based on time sequence data | |
| Chaturvedi et al. | An Innovative Approach of Early Diabetes Prediction using Combined Approach of DC based Bidirectional GRU and CNN | |
| Abidin et al. | Identification of disease based on symptoms by employing ML | |
| Malik et al. | Reviewing classification methods on health care | |
| Shih et al. | An early warning system for hemodialysis complications utilizing transfer learning from HD IoT dataset | |
| CN116230224A (en) | Method and system for predicting adverse events of heart failure based on time sequence model | |
| Liu et al. | Machine learning based early detection system of cardiac arrest | |
| Ramos et al. | Unsupervised learning approach for predicting sepsis onset in ICU patients | |
| CN114783587A (en) | Intelligent prediction system for severe acute kidney injury | |
| CN114613497A (en) | Intelligent medical auxiliary diagnosis method of patient sample based on GBDT sample level |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |