CN115099275B - Training method of arrhythmia diagnosis model based on artificial neural network - Google Patents
Training method of arrhythmia diagnosis model based on artificial neural network Download PDFInfo
- Publication number
- CN115099275B CN115099275B CN202210758714.3A CN202210758714A CN115099275B CN 115099275 B CN115099275 B CN 115099275B CN 202210758714 A CN202210758714 A CN 202210758714A CN 115099275 B CN115099275 B CN 115099275B
- Authority
- CN
- China
- Prior art keywords
- training
- data set
- verification
- neural network
- layer
- 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
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
-
- 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
-
- 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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Cardiology (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Pathology (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a training method of an arrhythmia diagnosis model based on an artificial neural network, and relates to the technical field of medical signal processing. Training N epochs on a preset deep learning model by using a training data set, and determining the optimal epochs with the minimum loss function; the loss function of each epoch is an average loss function obtained by carrying out k-fold cross training verification for M times for a training data set; and training the preset deep learning model by using the optimal epoch training data set to obtain a target heart rhythm diagnosis neural network model. According to the method, the average loss function obtained through M times of k-fold cross training verification is used for further obtaining the target heart rhythm diagnosis neural network model, accuracy and efficiency of training the target heart rhythm diagnosis neural network model can be improved, the target heart rhythm diagnosis neural network model is verified once, information leakage into the model due to multiple verification processes is prevented, and reliability and generalization performance of the model are reduced.
Description
Technical Field
The invention relates to the technical field of medical signal processing, in particular to a training method of an arrhythmia diagnosis model based on an artificial neural network.
Background
In recent years, deep learning has shown a good effect in electrocardiographic classification. Current popular artificial intelligence methods for diagnosing arrhythmias include Support Vector Machines (SVM), deep Learning (DL), etc. The electrocardio classification intelligent diagnosis system based on artificial intelligence can effectively reduce subjective uncertainty of an expert, thereby avoiding misdiagnosis. The traditional computer-aided electrocardio artificial intelligent diagnosis algorithm mainly comprises three steps of data preprocessing, feature extraction and electrocardio classification. However, the high ability of fully automatic feature extraction in deep learning makes the key step electrocardiogram heartbeat diagnosis in traditional ML easier. Deep learning is a series of presentation layers with automatic search processes to better data presentation, which are learned by training processes of artificial neural networks to help automatically extract features and learn data presentation. A recent study has shown that deep learning based arrhythmia diagnosis is more accurate and efficient than manual classification by experts.
However, the current arrhythmia diagnosis model based on the deep learning model has the following problems: the generalization is not high, and the accuracy and the efficiency cannot achieve a good balance point.
Disclosure of Invention
The invention aims to solve the problems of the background technology and provides a training method of an arrhythmia diagnosis model based on an artificial neural network.
The aim of the invention can be achieved by the following technical scheme:
the embodiment of the invention provides a training method of an arrhythmia diagnosis model based on an artificial neural network, which comprises the following steps:
acquiring a historical heart rhythm data set with labels, and dividing the historical heart rhythm data set into a training data set and a verification data set; the electrocardio rhythm type marked as heart rate data in the historical heart rate data set;
training N epochs on a preset deep learning model by using the training data set, and determining the epochs with the minimum loss function as the optimal epochs; the loss function of each epoch is an average loss function obtained by carrying out k-fold cross training verification for M times on the training data set; wherein N and M are preset values;
initializing the preset deep learning model by using the optimal epoch, and training the initialized preset deep learning model by using the training data set to obtain a target heart rhythm diagnosis neural network model;
and verifying the target heart rhythm diagnosis neural network model once by using the verification data set, and inputting the verification data set into the target heart rhythm diagnosis neural network model to obtain a performance evaluation result of the target heart rhythm diagnosis neural network model.
Optionally, the preset deep learning model includes 11 layers, which are respectively a 3-layer separable channel convolution, 1 normalization layer, 1 max pooling layer, 1 bidirectional LSTM layer, 1 flattening layer, 1 Dropout layer, 1 dense connection layer, 1 batch normalization layer, and 1 Softmax layer.
Optionally, the 3-layer separable channel convolution of the preset deep learning model uses a ReLu activation function, each layer having 32, 64, and 128 cores of size 3 in sequence; the parameters of the maximum pooling layer are 2*1 and the step length is 2; the bidirectional LSTM layer is 128 units; the flattening layer, the Dropout layer, the dense connecting layer, the batch normalization layer and the Softmax layer form a prediction model; the dense connectivity layer has 512 neurons, using the ReLu activation function; the 50% characteristic of the Dropout layer is set to zero; the preset deep learning model is compiled using an Adam optimizer and a classification cross entropy loss function.
Optionally, acquiring a labeled historical heart rhythm data set, dividing the historical heart rhythm data set into a training data set and a verification data set, including:
acquiring a historical heart rhythm data set with labels, and resampling the historical heart rhythm data set to obtain a first data set; the sampling frequency is 125Hz;
normalizing the amplitude of the first electrocardiosignal, and dividing heart rate data in the first data set into data fragments with preset length by using a window function with preset length to obtain a second data set;
according to the first derivative, a set of all local extrema in the second data set is obtained, and 0.9 of a normalized extremum is taken as an R candidate peak;
taking the median of all R-R time intervals as the nominal heartbeat period T of the time window, determining the final period of each R-R peak value as 1.2T, and filling the rest zero points to reach the same length to obtain a third data set;
and randomly dividing heart rate data in the third data set into a training data set and a verification data set according to a preset proportion.
Optionally, training the preset deep learning model by using the training data set for N epochs, and determining the epoch with the smallest loss function as the optimal epoch, including:
training N epochs on a preset deep learning model by using the training data set;
dividing the training data set into M target training subsets aiming at each epoch training, and performing M times of k-fold cross training verification on a preset deep learning model by using the M target training subsets to obtain an average loss function of the M times of k-fold cross training verification of the epoch;
aiming at ith training verification, taking an ith target training subset of M target training subsets as a verification subset, taking the rest M-1 target training subsets as training subsets, calculating the training weight of each training subset according to the number of each electrocardiographic rhythm type of each training subset, and obtaining a loss function of the ith training verification according to the training weights of the M-1 training subsets; the ith training verification is any one of M times of k-fold cross training verification;
and determining the epoch with the smallest average loss function as the optimal epoch.
Optionally, the training weights of the training subset are calculated as:
where Weight represents the training Weight, n_samples represents the number of training samples in the training subset, n_classes represents the number of types of cardiac rhythms, np.bincount (y) represents the number of types of the y-th cardiac rhythm in the training subset.
Optionally, training the preset deep learning model by using the training data set for N epochs, which specifically includes:
and training N epochs on a preset deep learning model by using the training data set, calculating the self-adaptive learning rate of each epochs, and if the loss function of 5 continuous epochs is not reduced, reducing the self-adaptive learning rate of the next epochs to half of the current self-adaptive learning rate.
The embodiment of the invention provides a training method of an arrhythmia diagnosis model based on an artificial neural network, which comprises the steps of obtaining a historical rhythm data set with labels, and dividing the historical rhythm data set into a training data set and a verification data set; the type of heart rhythm labeled as heart rate data in the historical heart rate dataset; training N epochs on a preset deep learning model by using a training data set, and determining the epochs with the minimum loss function as the optimal epochs; the loss function of each epoch is an average loss function obtained by carrying out k-fold cross training verification for M times for a training data set; wherein N and M are preset values; initializing a preset deep learning model by using an optimal epoch, and training the initialized preset deep learning model by using a training data set to obtain a target heart rhythm diagnosis neural network model; and performing one-time verification on the target heart rhythm diagnosis neural network model by using a verification data set, and inputting the verification data set into the target heart rhythm diagnosis neural network model to obtain a performance evaluation result of the target heart rhythm diagnosis neural network model.
According to the method, the average loss function is obtained through M times of k-fold cross training verification, so that the target heart rhythm diagnosis neural network model is obtained, the accuracy and efficiency of training the target heart rhythm diagnosis neural network model can be improved, the target heart rhythm diagnosis neural network model is verified once, information leakage into the model caused by a plurality of verification processes is prevented, and the reliability and generalization performance of the model are reduced.
Drawings
The invention is further described below with reference to the accompanying drawings.
Fig. 1 is a flowchart of a training method of an arrhythmia diagnosis model based on an artificial neural network according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The embodiment of the invention provides a training method of an arrhythmia diagnosis model based on an artificial neural network, referring to fig. 1, fig. 1 is a flowchart of the training method of the arrhythmia diagnosis model based on the artificial neural network, which can include the following steps:
s101, acquiring a historical heart rhythm data set with labels, and dividing the historical heart rhythm data set into a training data set and a verification data set.
S102, training N epochs on a preset deep learning model by using a training data set, and determining the epochs with the minimum loss function as the optimal epochs.
And S103, initializing a preset deep learning model by using the optimal epoch, and training the initialized preset deep learning model by using a training data set to obtain a target heart rhythm diagnosis neural network model.
And S104, performing primary verification on the target heart rhythm diagnosis neural network model by using the verification data set, and inputting the verification data set into the target heart rhythm diagnosis neural network model to obtain a performance evaluation result of the target heart rhythm diagnosis neural network model.
The type of cardiac rhythm labeled as heart rate data in the historical heart rate dataset.
The loss function of each epoch is an average loss function obtained by carrying out k-fold cross training verification for M times for a training data set; wherein N and M are preset values.
According to the training method for the arrhythmia diagnosis model based on the artificial neural network, provided by the embodiment of the invention, the average loss function is obtained through M times of k-fold cross training verification, so that the target arrhythmia diagnosis neural network model is obtained, the accuracy and efficiency of training the target arrhythmia diagnosis neural network model can be improved, the target arrhythmia diagnosis neural network model is verified once, information leakage into the model caused by a plurality of verification processes is prevented, and the reliability and generalization performance of the model are reduced.
In one implementation, the electrocardiographic rhythm types are classified into five types according to the AAMI EC57 standard.
The embodiment of the invention utilizes the MIT-BIH arrhythmia database to establish an electrocardiogram classification model. MIT-BIH arrhythmia database, each record collected about 48 (25 men: 22 women, age 23-89 years) complete 30-minute 2-lead electrocardiogram (electrodes placed on chest to obtain modified II and V1 leads), sampling frequency of 360Hz, another expert annotation file.
The arrhythmia database contains a plurality of heart beat types. According to the embodiment of the invention, the electrocardio heartbeats are divided into five groups according to the ANSI/AAMI EC57 standard and the labeling file. Referring to Table one, table one lists the definition and specification of five heart rhythms, and corresponding labels.
List one
In one implementation, N and M are preset values, which may be set by a technician according to actual situations, and are not limited herein. For example, N and M may be set to 100 and 10, respectively.
In one embodiment, the preset deep learning model includes 11 layers, respectively 3 separable channel convolutions, 1 normalization layer, 1 max pooling layer, 1 bi-directional LSTM layer, 1 flattening layer, 1 Dropout layer, 1 dense connection layer, 1 batch normalization layer, 1 Softmax layer.
In one implementation, the pre-set deep learning model includes 3 deep separable Convolutional Neural Network (CNN) models, and also includes a two-way long short-term memory (LSTM) model, effectively combining the velocity of CNN and the sequential sensitivity of Recurrent Neural Network (RNN).
In one embodiment, a 3-layer separable channel convolution of the preset deep learning model uses a ReLu activation function, each layer having 32, 64, and 128 cores of size 3 in sequence; the parameters of the maximum pooling layer are 2*1 and the step length is 2; the bidirectional LSTM layer is 128 units; the flattening layer, the dropout layer, the dense connecting layer, the batch normalization layer and the Softmax layer form a prediction model; the dense connectivity layer has 512 neurons, using the ReLu activation function; the 50% characteristic of the dropout layer is set to zero; the preset deep learning model is compiled using an Adam optimizer and a class cross entropy loss function.
In one implementation, the 50% characteristic of the dropout layer is set to zero, which can prevent overfitting.
In one embodiment, step S101 may include the steps of:
step one, acquiring a historical heart rhythm data set with labels, and resampling the historical heart rhythm data set to obtain a first data set, wherein the sampling frequency is 125Hz.
And secondly, normalizing the amplitude of the first electrocardiosignal, and dividing heart rate data in the first data set into data fragments with preset length by using a window function with preset length to obtain a second data set.
And thirdly, calculating the set of all local extremum values in the second data set according to the first derivative, and taking 0.9 of the normalized extremum value as an R candidate peak.
And step four, taking the median of all the R-R time intervals as the nominal heartbeat period T of the time window, determining the final period of each R-R peak value as 1.2T, and filling the rest zero points to reach the same length to obtain a third data set.
Fifthly, randomly dividing heart rate data in the third data set into a training data set and a verification data set according to a preset proportion.
In one implementation, the historical heart rate data set is typically 48 continuous heart rate signals with a duration of 30min, and may be divided into segments with a preset length by a window function, and then normalized, where the preset length may be set by a technician according to experience, and is not limited herein, and the preset length may be 10s, for example. A plurality of segments of electrocardiographic data of a predetermined length can be obtained as the second data set. The electrocardiosignal of the second data set has obvious QRS signal, the peak R is the highest, and the electrocardiosignal data sheet of the second data set needs to be continuously segmented according to the 1.2 times of the median of the R-R time interval to obtain a third data set.
In one embodiment, step S102 may include the steps of:
and step one, training N epochs on a preset deep learning model by using the training data set.
And secondly, dividing the training data set into M target training subsets aiming at each time of epoch training, and carrying out M times of k-fold cross training verification on a preset deep learning model by using the M target training subsets to obtain an average loss function of the M times of k-fold cross training verification of the epoch.
Thirdly, aiming at the ith training verification, taking the ith target training subset of M target training subsets as a verification subset, taking the rest M-1 target training subsets as training subsets, calculating the training weight of each training subset according to the number of each electrocardiographic rhythm type of each training subset, and obtaining the loss function of the ith training verification according to the training weights of the M-1 training subsets; the ith training verification is any one of M times of k-fold cross training verification;
and step four, determining the epoch with the smallest average loss function as the optimal epoch.
In one implementation, training weights are embedded in the loss function of the training data set to eliminate classification bias caused by the imbalance of the electrocardiographic heart beat type.
In one embodiment, the training weights for the training subset are calculated as:
where Weight represents the training Weight, n_samples represents the number of training samples in the training subset, n_classes represents the number of types of cardiac rhythms, np.bincount (y) represents the number of types of the y-th cardiac rhythm in the training subset.
In one implementation, the electrocardiographic rhythm types are classified into five types according to AAMI EC57 standard, i.e., n_class=5.
In one embodiment, training the preset deep learning model with the training data set for N epochs specifically includes:
and training N epochs on a preset deep learning model by using a training data set, calculating the self-adaptive learning rate of each epoch, and if the loss function of 5 continuous epochs is not reduced, reducing the self-adaptive learning rate of the next epochs to half of the current self-adaptive learning rate.
In one implementation, to evaluate the performance of the model, the following parameters are constructed:
wherein TP is true positive number, TN is true negative number, and FP and FN are false positive number and false negative number respectively. Acc represents accuracy, sen represents sensitivity, PPV represents positive rate, and F1 index is a performance index that combines both Sen and PPV.
Watch II
Training and testing the training method of the arrhythmia diagnosis model based on the artificial neural network provided by the embodiment of the invention on an MIT-BIH arrhythmia database. Referring to a second table, the second table is a performance evaluation table of a target cardiac arrhythmia diagnosis neural network model obtained by a training method of an arrhythmia diagnosis model based on an artificial neural network. The average accuracy of the model obtained by the method is 99.33%, the sensitivity is 93.67%, the positive predictive value is 89.85%, and the F1 score is 91.65%.
The foregoing describes one embodiment of the present invention in detail, but the description is only a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications within the scope of the present invention are intended to be covered by the present invention.
Claims (5)
1. A training method of an arrhythmia diagnosis model based on an artificial neural network, the method comprising:
acquiring a historical heart rhythm data set with labels, and dividing the historical heart rhythm data set into a training data set and a verification data set; the electrocardio rhythm type marked as heart rate data in the historical heart rate data set;
training N epochs on a preset deep learning model by using the training data set, and determining the epochs with the minimum loss function as the optimal epochs; the loss function of each epoch is an average loss function obtained by carrying out k-fold cross training verification for M times on the training data set; wherein N and M are preset values;
initializing the preset deep learning model by using the optimal epoch, and training the initialized preset deep learning model by using the training data set to obtain a target heart rhythm diagnosis neural network model;
performing primary verification on the target heart rhythm diagnosis neural network model by using the verification data set, and inputting the verification data set into the target heart rhythm diagnosis neural network model to obtain a performance evaluation result of the target heart rhythm diagnosis neural network model;
acquiring a marked historical heart rhythm data set, dividing the historical heart rhythm data set into a training data set and a verification data set, and comprising:
acquiring a historical heart rhythm data set with labels, and resampling the historical heart rhythm data set to obtain a first data set; the sampling frequency is 125Hz;
normalizing the amplitude of the first electrocardiosignal, and dividing heart rate data in the first data set into data fragments with preset length by using a window function with preset length to obtain a second data set;
according to the first derivative, a set of all local extrema in the second data set is obtained, and 0.9 of a normalized extremum is taken as an R candidate peak;
taking the median of all R-R time intervals as a nominal heartbeat period T of a time window, determining the final period of each R-R peak value as 1.2T, and filling the rest with zero points to reach the same length to obtain a third data set;
randomly dividing heart rate data in the third data set into a training data set and a verification data set according to a preset proportion;
training the N epochs on a preset deep learning model by using the training data set, and determining the epochs with the minimum loss function as the optimal epochs, wherein the method comprises the following steps:
training N epochs on a preset deep learning model by using the training data set;
dividing the training data set into M target training subsets aiming at each epoch training, and performing M times of k-fold cross training verification on a preset deep learning model by using the M target training subsets to obtain an average loss function of the M times of k-fold cross training verification of the epoch;
aiming at ith training verification, taking an ith target training subset of M target training subsets as a verification subset, taking the rest M-1 target training subsets as training subsets, calculating the training weight of each training subset according to the number of each electrocardiographic rhythm type of each training subset, and obtaining a loss function of the ith training verification according to the training weights of the M-1 training subsets; the ith training verification is any one of M times of k-fold cross training verification;
and determining the epoch with the smallest average loss function as the optimal epoch.
2. The training method of an artificial neural network-based arrhythmia diagnosis model according to claim 1, wherein the preset deep learning model comprises 11 layers, namely 3 layers of separable channel convolution, 1 normalization layer, 1 max pooling layer, 1 bidirectional LSTM layer, 1 flattening layer, 1 Dropout layer, 1 dense connection layer, 1 batch normalization layer and 1 Softmax layer.
3. The training method of an artificial neural network-based arrhythmia diagnosis model according to claim 2, wherein the 3-layer separable channel convolution of the preset deep learning model uses a ReLu activation function, each layer having 32, 64 and 128 cores of size 3 in sequence; the parameters of the maximum pooling layer are 2*1 and the step length is 2; the bidirectional LSTM layer is 128 units; the flattening layer, the Dropout layer, the dense connecting layer, the batch normalization layer and the Softmax layer form a prediction model; the dense connectivity layer has 512 neurons, using the ReLu activation function; the 50% characteristic of the Dropout layer is set to zero; the preset deep learning model is compiled using an Adam optimizer and a classification cross entropy loss function.
4. The training method of an arrhythmia diagnosis model based on an artificial neural network according to claim 1, wherein the training weights of the training subset are calculated by the following formula:
where Weight represents the training Weight, n_samples represents the number of training samples in the training subset, n_classes represents the number of types of cardiac rhythms, np.bincount (y) represents the number of types of the y-th cardiac rhythm in the training subset.
5. The training method of an arrhythmia diagnosis model based on an artificial neural network according to claim 1, wherein training of N epochs on a preset deep learning model by using the training data set specifically comprises:
and training N epochs on a preset deep learning model by using the training data set, calculating the self-adaptive learning rate of each epochs, and if the loss function of 5 continuous epochs is not reduced, reducing the self-adaptive learning rate of the next epochs to half of the current self-adaptive learning rate.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210758714.3A CN115099275B (en) | 2022-06-29 | 2022-06-29 | Training method of arrhythmia diagnosis model based on artificial neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210758714.3A CN115099275B (en) | 2022-06-29 | 2022-06-29 | Training method of arrhythmia diagnosis model based on artificial neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115099275A CN115099275A (en) | 2022-09-23 |
CN115099275B true CN115099275B (en) | 2023-07-07 |
Family
ID=83293887
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210758714.3A Active CN115099275B (en) | 2022-06-29 | 2022-06-29 | Training method of arrhythmia diagnosis model based on artificial neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115099275B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117281529B (en) * | 2023-11-23 | 2024-01-30 | 西南医科大学附属医院 | Electrocardiogram waveform data intelligent processing method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109589092A (en) * | 2018-10-08 | 2019-04-09 | 广州市本真网络科技有限公司 | Method and system are determined based on the Alzheimer's disease of integrated study |
JP2019063377A (en) * | 2017-10-03 | 2019-04-25 | 株式会社国際電気通信基礎技術研究所 | Biomarker for detecting disease state profile of depression in mammal and utilization of the same |
CN113657020A (en) * | 2021-07-13 | 2021-11-16 | 电子科技大学 | Deep cluster neural network model construction method for pneumatic data processing |
WO2021248187A1 (en) * | 2020-06-09 | 2021-12-16 | Annalise-Ai Pty Ltd | Systems and methods for automated analysis of medical images |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP3981003A1 (en) * | 2019-06-07 | 2022-04-13 | Leica Microsystems CMS GmbH | A system and method for training machine-learning algorithms for processing biology-related data, a microscope and a trained machine learning algorithm |
CN110824364B (en) * | 2019-10-24 | 2021-11-19 | 重庆邮电大学 | Lithium battery SOH estimation and RUL prediction method based on AST-LSTM neural network |
CN111370080B (en) * | 2020-03-05 | 2023-12-08 | 中国工程物理研究院上海激光等离子体研究所 | Radiation temperature inversion method based on artificial neural network algorithm |
CN111444657B (en) * | 2020-03-10 | 2023-05-02 | 五邑大学 | Method and device for constructing fatigue driving prediction model and storage medium |
US11763093B2 (en) * | 2020-04-30 | 2023-09-19 | Arizona Board Of Regents On Behalf Of Arizona State University | Systems and methods for a privacy preserving text representation learning framework |
WO2021259482A1 (en) * | 2020-06-25 | 2021-12-30 | PolyN Technology Limited | Analog hardware realization of neural networks |
CN113095302B (en) * | 2021-05-21 | 2023-06-23 | 中国人民解放军总医院 | Depth model for arrhythmia classification, method and device using same |
CN113257359A (en) * | 2021-06-08 | 2021-08-13 | 汕头大学 | CRISPR/Cas9 guide RNA editing efficiency prediction method based on CNN-SVR |
CN113344293B (en) * | 2021-06-29 | 2024-04-05 | 东南大学 | Photovoltaic power prediction method based on NCA-fusion regression tree model |
CN114093414A (en) * | 2021-11-22 | 2022-02-25 | 中国科学院合肥物质科学研究院 | ADMET property prediction method of ER alpha antagonist based on MMS _ ResNet _1d model |
CN114557705A (en) * | 2022-03-04 | 2022-05-31 | 莒县人民医院 | Resting electrocardio analysis method based on machine learning |
-
2022
- 2022-06-29 CN CN202210758714.3A patent/CN115099275B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019063377A (en) * | 2017-10-03 | 2019-04-25 | 株式会社国際電気通信基礎技術研究所 | Biomarker for detecting disease state profile of depression in mammal and utilization of the same |
CN109589092A (en) * | 2018-10-08 | 2019-04-09 | 广州市本真网络科技有限公司 | Method and system are determined based on the Alzheimer's disease of integrated study |
WO2021248187A1 (en) * | 2020-06-09 | 2021-12-16 | Annalise-Ai Pty Ltd | Systems and methods for automated analysis of medical images |
CN113657020A (en) * | 2021-07-13 | 2021-11-16 | 电子科技大学 | Deep cluster neural network model construction method for pneumatic data processing |
Non-Patent Citations (2)
Title |
---|
不同机器学习方法在抑郁症诊断和预后预测中的应用比较;花田甜;《中国优秀硕士学位论文电子期刊》;57-75页 * |
面向短QT综合征的药物作用建模与仿真研究;张恒贵等;《生物化学与生物物理进展》;523-535页 * |
Also Published As
Publication number | Publication date |
---|---|
CN115099275A (en) | 2022-09-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Alalawi et al. | Detection of cardiovascular disease using machine learning classification models | |
Tsipouras et al. | A framework for fuzzy expert system creation—application to cardiovascular diseases | |
CN106805965A (en) | A kind of electrocardiosignal sorting technique and device | |
Dabowsa et al. | A hybrid intelligent system for skin disease diagnosis | |
CN110111885B (en) | Attribute prediction method, attribute prediction device, computer equipment and computer readable storage medium | |
Xu et al. | I-vector-based patient adaptation of deep neural networks for automatic heartbeat classification | |
CN115099275B (en) | Training method of arrhythmia diagnosis model based on artificial neural network | |
Patidar et al. | Comparative analysis of machine learning algorithms for heart disease predictions | |
Arunachalam | Cardiovascular disease prediction model using machine learning algorithms | |
Chen et al. | Edge2Analysis: a novel AIoT platform for atrial fibrillation recognition and detection | |
Kumar et al. | Stroke disease prediction based on ECG signals using deep learning techniques | |
Pillai | Cardiac disease prediction with tabular neural network | |
Prusty et al. | Comparative analysis and prediction of coronary heart disease | |
Tiwaskar et al. | Comparison of prediction models for heart failure risk: a clinical perspective | |
CN113257410A (en) | Interrogation method based on traditional Chinese medicine clinical medical knowledge base and deep learning model | |
Alharbi et al. | Machine learning approach to predict cognitive performance using HRV | |
CN116649899A (en) | Electrocardiogram signal classification method based on attention mechanism feature fusion | |
Bashir et al. | Trigger learning and ECG parameter customization for remote cardiac clinical care information system | |
Yin et al. | An algorithm for locating pvc and spb in wearable ecgs | |
CN112022149A (en) | Atrial fibrillation detection method based on electrocardiosignals | |
Sharan | Detecting cardiac abnormalities using 12-lead ecg and deep learning | |
He et al. | Automatic classification of arrhythmias by residual network and BiGRU with attention mechanism | |
Wen et al. | ECG beat classification using GreyART network | |
Thilagamani | A survey on efficient heart disease prediction technique | |
Zhang | Machine Learning based Heart Disease Prediction Task |
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 |