CN111710386A - Quality control system for electrocardiogram diagnosis report - Google Patents
Quality control system for electrocardiogram diagnosis report Download PDFInfo
- Publication number
- CN111710386A CN111710386A CN202010367639.9A CN202010367639A CN111710386A CN 111710386 A CN111710386 A CN 111710386A CN 202010367639 A CN202010367639 A CN 202010367639A CN 111710386 A CN111710386 A CN 111710386A
- Authority
- CN
- China
- Prior art keywords
- electrocardiogram
- diagnosis
- conclusion
- neural network
- term
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 114
- 238000003908 quality control method Methods 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 55
- 230000000306 recurrent effect Effects 0.000 claims abstract description 44
- 238000003062 neural network model Methods 0.000 claims abstract description 28
- 238000012360 testing method Methods 0.000 claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims description 22
- 239000013598 vector Substances 0.000 claims description 20
- 238000000034 method Methods 0.000 claims description 11
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 125000004122 cyclic group Chemical group 0.000 claims description 4
- 230000001419 dependent effect Effects 0.000 claims 1
- 238000003759 clinical diagnosis Methods 0.000 abstract description 3
- 238000002565 electrocardiography Methods 0.000 description 16
- 230000006870 function Effects 0.000 description 9
- 230000033764 rhythmic process Effects 0.000 description 9
- 208000000418 Premature Cardiac Complexes Diseases 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000002861 ventricular Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 230000002349 favourable effect Effects 0.000 description 4
- 208000019622 heart disease Diseases 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 208000024172 Cardiovascular disease Diseases 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 208000010125 myocardial infarction Diseases 0.000 description 2
- 206010008190 Cerebrovascular accident Diseases 0.000 description 1
- 208000007177 Left Ventricular Hypertrophy Diseases 0.000 description 1
- 208000007888 Sinus Tachycardia Diseases 0.000 description 1
- 208000006011 Stroke Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000001746 atrial effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
-
- 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
- G06N3/045—Combinations of networks
-
- 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
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Epidemiology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The application relates to a quality control system for electrocardiogram diagnosis reports, which comprises: acquiring electrocardio diagnosis terms contained in a plurality of clinical electrocardiogram conclusions, and constructing a training set and a test set of a recurrent neural network model through preprocessing; taking the kth electrocardiogram diagnosis term contained in each electrocardiogram conclusion in the training set and the test set as input, taking the (k + 1) th electrocardiogram diagnosis term as output, and training a recurrent neural network; and predicting the (k + 1) th electrocardiogram diagnosis term of the central electrocardiogram conclusion of the electrocardiogram diagnosis report to be quality-controlled by using the trained recurrent neural network model, and judging whether the corresponding electrocardiogram diagnosis report has low-level errors. The present invention will help correct some low-level errors when the electrocardiograph gives a conclusion and help provide intelligent prompts for clinical diagnosis.
Description
Technical Field
The application belongs to the technical field of electrocardiogram diagnosis, and particularly relates to a quality control system for electrocardiogram diagnosis reports.
Background
With the increasing pace and pressure of life, heart diseases have become one of the major killers threatening the life and health of people. Cardiovascular diseases become a frequently-occurring disease and a common disease in China, and according to survey data published by Ministry of health, the heart disease prevalence rate in China is high and gradually increases year by year, the onset age of coronary heart disease and myocardial infarction tends to be younger, and the occurrence of myocardial infarction and cerebral apoplexy is rare in about thirty years old.
Accurate analysis and diagnosis of electrocardiograms plays a key role in cardiovascular disease in order to discover and treat heart disease early. The electrocardiogram reflects the health condition of the heart of a human body, has the advantages of non-invasiveness and low cost, can accurately diagnose the heart and other diseases, provides a reliable basis for rescue treatment, and is widely used for heart disease examination in clinic.
A great deal of information on the electrical activity of the heart is available from the electrocardiogram, so clinically, there are many descriptive terms for abnormalities in the electrocardiogram, such as ventricular premature beat, atrial premature beat, first-order block, ST-T change, left ventricular hypertrophy, etc., and the electrocardiogram conclusion is often a combination of these diagnostic terms.
These diagnostic terms need to be written in a prescribed order in the electrocardiographic diagnostic conclusion, some may appear in the same electrocardiographic diagnostic conclusion at the same time, while some electrocardiographic diagnostic terms are mutually exclusive and cannot be used to describe a diagnostic conclusion of an electrocardiogram at the same time, for example: 1. normal electrocardiogram; ST segment is raised.
In the prior art, when the quality of the electrocardiogram diagnosis report is controlled, the description of the electrocardiogram report and the diagnosis result are evaluated according to the diagnosis standard of the electrocardiogram, and the low-level errors that mutually exclusive diagnosis terms or writing sequence in a diagnosis conclusion do not meet the standard can not be detected.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the quality control system of the electrocardiogram diagnosis report is provided for solving the problem that low-level errors in the electrocardiogram diagnosis report cannot be obtained in the prior art.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a quality control system for electrocardiographic diagnostic reporting, comprising:
the data acquisition module is used for acquiring electrocardio diagnosis terms contained in a plurality of clinical electrocardiogram conclusions;
the preprocessing module is used for preprocessing the electrocardio diagnosis terms and constructing a training set and a testing set of a recurrent neural network model according to the preprocessed electrocardiogram conclusion data meeting the writing standard requirements;
the model training module is used for taking the current electrocardiogram diagnosis term of each electrocardiogram conclusion in the training set and the testing set as input, taking the next adjacent electrocardiogram diagnosis term as output, training a recurrent neural network and storing a trained recurrent neural network model;
the prediction module is used for predicting the probability that all the next possible electrocardio diagnosis terms meet the writing standard by taking the current electrocardio diagnosis term of the electrocardiogram conclusion in the electrocardio diagnosis report to be quality controlled as input by utilizing the trained recurrent neural network model;
and the judging module is used for comparing the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled with all the next possible electrocardiogram diagnosis terms to be predicted, and if the probability that the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled meets the writing specification is smaller than the preset value, judging that the corresponding electrocardiogram diagnosis report has low-level errors.
The invention has the beneficial effects that: the invention predicts the writing sequence of the diagnosis terms and the rejection conditions among the diagnosis terms in the electrocardiogram conclusion by training the neural network model, is favorable for correcting some low-level errors when the electrocardiogram doctor gives the conclusion, and is favorable for providing intelligent prompt for clinical diagnosis.
Drawings
The technical solution of the present application is further explained below with reference to the drawings and the embodiments.
Fig. 1 is a flowchart of a working process of a quality control system according to an embodiment of the present application;
fig. 2 is a low-level error determination flowchart of an embodiment of the present application.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The technical solutions of the present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Examples
The embodiment provides a quality control system for electrocardiogram diagnosis report, comprising:
the data acquisition module is used for acquiring electrocardio diagnosis terms contained in a plurality of clinical electrocardiogram conclusions;
the preprocessing module is used for preprocessing the electrocardio diagnosis terms and constructing a training set and a testing set of a recurrent neural network model according to the preprocessed electrocardiogram conclusion data meeting the writing standard requirements;
the model training module is used for taking the current electrocardiogram diagnosis term of each electrocardiogram conclusion in the training set and the testing set as input, taking the next adjacent electrocardiogram diagnosis term as output, training a recurrent neural network and storing a trained recurrent neural network model;
the prediction module is used for predicting the probability that all the next possible electrocardio diagnosis terms meet the writing standard by taking the current electrocardio diagnosis term of the electrocardiogram conclusion in the electrocardio diagnosis report to be quality controlled as input by utilizing the trained recurrent neural network model;
and the judging module is used for comparing the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled with all the next possible electrocardiogram diagnosis terms to be predicted, and if the probability that the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled meets the writing specification is smaller than the preset value, judging that the corresponding electrocardiogram diagnosis report has low-level errors.
Generally, an electrocardiograph diagnosis report provided by an electrocardiograph needs to be subjected to quality control, and the quality control system of the embodiment is mainly used for detecting low-level errors in an electrocardiograph conclusion, wherein the low-level errors mainly refer to whether the writing sequence of the electrocardiograph diagnosis conclusion does not meet the specification or not and whether mutually exclusive electrocardiograph diagnosis terms appear in the same electrocardiograph conclusion or not.
Fig. 1 is a flowchart illustrating a working process of a quality control system according to an embodiment of the present invention. According to the embodiment of the invention, through training the recurrent neural network, the kth electrocardiogram diagnosis term written in sequence in the central electrocardiogram conclusion of the electrocardio diagnosis report to be controlled is input into the trained recurrent neural network model, so that the kth +1 th all possible electrocardio diagnosis terms are predicted, and through comparison, if the probability that the kth +1 st electrocardio diagnosis term of the central electrocardiogram conclusion of the electrocardio diagnosis report to be controlled meets the writing specification is smaller than the preset value, the corresponding electrocardio diagnosis report is judged to have low-level errors.
The embodiment of the invention predicts whether mutually exclusive electrocardio diagnosis terms exist in the electrocardiogram conclusion or not by training the Recurrent Neural Network (RNN), is favorable for correcting some low-level errors when the electrocardiogram doctor gives the conclusion, and is favorable for providing intelligent prompt for clinical diagnosis.
Optionally, the preprocessing module in this embodiment further includes:
the coding unit is used for adopting one-hot coding for each electrocardiogram diagnosis term contained in the electrocardiogram conclusion to obtain electrocardiogram conclusion data which are subjected to the one-hot coding and meet the writing specification requirement;
and the data dividing unit is used for dividing the encoded electrocardiogram conclusion data into a training set and a test set, taking the one-hot code corresponding to the kth diagnostic term in the electrocardiogram conversion conclusion as input data of the training set and the test set, and taking the one-hot code corresponding to the (k + 1) th diagnostic term in the electrocardiogram conversion conclusion as output data of the training set and the test set.
In the embodiment, 30 ten thousand pieces of clinical electrocardiogram conclusion data given according to the Shanghai electrocardiogram quality control manual are collected to obtain N common electrocardiogram diagnosis terms, the common electrocardiogram diagnosis terms are coded by adopting a one-hot coding method, and the electrocardiogram conclusion data meeting the writing standard requirements are formed.
The specific encoding method of the encoding unit of the embodiment is as follows:
s11, numbering each electrocardio diagnosis term in sequence by using Arabic numerals;
and S12, encoding the electrocardio diagnostic term into a 1xN vector, wherein the value of the element position corresponding to the corresponding electrocardio diagnostic term number in the 1xN vector is 1, and the values of the other element positions are 0.
In this embodiment, N electrocardiographic diagnostic terms are numbered in an arabic number order, as shown in table 1, this embodiment lists the numbers of common electrocardiographic diagnostic terms, and adds the number of a conclusion ending symbol.
After one-hot encoding is adopted, each electrocardiographic diagnostic term can be represented by a vector with the size of 1 × N, the element value of the position corresponding to the position j except the electrocardiographic diagnostic term in the 1 × N vector is 1, and the element values of the rest positions are 0, that is:
i denotes the element position sequence number in the 1 × N vector, j denotes the number of the corresponding electrocardiographic diagnostic term, i 1,2, 3.
Taking the common ECG diagnostic terminology listed in Table 1 as an example, when the "sinus rhythm" is subjected to one-hot encoding, the "sinus rhythm" is numbered 2, and thus, the one-hot encoding of the sinus rhythm is obtained as [ 010 … 0], and if the "normal electrocardiogram" is subjected to one-hot encoding, the "normal electrocardiogram" is obtained as [ 100 … 0]
TABLE 1 common ECG diagnostic term numbering
Optionally, in this embodiment, each electrocardiographic diagnostic term in the electrocardiographic conclusion is encoded by one-hot, the one-hot of each electrocardiographic diagnostic term constitutes a 1 × N vector, the obtained encoded electrocardiographic conclusion data is an M × N matrix, and the one-hot of the last action end symbol in the matrix is encoded. M is the number of the ECG diagnosis terms in the electrocardiogram conclusion, and N is the number of the obtained ECG diagnosis terms.
And according to the electrocardiogram diagnosis terms contained in the electrocardiogram conclusion, converting the electrocardiogram conclusion into a matrix formed by one-hot codes corresponding to each electrocardiogram diagnosis term, and obtaining the encoded electrocardiogram conclusion data.
For example, the electrocardiogram concludes that: 1. sinus rhythm, 2. ventricular premature beat. The electrocardiogram conclusion includes two electrocardiogram diagnosis terms, and the coding mode of the present embodiment plus the coding of the conclusion ending character, then the coding of the electrocardiogram conclusion is a 3 × N matrix, that is:
the first row of the matrix represents the code for the first ECG diagnostic term "sinus rhythm" in the ECG conclusion, the second row represents the code for the second ECG diagnostic term "ventricular premature beat", and the third row represents the code for the "conclusion terminator".
The embodiment divides the obtained encoded electrocardiogram data into two groups, wherein one group of data is used as a training set, and the other group of data is used as a test set, for example, the data of 3/5 can be divided into the training set, and the data of 2/5 can be used as the test set.
If M electrocardio diagnosis terms are contained in the electrocardiogram conclusion, the kth electrocardio diagnosis term of the electrocardiogram conclusion is used as input data of a training set and a test set, the kth +1 electrocardio diagnosis term of the electrocardiogram conclusion is used as output data of the training set and the test set, and k is 1,2,3, … … and M-1. For example, if the electrocardiogram concludes that: 1. sinus rhythm, 2. ventricular premature beat, the input data is the one-hot code of sinus rhythm, and the output data is the one-hot code of ventricular premature beat.
Optionally, the specific implementation manner of the model training module of this embodiment is as follows:
s21, forming a 1XN vector of one-hot coding by the input data and the output data in the training set;
s22, inputting the input data in the training set to the input layer of the recurrent neural network model, and inputting the output data in the training set to the output layer of the recurrent neural network model;
s23, training the recurrent neural network model, stopping training when the loss function is smaller than the set threshold value, and storing the training model;
s24, testing the accuracy of the training model by adopting a test set, and adjusting the related network parameters of the training model;
and S25, repeating the steps S21-S24 until the accuracy of the recurrent neural network reaches a preset value.
Concluding with one of the electrocardiograms: 1. sinus rhythm, 2. ventricular premature beat as an example, which includes two diagnostic terms for electrocardiography, then, when training the recurrent network model, the one-hot code of "sinus rhythm" is input to the input layer of the recurrent neural network model, and the one-hot code of "ventricular premature beat" is input to the output layer of the recurrent neural network model.
After data is input, a cyclic neural network model is trained by adopting an existing training algorithm, for example, a gradient descent method, when the loss function of the neural network model is smaller than a set threshold value, the training is stopped, the trained cyclic network model is stored, then a test set is adopted to test the accuracy of the training model, relevant network parameters of the training model are adjusted, and the training and testing processes are repeated until the accuracy of the cyclic neural network reaches a preset value.
Optionally, the recurrent neural network model in this embodiment is:
y=Softmax(wya·a1+by)
a1=ReLU(waa·a0+wax·x+ba)
wherein x denotes the input of the recurrent neural network, y denotes the output of the recurrent neural network, a0And a1Representing hidden layer states of a recurrent neural network, waaIs the weight matrix with the last value of the hidden layer as the input of this time, WaxIs a weight matrix of the input layer to the hidden layer, baIs the data deviation of the hidden layer, WyaIs a weight matrix from the hidden layer to the output layer, byIs the data deviation of the output layer, w thereinaa,wax,wya,ba,byAre all the parameters of the recurrent neural network to be trained.
The loss function of the recurrent neural network of the embodiment adopts a cross entropy function.
As a further preferred embodiment, the present example waaA structured parametric model is used, namely:
wherein p is1~pn、q1~qmFor the weight matrix w to be trainedaaThe elements in the matrix are obtained by training a recurrent neural network.
In the embodiment, a structured parameter model is adopted, so that the requirement of the recurrent neural network on the number of training samples can be reduced. Compared with the prior artThe weight matrix is adopted, each element in the matrix needs to be trained, and for p1~pnConcerning the element values, only p in the first row needs to be trained in this embodiment1~pnThe n elements of (2) are automatically obtained with respect to the elements following the second line of parameters. And q is1~qmIs to implement the matrix expansion, if used, only q needs to be trained1~qmAnd the related weight matrix involved in the prior art, if an m × n matrix needs to be obtained, m × n times need to be trained.
Optionally, as shown in fig. 2, a specific implementation method for determining that a low-level error exists in the electrocardiographic diagnosis report to be quality controlled by the determination module of this embodiment is as follows:
s31, inputting one-hot codes corresponding to the kth electrocardiogram diagnosis term in the electrocardiogram conclusion of the quality control electrocardiogram diagnosis report into the trained recurrent neural network model, wherein the recurrent neural network model outputs a 1xN vector representing the kth +1 electrocardiogram diagnosis term, and k is less than or equal to M-1;
s32, determining element values at positions corresponding to the number of the (k + 1) th electrocardio diagnosis term in the electrocardiogram conclusion of the electrocardio diagnosis report to be quality controlled in the 1xN vector, wherein the element values in the 1xN vector represent the probability that the electrocardio diagnosis term with the corresponding number meets writing specifications;
s33, if the element value at the corresponding position is smaller than the preset value, the corresponding electrocardiogram diagnosis conclusion is judged to have mutually exclusive diagnosis terms or the writing sequence of the electrocardiogram diagnosis terms does not accord with the standard, and the corresponding electrocardiogram diagnosis report to be controlled has low-level errors;
and S34, if not, adding 1 to the value of k, and repeating the steps until k is M-1.
In this embodiment, the trained RNN model is used to predict the (k + 1) th electrocardiographic diagnostic term in the electrocardiographic conclusion, and determine whether there is a low-level error. Suppose that M ECG diagnostic terms are included in the ECG conclusion to be predicted, and the kth ECG conclusion (where k is the number of K) is included in the ECG conclusion<M-1) using one-hot code of electrocardio diagnosis term as input, utilizing RNN model to calculate output and obtaining a direction with length NQuantity I1×N。
Since the present embodiment uses the softmax excitation function, the obtained vector I1×NAll the element values in (1) are [0]And the sum of all element values is 1, the vector I1×NThe element values in (1) represent the probability of the electrocardiographic diagnostic term under the corresponding number.
According to the output vector I1×NDetermining the element value at the position corresponding to the number j of the (k + 1) th ECG diagnostic term in the electrocardiogram conclusion to be controlledj=1,2,3,…,N。
If it is notIf the current value is less than the preset value, the fact that low-level errors possibly exist in the electrocardio diagnosis report to be controlled is prompted, the value of the preset value can be 0.001-0.01, and the value of the preset value can be 0.01.
If it is notAnd if the k value is larger than the preset value, adding 1 to the value of k, (the initial value of k is 1), and repeating the calculation process until k is equal to M-1.
For example, if an ecg result of a quality control ecg diagnosis report includes 4 ecg diagnosis terms, where M is 4, the 1 st ecg diagnosis term is first input and then passed through the RNN model, and a 1 × N vector representing all possible 2 nd ecg diagnosis terms is output.
And searching the number of the 2 nd electrocardiogram diagnosis term in the electrocardiogram conclusion of the quality control electrocardiogram diagnosis report, and determining the element value at the position corresponding to the number of the 2 nd electrocardiogram diagnosis term in the 1xN vector.
If the 2 nd ECG diagnostic term is "sinus tachycardia," which corresponds to a number 3 according to Table 1, then the output I is looked up1×NThe corresponding element value at the 3 rd element position.
If I1×NIn the vector, the 3 rdA corresponding element value of 0 at the element position is considered to be a low level error. If the element value corresponding to the 3 rd element position is 0.5, the low-level error does not exist, and the code of the 2 nd ECG diagnostic term in the electrocardiogram conclusion to be controlled is continuously input until all the ECG diagnostic terms are detected.
Further optionally, the embodiment sends out an alarm prompt tone when the electrocardio diagnosis report to be controlled has low-level errors.
This has the advantage of facilitating the ability of the physician to timely detect low-level errors in the report and to timely address them.
In light of the foregoing description of the preferred embodiments according to the present application, it is to be understood that various changes and modifications may be made without departing from the spirit and scope of the invention. The technical scope of the present application is not limited to the contents of the specification, and must be determined according to the scope of the claims.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Claims (10)
1. A quality control system for electrocardiographic diagnostic reporting, comprising:
the data acquisition module is used for acquiring electrocardio diagnosis terms contained in a plurality of clinical electrocardiogram conclusions;
the preprocessing module is used for preprocessing the electrocardio diagnosis terms and constructing a training set and a testing set of a recurrent neural network model according to the preprocessed electrocardiogram conclusion data meeting the writing standard requirements;
the model training module is used for taking the current electrocardiogram diagnosis term of each electrocardiogram conclusion in the training set and the testing set as input, taking the next adjacent electrocardiogram diagnosis term as output, training a recurrent neural network and storing a trained recurrent neural network model;
the prediction module is used for predicting the probability that all the next possible electrocardio diagnosis terms meet the writing standard by taking the current electrocardio diagnosis term of the electrocardiogram conclusion in the electrocardio diagnosis report to be quality controlled as input by utilizing the trained recurrent neural network model;
and the judging module is used for comparing the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled with all the next possible electrocardiogram diagnosis terms to be predicted, and if the probability that the next electrocardiogram diagnosis term of the electrocardiogram conclusion of the electrocardiogram diagnosis report center to be controlled meets the writing specification is smaller than the preset value, judging that the corresponding electrocardiogram diagnosis report has low-level errors.
2. The system of claim 1, wherein the preprocessing module further comprises:
the coding unit is used for adopting one-hot coding for each electrocardiogram diagnosis term contained in the electrocardiogram conclusion to obtain electrocardiogram conclusion data which are subjected to the one-hot coding and meet the writing specification requirement;
and the data dividing unit is used for dividing the encoded electrocardiogram conclusion data into a training set and a test set, taking the one-hot code corresponding to the kth diagnostic term in the electrocardiogram conversion conclusion as input data of the training set and the test set, and taking the one-hot code corresponding to the (k + 1) th diagnostic term in the electrocardiogram conversion conclusion as output data of the training set and the test set.
3. The system of claim 2, wherein the encoding unit is further configured to:
and adopting one-hot codes for each electrocardiogram diagnosis term in the electrocardiogram conclusion, wherein the one-hot codes of each electrocardiogram diagnosis term form a 1XN vector, the obtained electrocardiogram conclusion data is an M XN matrix, the one-hot codes of the last behavior end symbol in the matrix are M, the M is the number of the electrocardiogram conclusion center electrocardiogram diagnosis terms, and the N is the number of the obtained electrocardiogram diagnosis terms.
4. The system of claim 3, wherein the step of encoding each ECG diagnostic term contained in the ECG conclusion by one-hot encoding comprises:
numbering each electrocardiogram diagnosis term by adopting an Arabic numeral sequence;
and encoding an electrocardiogram diagnostic term into a 1xN vector, wherein the value of the element position corresponding to the corresponding electrocardiogram diagnostic term number in the 1xN vector is 1, and the values of the other element positions are 0.
5. The system of claim 4, wherein the step of training the recurrent neural network model by the model training module comprises:
forming 1XN vectors of one-hot coding by input data and output data in a training set;
inputting the input data in the training set to an input layer of the recurrent neural network model, and inputting the output data in the training set to an output layer of the recurrent neural network model;
training the cyclic neural network model, stopping training when the loss function is smaller than a set threshold value, and storing the training model;
testing the accuracy of the training model by adopting a test set, and adjusting related network parameters of the training model;
and repeating the steps until the accuracy of the recurrent neural network reaches a preset value.
6. The system of claim 5, wherein the step of determining that the corresponding ECG diagnostic report has a low-level error comprises:
if the element values at the corresponding positions are smaller than the preset values, judging that mutually exclusive diagnosis terms exist in the corresponding electrocardiogram diagnosis conclusion or the writing sequence of the electrocardiogram diagnosis terms does not meet the specification, and judging that a corresponding electrocardiogram diagnosis report to be controlled has low-level errors;
otherwise, add 1 to the value of k and repeat the above steps until k equals M-1.
7. The system of claim 1, wherein the recurrent neural network model is:
y=Softmax(wya·a1+by)
a1=ReLU(waa·a0+wax·x+ba)
wherein x denotes the input of the recurrent neural network, y denotes the output of the recurrent neural network, a0And a1Representing hidden layer states of a recurrent neural network, waaIs the weight matrix with the last value of the hidden layer as the input of this time, WaxIs a weight matrix of the input layer to the hidden layer, baIs the data deviation of the hidden layer, WyaIs a weight matrix from the hidden layer to the output layer, byIs the data deviation of the output layer, w thereinaa,wax,wya,ba,byAre all the parameters of the recurrent neural network to be trained.
9. The system of any one of claims 1-8, further comprising an alarm module for sounding an alarm when the ECG diagnostic report to be controlled has a low-level error.
10. The method of claim 6, wherein the predetermined value is 0.001-0.01.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010367639.9A CN111710386A (en) | 2020-04-30 | 2020-04-30 | Quality control system for electrocardiogram diagnosis report |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010367639.9A CN111710386A (en) | 2020-04-30 | 2020-04-30 | Quality control system for electrocardiogram diagnosis report |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111710386A true CN111710386A (en) | 2020-09-25 |
Family
ID=72536845
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010367639.9A Pending CN111710386A (en) | 2020-04-30 | 2020-04-30 | Quality control system for electrocardiogram diagnosis report |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111710386A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117133449A (en) * | 2023-10-26 | 2023-11-28 | 纳龙健康科技股份有限公司 | Electrocardiogram analysis system, electrocardiogram analysis model construction, training method, and medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107451417A (en) * | 2017-09-08 | 2017-12-08 | 北京蓬阳丰业医疗设备有限公司 | Dynamic ECG analysis intelligent diagnosis system and method |
CN108095716A (en) * | 2017-11-21 | 2018-06-01 | 郑州鼎创智能科技有限公司 | A kind of electrocardiograph signal detection method based on confidence rule base and deep neural network |
CN108257650A (en) * | 2017-12-26 | 2018-07-06 | 宁波市科技园区明天医网科技有限公司 | A kind of intelligent correction method applied to medical technologies audit report |
WO2018188981A1 (en) * | 2017-04-12 | 2018-10-18 | Koninklijke Philips N.V. | Drawing conclusions from free form texts with deep reinforcement learning |
CN109480827A (en) * | 2018-12-18 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | Vectorcardigram classification method and device |
CN109871851A (en) * | 2019-03-06 | 2019-06-11 | 长春理工大学 | A kind of Chinese-character writing normalization determination method based on convolutional neural networks algorithm |
WO2019205319A1 (en) * | 2018-04-25 | 2019-10-31 | 平安科技(深圳)有限公司 | Commodity information format processing method and apparatus, and computer device and storage medium |
CN110432892A (en) * | 2019-08-05 | 2019-11-12 | 苏州米特希赛尔人工智能有限公司 | Machine learning ECG Automatic Diagnosis System |
CN110705262A (en) * | 2019-09-06 | 2020-01-17 | 宁波市科技园区明天医网科技有限公司 | Improved intelligent error correction method applied to medical skill examination report |
CN110797101A (en) * | 2019-10-28 | 2020-02-14 | 腾讯医疗健康(深圳)有限公司 | Medical data processing method, device, readable storage medium and computer equipment |
CN110942825A (en) * | 2019-11-26 | 2020-03-31 | 北京华医共享医疗科技有限公司 | Electrocardiogram diagnosis method based on combination of convolutional neural network and cyclic neural network |
-
2020
- 2020-04-30 CN CN202010367639.9A patent/CN111710386A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018188981A1 (en) * | 2017-04-12 | 2018-10-18 | Koninklijke Philips N.V. | Drawing conclusions from free form texts with deep reinforcement learning |
CN107451417A (en) * | 2017-09-08 | 2017-12-08 | 北京蓬阳丰业医疗设备有限公司 | Dynamic ECG analysis intelligent diagnosis system and method |
CN108095716A (en) * | 2017-11-21 | 2018-06-01 | 郑州鼎创智能科技有限公司 | A kind of electrocardiograph signal detection method based on confidence rule base and deep neural network |
CN108257650A (en) * | 2017-12-26 | 2018-07-06 | 宁波市科技园区明天医网科技有限公司 | A kind of intelligent correction method applied to medical technologies audit report |
WO2019205319A1 (en) * | 2018-04-25 | 2019-10-31 | 平安科技(深圳)有限公司 | Commodity information format processing method and apparatus, and computer device and storage medium |
CN109480827A (en) * | 2018-12-18 | 2019-03-19 | 武汉中旗生物医疗电子有限公司 | Vectorcardigram classification method and device |
CN109871851A (en) * | 2019-03-06 | 2019-06-11 | 长春理工大学 | A kind of Chinese-character writing normalization determination method based on convolutional neural networks algorithm |
CN110432892A (en) * | 2019-08-05 | 2019-11-12 | 苏州米特希赛尔人工智能有限公司 | Machine learning ECG Automatic Diagnosis System |
CN110705262A (en) * | 2019-09-06 | 2020-01-17 | 宁波市科技园区明天医网科技有限公司 | Improved intelligent error correction method applied to medical skill examination report |
CN110797101A (en) * | 2019-10-28 | 2020-02-14 | 腾讯医疗健康(深圳)有限公司 | Medical data processing method, device, readable storage medium and computer equipment |
CN110942825A (en) * | 2019-11-26 | 2020-03-31 | 北京华医共享医疗科技有限公司 | Electrocardiogram diagnosis method based on combination of convolutional neural network and cyclic neural network |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117133449A (en) * | 2023-10-26 | 2023-11-28 | 纳龙健康科技股份有限公司 | Electrocardiogram analysis system, electrocardiogram analysis model construction, training method, and medium |
CN117133449B (en) * | 2023-10-26 | 2024-01-12 | 纳龙健康科技股份有限公司 | Electrocardiogram analysis system, electrocardiogram analysis model construction, training method, and medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5501229A (en) | Continuous monitoring using a predictive instrument | |
US20230131876A1 (en) | Systems and methods of identity analysis of electrocardiograms | |
CN111772619B (en) | Heart beat identification method based on deep learning, terminal equipment and storage medium | |
Lee et al. | Mining biosignal data: coronary artery disease diagnosis using linear and nonlinear features of HRV | |
CN107837082A (en) | Electrocardiogram automatic analysis method and device based on artificial intelligence self study | |
KR102208759B1 (en) | Method for generating deep-learning model for diagnosing health status and pathology symptom based on biosignal | |
CN111329469A (en) | Arrhythmia prediction method | |
KR102483693B1 (en) | Method and apparatus of explainable multi electrocardiogram arrhythmia diagnosis | |
CN116782829A (en) | Method and apparatus for converting bioelectric signal data into numerical vector and method and apparatus for analyzing disease using the same | |
CN111528832A (en) | Arrhythmia classification method and validity verification method thereof | |
CN115153581A (en) | Electrocardiosignal characteristic waveform detection method and device and storage medium | |
CN111710386A (en) | Quality control system for electrocardiogram diagnosis report | |
CN112545525B (en) | Electrocardiogram data classification method, device and system | |
Deng et al. | ST-Net: Synthetic ECG tracings for diagnosing various cardiovascular diseases | |
EP4216232A1 (en) | Methods and system for cardiac arrhythmia prediction using transformer-based neural networks | |
KR102560516B1 (en) | An apparatus and method for predicting heart disease based on deep learning models using ECG data and body information | |
CN111710387A (en) | Quality control method for electrocardiogram diagnosis report | |
CN112022140B (en) | Automatic diagnosis method and system for diagnosis conclusion of electrocardiogram | |
CN116662742A (en) | Brain electrolysis code method based on hidden Markov model and mask empirical mode decomposition | |
Magrupov et al. | A Technique for Classifying the ECG Signal into Various Possible States of the Cardiovascular System | |
CN112842355B (en) | Electrocardiosignal heart beat detection and identification method based on deep learning target detection | |
Bashir et al. | Highlighting the current issues with pride suggestions for improving the performance of real time cardiac health monitoring | |
Dolganov | Application of the evolutionary programming for indirect evaluation of blood pressure based on heart rate variability parameters | |
CN113012807B (en) | ECG signal marking method, system, device and computer readable storage medium | |
CN111084621A (en) | QRS wave group form identification method and device based on depth self-encoder |
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 |