CN111798975A - Disease diagnosis system, apparatus and medium based on cyclic time convolution network - Google Patents
Disease diagnosis system, apparatus and medium based on cyclic time convolution network Download PDFInfo
- Publication number
- CN111798975A CN111798975A CN202010483848.XA CN202010483848A CN111798975A CN 111798975 A CN111798975 A CN 111798975A CN 202010483848 A CN202010483848 A CN 202010483848A CN 111798975 A CN111798975 A CN 111798975A
- Authority
- CN
- China
- Prior art keywords
- diagnosis
- disease
- vector
- convolution network
- predicted
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 131
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 131
- 238000003745 diagnosis Methods 0.000 title claims abstract description 99
- 125000004122 cyclic group Chemical group 0.000 title claims description 25
- 239000013598 vector Substances 0.000 claims abstract description 70
- 238000012549 training Methods 0.000 claims abstract description 36
- 238000010276 construction Methods 0.000 claims abstract description 9
- 238000000034 method Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 18
- 230000015654 memory Effects 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 230000007774 longterm Effects 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 abstract description 5
- 238000012360 testing method Methods 0.000 description 7
- 238000003759 clinical diagnosis Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 206010020772 Hypertension Diseases 0.000 description 2
- 206010058467 Lung neoplasm malignant Diseases 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 201000005202 lung cancer Diseases 0.000 description 2
- 208000020816 lung neoplasm Diseases 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 208000030507 AIDS Diseases 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 206010008111 Cerebral haemorrhage Diseases 0.000 description 1
- 206010061818 Disease progression Diseases 0.000 description 1
- 208000005016 Intestinal Neoplasms Diseases 0.000 description 1
- 206010040047 Sepsis Diseases 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000005750 disease progression Effects 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 201000002313 intestinal cancer Diseases 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004984 smart glass Substances 0.000 description 1
- 238000012795 verification Methods 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
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a disease diagnosis system, equipment and medium based on a cycle time convolution network, comprising: the prediction model construction and training module is used for extracting a known disease course vector of a known disease diagnosis result according to the historical electronic medical record, and taking the known disease course vector as a cycle time convolution network prediction model which is trained and constructed by a training set; and the disease diagnosis prediction module is used for extracting a disease course vector to be detected according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be detected into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted. A cycle time convolution network prediction model constructed based on a time convolution network fusion long-time cycle network establishes a relation between a disease diagnosis result and historical electronic medical record diagnosis information, retains long-time historical information, performs long-range modeling prediction on sequence data, and realizes refined prediction on long-sequence data information.
Description
Technical Field
The invention relates to the technical field of medical data mining, in particular to a disease diagnosis system, equipment and medium based on a cycle time convolutional network.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the increasing of medical record data, intelligent disease diagnosis can play a good auxiliary role for doctors, at present, researchers often select data from Electronic Medical Records (EMR) to perform further diagnosis and prediction, the EMR is a complete and detailed clinical information resource generated and recorded by residents in the course of each visit of a medical institution, almost covers all past and present medical information of patients, and the inventor finds that part of patients have longer disease course sequences, the traditional prediction model cannot always keep long-term historical information, long-term modeling is difficult to perform on sequence data, the prediction model consumes long time, the precision is low, and great difficulty is caused for later diagnosis prediction; therefore, EMR data processing and disease diagnosis based on EMR data processing have become an inherent challenge, and relying on traditional machine learning methods alone is not enough to perform refined representation of long-range data information, and at present, a multi-disease diagnosis system capable of capturing long-range time dependence and further performing refined prediction has not appeared.
Disclosure of Invention
In order to solve the problems, the invention provides a disease diagnosis system, equipment and medium based on a cycle time convolution network, a cycle time convolution network prediction model constructed based on a time convolution network fusion long-time cycle network is used for establishing a relation between a disease diagnosis result and historical electronic medical record diagnosis information, long-time historical information is reserved, long-range modeling prediction is carried out on sequence data, and refined prediction is carried out on long-sequence data information.
In order to achieve the purpose, the invention adopts the following technical scheme:
in a first aspect, the present invention provides a disease diagnosis system based on a cyclic time convolution network, comprising:
the prediction model construction and training module is used for extracting a known disease course vector of a known disease diagnosis result according to the historical electronic medical record, and taking the known disease course vector as a cycle time convolution network prediction model which is trained and constructed by a training set;
and the disease diagnosis prediction module is used for extracting a disease course vector to be detected according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be detected into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
In a second aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the steps of:
extracting known disease course vectors of known disease diagnosis results according to the historical electronic medical records, and training and constructing a cycle time convolution network prediction model by using the known disease course vectors as a training set;
and extracting a disease course vector to be predicted according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be predicted into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
In a third aspect, the present invention provides a computer readable storage medium storing computer instructions that, when executed by a processor, perform the steps of:
extracting known disease course vectors of known disease diagnosis results according to the historical electronic medical records, and training and constructing a cycle time convolution network prediction model by using the known disease course vectors as a training set;
and extracting a disease course vector to be predicted according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be predicted into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
Compared with the prior art, the invention has the beneficial effects that:
according to the long-term historical medical record diagnosis information, diagnosis information related to a disease diagnosis result is extracted, and a known disease course vector training set is constructed; a cycle time convolution network prediction model is constructed based on a time convolution network fusion long-time cycle network, the model is trained by a known course vector training set, so that various disease diagnosis prediction results are linked with historical disease diagnosis of patients, the disease test accuracy is high, the reliability is strong, the stability is high, and the refined prediction of long-sequence medical record data information is realized.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a schematic diagram of a disease diagnosis system based on a cyclic time convolution network according to embodiment 1 of the present invention;
FIG. 2 is a schematic diagram of a Skip-gram encoding module according to embodiment 1 of the present invention;
fig. 3 is a schematic diagram of a convolution process of the feature extraction module provided in embodiment 1 of the present invention;
FIG. 4 is a schematic diagram of a long term memory network according to embodiment 1 of the present invention;
fig. 5 is a structural diagram of a cyclic time convolution network according to embodiment 1 of the present invention;
FIG. 6 is a graph showing an iterative comparison of the recall ratio of the method of the present invention compared to other methods.
The specific implementation mode is as follows:
the invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, and it should be understood that the terms "comprises" and "comprising", and any variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example 1
As shown in fig. 1, the present embodiment provides a disease diagnosis system based on a cyclic time convolution network, including:
the prediction model construction and training module is used for extracting a known disease course vector of a known disease diagnosis result according to the historical electronic medical record, and taking the known disease course vector as a cycle time convolution network prediction model which is trained and constructed by a training set;
and the disease diagnosis prediction module is used for extracting a disease course vector to be detected according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be detected into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
The prediction model building and training module comprises:
a training data acquisition sub-module configured to acquire historical patient visit electronic medical records of known disease diagnosis results;
as will be appreciated, the electronic medical record includes medication records, disease diagnosis records, physiological indicators, assay results, non-textual records, and the like.
The training data preprocessing submodule is configured to preprocess the collected historical electronic medical record information for treatment to obtain preprocessed historical electronic medical record information for treatment;
the training data extraction sub-module is configured to extract a known disease course sequence of each known disease diagnosis result patient from the preprocessed historical electronic medical record information;
it will be appreciated that the disease progression sequence includes patient historical disease diagnostic information, such as [ heart disease, hypertension, lung cancer ], etc.;
and the training data coding sub-module is configured to code and represent each known disease diagnosis result according to the extracted known disease course sequence to obtain a plurality of known disease course vectors of the known disease diagnosis results.
In this embodiment, the training data preprocessing sub-module includes:
the data screening unit is configured to perform primary screening on the acquired historical electronic medical record information, specifically screen historical disease diagnosis results and diagnosis time, and delete non-required data; deleting a plurality of preset diagnosis results of the non-important diseases from the primary screening result, and keeping the diagnosis results and the diagnosis time of the important diseases;
it is to be understood that the non-demand data includes: test number, name, race, examination department, etc.;
a data fitting unit configured to fit missing values of the diagnosis time;
and a format conversion unit configured to perform format conversion on both the diagnosis result and the diagnosis time of the important disease.
In this embodiment, the training data encoding submodule includes:
establishing a known disease course sequence of the patient corresponding to the historical disease diagnosis result as a first sample, such as patient 1: [ AIDS, hypertension, lung cancer ]; patient 2[ sepsis, intestinal cancer, coronary heart disease, cerebral hemorrhage ], etc.;
coding the first sample by adopting a Skip-gram method to obtain a plurality of known disease course vectors corresponding to historical disease diagnosis results, such as [0.3682,0.3452,0.5835 ]; [0.8735,0.4721,0.4524,0.5423].
A plurality of known disease course vectors corresponding to the historical disease diagnosis result are used as second samples, and the second samples are randomly divided into a training set and a testing set; the training set is used to train a predictive model.
Further, using the Skip-gram method, the selected target vocabulary, i.e., the vocabulary for predicting the occurrence of a disease using a given word, the Skip-gram selecting the target vocabulary and predicting its neighborhood, as shown in fig. 2;
continuously selecting a current target vocabulary by a sliding window mode in a diagnosis sequence and predicting, wherein a target task is a log-likelihood function shown in an optimization formula (1):
wherein T is the length of the disease diagnosis vector to be trained, C is the size of the sliding window, CtIs omegatThe neighbor diagnosis word set of (1); omegatFor the current diagnosis, ωcIs omegatNeighbor diagnosis of (1);
if the function defining the similarity of two diagnoses is s, then the softmax function can be used to define p (ω [. omega. ])c|ωt) As shown in formula (2):
wherein, W is the total number of clinical diagnoses, and s is the similarity of clinical diagnosis words; omegatFor current clinical diagnosis, omegacIs omegatClinical diagnosis of neighbors.
Further, the word vector inner product defines a similarity function, as shown in equation (3):
wherein, ω istFor current clinical diagnosis, omegacIs omegatIn the neighbor clinical diagnosis of (1), v is ωcIs represented by a vector of (u) is ωtVector representation of ctIs omegatA set of context words.
Regarding the words at the position t, regarding all the neighbor diagnosis concept words as positive examples, and randomly extracting negative examples from the diagnosis concept set table for training;
for the selected neighbor position c, solving the objective function shown in the formula (4) to obtain an optimal solution, and obtaining a final vector representation:
where T is the length of the medical diagnosis vector, c is the context word, ωtFor central diagnosis of concept words, omegacIs omegatA context diagnosis concept word of CtIs omegatS is a conceptual word similarity function, NtIs a set of words.
In the prediction model building and training module, the method further comprises:
the cyclic time convolution network construction sub-module is configured to construct a cyclic time convolution network prediction model based on the time convolution network fusion long-time cyclic network;
and the cycle time convolution network training sub-module is configured to train the cycle time convolution network prediction model by using the known disease course vectors of a plurality of known disease diagnosis results to obtain the trained cycle time convolution network prediction model.
And the cycle time convolution network testing sub-module is configured to test the trained cycle time convolution network prediction model.
As shown in fig. 3, in this embodiment, the building of the cyclic time convolutional network prediction model based on the time convolutional network fusion long-and-short cyclic network includes feature extraction of a training set, where the feature extraction includes: a convolution unit and a residual connecting unit;
the convolution unit takes the known disease course vector of the second sample as input, i.e. X ═ X1,...,xT) (ii) a A convolution kernel is defined as F ═ (0.., F., F)k-1) Performing feature extraction on time series data acquired from input data, in the case of xiWhen convolution is carried out, only the previous information is operated, the output at the time t is ensured to be only convoluted with the elements of the current layer and the previous layer, and the low-dimensional feature is output, as shown in FIG. 4;
meanwhile, the receptive field is enlarged by setting the hyper-parameter expansion rate, and a hole is injected on the basis of a common convolution method, so that the purpose of widening the receptive field is achieved, long-term memory is kept, the data volume is reduced, the calculation complexity is reduced, meanwhile, high-dimensional characteristic expression containing long-term historical information is output, and the convolution process is as shown in formula (5):
where k is the convolution kernel size, s is the element in the sequence, X is the input code, d is the expansion ratio, and f (X) represents the convolution operation on the X sequence to obtain the signature sequence G ═ { G ═ G1,G2,G3,...,Gn-k}；
And (3) carrying out weight normalization and activation function operation on the characteristic sequence, as shown in formula (6):
H(X)＝ReLU(WeightedNorm(G)) (6)
the residual error connection unit firstly performs one-dimensional convolution on the identity mapping X to ensure that the input and output sizes are the same;
then adding the initial input X of the model with the output of the previous layer, directly skipping the middle layers, integrating the deep layer network and the shallow layer network as the output of the whole residual module, and solving the problems of gradient disappearance and gradient explosion generated when the network is too deep, as shown in formula (7):
Z＝Activation(X+H(X)) (7)
the cyclic time convolution network construction sub-module also comprises a sequence prediction unit which adopts an LSTM model to carry out sequence modeling prediction;
LSTM (Long Short-Term Memory) is a Long Short-Term Memory network, a time recurrent neural network, suitable for processing and predicting important events with relatively Long intervals and delays in time series.
The sequence prediction unit receives the sequence feature vector Z ═ Z output by the residual connection unit1,z2,z3,zt.. } as input, the history information is updated and the sequence is modeled and predicted by using input gate, forgetting gate, output gate and memory gate controlled by 3 gate control units respectively, which are respectively used for reading, resetting and updating the history information, and the structure is shown in fig. 5:
wherein the input gate is defined as: i.e. it＝σ(Wzizt+Whiht-1+Wciht-1+bi) (8)
The definition of forgetting gate is: f. oft＝σ(Wzfzt+Whfht-1+Wcfht-1+bf) (9)
The definition of the memory cell is: c. Ct＝ftct-1+ittanh(Wzcxt+Whcxt-1+bc) (10)
The definition of the output gate is: ot＝σ(Wxozt+Whoht-1+Wcoht-1+bo) (11)
The hidden state is defined as: h ist＝ottanh(ct) (12)
In the above formulas, it,ft,otRespectively an input gate, a forgetting gate, an output gate, ctIs a memory cell, ztAs an input vector, htIn a hidden state, W corresponds to the weight of each layer of data input and output, b is various offsets, and σ isAnd the value range of the logistic id function is (0, 1).
Setting the loss function as a negative log-likelihood loss function, as shown in equation (13):
wherein, yiAs a true category, yi' is the prediction probability value, and M is the number of categories.
In the embodiment, the features are extracted by a convolution method, and the LSTM model is utilized for serialization modeling, so that the long-range time dependence can be captured, and refined prediction can be performed;
in this embodiment, a cyclic time convolutional network prediction model is trained based on a training set until the model converges; the deep learning model based on the cycle time convolutional network enables various disease diagnosis prediction results to be connected with historical disease diagnosis of patients, and has the advantages of being high in testing accuracy, strong in reliability and stable.
The disease diagnosis and prediction module comprises:
the data acquisition sub-module to be predicted is configured to acquire electronic medical record information of a patient to be predicted;
the data to be predicted extraction submodule is configured to extract a disease course vector of a patient from electronic medical record information of the patient to be predicted;
and the data to be predicted prediction result output submodule is configured to input the extracted disease course vector of the patient to be predicted into a pre-trained cycle time convolution network prediction model and output a disease diagnosis prediction result of the patient to be predicted.
In this example, the data used had a total admission report of 7317 patients, each patient had more than 2 admission records, ICD9 codes for their diagnostic records were extracted for each patient's course information, and a total of 15820 admission records were counted; building a multi-level nested list to predict diagnostic categories in the next visit, the nested list is as follows:
Hp b＝[[d1,d2...df],[d1,d2...dg],...,[d1,d2...dh]]
wherein, p and b are the number of patients and the admission times respectively, namely p is 7317, b is 15820; d is clinical diagnosis; f, g and h are the number of diagnosis results of different patients.
This example extracts and feeds the diagnostic record for each patient into the Skip-gram model in chronological order. In order to accurately represent the disease concept to a greater extent and reduce the operation time, the window size is set to be 5, and 128-dimensional vectors are selected for vector representation;
in this embodiment, to preserve the diagnosis granularity information to a greater extent, the disease diagnoses appearing in the diagnoses are classified into 283 groups according to the CCS diagnosis classifier, 70% of sequences are used for training, 10% of data are set as a parameter adjustment verification set, 20% of data are preserved for testing, and a random gradient descent (SGD) method is used for training units every 100 epochs; since the research problem is essentially a sequence prediction problem, the Recall @ k method and the Accuracy @ k method are used for evaluation.
The Recall @ k method is as follows: if there are m in the first k records for a certain patient sequence predicted by the modeliA correct diagnosis record and n in the sequence of the original disease course of the patientiThe right case records
Calculating the value of each patient, and averaging, wherein the average value is the recall rate of the model; the results obtained are shown in table 1:
TABLE 1 comparison of model recall rates at different k values before and after optimization
The AUC @ k (mean average precision) method is as follows: giving k prediction records, if the real diagnosis data are in the k records, recording the result as 1, otherwise, recording the result as 0; the results obtained are shown in Table 2:
TABLE 2 comparison of AUC values before and after optimization for different k values
It can be seen that according to the experimental result, when k is 20, the model preprocessed by the Skip-gram model obtains the result with the highest precision, and the effectiveness of the method is proved; meanwhile, with the increase of the k value selection, effective results are obtained continuously; the Recall rate was improved by 14.14% compared to the original LSTM model, and the iterative process for Recall @20 is shown in fig. 6.
In order to verify the robustness and reliability of the model, 50-bit patient electronic medical record data with 3 hospital admissions are randomly extracted, 50 pieces of patient information are converted into vector codes according to a disease vector coded by a Skip-gram method by adopting the disease diagnosis system based on deep learning of the embodiment, and the vector codes are imported into a cyclic time convolution network prediction model for prediction, so that the result shows that the next illness state of 44 patients can be predicted correctly, the accuracy reaches 88%, the result is not lower than the model prediction range, and the cyclic convolution network model is verified to have certain reliability and robustness.
a disease diagnosis method based on a cyclic time convolutional network, comprising:
extracting known disease course vectors of known disease diagnosis results according to the historical electronic medical records, and training and constructing a cycle time convolution network prediction model by using the known disease course vectors as a training set;
and extracting a disease course vector to be predicted according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be predicted into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
In further embodiments, there is also provided:
an electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the method described in embodiment 2. For brevity, no further description is provided herein.
The electronic device may be a mobile terminal and a non-mobile terminal, the non-mobile terminal includes a desktop computer, and the mobile terminal includes a Smart Phone (such as an Android Phone and an IOS Phone), Smart glasses, a Smart watch, a Smart bracelet, a tablet computer, a notebook computer, a personal digital assistant, and other mobile internet devices capable of performing wireless communication.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and so on. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may include both read-only memory and random access memory, and may provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.
A computer readable storage medium storing computer instructions which, when executed by a processor, perform the method described in embodiment 2.
The method in embodiment 2 may be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software modules may be located in ram, flash, rom, prom, or eprom, registers, among other storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor. To avoid repetition, it is not described in detail here.
Those of ordinary skill in the art will appreciate that the various illustrative elements, i.e., algorithm steps, described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the above embodiments, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is merely a division of one logic function, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some interfaces, and may be in an electrical, mechanical or other form.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A cycle time convolutional network-based disease diagnostic system, comprising:
the prediction model construction and training module is used for extracting a known disease course vector of a known disease diagnosis result according to the historical electronic medical record, and taking the known disease course vector as a cycle time convolution network prediction model which is trained and constructed by a training set;
and the disease diagnosis prediction module is used for extracting a disease course vector to be detected according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be detected into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
2. The disease diagnosis system based on the cyclic time convolution network as claimed in claim 1, wherein the acquired historical electronic medical record is preprocessed, specifically:
screening known disease diagnosis results and diagnosis time, deleting preset non-required data and non-important diagnosis results, fitting missing values of the diagnosis time, and performing format conversion on important diagnosis results and the diagnosis time.
3. The system of claim 1, wherein the extracting the known course vector of the known disease diagnosis result comprises:
and establishing a known disease course sequence corresponding to the known disease diagnosis result of the patient as a first sample, and coding the first sample by adopting a Skip-gram method to obtain a plurality of known disease course vectors corresponding to the known disease diagnosis result.
4. The disease diagnosis system based on the cyclic time convolution network as claimed in claim 3, characterized in that a Skip-gram method is adopted to predict the disease vocabulary in its neighborhood according to the target vocabulary;
defining a similarity function by using the word vector inner product, solving a target function according to the similarity function to obtain the optimal predicted value of a target word selected by a sliding window in a disease diagnosis vector, and obtaining a known disease course vector;
the objective function is:
where T is the length of the disease diagnosis vector, c is the context word, ωtFor central diagnosis of concept words, omegacIs omegatA context diagnosis concept word of CtIs omegatS is a conceptual word similarity function, NtIs a set of words.
5. The system of claim 1, wherein the cyclic time convolutional network prediction model is constructed by fusing long and short cyclic networks based on a time convolutional network.
6. The system according to claim 1, wherein the construction of the prediction model of the cyclic time convolutional network comprises a convolution operation on a known disease course vector, specifically:
defining a convolution kernel, and performing feature extraction on the extracted known disease course vector to obtain low-dimensional features;
and setting a hyper-parameter expansion rate, performing convolution operation on the low-dimensional features, outputting a feature sequence containing long-term historical information, and performing weight normalization and data processing of an activation function on the feature sequence.
7. The disease diagnosis system based on the cyclic time convolution network as claimed in claim 6, wherein the construction of the cyclic time convolution network prediction model includes a residual connecting operation, specifically, adding a known disease course vector to the feature sequence after data processing to obtain a sequence feature vector.
8. The disease diagnosis system based on the cyclic time convolution network as claimed in claim 6, wherein the construction of the cyclic time convolution network prediction model includes performing a sequence modeling prediction by using an LSTM model, specifically performing an input gate, a forgetting gate and an output gate processing by using a sequence feature vector as an input, and finally performing a modeling prediction on the sequence feature vector through a memory unit.
9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the steps of:
extracting known disease course vectors of known disease diagnosis results according to the historical electronic medical records, and training and constructing a cycle time convolution network prediction model by using the known disease course vectors as a training set;
and extracting a disease course vector to be predicted according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be predicted into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
10. A computer readable storage medium storing computer instructions that, when executed by a processor, perform the steps of:
extracting known disease course vectors of known disease diagnosis results according to the historical electronic medical records, and training and constructing a cycle time convolution network prediction model by using the known disease course vectors as a training set;
and extracting a disease course vector to be predicted according to the electronic medical record information of the patient to be predicted, and inputting the disease course vector to be predicted into the trained cycle time convolution network prediction model to obtain a disease diagnosis prediction result of the patient to be predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010483848.XA CN111798975A (en) | 2020-06-01 | 2020-06-01 | Disease diagnosis system, apparatus and medium based on cyclic time convolution network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010483848.XA CN111798975A (en) | 2020-06-01 | 2020-06-01 | Disease diagnosis system, apparatus and medium based on cyclic time convolution network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111798975A true CN111798975A (en) | 2020-10-20 |
Family
ID=72806051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010483848.XA Pending CN111798975A (en) | 2020-06-01 | 2020-06-01 | Disease diagnosis system, apparatus and medium based on cyclic time convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111798975A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112201350A (en) * | 2020-11-11 | 2021-01-08 | 北京嘉和海森健康科技有限公司 | Intelligent triage method and device and electronic equipment |
| CN112951362A (en) * | 2021-02-23 | 2021-06-11 | 上海商汤智能科技有限公司 | Medicine recommendation method, device, equipment and storage medium |
| CN113066572A (en) * | 2021-03-03 | 2021-07-02 | 山东师范大学 | Traditional Chinese medicine auxiliary diagnosis system and method for enhancing local feature extraction |
| CN113707322A (en) * | 2021-08-31 | 2021-11-26 | 平安科技（深圳）有限公司 | Training method and device of information prediction model, computer equipment and storage medium |
| CN114582511A (en) * | 2022-05-07 | 2022-06-03 | 中国人民解放军总医院第八医学中心 | Bronchiectasis acute exacerbation early warning method, device, equipment and medium |
| CN114758768A (en) * | 2020-12-28 | 2022-07-15 | 中国科学院沈阳计算技术研究所有限公司 | Cervical cancer diagnosis and treatment project prediction method based on Bi-GRU |
| CN114864093A (en) * | 2022-07-04 | 2022-08-05 | 北京鹰瞳科技发展股份有限公司 | Apparatus, method and storage medium for disease prediction based on fundus image |
| CN116959715A (en) * | 2023-09-18 | 2023-10-27 | 之江实验室 | Disease prognosis prediction system based on time sequence evolution process explanation |
| CN117271804A (en) * | 2023-11-21 | 2023-12-22 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109805898A (en) * | 2019-03-22 | 2019-05-28 | 中国科学院重庆绿色智能技术研究院 | Critical illness Mortality Prediction method based on attention mechanism timing convolutional network algorithm |
| CN110472779A (en) * | 2019-07-30 | 2019-11-19 | 东莞理工学院 | A kind of power-system short-term load forecasting method based on time convolutional network |
| CN110596506A (en) * | 2019-10-30 | 2019-12-20 | 福州大学 | Converter fault diagnosis method based on time convolution network |
| CN110909926A (en) * | 2019-11-18 | 2020-03-24 | 中国计量大学 | TCN-LSTM-based solar photovoltaic power generation prediction method |
-
2020
- 2020-06-01 CN CN202010483848.XA patent/CN111798975A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109805898A (en) * | 2019-03-22 | 2019-05-28 | 中国科学院重庆绿色智能技术研究院 | Critical illness Mortality Prediction method based on attention mechanism timing convolutional network algorithm |
| CN110472779A (en) * | 2019-07-30 | 2019-11-19 | 东莞理工学院 | A kind of power-system short-term load forecasting method based on time convolutional network |
| CN110596506A (en) * | 2019-10-30 | 2019-12-20 | 福州大学 | Converter fault diagnosis method based on time convolution network |
| CN110909926A (en) * | 2019-11-18 | 2020-03-24 | 中国计量大学 | TCN-LSTM-based solar photovoltaic power generation prediction method |
Non-Patent Citations (1)
| Title |
|---|
| 王露潼等: "基于FT-LSTM 模型的临床事件诊断序列预测研究", 计算机应用研究, pages 2961 - 2965 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112201350A (en) * | 2020-11-11 | 2021-01-08 | 北京嘉和海森健康科技有限公司 | Intelligent triage method and device and electronic equipment |
| CN114758768A (en) * | 2020-12-28 | 2022-07-15 | 中国科学院沈阳计算技术研究所有限公司 | Cervical cancer diagnosis and treatment project prediction method based on Bi-GRU |
| CN112951362A (en) * | 2021-02-23 | 2021-06-11 | 上海商汤智能科技有限公司 | Medicine recommendation method, device, equipment and storage medium |
| CN113066572A (en) * | 2021-03-03 | 2021-07-02 | 山东师范大学 | Traditional Chinese medicine auxiliary diagnosis system and method for enhancing local feature extraction |
| CN113066572B (en) * | 2021-03-03 | 2023-06-16 | 山东师范大学 | Traditional Chinese medicine auxiliary diagnosis system and method for enhancing local feature extraction |
| CN113707322B (en) * | 2021-08-31 | 2023-09-19 | 平安科技（深圳）有限公司 | Training method and device for information prediction model, computer equipment and storage medium |
| CN113707322A (en) * | 2021-08-31 | 2021-11-26 | 平安科技（深圳）有限公司 | Training method and device of information prediction model, computer equipment and storage medium |
| CN114582511A (en) * | 2022-05-07 | 2022-06-03 | 中国人民解放军总医院第八医学中心 | Bronchiectasis acute exacerbation early warning method, device, equipment and medium |
| CN114864093A (en) * | 2022-07-04 | 2022-08-05 | 北京鹰瞳科技发展股份有限公司 | Apparatus, method and storage medium for disease prediction based on fundus image |
| CN114864093B (en) * | 2022-07-04 | 2022-11-01 | 北京鹰瞳科技发展股份有限公司 | Apparatus, method and storage medium for disease prediction based on fundus image |
| CN116959715A (en) * | 2023-09-18 | 2023-10-27 | 之江实验室 | Disease prognosis prediction system based on time sequence evolution process explanation |
| CN116959715B (en) * | 2023-09-18 | 2024-01-09 | 之江实验室 | Disease prognosis prediction system based on time sequence evolution process explanation |
| CN117271804A (en) * | 2023-11-21 | 2023-12-22 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
| CN117271804B (en) * | 2023-11-21 | 2024-03-01 | 之江实验室 | Method, device, equipment and medium for generating common disease feature knowledge base |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111798975A (en) | Disease diagnosis system, apparatus and medium based on cyclic time convolution network | |
| CN110444263B (en) | Disease data processing method, device, equipment and medium based on federal learning | |
| El-Kenawy et al. | Meta-heuristic optimization and keystroke dynamics for authentication of smartphone users | |
| Singh et al. | Deep ConvLSTM with self-attention for human activity decoding using wearable sensors | |
| US11694123B2 (en) | Computer based object detection within a video or image | |
| US10242293B2 (en) | Method and program for computing bone age by deep neural network | |
| Alanazi | Identification and prediction of chronic diseases using machine learning approach | |
| Zhu et al. | Human identification for activities of daily living: A deep transfer learning approach | |
| US11915127B2 (en) | Prediction of healthcare outcomes and recommendation of interventions using deep learning | |
| Shorfuzzaman et al. | Artificial Neural Network‐Based Deep Learning Model for COVID‐19 Patient Detection Using X‐Ray Chest Images | |
| CN108417272B (en) | Similar case recommendation method and device with time sequence constraint | |
| KR20190030876A (en) | Method for prediting health risk | |
| CN111785366B (en) | Patient treatment scheme determination method and device and computer equipment | |
| Chowdhury et al. | Heart disease prognosis using machine learning classification techniques | |
| CN112132624A (en) | Medical claims data prediction system | |
| Rajliwall et al. | Cardiovascular risk prediction based on XGBoost | |
| CN112951362A (en) | Medicine recommendation method, device, equipment and storage medium | |
| CN116364299A (en) | Disease diagnosis and treatment path clustering method and system based on heterogeneous information network | |
| CN112447270A (en) | Medication recommendation method, device, equipment and storage medium | |
| Shirazi et al. | Deep learning in the healthcare industry: theory and applications | |
| Slim et al. | IoT System Based on parameter optimization of Deep Learning using Genetic Algorithm. | |
| Ju et al. | 3D-CNN-SPP: A patient risk prediction system from electronic health records via 3D CNN and spatial pyramid pooling | |
| Agrawal et al. | ElectroCardioGram (ECG)-based user authentication using deep learning algorithms | |
| Mogan et al. | Advances in vision-based gait recognition: From handcrafted to deep learning | |
| Yao et al. | Graph Kernel prediction of drug prescription |
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 |