CN112365976A - Compound disease clinical path construction method and system based on transfer learning - Google Patents
Compound disease clinical path construction method and system based on transfer learning Download PDFInfo
- Publication number
- CN112365976A CN112365976A CN202011273670.2A CN202011273670A CN112365976A CN 112365976 A CN112365976 A CN 112365976A CN 202011273670 A CN202011273670 A CN 202011273670A CN 112365976 A CN112365976 A CN 112365976A
- Authority
- CN
- China
- Prior art keywords
- data set
- disease
- neural network
- time domain
- medical
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention relates to a composite disease clinical path construction method and a system based on transfer learning, wherein the method comprises the following steps: acquiring multi-source heterogeneous medical data of a compound disease species generated by a non-clinical path; structuring the ICD according to the ICD classification to form a first data set; performing feature extraction and clustering on the first data set according to the sign information, the medical staging, the diagnosis information and the treatment evaluation to form a second data set; randomly extracting two disease data sets from the second data set, training a time domain convolution neural network to learn one disease data set, and adjusting the time domain convolution neural network through MMD (mass-median-variance) for transfer learning of the other disease data set; and finally, fusing according to the output result to obtain a clinical path of the compound disease species. The invention improves the performance and interpretability of the clinical path structure of the compound disease species and avoids gradient explosion through the simplification of a data set and the integration of a time domain convolution neural network and a generating type antagonistic neural network.
Description
Technical Field
The invention relates to the field of medical information and deep learning, in particular to a composite disease clinical path construction method and system based on transfer learning.
Background
Clinical Pathways (CP) refer to a programmed and standardized diagnosis and treatment plan with strict work sequence and accurate time requirements, which is made based on evidence-based medicine and with the purpose of expected treatment effect and cost control, for diseases or surgical operations corresponding to a certain International Classification of Di diseases (ICD).
As the clinical path development in China is in the initial stage, the clinical path is mainly established by means of traditional expert evaluation and evaluation, the time consumption is long, the cost is high, and the degree of variation is large in clinical practice. So far, the clinical routes established by the national defense council are only about 1200, which are far from the 40000 diseases recorded by ICD-10/IC9-9 in China, and the clinical routes of the national defense council are far from the requirements of clinical practice. The national clinical pathway management standard includes applicable subjects, diagnosis basis, entry pathway standard, standard length of stay in hospital, examination items during stay in hospital, preparation before stay in hospital, selection of treatment plan, discharge standard, variation and reason analysis. There is no specific guidance regarding the choice of treatment protocol, the specific medication and the cost instructions.
The data sources for data mining of clinical paths are usually Information software commonly used in hospitals, and Information Systems commonly used in hospitals include Hospital Information Systems (HIS), electronic medical Record Systems (EMR), Laboratory Information management Systems (LIS), Picture Archiving and Communication Systems (PACS) Systems. The LIS system is used for examining and testing departments, and after the experimental instruments for examination are linked with a computer, data of a plurality of examination and testing items such as blood routine examination and the like can be acquired and analyzed to generate a detection report. The PACS system is widely applied to the imaging department, can store daily medical images such as CT, nuclear magnetic resonance and the like in an electronic mode, and is accompanied with an auxiliary array diagnosis function. Whereas the test results, image results, produced in LIS systems and PACS systems are often presented and utilized in HIS systems and EMR systems.
In the prior art, a multi-component model or a graph model is established for medical data, patient information, diagnosis information, examination information and expense information to construct a clinical path; there are also some methods of mining text data using natural language processing or machine learning, expert knowledge bases or medical knowledge bases, and mining the inherent association to construct a clinical pathway, which, although some effects are achieved, still have several problems:
1. although the multi-component model or the graph model attaches importance to the intrinsic correlation of the patient sign information, the treatment means and the drugs, and even utilizes reasoning learning, the multi-component model or the graph model ignores the time sequence and the variability of the patient information and the diagnosis information; that is, the final treatment effect of the same patient under the condition of the same physical sign information, treatment means and medicine of the patient may also be different;
2. although a certain effect can be obtained by mining the text information of the medical data by using a clustering method of natural language processing or machine learning, the interpretability of the medical data is questioned and the generalization is poor;
3. an effective clinical path construction method for a clinical path of a compound disease species does not exist so far;
in conclusion, the existing clinical path construction method model is lack of time sequence, cause and effect connection among data and low in interpretability.
Disclosure of Invention
The invention provides a composite disease category clinical path construction method based on transfer learning, aiming at the technical problems of lack of time sequence, too deep model layer number, weak generalization, large calculated amount and poor interpretability in the existing composite disease category clinical path construction technology, and comprising the following steps:
the method comprises the steps of obtaining multi-source heterogeneous medical data of a compound disease species generated by a non-clinical path, preprocessing and structuring the multi-source heterogeneous medical data according to an ICD classified clinical path template, and obtaining a first data set consisting of a plurality of single disease species data sets derived from the same compound disease species;
clustering and feature extracting are carried out on the first data set according to patient sign information, medical staging, diagnosis information and treatment evaluation of the same disease type, and features of the patient sign information, the medical staging, the diagnosis information and the treatment evaluation are labeled to form a second data set;
randomly extracting a disease category data set from the second data set, and recording the disease category data set as a first disease category data set; taking the human body characteristic information label and the medical staging label of the first disease type data set as input labels, taking the medical staging and diagnosis information as output labels to train the time domain convolutional neural network until the error is lower than a threshold value, and obtaining a first time domain convolutional neural network;
randomly extracting another disease category data set from the second data set, and recording the data set as a second disease category data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and fusing and unstructured output of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
In some embodiments of the present invention, the obtaining of multi-source heterogeneous medical data of a compound disease species generated by a non-clinical pathway, and preprocessing and structuring the multi-source heterogeneous medical data according to an ICD-classified clinical pathway template to obtain a first data set composed of a plurality of single disease species data sets derived from a same compound disease species includes the following steps:
and structuring the data of the same disease category according to the disease category classification of the ICD and the fields of the standard template which accord with the disease category, cleaning and screening texts, data and images in the data set according to the field sequence and format of the data, and removing the parts of the data which cannot be identified or are wrong.
In some embodiments of the present invention, the clustering and feature extracting the first data set according to the patient sign information, medical stage, diagnosis information, and treatment evaluation of the same disease type, and labeling the features of the patient sign information, medical stage, diagnosis information, and treatment evaluation to form a second data set includes the following steps:
and clustering and extracting the characteristic of the patient sign information, the medical stage, the diagnosis information and the treatment evaluation of the same disease type in the first data set according to Euclidean distance and K-means.
In some embodiments of the invention, the first time domain convolutional neural network comprises at least two convolutional concealment layers, the output of at least one convolutional concealment layer is determined by a set number of the latest tag data, at least one residual block, and the output of one convolutional concealment layer is determined by all tag data.
In some embodiments of the invention, the maximum mean difference algorithm is represented as:
the MMD [ F, X, Y ] represents the distance between a first disease type data set and a second disease type data set in Hilbert space, F is a continuous function set on a sample space, X, Y represents the first disease type data set and the second disease type data set respectively, k () represents a kernel function of a sample of the first disease type data set or a sample of the second disease type data set, m represents the number of samples of the first disease type data set, n represents the number of samples of the second disease type data set, and i and j are serial numbers of any two different samples of the first disease type data set or the second disease type data set.
In some embodiments of the present invention, before the fusing and the unstructured performing the fusion and the unstructured on the outputs of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain the composite disease clinical path, the method further includes:
and constructing a generating type antagonistic neural network according to the first time domain convolutional neural network and the second time domain convolutional neural network. Furthermore, the second time domain convolutional neural network is used as a generating network of the generating type antagonistic neural network, and the first time domain convolutional neural network is used as a discriminating network of the generating type antagonistic neural network.
In a second aspect of the invention, a composite disease clinical path construction system based on transfer learning is provided, which comprises an acquisition module, an extraction module, a first training module, a second training module and an output module, wherein the acquisition module is used for acquiring multi-source heterogeneous medical data of a composite disease generated by a non-clinical path, and preprocessing and structuring the heterogeneous medical data according to a clinical path template classified by an ICD (interface control document) to obtain a first data set consisting of a plurality of single disease data sets;
the extraction module is used for clustering and extracting the characteristics of the first data set according to the patient sign information, the medical stage, the diagnosis information and the treatment evaluation of the same disease type, and marking the characteristics of the patient sign information, the medical stage, the diagnosis information and the treatment evaluation to form a second data set
The first training module is used for randomly extracting a disease type data set from the second data set and recording the disease type data set as a first disease type data set; and training the time domain convolutional neural network by using the human body characteristic information label and the medical staging label of the first disease data set as input labels and using the medical staging and diagnosis information as output labels until the error is lower than a threshold value to obtain a first time domain convolutional neural network
The second training module is used for randomly extracting another disease type data set from the second data set and recording the another disease type data set as a second disease type data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and the output module is used for fusing and unstructured output of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
The present invention provides an electronic device, comprising: one or more processors; a storage device, configured to store one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the method for constructing a composite disease clinical pathway based on transfer learning according to the first aspect of the present invention.
The present invention provides a computer readable medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method for constructing a clinical pathway of a compound disease based on transfer learning according to the first aspect of the present invention.
Has the advantages that:
1. the invention fully considers the complexity, multi-source heterogeneity and time sequence of the medical data, and simplifies the medical data into four major parts: sign information, medical staging, diagnosis information and treatment evaluation, and reduces complex preprocessing, feature extraction and clustering operations. The body state of a patient when the patient is diagnosed is standardized through standard medical staging, the relation between diagnostic information and sign information is enhanced on a data source, and the interpretability of a data set is improved;
2. in consideration of the time sequence of sign information, medical staging, diagnosis information and treatment evaluation, the invention adopts a time domain convolution neural network; compared with recurrent neural networks such as RNNs (recurrent neural networks), LSTM (local finite state machine) and the like, the convolutional network has better performance, avoids common defects of recurrent models, such as gradient explosion/disappearance problems or lack of memory retention, and also improves the interpretability of the models. In addition, the output can be calculated in parallel by using the convolution network, and the convergence and the processing speed are faster compared with a recurrent neural network.
3. The distribution distance of the two disease data sets is measured according to the MMD, and the time convolution network is adjusted by utilizing the MMD, so that the accuracy of transfer learning is improved.
4. In order to further improve the accuracy of data output, a generative type antagonistic neural network is constructed, and the generalization capability of the model is improved.
Drawings
FIG. 1 is a basic flow diagram of a composite disease clinical pathway construction method based on transfer learning in some embodiments of the present invention;
FIG. 2 is a diagram illustrating a relationship between a first data set and a second data set;
FIG. 3 is a schematic diagram of a composite disease clinical pathway construction system based on transfer learning according to some embodiments of the present invention;
fig. 4 is a schematic structural diagram of an electronic device in some embodiments of the invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1 and fig. 2, a composite disease clinical pathway construction method based on transfer learning includes the following steps: s101, multi-source heterogeneous medical data of a compound disease species generated by a non-clinical path are obtained, preprocessing and structuring are carried out on the multi-source heterogeneous medical data according to an ICD classified clinical path template, and a first data set consisting of a plurality of single disease species data sets derived from the same compound disease species is obtained;
it should be noted that, because the pathological analysis of the compound disease species is relatively complex, in order to simplify the model, the present invention divides the compound disease species into two main categories according to the clinical symptoms and the actual diagnosis and treatment experience of the compound disease species: can be decomposed into a plurality of combined disease species related to a single disease species, and can not be decomposed into the disease species with a plurality of clinical symptoms of the disease species. For example, the clinical symptoms of a viral pneumonia manifest as: dividing the cold, fever and pneumonia into disease types combined by the cold, fever and pneumonia according to actual diagnosis and treatment records; it is understood that as medicine progresses, the division will be classified into a single disease species or other disease species by multiple disease species;
illustratively, tuberculosis is a chronic basic disease of the lung that is compromised in immune function and is often associated with multiple mixed bacterial infections. Pulmonary tuberculosis is one of infectious diseases threatening human health, diabetes is an important related disease of pulmonary tuberculosis, a diabetic patient is a susceptible patient of pulmonary tuberculosis, and the pulmonary tuberculosis is one of important common reasons for inducing and aggravating complications of diabetes, the condition of pulmonary tuberculosis can be aggravated when the pulmonary tuberculosis and the pulmonary tuberculosis coexist, the treatment time is prolonged, the pulmonary tuberculosis can become a chronic infection source, and the pulmonary tuberculosis can aggravate the diabetes and induce various complications of the pulmonary tuberculosis. The two diseases coexist and affect each other, and the disease condition is complicated. Among deep fungal infections, pulmonary fungal infections are most common, since atmospheric and environmental fungi can subsequently be drawn into the lungs. The common inducement of pulmonary fungal infection is that the immunity of the organism is reduced due to long-term use of powerful broad-spectrum antibacterial drugs, glucocorticoid and the like, and inappropriate use of a large amount of antibacterial drugs leads to fungal infection due to long-term unreasonable use of the antibacterial drugs, and diabetes further aggravates the infection.
S102, clustering and feature extracting are carried out on the first data set according to patient sign information, medical staging, diagnosis information and treatment evaluation of the same disease, and features of the patient sign information, the medical staging, the diagnosis information and the treatment evaluation are labeled to form a second data set;
it should be noted that the patient sign information includes the age, sex, sign of the patient, and various physical and chemical indexes in the body temperature, pulse, respiration, blood pressure, blood routine examination, etc.; medical staging: the division into duration and severity stages of a patient suffering from a complex disease species is determined on the basis of routine examination. For example: staging of tumors, which describes the severity and extent of involvement of malignant tumors based on the primary tumor and the extent of dissemination in the individual; diabetes is divided into 1-4 stages according to the occurrence and development of nephropathy, wherein the 1 stage shows that the volume of the kidney is increased, the 2 stage shows that basement membrane of glomerular capillaries is thickened, the 3 stage shows that the microalbuminuria is achieved, and the 4 stage shows that the 24-hour urine protein is more than 0.5 g. In addition, hypertension, heart disease and leukemia are classified similarly, and the present invention is not repeated. The diagnosis information comprises information such as medium-long term medical advice, short term medical advice, nursing grade, medicine use, examination means and the like of each treatment day listed in the standard clinical path form; the treatment evaluation reference dimension comprises cost information, the discharge state (physical sign information or direct feedback) of a patient, the feeling of the patient during hospitalization and the like, and the treatment evaluation is obtained comprehensively by evaluating each dimension through the decision tree model.
S103, randomly extracting a disease type data set from the second data set, and recording the disease type data set as a first disease type data set; taking the human body characteristic information label and the medical staging label of the first disease type data set as input labels, taking the medical staging and diagnosis information as output labels to train the time domain convolutional neural network until the error is lower than a threshold value, and obtaining a first time domain convolutional neural network;
s104, randomly extracting another disease type data set from the second data set, and recording the disease type data set as a second disease type data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and S105, fusing and unstructured output of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
In step S101 according to some embodiments of the present invention, the obtaining multi-source heterogeneous medical data of a compound disease category generated by a non-clinical pathway, and preprocessing and structuring the multi-source heterogeneous medical data according to a clinical pathway template classified by an ICD to obtain a first data set composed of a plurality of single disease category data sets derived from a same compound disease category includes:
and structuring the data of the same disease category according to the disease category classification of the ICD and the fields of the standard template which accord with the disease category, cleaning and screening texts, data and images in the data set according to the field sequence and format of the data, and removing the parts of the data which cannot be identified or are wrong.
In step S102 of some embodiments of the present invention, the clustering and feature extracting the first data set according to the patient sign information, the medical stage, the diagnosis information, and the treatment evaluation of the same disease type, and labeling the features of the patient sign information, the medical stage, the diagnosis information, and the treatment evaluation to form the second data set includes the following steps:
and clustering and/or characteristic extraction are carried out on the patient sign information, the medical stage, the diagnosis information and the treatment evaluation of the same disease in the first data set according to Euclidean distance and K-means. It will be appreciated that other machine learning methods may also be employed for feature extraction and clustering: SVM, decision tree, KNN, embedding operations, etc.
In some embodiments of the invention, the first time domain Convolutional neural Network (TCN) comprises at least two Convolutional concealment layers, an output of at least one Convolutional concealment layer is determined by a set number of the latest tag data, at least one residual block, and an output of one Convolutional concealment layer is determined by all tag data. The label data refers to the labeled label data in the second data set.
Specifically, the residual module ensures that the dimensions of input data and output data are consistent by utilizing a Zero-padding method. For a single residual block, the input it accepts is the output of the previous block (the first block accepts the source data input), and this data is used in two places: one is used to compute the residual block result and the other is summed with the residual block result via one-dimensional convolution as the output of the present module. The part for calculating the residual error is firstly calculated through a hole causal convolution (DiatedCasualConv), historical information contained in input data is calculated, namely the preamble label information in the invention, then the historical information data is subjected to weight regularization (Weightnorm) and nonlinear transformation (ReLU) processes, the result is controlled within a reasonable range, and finally the part of the result is randomly zeroed through a random inactivation layer (Dropout), so that the interdependence between modules is reduced. And extracting the point-of-interest data related to time from the part passing through the one-dimensional convolution layer (1 multiplied by 1Conv), and performing residual connection (+) with the historical information calculated by the residual block to obtain the corrected data as the output of the current module. Thus, after a plurality of residual error modules are stacked and corrected, the output data TCN contains the required time interest point probability information for subsequent calculation.
In some embodiments of the invention, the maximum mean difference algorithm is represented as:
the MMD [ F, X, Y ] represents the distance between a first disease type data set and a second disease type data set in Hilbert space, F is a continuous function set on a sample space, X, Y represents the first disease type data set and the second disease type data set respectively, k () represents a kernel function of a sample of the first disease type data set or a sample of the second disease type data set, m represents the number of samples of the first disease type data set, n represents the number of samples of the second disease type data set, and i and j are serial numbers of any two different samples of the first disease type data set or the second disease type data set.
In some embodiments of the present invention, before the fusing and the unstructured performing the fusion and the unstructured on the outputs of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain the composite disease clinical path, the method further includes: and constructing a generating type antagonistic neural network according to the first time domain convolutional neural network and the second time domain convolutional neural network. Furthermore, the second time domain convolutional neural network is used as a generating network of the generating type antagonistic neural network, and the first time domain convolutional neural network is used as a discriminating network of the generating type antagonistic neural network.
It can be understood that the above fusing and unstructured the outputs of the first time domain convolutional neural network and the second time domain convolutional neural network specifically include: because the output of the first time domain convolutional neural network and the second time domain convolutional neural network is a multidimensional vector, the multidimensional vector is matched with the multi-element heterogeneous medical data in the first data set to obtain specific data such as patient sign information, medical staging, diagnostic information, treatment evaluation and the like corresponding to at least two single disease species belonging to the same composite disease species, then the two data are clustered and deduplicated, and the medical staging, diagnostic information and the like are output according to a standard clinical path template, so that the clinical path of the composite disease species is obtained.
Referring to fig. 3, in a second aspect of the present invention, a composite disease clinical pathway construction system 1 based on transfer learning is provided, including an obtaining module 11, an extracting module 12, a first training module 13, a second training module, 14, and an output module 15, where the obtaining module 11 is configured to obtain multi-source heterogeneous medical data of a composite disease generated by a non-clinical pathway, and pre-process and structure the heterogeneous medical data according to a clinical pathway template classified by an ICD to obtain a first data set composed of a plurality of single disease data sets;
the extraction module 12 is configured to cluster and extract features of the first data set according to patient sign information, medical staging, diagnosis information, and treatment evaluation of the same disease type, and label features of the patient sign information, medical staging, diagnosis information, and treatment evaluation to form a second data set;
the first training module 13 is configured to randomly extract a disease category data set from the second data set, and record the disease category data set as a first disease category data set; taking the human body characteristic information label and the medical staging label of the first disease type data set as input labels, taking the medical staging and diagnosis information as output labels to train the time domain convolutional neural network until the error is lower than a threshold value, and obtaining a first time domain convolutional neural network;
the second training module 14 is configured to randomly extract another disease category data set from the second data set, and record the another disease category data set as the second disease category data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and the output module 15 is configured to perform fusion and unstructured on the outputs of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
Referring to fig. 4, in a third aspect of the present invention, there is provided an electronic apparatus comprising: one or more processors; a storage device, configured to store one or more programs, when the one or more programs are executed by the one or more processors, cause the one or more processors to implement the map information representation method based on the graph structure provided in the first aspect of the present invention.
The electronic device 500 may include a processing means (e.g., central processing unit, graphics processor, etc.) 501 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)502 or a program loaded from a storage means 508 into a Random Access Memory (RAM) 503. In the RAM503, various programs and data necessary for the operation of the electronic apparatus 500 are also stored. The processing device 501, the ROM502, and the RAM503 are connected to each other through a bus 504. An input/output (I/O) interface 505 is also connected to bus 504.
The following devices may be connected to the I/O interface 505 in general: input devices 506 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a Liquid Crystal Display (LCD), speakers, vibrators, and the like; a storage device 508 including, for example, a hard disk; and a communication device 509. The communication means 509 may allow the electronic device 500 to communicate with other devices wirelessly or by wire to exchange data. While fig. 4 illustrates an electronic device 500 having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided. Each block shown in fig. 4 may represent one device or may represent multiple devices as desired.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication means 509, or installed from the storage means 508, or installed from the ROM 502. The computer program, when executed by the processing device 501, performs the above-described functions defined in the methods of embodiments of the present disclosure. It should be noted that the computer readable medium described in the embodiments of the present disclosure may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In embodiments of the disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In embodiments of the present disclosure, however, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.
The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device. The computer readable medium carries one or more computer programs which, when executed by the electronic device, cause the electronic device to:
computer program code for carrying out operations for embodiments of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, Python, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A composite disease clinical path construction method based on transfer learning is characterized by comprising the following steps:
the method comprises the steps of obtaining multi-source heterogeneous medical data of a compound disease species generated by a non-clinical path, preprocessing and structuring the multi-source heterogeneous medical data according to an ICD classified clinical path template, and obtaining a first data set consisting of a plurality of single disease species data sets derived from the same compound disease species;
clustering and feature extracting are carried out on the first data set according to patient sign information, medical staging, diagnosis information and treatment evaluation of the same disease type, and features of the patient sign information, the medical staging, the diagnosis information and the treatment evaluation are labeled to form a second data set;
randomly extracting a disease category data set from the second data set, and recording the disease category data set as a first disease category data set; taking the human body characteristic information label and the medical staging label of the first disease type data set as input labels, taking the medical staging and diagnosis information as output labels to train the time domain convolutional neural network until the error is lower than a threshold value, and obtaining a first time domain convolutional neural network;
randomly extracting another disease category data set from the second data set, and recording the data set as a second disease category data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and fusing and unstructured output of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
2. The composite disease clinical pathway construction method based on transfer learning of claim 1, wherein the step of obtaining multi-source heterogeneous medical data of a composite disease generated by a non-clinical pathway, and preprocessing and structuring the multi-source heterogeneous medical data according to an ICD-classified clinical pathway template to obtain a first data set consisting of a plurality of single disease data sets derived from the same composite disease comprises the steps of:
and structuring the data of the same disease category according to the disease category classification of the ICD and the fields of the standard template which accord with the disease category, cleaning and screening texts, data and images in the data set according to the field sequence and format of the data, and removing the parts of the data which cannot be identified or are wrong.
3. The composite disease clinical pathway construction method based on transfer learning of claim 1, wherein the step of clustering and feature extracting the first data set according to patient sign information, medical stages, diagnosis information and treatment evaluation of the same disease, and labeling the features of the patient sign information, medical stages, diagnosis information and treatment evaluation to form a second data set comprises the following steps:
and clustering and extracting the characteristic of the patient sign information, the medical stage, the diagnosis information and the treatment evaluation of the same disease type in the first data set according to Euclidean distance and K-means.
4. The method according to claim 1, wherein the first time domain convolutional neural network comprises at least two convolutional concealment layers, the output of at least one convolutional concealment layer is determined by a set number of latest label data, at least one residual module, and the output of one convolutional concealment layer is determined by all label data.
5. The composite disease clinical pathway construction method based on transfer learning of claim 1, wherein the maximum mean difference algorithm is expressed as:
the MMD [ F, X, Y ] represents the distance between a first disease type data set and a second disease type data set in Hilbert space, F is a continuous function set on a sample space, X, Y represents the first disease type data set and the second disease type data set respectively, k () represents a kernel function of a sample of the first disease type data set or a sample of the second disease type data set, m represents the number of samples of the first disease type data set, n represents the number of samples of the second disease type data set, and i and j are serial numbers of any two different samples of the first disease type data set or the second disease type data set.
6. The method for constructing a composite disease clinical pathway based on transfer learning of claim 1, wherein before the fusion and the unstructured processing of the outputs of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain the composite disease clinical pathway, the method further comprises:
and constructing a generating type antagonistic neural network according to the first time domain convolutional neural network and the second time domain convolutional neural network.
7. The method for constructing a composite disease clinical pathway based on transfer learning of claim 6, wherein the step of constructing a generative antagonistic neural network according to the first and second time domain convolutional neural networks comprises the following steps:
and taking the second time domain convolutional neural network as a generating network of the generating type antagonistic neural network, and taking the first time domain convolutional neural network as a judging network of the generating type antagonistic neural network.
8. A composite disease clinical path construction system based on transfer learning is characterized by comprising an acquisition module, an extraction module, a first training module, a second training module and an output module,
the acquisition module is used for acquiring multi-source heterogeneous medical data of a compound disease type generated by a non-clinical path, and preprocessing and structuring the heterogeneous medical data according to a clinical path template classified by an ICD (interface control document) to obtain a first data set consisting of a plurality of single disease type data sets;
the extraction module is used for clustering and extracting the characteristics of the first data set according to the patient sign information, the medical stage, the diagnosis information and the treatment evaluation of the same disease type, and marking the characteristics of the patient sign information, the medical stage, the diagnosis information and the treatment evaluation to form a second data set
The first training module is used for randomly extracting a disease type data set from the second data set and recording the disease type data set as a first disease type data set; and training the time domain convolutional neural network by using the human body characteristic information label and the medical staging label of the first disease data set as input labels and using the medical staging and diagnosis information as output labels until the error is lower than a threshold value to obtain a first time domain convolutional neural network
The second training module is used for randomly extracting another disease type data set from the second data set and recording the another disease type data set as a second disease type data set; taking the human body information label and the medical staging label of the second disease category data set as input labels, and taking the medical staging and diagnosis information as output labels; adjusting and training the first time domain convolutional neural network according to a maximum mean difference algorithm until the error is lower than a threshold value to obtain a second time domain convolutional neural network;
and the output module is used for fusing and unstructured output of the first time domain convolutional neural network and the second time domain convolutional neural network to obtain a composite disease clinical path.
9. An electronic device, comprising: one or more processors; storage means for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to carry out the method according to any one of claims 1-7.
10. A computer-readable medium, on which a computer program is stored, wherein the computer program, when being executed by a processor, carries out the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011273670.2A CN112365976B (en) | 2020-11-14 | 2020-11-14 | Composite disease species clinical path construction method and system based on transfer learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011273670.2A CN112365976B (en) | 2020-11-14 | 2020-11-14 | Composite disease species clinical path construction method and system based on transfer learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112365976A true CN112365976A (en) | 2021-02-12 |
| CN112365976B CN112365976B (en) | 2023-08-11 |
Family
ID=74514828
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011273670.2A Active CN112365976B (en) | 2020-11-14 | 2020-11-14 | Composite disease species clinical path construction method and system based on transfer learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112365976B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114743660A (en) * | 2022-05-20 | 2022-07-12 | 北京嘉和美康信息技术有限公司 | Clinical path maintenance method and device, electronic equipment and storage medium |
| CN114783597A (en) * | 2022-06-23 | 2022-07-22 | 瀚依科技(杭州)有限公司 | Method and device for diagnosing multi-class diseases, electronic equipment and storage medium |
| CN117809792B (en) * | 2024-02-28 | 2024-05-03 | 神州医疗科技股份有限公司 | Method and system for structuring disease seed data during cross-disease seed migration |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915560A (en) * | 2015-06-11 | 2015-09-16 | 万达信息股份有限公司 | Method for disease diagnosis and treatment scheme based on generalized neural network clustering |
| CN106021967A (en) * | 2016-07-12 | 2016-10-12 | 上海市浦东新区周浦医院 | Data processing method for clinical pathway quality evaluation |
| CN106339577A (en) * | 2016-08-15 | 2017-01-18 | 成都中医药大学 | Traditional Chinese medicine characteristic specialized medical treatment information method |
| CN108538349A (en) * | 2018-03-28 | 2018-09-14 | 深圳市前海安测信息技术有限公司 | Hospital clinical path based on clinical path variance analysis adjusts system and method |
| CN109192299A (en) * | 2018-08-13 | 2019-01-11 | 中国科学院计算技术研究所 | A kind of medical analysis auxiliary system based on convolutional neural networks |
| CN109727680A (en) * | 2018-12-28 | 2019-05-07 | 上海列顿信息科技有限公司 | A kind of region clinical path management system based on big data technology |
| CN109830303A (en) * | 2019-02-01 | 2019-05-31 | 上海众恒信息产业股份有限公司 | Clinical data mining analysis and aid decision-making method based on internet integration medical platform |
-
2020
- 2020-11-14 CN CN202011273670.2A patent/CN112365976B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915560A (en) * | 2015-06-11 | 2015-09-16 | 万达信息股份有限公司 | Method for disease diagnosis and treatment scheme based on generalized neural network clustering |
| CN106021967A (en) * | 2016-07-12 | 2016-10-12 | 上海市浦东新区周浦医院 | Data processing method for clinical pathway quality evaluation |
| CN106339577A (en) * | 2016-08-15 | 2017-01-18 | 成都中医药大学 | Traditional Chinese medicine characteristic specialized medical treatment information method |
| CN108538349A (en) * | 2018-03-28 | 2018-09-14 | 深圳市前海安测信息技术有限公司 | Hospital clinical path based on clinical path variance analysis adjusts system and method |
| CN109192299A (en) * | 2018-08-13 | 2019-01-11 | 中国科学院计算技术研究所 | A kind of medical analysis auxiliary system based on convolutional neural networks |
| CN109727680A (en) * | 2018-12-28 | 2019-05-07 | 上海列顿信息科技有限公司 | A kind of region clinical path management system based on big data technology |
| CN109830303A (en) * | 2019-02-01 | 2019-05-31 | 上海众恒信息产业股份有限公司 | Clinical data mining analysis and aid decision-making method based on internet integration medical platform |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114743660A (en) * | 2022-05-20 | 2022-07-12 | 北京嘉和美康信息技术有限公司 | Clinical path maintenance method and device, electronic equipment and storage medium |
| CN114783597A (en) * | 2022-06-23 | 2022-07-22 | 瀚依科技(杭州)有限公司 | Method and device for diagnosing multi-class diseases, electronic equipment and storage medium |
| CN114783597B (en) * | 2022-06-23 | 2022-11-29 | 瀚依科技(杭州)有限公司 | Method and device for diagnosing multi-class diseases, electronic equipment and storage medium |
| CN117809792B (en) * | 2024-02-28 | 2024-05-03 | 神州医疗科技股份有限公司 | Method and system for structuring disease seed data during cross-disease seed migration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112365976B (en) | 2023-08-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106934235B (en) | Patient's similarity measurement migratory system between a kind of disease areas based on transfer learning | |
| CN111316281B (en) | Semantic classification method and system for numerical data in natural language context based on machine learning | |
| US11037070B2 (en) | Diagnostic test planning using machine learning techniques | |
| CN110136837B (en) | Medical data processing platform | |
| US20130231953A1 (en) | Method, system and computer program product for aggregating population data | |
| US20180211008A1 (en) | Assist Selection of Provider/Facility for Surgical Procedures Based on Frequency of Procedure, History of Complications, and Cost | |
| CN112365976A (en) | Compound disease clinical path construction method and system based on transfer learning | |
| Mhamdi et al. | Artificial intelligence for cardiac diseases diagnosis and prediction using ECG images on embedded systems | |
| Al-Mualemi et al. | A deep learning-based sepsis estimation scheme | |
| Aslam et al. | Explainable computational intelligence model for antepartum fetal monitoring to predict the risk of IUGR | |
| WO2019220833A1 (en) | Diagnosis assistance system and diagnosis assistance device | |
| Baghdadi et al. | Reliable sarcoidosis detection using chest X-rays with efficientnets and stain-normalization techniques | |
| Andayeshgar et al. | Developing graph convolutional networks and mutual information for arrhythmic diagnosis based on multichannel ECG signals | |
| Mertes et al. | A deep learning approach for the assessment of signal quality of non-invasive foetal electrocardiography | |
| CN117292846A (en) | Construction method and device of intestinal microorganism knowledge graph | |
| Jatmiko et al. | A review of big data analytics in the biomedical field | |
| Boudet et al. | Use of Deep Learning to Detect the Maternal Heart Rate and False Signals on Fetal Heart Rate Recordings | |
| KR102072815B1 (en) | Classification method for state of health based on microbiome and classification apparatus | |
| CN112184295A (en) | Health maintenance service determination method and device, electronic equipment and storage medium | |
| Agoston | Big data, artificial intelligence, and machine learning in neurotrauma | |
| Yang et al. | Automatic classification method of arrhythmias based on 12-lead electrocardiogram | |
| Nair et al. | An exploratory study on big data processing: A case study from a biomedical informatics | |
| Bae et al. | Transfer learning for predicting conversion from mild cognitive impairment to Dementia of Alzheimer’s type based on 3D-convolutional neural network | |
| WO2023110477A1 (en) | A computer implemented method and a system | |
| CN115146179A (en) | Information recommendation method and device based on cross-domain medical data and computer equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |