CN117316457A - Intelligent management method and system for physiological index prediction in dialysis - Google Patents
Intelligent management method and system for physiological index prediction in dialysis Download PDFInfo
- Publication number
- CN117316457A CN117316457A CN202311372032.XA CN202311372032A CN117316457A CN 117316457 A CN117316457 A CN 117316457A CN 202311372032 A CN202311372032 A CN 202311372032A CN 117316457 A CN117316457 A CN 117316457A
- Authority
- CN
- China
- Prior art keywords
- data
- patient
- physiological
- dialysis
- prediction
- 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
- 238000000502 dialysis Methods 0.000 title claims abstract description 82
- 238000007726 management method Methods 0.000 title claims abstract description 20
- 238000012544 monitoring process Methods 0.000 claims abstract description 31
- 238000011282 treatment Methods 0.000 claims abstract description 21
- 238000000034 method Methods 0.000 claims abstract description 18
- 238000005457 optimization Methods 0.000 claims abstract description 18
- 238000013507 mapping Methods 0.000 claims abstract description 16
- 238000012502 risk assessment Methods 0.000 claims abstract description 11
- 201000010099 disease Diseases 0.000 claims abstract description 7
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 7
- 238000001631 haemodialysis Methods 0.000 claims description 16
- 230000000322 hemodialysis Effects 0.000 claims description 16
- 238000012800 visualization Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 13
- 239000008280 blood Substances 0.000 claims description 13
- 210000004369 blood Anatomy 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 230000003993 interaction Effects 0.000 claims description 10
- 230000001788 irregular Effects 0.000 claims description 10
- 238000004891 communication Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000004140 cleaning Methods 0.000 claims description 7
- 230000008030 elimination Effects 0.000 claims description 7
- 238000003379 elimination reaction Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 230000003068 static effect Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 230000036772 blood pressure Effects 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 claims description 5
- 239000003814 drug Substances 0.000 claims description 4
- 230000035945 sensitivity Effects 0.000 claims description 4
- 238000012706 support-vector machine Methods 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 230000017531 blood circulation Effects 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 claims description 3
- 229940079593 drug Drugs 0.000 claims description 3
- 238000013526 transfer learning Methods 0.000 claims description 2
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000035488 systolic blood pressure Effects 0.000 description 4
- 208000001953 Hypotension Diseases 0.000 description 3
- 230000004872 arterial blood pressure Effects 0.000 description 3
- 208000020832 chronic kidney disease Diseases 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000036543 hypotension Effects 0.000 description 3
- 206010070476 Haemodialysis complication Diseases 0.000 description 2
- DGAQECJNVWCQMB-PUAWFVPOSA-M Ilexoside XXIX Chemical compound C[C@@H]1CC[C@@]2(CC[C@@]3(C(=CC[C@H]4[C@]3(CC[C@@H]5[C@@]4(CC[C@@H](C5(C)C)OS(=O)(=O)[O-])C)C)[C@@H]2[C@]1(C)O)C)C(=O)O[C@H]6[C@@H]([C@H]([C@@H]([C@H](O6)CO)O)O)O.[Na+] DGAQECJNVWCQMB-PUAWFVPOSA-M 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000002457 bidirectional effect Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 229910052708 sodium Inorganic materials 0.000 description 2
- 239000011734 sodium Substances 0.000 description 2
- 230000004083 survival effect Effects 0.000 description 2
- 208000030090 Acute Disease Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- ZLMJMSJWJFRBEC-UHFFFAOYSA-N Potassium Chemical compound [K] ZLMJMSJWJFRBEC-UHFFFAOYSA-N 0.000 description 1
- XSQUKJJJFZCRTK-UHFFFAOYSA-N Urea Chemical compound NC(N)=O XSQUKJJJFZCRTK-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 208000007502 anemia Diseases 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000004202 carbamide Substances 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000004087 circulation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 208000028208 end stage renal disease Diseases 0.000 description 1
- 201000000523 end stage renal failure Diseases 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000003734 kidney Anatomy 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 229910052700 potassium Inorganic materials 0.000 description 1
- 239000011591 potassium Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000009256 replacement therapy Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
-
- 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
- G16H80/00—ICT specially adapted for facilitating communication between medical practitioners or patients, e.g. for collaborative diagnosis, therapy or health monitoring
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Epidemiology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- External Artificial Organs (AREA)
Abstract
The invention discloses a physiological index prediction intelligent management method and system in dialysis, which relate to the field of medical monitoring, and the method comprises the following steps: step 1, physiological data of a patient are acquired remotely in real time and are accessed into a hospital electronic information system; step 2, predicting future physiological characteristics of the patient by using a CNN-TIME-LSTM network model in combination with historical data of the patient; step 3, mapping future physiological characteristics into the probability of the interval [0,1] through a membership function, and predicting the complication disease risk of the patient; step 4, transmitting the complication risk to a hospital, remotely adjusting dialysis treatment parameters of a patient, and providing medical emergency rescue when necessary; and 5, comparing actual data of the patient with predicted disease risks, and adjusting model parameters suitable for the patient. The system comprises: the system comprises a multidimensional information feature selection module, a physiological sign prediction module in dialysis, a complication risk assessment module in dialysis, a remote terminal monitoring module and a prediction model optimization module.
Description
Technical Field
The invention relates to the field of medical monitoring, in particular to a physiological index prediction intelligent management method and system in dialysis.
Background
The number of patients undergoing dialysis for chronic kidney disease is continuously increasing at a rate of 5% per year worldwide, and the focus of research and study in the medical field has long been to overcome the problems of end-stage renal disease treatment and improve the quality of life of patients. Hemodialysis is a key way of kidney replacement therapy for patients with acute and chronic kidney disease. In recent years, home dialysis is gradually rising due to the problems of huge central hemodialysis cost, clinical resource tension, psychological stress caused by centralized dialysis, and the like. Meanwhile, according to researches, longer or more frequent dialysis at home, including night sleep dialysis, can obtain higher clinical benefit. Therefore, development of a home hemodialysis machine suitable for home safety dialysis is urgent. In order to further ensure the safety of dialysis without on-line supervision of medical staff, the household hemodialysis machine needs to monitor the abnormal trend of the physiological index of the patient in advance so as to predict possible complications and remotely transmit to the medical staff for remote management and control.
Cao Hui et al in China patent application "on-line monitoring device for blood permeation by fusing fuzzy neural network and rule type expert system" (publication number: CN 103984861B), use fuzzy logic, with neural network as core, use expert system in combination with actual condition of blood permeation path, realize the construction of complication diagnosis system of blood permeation instrument.
JINGSONG LI et al in patent application No. System for performing long-term hazard prediction on hemodialysis complications on basis of convolutional survival network (publication No. WO2022166158A 1) deal with multidimensional hemodialysis timing characteristics based on a convolved survival network, and a system for long-term risk prediction of hemodialysis complications.
Chen Jing et al in China patent application "remote dialysis monitoring System" (publication No. CN 112755284A) transmit acquired physiological data of patients to dialysis terminals, remotely control the work of the corresponding dialysis terminals, ensure the safety of dialysis treatment of home hemodialysis patients, and prevent serious consequences.
The related art now has the following disadvantages:
1. most of the prediction parameters in the dialysis prediction system only consider blood pressure prediction, but do not consider other necessary physiological parameters and treatment evaluation parameters, such as sodium blood concentration, potassium blood concentration, blood volume, urea removal rate, etc.
2. The existing physiological parameter prediction method directly maps the input physiological signals into output without explicitly modeling potential time dependence, so that the prediction accuracy of the model is low and safe remote monitoring cannot be realized.
3. The data communication interface of the conventional dialysis remote monitoring system is single, and the system is not really connected to an electronic information platform of a hospital, and different information systems adopted by different hospitals can have different architectures and heterogeneous data during interconnection.
4. The dialysis physiological index prediction system, the dialysis remote monitoring system and the dialysis parameter control system do not form a closed loop, and the coupling degree of the system is low.
Accordingly, those skilled in the art have been directed to developing a new method and system for managing physiological indicators in dialysis that addresses the above-identified deficiencies in the art.
Disclosure of Invention
In view of the above-mentioned drawbacks of the prior art, the present invention aims to solve the problems of single prediction parameters and data communication interfaces, lack of consideration of time dependence and low coupling degree of the whole system existing in the existing dialysis prediction system.
In order to achieve the above purpose, the invention provides a physiological index prediction intelligent management method in dialysis, which comprises the following steps:
step 1, physiological data of a patient are acquired remotely in real time and are accessed into a hospital electronic information system;
step 2, predicting future physiological characteristics of the patient by using a CNN-TIME-LSTM network model based on irregular TIME intervals in combination with historical data of the patient;
step 3, mapping the future physiological characteristics into the probability of interval [0,1] through a membership function, and predicting the complication disease risk of the patient;
step 4, transmitting the complication risk to a hospital, remotely adjusting dialysis treatment parameters of the patient, and providing medical emergency rescue when necessary;
and 5, comparing the actual physiological data of the patient with the predicted future physiological characteristics and the complication disease risk, and adjusting model parameters suitable for the patient according to the comparison result.
Further, the CNN-TIME-LSTM network model in the step 2 includes:
a CNN layer capturing a dependency of local static features from the patient data;
LSTM layer, which incorporates time gate T t The time difference delta between the data is calculated t ＝t m+1 -t m Into the model, whereby said time gate T t Input gate i t Forgetting door f t Status update c t Output door o t Output h t Is rewritten as:
T t ＝σ t (x t W xt +σ Δt (Δ t W tt )+b t )
h t ＝o t tanh(c t )
wherein sigma t 、σ Δt 、σ f 、σ c 、σ h Representing the activation functions selected by different gates, W represents the weight parameters of the different gates, b represents the offset parameters of the different gates, h t-1 The output of the last time period is indicated,representing the product between the two matrices.
Further, the physiological data acquired in real time in the step 1 comprises individual characteristic data similar to blood pressure, pulse, blood temperature and blood flow; the history data in the step 2 includes medication data, diagnosis data, examination data, and medical result data, which are acquired from the hospital electronic information system according to the ID of the patient; at least 70 features are included in the physiological data and the historical data to form a feature set.
Further, the collected physiological data is subjected to preliminary examination, null values are deleted, data bars containing obvious errors are preliminarily corrected, a support vector machine is used as a filter based on recursive feature elimination, the importance of each feature in the feature set is calculated in a recursive mode, the feature with the lowest weight is removed, the feature set is updated, and feature selection is completed; and finally, mapping the updated features in the feature set to a fixed [0,1] interval.
Further, in said step 4, data is stored in a shared mode in combination with a blockchain-based chain up-and-down storage, wherein said physiological data and said historical data of said patient are stored under the chain, and a data sharing log and personalized model parameters of a specific patient are stored on the chain;
meanwhile, the dialysis treatment data are classified according to sensitivity, encrypted by using a ciphertext policy attribute-based hierarchical encryption algorithm, and different access control policies are set on the basis of an Ethernet intelligent contract.
The invention also provides a physiological index prediction intelligent management system in dialysis, which comprises the following steps:
the multi-dimensional information feature selection module comprises a step of cleaning and preprocessing physiological data of a patient acquired in real time, and a step of accessing the preprocessed physiological data into a hospital electronic information system;
the in-dialysis physiological sign prediction module is connected with the multi-dimensional information feature selection module, acquires the preprocessed physiological data, and predicts future physiological features of the patient by using a CNN-TIME-LSTM network model based on irregular TIME intervals in combination with the historical data of the patient;
the in-dialysis complication risk assessment module is connected with the in-dialysis physiological sign prediction module, and predicts the complication risk of the patient by mapping the input future physiological features into the probability of intervals [0,1] through membership functions according to a pre-designed reasoning rule;
the remote terminal monitoring module is connected with the complication risk assessment module in dialysis, transmits the complication risk to a hospital, remotely adjusts dialysis treatment parameters of the patient, and provides medical emergency rescue when necessary; the remote terminal monitoring module comprises a 5G transmission service, a blockchain storage encryption service and a man-machine interaction service;
and the prediction model optimization module is based on a FedMD framework, performs model retraining according to domain differences between the CNN-TIME-LSTM network model and data accumulated by the patient for a long TIME, and adjusts model parameters suitable for a specific patient on private data of the specific patient.
Further, the cleaning the physiological data in the multidimensional information feature selection module includes: re-examining and checking the physiological data, deleting repeated information, correcting errors, and ensuring data consistency;
preprocessing the physiological data includes: performing feature selection, performing correlation map analysis on the multi-dimensional features to eliminate co-linear features, realizing recursive feature elimination based on an RFECV method, omitting features with small influence on a prediction target, and reducing the model scale under the condition of not influencing the precision; feature scaling is performed to map values to a fixed [0,1] range, speeding up the random gradient descent rate of model training.
Further, the CNN-TIME-LSTM network model in the in-dialysis physiological sign prediction module comprises:
a CNN layer capturing a dependency of local static features from the patient data;
LSTM layer, which incorporates time gate T t The time difference delta between the data is calculated t ＝t m+1 -t m Into the model, whereby said time gate T t Input gate i t Forgetting door f t Status update c t Output door o t Output h t Is rewritten as:
T t ＝σ t (x t W xt +σ Δt (Δ t W tt )+b t )
h t ＝o t tanh(c t )
wherein sigma t 、σ Δt 、σ f 、σ c 、σ h Representing the activation functions selected by different gates, W represents the weight parameters of the different gates, b represents the offset parameters of the different gates, h t-1 The output of the last time period is indicated,representing the product between the two matrices.
Further, the 5G transmission service in the remote terminal monitoring module uses a 4G/5G/WIFI network as a communication technology; the blockchain storage encryption service adopts a sharing mode storage data based on the combination of the blockchain-based up-chain and down-chain storage, wherein the physiological data and the historical data of the patient are stored under the chain, the data sharing log and the personalized model parameters of the specific patient are stored on the chain, and the access control of the patient data is realized based on a ciphertext policy attribute-based hierarchical encryption algorithm; the man-machine interaction service comprises a visualization of monitoring data, a visualization of a prediction result, a visualization of a complication risk and a remote control interface of the hemodialysis machine, wherein the visualization of the monitoring data, the visualization of the prediction result and the visualization of the complication risk are realized on the man-machine interaction APP.
Further, in the predictive model optimization module, when the private data D of the particular patient private When enough, the private data D can be obtained based on transfer learning private Training to obtain a new mapping relation f k The method comprises the steps of carrying out a first treatment on the surface of the The prediction model optimization module optimizes the initial shared data set D share Input to the new mapping relation f k Average score is obtained by averagingAs a global consensus; the predictive model optimization module first uses blockchain to base the shared dataset D on the patient-specific chain share Learning the global consensus->Causing the new mapping relation f k First approach to the global consensus->And then based on the private data D private Retraining optimization f k Finally, a physiological index prediction model adapted to the specific patient is obtained.
The intelligent management method and system for predicting the physiological index in dialysis provided by the invention have at least the following technical effects:
the management method and the management system provided by the invention can be embedded and applied to a home hemodialysis machine, ensure the dialysis sufficiency and safety of a home hemodialysis machine, realize high-precision physiological index prediction and risk probability assessment of complications, and simultaneously connect a blood purification center of a hospital so as to facilitate remote management and control of medical staff, thereby further ensuring the safety of home dialysis.
The conception, specific structure, and technical effects of the present invention will be further described with reference to the accompanying drawings to fully understand the objects, features, and effects of the present invention.
Drawings
FIG. 1 is a schematic diagram of a system flow according to a preferred embodiment of the present invention;
FIG. 2 is a diagram of the CNN-TIME-LSTM network in the embodiment shown in FIG. 1;
FIG. 3 is a LSTM internal structure diagram of the CNN-TIME-LSTM network join TIME gate of FIG. 2;
fig. 4 is a schematic diagram of a remote terminal monitoring module in the embodiment shown in fig. 1.
Detailed Description
The following description of the preferred embodiments of the present invention refers to the accompanying drawings, which make the technical contents thereof more clear and easy to understand. The present invention may be embodied in many different forms of embodiments and the scope of the present invention is not limited to only the embodiments described herein.
The embodiment of the invention provides a physiological index prediction intelligent management method and system in dialysis. And secondly, sending the data subjected to the feature extraction into a physiological sign prediction module in dialysis to output a physiological information prediction result of the patient after a period of time in the future. And then, the complication risk assessment module in dialysis is combined with the patient history data and the future physiological information prediction result to assess the incidence probability of the complications and is transmitted to the remote terminal monitoring module, so that medical staff can prescribe more proper dialysis according to the transmission result, and the safety of home dialysis is ensured. Finally, under the condition that the data volume is large enough, the prediction model optimization module can timely adjust model parameters suitable for individuals of patients according to errors of physiological data prediction results and actually monitored data, so that model prediction accuracy is improved, and personalized dialysis prediction is realized.
Example 1
As shown in fig. 1, an embodiment of the present invention provides a method for intelligently managing physiological index prediction in dialysis, which includes the following steps:
step 1, physiological data of a patient are acquired remotely in real time and are accessed into a hospital electronic information system;
step 2, predicting future physiological characteristics of the patient by using a CNN-TIME-LSTM network model based on irregular TIME intervals in combination with historical data of the patient;
step 3, mapping future physiological characteristics into the probability of the interval [0,1] through a membership function, and predicting the complication disease risk of the patient;
step 4, transmitting the complication risk to a hospital, remotely adjusting dialysis treatment parameters of a patient, and providing medical emergency rescue when necessary;
and 5, comparing the actual physiological data of the patient with predicted future physiological characteristics and complications and risks, and adjusting model parameters suitable for the patient according to the comparison result.
In particular, a step 1 comprises a step of washing and preprocessing the acquired physiological data. And the data cleaning is to recheck and check the data, delete repeated information, correct the existing errors and ensure the consistency of the data. The data preprocessing comprises the following steps: performing feature selection, performing correlation map analysis on the multi-dimensional features to eliminate co-linear features, realizing recursive feature elimination based on an RFECV method, omitting features with small influence on a prediction target, and reducing the model scale under the condition of not influencing the precision; feature scaling is performed to map values to a fixed [0,1] range, speeding up the random gradient descent rate of model training.
In particular, the acquired home dialysis patient data includes five major categories of individual characteristic data, medication data, diagnostic data, examination data, and medical outcome data, with at least 70 characteristics. The individual characteristic data are physiological data of a patient, such as blood pressure, pulse, blood temperature, blood flow and the like, continuously collected by the dialysis machine intelligent sensor network in real time. Other data is obtained from the hospital electronic information system based on the patient ID. Because the number of the features is too large, in order to remove irrelevant features, reduce the risk of overfitting and reduce the dimension disaster, the feature set needs to be selected and divided, so that the number of key features is reduced, the accuracy of a model is improved, and the running time is reduced.
In particular, the multidimensional information feature selection module performs a preliminary review of the collected data, removes null values and preliminarily corrects data strips containing significant errors.
In particular, the multidimensional information feature selection module uses a support vector machine as a filter based on recursive feature elimination, calculates the importance of each feature, removes the feature with the lowest weight, and updates the feature set. The filter then performs a new round of training learning and rating calculations on the feature set until the ratings of all feature importance are recursively completed. And the support vector machine filter performs cross verification on each feature subset, and outputs the importance degree sequence of each feature and the number of the optimal features to complete feature selection.
In particular, the multidimensional information feature selection module maps the feature with non-uniform numerical range to a fixed [0,1] interval, so that the random gradient descent rate of model training is accelerated.
Example 2
On the basis of the embodiment 1, since the time interval for monitoring and uploading the physiological data of the patient is not fixed, while the traditional RNN circulation time network can realize modeling of the time sequence, the influence of the time interval on the result is not considered, the data sequence with fixed time interval is better represented, and the prediction task of uneven time interval is lower represented. Therefore, a CNN-TIME-LSTM network structure based on irregular TIME intervals is adopted.
The CNN-TIME-LSTM network structure based on irregular TIME intervals, wherein a local CNN layer captures the dependency relationship of local static characteristics, the LSTM layer models dynamic TIME information of irregular trend in TIME sequence components, a TIME gate mechanism is introduced to solve the problem of unfixed TIME intervals of TIME sequence data, and finally regression output is obtained through full-connection layer prediction, so that accurate prediction of future TIME physiological characteristics of a patient is realized, wherein the physiological characteristics comprise blood pressure, sodium concentration, relative blood volume and the like.
As shown in fig. 2, the CNN-TIME-LSTM network design procedure based on irregular TIME intervals is as follows. Firstly, introducing a CNN layer, and utilizing alternation of a convolution layer and an activation layer to realize effective extraction of local features of data and reduction of local feature dimensions, so as to reduce the number of weights and complexity of a model, namely capturing the dependency relationship of local static features from the patient data. As shown in FIG. 3, a time gate T is added to the LSTM layer t The time difference delta between the data is calculated t ＝t m+1 -t m Into the model, whereby said time gate T t Input gate i t Forgetting door f t Shape and shape ofState update c t Output door o t Output h t Is rewritten as:
T t ＝σ t (x t W xt +σ Δt (Δ t W tt )+b t )
h t ＝o t tanh(c t )
wherein sigma t 、σ Δt 、σ f 、σ c 、σ h Representing the activation functions selected by different gates, W represents the weight parameters of the different gates, b represents the offset parameters of the different gates, h t-1 The output of the last time period is indicated,representing the product between the two matrices.
In particular, the data selects three time windows (total duration of at least 30 minutes) as input data, and outputs physiological characteristic prediction results for the next time window. The training data was converted into a time window training set (70%), a validation set (10%) and a test set (20%), and the predictive ability of the model was evaluated using Root Mean Square Error (RMSE), mean Absolute Error (MAE) and Mean Absolute Percent Error (MAPE).
Example 3
Based on embodiments 1 and 2, step 3 takes the output of step 2 as input, designs inference rules according to medical prior knowledge, maps the input future physiological characteristics to the probability of interval [0,1] through membership functions, and reflects the possibility of the complication, wherein the type of the complication comprises hypotension, hypertension, unbalanced syndrome, anemia and the like in dialysis.
Taking hypotension as an example, three types of hypotension in dialysis are defined in medicine: a systolic blood pressure <90mmHg; 1 hour systolic blood pressure decrease from the start time > =20 mmHg or mean arterial blood pressure decrease > =10 mmHg; 1 hour systolic blood pressure decrease > = 20mmHg or mean arterial blood pressure decrease > = 10mmHg from the predicted time. However, due to patient body specificity, clinical definitions may vary somewhat, and thus the complication risk assessment module will take into account the patient's physical state self-assessment. The method comprises the following specific steps: the complication risk assessment module compares the predicted systolic pressure or mean arterial pressure with the historical value, and carries out fuzzy reasoning assessment by combining the self-evaluation state of the body fed back by the patient through the remote terminal monitoring module according to the clinical definition of the complication to obtain the illness probability and map the illness probability to the [0,1] interval, wherein the numerical value is closer to 1, so that the illness risk is larger.
In step 4, storing data in a sharing mode based on combination of chain up-chain and down-chain storage of a blockchain, wherein physiological data and historical data of a patient are stored under the chain, and a data sharing log and personalized model parameters of a specific patient are stored on the chain; meanwhile, the dialysis treatment data are classified according to sensitivity, encrypted by using a ciphertext policy attribute-based hierarchical encryption algorithm, and different access control policies are set on the basis of an Ethernet intelligent contract.
Example 4
The embodiment of the invention provides a physiological index prediction intelligent management system in dialysis, which comprises any one of the physiological index prediction intelligent management methods in dialysis provided in embodiments 1 to 3, and further comprises the following steps:
the multidimensional information feature selection module comprises a step of cleaning and preprocessing physiological data of a patient acquired in real time, and a step of accessing the preprocessed physiological data into a hospital electronic information system;
the physiological sign prediction module in dialysis is connected with the multidimensional information feature selection module, acquires preprocessed physiological data, and predicts future physiological features of a patient by combining historical data of the patient and using a CNN-TIME-LSTM network model based on irregular TIME intervals; wherein the CNN-TIME-LSTM network model is identical to the model in example 2;
the in-dialysis complication risk assessment module is connected with the in-dialysis physiological sign prediction module, maps the input future physiological features into the probability of interval [0,1] through the membership function according to a pre-designed reasoning rule, and predicts the complication risk of the patient;
the remote terminal monitoring module is connected with the complication risk assessment module in dialysis, transmits the complication risk to a hospital, remotely adjusts dialysis treatment parameters of a patient, and provides medical emergency rescue when necessary; the remote terminal monitoring module comprises a 5G transmission service, a blockchain storage encryption service and a man-machine interaction service, as shown in fig. 4;
the prediction model optimization module is based on a FedMD framework, performs model retraining according to domain differences between a CNN-TIME-LSTM network model and data accumulated by a patient for a long TIME, and adjusts model parameters suitable for the specific patient on private data of the specific patient.
In particular, the cleaning of physiological data in the multidimensional information feature selection module includes: rechecking and checking the physiological data, deleting repeated information, correcting errors, and ensuring data consistency; preprocessing physiological data includes: performing feature selection, performing correlation map analysis on the multi-dimensional features to eliminate co-linear features, realizing recursive feature elimination based on an RFECV method, omitting features with small influence on a prediction target, and reducing the model scale under the condition of not influencing the precision; feature scaling is performed to map values to a fixed [0,1] range, speeding up the random gradient descent rate of model training.
Example 5
Based on embodiment 4, the remote terminal monitoring module in step S4 includes a 5G transmission service, a blockchain storage encryption service, and a man-machine interaction service. The 5G transmission service uses a 4G/5G/WIFI network as a communication technology, so that real-time bidirectional transmission of patient vital signs and medical staff decision data is effectively ensured. The block chain storage encryption service adopts a shared mode storage data combining medical data block chain up-chain and down-chain storage, and realizes access control on hemodialysis data of a patient based on a ciphertext policy attribute-based hierarchical encryption algorithm. The module relates to man-machine interaction APP, and comprises visualization of monitoring data, visualization of a prediction result, visualization of complication risk and a remote control interface of a hemodialysis machine.
Particularly, the remote terminal monitoring module is compatible with electronic information systems of different hospitals, risk probability obtained in the complication risk assessment module in dialysis is transmitted to a blood purification center where a patient is located, medical staff can prescribe more proper dialysis according to transmission results, a household hemodialysis instrument of the patient is controlled remotely, and medical institutions are called for rescue before necessary, so that the completeness and safety of dialysis are guaranteed.
In particular, the remote terminal monitoring module preferably adopts 5G as the communication technology, and then selects 4G, WIFI as the communication technology. The 5G network has the characteristics of ultra-low time delay, mass connection and ultra-large bandwidth, and can effectively ensure real-time bidirectional transmission of patient vital signs and medical personnel decision data.
In particular, there are differences in information systems used by hospitals at different periods and stages, and currently, the mainstream hospital information systems include a hospital information system HIS, a laboratory information system LIS, a clinical information system CIS, and the like. In order to prevent the data heterogeneous problem possibly occurring in data transmission, a multi-interface is designed to realize the compatibility of a remote terminal monitoring module and different hospital electronic information systems, and based on the design of an enterprise service bus ESB, an application service assembly between an assembly layer and a Web service layer is constructed to realize the conversion of mainstream communication protocols such as HTTP/HTTPS, XML, DICOM, JMS, HL7 and the like.
In particular, the remote terminal monitoring module adopts a hemodialysis data sharing platform based on a blockchain to realize information sharing and information transmission transactions of the household hemodialysis instrument and the hospital electronic information system. The data storage adopts a shared mode of combining up-chain and down-chain to store data, the original dialysis treatment data and the hospital data of the patient are stored under the chain, and the data sharing log and the personalized model parameters of the specific patient are stored on the chain. Meanwhile, the dialysis treatment data are classified according to sensitivity, the ciphertext policy attribute-based hierarchical encryption algorithm is used for encrypting the patient data, the access control policy of the medical data is set based on the Ethernet intelligent contract, the information of the data and the corresponding encryption and decryption information are deployed on the blockchain, fine granularity access control of the patient and medical staff on the medical data is ensured, and a hospital blood purification center can control access and modification of specific staff on the dialysis treatment data.
Particularly, the remote terminal monitoring module is designed to comprise an intelligent interaction platform with functions of monitoring visualization, dialysis controllability and the like, the client uses NW.js or Electron as a basic development frame, spans multiple platforms such as a tablet computer, a mobile phone and the like, integrates the Html/JS language to complete the development of a UI interface and design an interface of a blockchain data sharing platform, and realizes the functions of visual display of patient data, risk early warning prompt, active treatment course management of doctors and patient feedback callback in the hemodialysis treatment process.
Example 6
Based on embodiments 4 and 5, in the case that the data amount is large enough, the prediction model optimization module can timely adjust model parameters suitable for the individual patient according to the error between the prediction result of the physiological data and the actually monitored data.
Particularly, the prediction model optimization module is based on a FedMD framework, retrains according to the domain difference between the trained global source model and the data accumulated by a specific patient for a long time, and fine-adjusts the physiological index prediction model on private data so as to improve model precision and individuation degree.
When private data D of a specific patient private When enough, the private data D can be obtained based on the migration learning private Training to obtain a new mapping relation f k The method comprises the steps of carrying out a first treatment on the surface of the The prediction model optimization module optimizes the initial shared data set D share Input to newMapping relation f of (2) k Average score is obtained by averagingAs a global consensus; the predictive model optimization module first uses blockchain to base the shared dataset D on a patient-specific chain share Learning global consensus->Make new mapping relation f k First approach to global consensus->Then based on private data D private Retraining optimization f k Finally, a physiological index prediction model suitable for a specific patient is obtained, individuation and high-precision prediction are realized, and an index with higher reliability is provided for future complication risk probability evaluation.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention without requiring creative effort by one of ordinary skill in the art. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.
Claims (10)
1. The intelligent management method for predicting the physiological index in dialysis is characterized by comprising the following steps of:
step 1, physiological data of a patient are acquired remotely in real time and are accessed into a hospital electronic information system;
step 2, predicting future physiological characteristics of the patient by using a CNN-TIME-LSTM network model based on irregular TIME intervals in combination with historical data of the patient;
step 3, mapping the future physiological characteristics into the probability of interval [0,1] through a membership function, and predicting the complication disease risk of the patient;
step 4, transmitting the complication risk to a hospital, remotely adjusting dialysis treatment parameters of the patient, and providing medical emergency rescue when necessary;
and 5, comparing the actual physiological data of the patient with the predicted future physiological characteristics and the complication disease risk, and adjusting model parameters suitable for the patient according to the comparison result.
2. The method for intelligently managing physiological index prediction in dialysis according to claim 1, wherein said CNN-TIME-LSTM network model in step 2 comprises:
a CNN layer capturing a dependency of local static features from the patient data;
LSTM layer, which incorporates time gate T t The time difference delta between the data is calculated t ＝t m+1 -t m Into the model, whereby said time gate T t Input gate i t Forgetting door f t Status update c t Output door o t Output h t Is rewritten as:
T t ＝σ t (x t W xt +σ Δt (Δ t W tt )+b t )
h t ＝o t tanh(c t )
wherein sigma t 、σ Δt 、σ f 、σ c 、σ h Representing the activation functions selected by different gates, W represents the weight parameters of the different gates, b represents the offset parameters of the different gates, h t-1 The output of the last time period is indicated,representing the product between the two matrices.
3. The method for intelligently managing physiological index prediction in dialysis according to claim 1, wherein the physiological data collected in real time in step 1 includes individual characteristic data similar to blood pressure, pulse, blood temperature, blood flow; the history data in the step 2 includes medication data, diagnosis data, examination data, and medical result data, which are acquired from the hospital electronic information system according to the ID of the patient; at least 70 features are included in the physiological data and the historical data to form a feature set.
4. The method for intelligently managing physiological index prediction in dialysis according to claim 3, wherein the collected physiological data is subjected to preliminary examination, null values are deleted, data bars containing obvious errors are preliminarily corrected, a support vector machine is used as a filter based on recursive feature elimination, the importance of each feature in the feature set is calculated in a recursive manner, the feature with the lowest weight is removed, the feature set is updated, and feature selection is completed; and finally, mapping the updated features in the feature set to a fixed [0,1] interval.
5. The method of claim 1, wherein in step 4, data is stored in a shared mode based on a combination of blockchain-based on-chain off-chain storage, wherein the physiological data and the historical data of the patient are stored off-chain, and wherein data sharing logs and personalized model parameters of a specific patient are stored on-chain;
meanwhile, the dialysis treatment data are classified according to sensitivity, encrypted by using a ciphertext policy attribute-based hierarchical encryption algorithm, and different access control policies are set on the basis of an Ethernet intelligent contract.
6. An intelligent management system for predicting physiological indexes in dialysis is characterized by comprising:
the multi-dimensional information feature selection module comprises a step of cleaning and preprocessing physiological data of a patient acquired in real time, and a step of accessing the preprocessed physiological data into a hospital electronic information system;
the in-dialysis physiological sign prediction module is connected with the multi-dimensional information feature selection module, acquires the preprocessed physiological data, and predicts future physiological features of the patient by using a CNN-TIME-LSTM network model based on irregular TIME intervals in combination with the historical data of the patient;
the in-dialysis complication risk assessment module is connected with the in-dialysis physiological sign prediction module, and predicts the complication risk of the patient by mapping the input future physiological features into the probability of intervals [0,1] through membership functions according to a pre-designed reasoning rule;
the remote terminal monitoring module is connected with the complication risk assessment module in dialysis, transmits the complication risk to a hospital, remotely adjusts dialysis treatment parameters of the patient, and provides medical emergency rescue when necessary; the remote terminal monitoring module comprises a 5G transmission service, a blockchain storage encryption service and a man-machine interaction service;
and the prediction model optimization module is based on a FedMD framework, performs model retraining according to domain differences between the CNN-TIME-LSTM network model and data accumulated by the patient for a long TIME, and adjusts model parameters suitable for a specific patient on private data of the specific patient.
7. The in-dialysis physiological index prediction intelligent management system of claim 6, wherein the cleaning of the physiological data in the multi-dimensional information feature selection module comprises: re-examining and checking the physiological data, deleting repeated information, correcting errors, and ensuring data consistency;
preprocessing the physiological data includes: performing feature selection, performing correlation map analysis on the multi-dimensional features to eliminate co-linear features, realizing recursive feature elimination based on an RFECV method, omitting features with small influence on a prediction target, and reducing the model scale under the condition of not influencing the precision; feature scaling is performed to map values to a fixed [0,1] range, speeding up the random gradient descent rate of model training.
8. The in-dialysis physiological index prediction intelligent management system of claim 6, wherein the CNN-TIME-LSTM network model in the in-dialysis physiological index prediction module comprises:
a CNN layer capturing a dependency of local static features from the patient data;
LSTM layer, which incorporates time gate T t The time difference delta between the data is calculated t ＝t m+1 -t m Into the model, whereby said time gate T t Input gate i t Forgetting door f t Status update c t Output door o t Output h t Is rewritten as:
T t ＝σ t (x t W xt +σ Δt (Δ t W tt )+b t )
h t ＝o t tanh(c t )
wherein sigma t 、σ Δt 、σ f 、σ c 、σ h Representing the activation functions selected by different gates, W represents the weight parameters of the different gates, b represents the offset parameters of the different gates, h t-1 The output of the last time period is indicated,representing the product between the two matrices.
9. The in-dialysis physiological index prediction intelligent management system of claim 6, wherein said 5G transmission service in said remote terminal monitoring module uses a 4G/5G/WIFI network as a communication technology; the blockchain storage encryption service adopts a sharing mode storage data based on the combination of the blockchain-based up-chain and down-chain storage, wherein the physiological data and the historical data of the patient are stored under the chain, the data sharing log and the personalized model parameters of the specific patient are stored on the chain, and the access control of the patient data is realized based on a ciphertext policy attribute-based hierarchical encryption algorithm; the man-machine interaction service comprises a visualization of monitoring data, a visualization of a prediction result, a visualization of a complication risk and a remote control interface of the hemodialysis machine, wherein the visualization of the monitoring data, the visualization of the prediction result and the visualization of the complication risk are realized on the man-machine interaction APP.
10. The in-dialysis physiological index prediction intelligent of claim 6Management system, characterized in that in said predictive model optimization module, when said private data D of said specific patient private When enough, the private data D can be obtained based on transfer learning private Training to obtain a new mapping relation f k The method comprises the steps of carrying out a first treatment on the surface of the The prediction model optimization module optimizes the initial shared data set D share Input to the new mapping relation f k Average score is obtained by averagingAs a global consensus; the predictive model optimization module first uses blockchain to base the shared dataset D on the patient-specific chain share Learning the global consensus->Causing the new mapping relation f k First approach to the global consensus->And then based on the private data D private Retraining optimization f k Finally, a physiological index prediction model adapted to the specific patient is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311372032.XA CN117316457A (en) | 2023-10-23 | 2023-10-23 | Intelligent management method and system for physiological index prediction in dialysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311372032.XA CN117316457A (en) | 2023-10-23 | 2023-10-23 | Intelligent management method and system for physiological index prediction in dialysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117316457A true CN117316457A (en) | 2023-12-29 |
Family
ID=89237235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311372032.XA Pending CN117316457A (en) | 2023-10-23 | 2023-10-23 | Intelligent management method and system for physiological index prediction in dialysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117316457A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117637152A (en) * | 2024-01-17 | 2024-03-01 | 中国人民解放军总医院 | Method and system for predicting sodium blood fluctuation |
| CN117936054A (en) * | 2024-03-25 | 2024-04-26 | 四川互慧软件有限公司 | Emergency emergency treatment quality control index flexible display statistical method based on big data |
-
2023
- 2023-10-23 CN CN202311372032.XA patent/CN117316457A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117637152A (en) * | 2024-01-17 | 2024-03-01 | 中国人民解放军总医院 | Method and system for predicting sodium blood fluctuation |
| CN117936054A (en) * | 2024-03-25 | 2024-04-26 | 四川互慧软件有限公司 | Emergency emergency treatment quality control index flexible display statistical method based on big data |
| CN117936054B (en) * | 2024-03-25 | 2024-05-17 | 四川互慧软件有限公司 | Emergency emergency treatment quality control index flexible display statistical method based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117316457A (en) | Intelligent management method and system for physiological index prediction in dialysis | |
| US20200221990A1 (en) | Systems and methods for assessing and evaluating renal health diagnosis, staging, and therapy recommendation | |
| Lasierra et al. | A three stage ontology-driven solution to provide personalized care to chronic patients at home | |
| CN109119130A (en) | A kind of big data based on cloud computing is health management system arranged and method | |
| CN110111886A (en) | A kind of intelligent interrogation system and method based on XGBoost disease forecasting | |
| US20200402662A1 (en) | System for integrating data for clinical decisions | |
| Basha et al. | A soft computing approach to provide recommendation on PIMA diabetes | |
| Habib et al. | A neutrosophic clinical decision-making system for cardiovascular diseases risk analysis | |
| Gupta et al. | A novel approach based on computing with words for monitoring the heart failure patients | |
| US20230053474A1 (en) | Medical care system for assisting multi-diseases decision-making and real-time information feedback with artificial intelligence technology | |
| Rghioui et al. | Diabetic patients monitoring and data classification using IoT application | |
| Choi et al. | Longitudinal Healthcare Data Management Platform of Healthcare IoT Devices for Personalized Services. | |
| US20210193318A1 (en) | System for increased data optimization for genetic data | |
| KR102033696B1 (en) | System for predicting thyrotoxicosis using a wearable device and computer program using the same | |
| CN117475447A (en) | Intelligent severe infection early warning and auxiliary decision making system based on multi-modal data fusion federal learning | |
| US20150234997A1 (en) | Task optimization in remote health monitoring systems | |
| US11915812B2 (en) | System for integrating data for clinical decisions including multiple engines | |
| US11810652B1 (en) | Computer decision support for determining surgery candidacy in stage four chronic kidney disease | |
| Neves et al. | Causal dynamic Bayesian networks for the management of glucose control in gestational diabetes | |
| Teixeira et al. | A goal-based framework for supporting medical assistance: The case of chronic diseases | |
| Zhang | Real-time analysis of physiological data and development of alarm algorithms for patient monitoring in the intensive care unit | |
| US20240120107A1 (en) | Data processing system and method for predicting a score representative of a probability of a sepsis for a patient | |
| Mittal | Efficiency Improvement of IoT in Healthcare Using Embedded Fuzzy Technique | |
| Lu et al. | A Deep Learning Approach of Blood Glucose Predictive Monitoring for Women with Gestational Diabetes | |
| Kumar et al. | Fuzzy logic applications in healthcare: A review-based study |
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 |