CN117457192A - Intelligent remote diagnosis method and system - Google Patents
Intelligent remote diagnosis method and system Download PDFInfo
- Publication number
- CN117457192A CN117457192A CN202311774882.2A CN202311774882A CN117457192A CN 117457192 A CN117457192 A CN 117457192A CN 202311774882 A CN202311774882 A CN 202311774882A CN 117457192 A CN117457192 A CN 117457192A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- disease
- generating
- graph
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000004171 remote diagnosis Methods 0.000 title claims abstract description 24
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 140
- 201000010099 disease Diseases 0.000 claims abstract description 134
- 238000005516 engineering process Methods 0.000 claims abstract description 85
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 70
- 230000036541 health Effects 0.000 claims abstract description 53
- 238000000605 extraction Methods 0.000 claims abstract description 44
- 238000011282 treatment Methods 0.000 claims abstract description 34
- 238000003745 diagnosis Methods 0.000 claims abstract description 33
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 22
- 238000012800 visualization Methods 0.000 claims abstract description 12
- 238000003909 pattern recognition Methods 0.000 claims abstract description 7
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 claims abstract 4
- 238000004458 analytical method Methods 0.000 claims description 43
- 238000012545 processing Methods 0.000 claims description 37
- 238000013135 deep learning Methods 0.000 claims description 27
- 230000006978 adaptation Effects 0.000 claims description 26
- 238000010606 normalization Methods 0.000 claims description 25
- 238000012098 association analyses Methods 0.000 claims description 24
- 238000005065 mining Methods 0.000 claims description 22
- 238000012549 training Methods 0.000 claims description 21
- 230000004927 fusion Effects 0.000 claims description 19
- 238000010586 diagram Methods 0.000 claims description 17
- 238000005457 optimization Methods 0.000 claims description 17
- 208000024891 symptom Diseases 0.000 claims description 17
- 238000013526 transfer learning Methods 0.000 claims description 17
- 238000004140 cleaning Methods 0.000 claims description 15
- 238000013507 mapping Methods 0.000 claims description 15
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000001514 detection method Methods 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 13
- 238000012300 Sequence Analysis Methods 0.000 claims description 12
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000007689 inspection Methods 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 238000007418 data mining Methods 0.000 claims description 8
- 238000013499 data model Methods 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 8
- 238000013461 design Methods 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 7
- 238000010801 machine learning Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000011425 standardization method Methods 0.000 claims description 6
- 238000011269 treatment regimen Methods 0.000 claims description 6
- 238000002790 cross-validation Methods 0.000 claims description 5
- 230000010354 integration Effects 0.000 claims description 5
- 230000007787 long-term memory Effects 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000001133 acceleration Effects 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 230000002452 interceptive effect Effects 0.000 claims description 4
- 230000008376 long-term health Effects 0.000 claims description 4
- 230000006403 short-term memory Effects 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 4
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 2
- 230000002787 reinforcement Effects 0.000 claims 1
- 238000007405 data analysis Methods 0.000 abstract description 8
- 238000013508 migration Methods 0.000 abstract description 3
- 230000005012 migration Effects 0.000 abstract description 3
- 238000013459 approach Methods 0.000 abstract description 2
- 208000035475 disorder Diseases 0.000 description 5
- 238000012731 temporal analysis Methods 0.000 description 5
- 238000000700 time series analysis Methods 0.000 description 5
- 238000013136 deep learning model Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000008676 import Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 238000009472 formulation Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 239000000203 mixture Substances 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
- 238000005728 strengthening Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000002759 z-score normalization Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 206010061818 Disease progression Diseases 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000011263 association rule mining technique Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005750 disease progression Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000009533 lab test Methods 0.000 description 1
- 230000009021 linear effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 238000013450 outlier detection Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000011321 prophylaxis Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000001225 therapeutic effect Effects 0.000 description 1
- 238000007794 visualization technique 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
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—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
- G06N3/045—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
- G06N3/04—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
- G06N3/096—Transfer learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/60—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices
- G16H40/67—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices for remote operation
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Pathology (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- General Business, Economics & Management (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention relates to the technical field of medical data analysis, in particular to an intelligent remote diagnosis method and system, comprising the following steps: based on the medical image data, a deep convolutional neural network including feature extraction and pattern recognition is employed to generate feature extracted image data. In the invention, the association rule learning is carried out by using the Apriori algorithm, the association between the image characteristics and the disease types is accurately analyzed, a more scientific basis is provided for diagnosis, the future health trend of a patient is effectively predicted by combining the time sequence prediction carried out by ARIMA and LSTM models, so that important roles in early warning and intervention are played, the models are personalized and adjusted by the migration learning technology, the individual differences of different patients are better adapted, more accurate diagnosis is provided, a relational network among diseases is constructed by using the map database technology of Neo4j, the visualization and interpretation of data are enhanced, and a new approach is provided for revealing the potential relation and treatment opportunity among diseases.
Description
Technical Field
The invention relates to the technical field of medical data analysis, in particular to an intelligent remote diagnosis method and system.
Background
The intelligent remote diagnosis method belongs to the technical field of medical data analysis, and the field is focused on the deep analysis of medical data by utilizing computer technology, particularly artificial intelligence and machine learning algorithm to assist diagnosis and treatment decision. The processing and analysis of various types of medical data, ranging from electronic health records, medical images to genomic data, and the like, is contemplated. Through advanced data analysis techniques, such as natural language processing, image recognition, predictive modeling, etc., medical data analysis can reveal disease patterns, assist in the formulation of personalized treatment plans, and improve the overall efficiency and effectiveness of medical services.
The intelligent remote diagnosis method is a technical means combining remote medical treatment and intelligent data analysis, and aims to provide accurate medical diagnosis in a remote mode. In this way, a doctor or medical professional can remotely access the patient's medical data, including medical images, laboratory test results, and electronic health records. Intelligent algorithms, such as machine learning and deep learning models, are applied to analyze these data to identify signs and patterns of disease, improving the accuracy of the diagnosis. The method aims to make the medical resource more popular and convenient, especially for patients in remote geographical locations or in areas with insufficient medical resources, and improves the diagnosis efficiency and accuracy. Through intelligent remote diagnosis, the medical institution can optimize resource allocation, reduce unnecessary face-to-face diagnosis and treatment, and increase flexibility and accessibility of medical services.
The traditional medical diagnostic methods have some drawbacks. Traditional diagnosis often relies on experience and visual judgment of doctors, and lacks accurate algorithm support, which leads to inconsistency and variability of diagnosis results. In addition, traditional methods are often not accurate enough in predicting future health trends of patients, and lack effective early warning mechanisms. Traditional diagnostic methods often ignore patient-to-patient variability and lack personalized diagnostic protocols. The traditional method is generally rough in construction and analysis of a disease relation network, and is difficult to deeply mine complex relations among diseases, so that development of comprehensive diagnosis and treatment strategies is limited.
Disclosure of Invention
The invention aims to solve the defects in the prior art and provides an intelligent remote diagnosis method and system.
In order to achieve the above purpose, the present invention adopts the following technical scheme: an intelligent remote diagnosis method, comprising the steps of:
s1: based on the medical image data, generating feature extraction image data by adopting a deep convolutional neural network comprising feature extraction and pattern recognition;
s2: extracting image data based on the features, analyzing the relevance of the image features and the disease types by using an association rule learning algorithm comprising an Apriori algorithm, and generating an association analysis report;
S3: based on the association analysis report, a time sequence prediction model comprising ARIMA and LSTM models is applied to predict future health trend of the patient, and a health risk prediction report is generated;
s4: based on the health risk prediction report, personalized adjustment is carried out on the model by using a transfer learning technology, and a personalized adaptation model is generated;
s5: based on the personalized adaptation model, carrying out data standardization processing comprising data standardization and format unification to generate a standardized medical data set;
s6: constructing a relationship network among symptoms by utilizing a graph database technology comprising Neo4j based on the standardized medical data set, and generating a disease relationship network graph;
s7: based on the disease relation network diagram, analyzing potential links and treatment opportunities among diseases by adopting a diagram mining technology, and generating a comprehensive diagnosis and treatment scheme report;
the feature extraction image data comprises key image marks and abnormal region identification results, the association analysis report comprises disease prediction rules and feature association diagrams, the health risk prediction report comprises short-term and long-term health trend predictions, the personalized adaptation model is specifically a diagnosis model adjusted for individual cases, the standardized medical data set comprises uniformly formatted health records, and the comprehensive diagnosis and treatment scheme report specifically comprises a treatment strategy of a target disease mode.
As a further aspect of the present invention, based on medical image data, the step of generating feature extraction image data using a deep convolutional neural network including feature extraction and pattern recognition specifically includes:
s101: based on medical image data, performing image preprocessing including image normalization and denoising processing by adopting a deep learning frame TensorFlow, and generating preprocessed image data;
s102: based on the preprocessed image data, performing feature extraction by using a convolutional neural network to generate a feature mapping set;
s103: based on the feature mapping set, an activation function ReLU and batch normalization are applied to strengthen the nonlinear expression capacity and stability of the model, and optimized feature data are generated;
s104: based on the optimized feature data, mapping the features to an output layer through a full connection layer, and performing feature integration to generate feature extraction image data;
the deep learning framework comprises data flow graph construction, automatic differentiation and GPU acceleration, the convolutional neural network comprises a plurality of convolutional layers and a pooling layer, the convolutional neural network is used for extracting spatial features of images, the batch normalization specifically refers to standardized processing on output of each layer, model training is accelerated, and the full-connection layer specifically refers to a layer in which neurons in the network are connected with each neuron in the previous layer and is used for integrating the learned features.
As a further aspect of the present invention, the step of generating a correlation analysis report by analyzing the correlation between the image feature and the disease type using a correlation rule learning algorithm including an Apriori algorithm based on the feature extraction image data specifically includes:
s201: extracting image data based on the features, performing data format conversion and cleaning by adopting a data mining tool Python, and generating a cleaned feature data set;
s202: based on the cleaned characteristic data set, adopting an Apriori algorithm to learn association rules, and generating a frequent item set;
s203: based on the frequent item set, the Apriori algorithm is utilized again, and a strong association rule is found out through calculating confidence coefficient and lifting degree indexes, so that an association rule set is generated;
s204: based on the association rule set, displaying the relationship between the association rule and the disease type through a visualization tool, and generating an association analysis report;
the Apriori algorithm specifically sets a support threshold through the process of iteratively searching frequent item sets.
As a further aspect of the present invention, based on the association analysis report, applying a time series prediction model including ARIMA and LSTM models, predicting a future health trend of the patient, the step of generating a health risk prediction report specifically includes:
S301: based on the association analysis report, adopting an autoregressive integral moving average model to perform time sequence analysis, and generating a preliminary time sequence analysis result;
s302: based on the preliminary time sequence analysis result, carrying out deep learning time sequence prediction by adopting a long-term and short-term memory network, and generating a deep learning time sequence prediction result;
s303: generating a fusion prediction result by adopting a model fusion technology based on the preliminary time sequence analysis result and the deep learning time sequence prediction result;
s304: based on the fusion prediction result, a performance evaluation method is adopted, wherein the performance evaluation method comprises MSE and R is a decision coefficient, and a health risk prediction report is generated;
the autoregressive integral moving average model comprises autoregressive, differential and moving average, the long-term memory network comprises a forgetting gate, an input gate and an output gate, and the model fusion technology comprises weighted average and error correction.
As a further scheme of the present invention, based on the health risk prediction report, the model is personalized adjusted by using a migration learning technology, and the step of generating a personalized adaptive model specifically includes:
s401: based on the health risk prediction report, adopting a transfer learning technology to generate a pre-training model selection result;
S402: performing model fine adjustment based on the pre-training model selection result, including network level adjustment and parameter optimization, and generating a fine-tuned personalized model;
s403: based on the fine-tuned personalized model, a cross verification technology is adopted to generate a model performance verification result;
s404: based on the model performance verification result, model optimization and adjustment are carried out to generate a personalized adaptation model;
the transfer learning technique includes pre-training model selection and model initialization, and the cross-validation technique includes data segmentation and multiple rounds of validation.
As a further aspect of the present invention, based on the personalized adaptive model, the step of performing data normalization processing including data normalization and format unification to generate a normalized medical data set specifically includes:
s501: based on the personalized adaptation model, cleaning by adopting a data preprocessing algorithm, including abnormal value detection and processing and missing value filling, and generating a preprocessed original medical data set;
s502: based on the preprocessed original medical data set, carrying out standardization processing on the data by adopting a Z score standardization method to generate a Z score standardization medical data set;
S503: based on the Z-score standardized medical data set, carrying out format unification processing, verifying data format consistency, and generating a standardized medical data set with uniform format;
s504: based on the standardized medical data set with uniform format, executing data quality inspection, ensuring the accuracy and consistency of the data, and obtaining a final standardized medical data set;
the data preprocessing algorithm comprises the steps of detecting abnormal values by using a median filling and IQR method, the Z-score standardization is specifically used for converting data points by using the mean value and standard deviation of data, the format unification comprises data type conversion and unification time format, and the data quality inspection comprises integrity inspection and consistency inspection.
As a further aspect of the present invention, the step of constructing a relationship network between disorders based on the standardized medical data set using a graph database technology including Neo4j, and generating a disease relationship network graph specifically includes:
s601: based on the standardized medical data set, a relationship extraction algorithm is adopted to identify potential relationships among symptoms, and symptom relationship metadata is generated;
s602: based on the condition relation metadata, converting the metadata into a structure of a matching graph database by adopting a graph data modeling method, and generating a graph database applicable data model;
S603: constructing a relationship network among symptoms by using Neo4j graph database technology based on the graph database applicable data model to obtain a symptom relationship network database;
s604: based on the disease relation network database, a network visualization technology is applied to carry out graphical interface design, and a disease relation network diagram is generated;
the relation extraction algorithm comprises entity identification and relation mining, the graph data modeling specifically refers to using nodes of a graph to represent entities, edges to represent relations among the entities, the Neo4j graph database technology comprises creating a graph structure, establishing an index and applying a Cypher query language, and the network visualization technology comprises using a graph layout algorithm and interactive design elements.
As a further aspect of the present invention, based on the disease relationship network graph, the potential links and treatment opportunities between diseases are analyzed by using graph mining technology, and the steps of generating a comprehensive diagnosis and treatment plan report are specifically as follows:
s701: based on the disease relation network graph, identifying key disease nodes by adopting a graph centrality analysis algorithm, and generating a key disease node set;
s702: based on the key disease node set, a community detection algorithm is applied to reveal the group structures among diseases, and a disease group structure analysis result is generated;
S703: based on the disease group structure analysis result, analyzing the association mode among diseases by using an association rule mining technology, and generating an inter-disease association mode analysis result;
s704: based on the analysis result of the inter-disease association mode, combining a clinical knowledge base, and adopting a decision support system technology to generate a comprehensive diagnosis and treatment scheme report;
the graph centrality analysis algorithm comprises degree centrality, medium number centrality and proximity centrality, the community detection algorithm comprises modularized optimization and spectral clustering, the association rule mining technology comprises an Apriori algorithm and confidence calculation, and the decision support system technology comprises an expert system and a machine learning prediction model.
The intelligent remote diagnosis system is used for executing the intelligent remote diagnosis method and comprises a feature extraction module, a relevance analysis module, a health trend prediction module, a personalized adjustment module, a data standardization module and a disease relation network construction module.
As a further scheme of the invention, the feature extraction module performs image preprocessing by adopting a TensorFlow deep learning framework based on medical image data, performs feature extraction by utilizing a convolutional neural network, and generates a feature mapping set by using a ReLU activation function and the nonlinear expression capacity of a batch normalization strengthening model;
The relevance analysis module uses a Python data mining tool to perform data format conversion and cleaning based on the feature mapping set, and adopts an Apriori algorithm to learn the relevance rule of the image features and the disease types so as to generate a relevance analysis report;
the health trend prediction module performs deep learning time sequence prediction by using ARIMA and LSTM time sequence prediction models based on the association analysis report, and combines a model fusion technology to strengthen the prediction capability so as to generate a health risk prediction report;
the personalized adjustment module adopts a transfer learning technology to carry out model fine adjustment based on a health risk prediction report, and comprises network level adjustment and parameter optimization to generate a personalized adaptation model;
the data standardization module performs data cleaning and outlier processing based on the personalized adaptation model, performs standardization processing on the data by using a Z score standardization method, and generates a standardized medical data set;
the disease relation network construction module is used for constructing a relation network among diseases by using Neo4j graph database technology based on a standardized medical data set, analyzing potential relations among diseases by using graph mining technology, and generating a disease relation network graph.
Compared with the prior art, the invention has the advantages and positive effects that:
In the invention, the association rule learning is carried out by using the Apriori algorithm, the association between the image characteristics and the disease types is accurately analyzed, and a more scientific basis is provided for diagnosis. The time sequence prediction by combining ARIMA and LSTM models effectively predicts the future health trend of the patient, thereby playing an important role in early warning and intervention. The model is subjected to personalized adjustment through a transfer learning technology, so that the model is better suitable for individual differences of different patients, and more accurate diagnosis is provided. The map database technology of Neo4j is used for constructing a relational network between diseases, so that the visualization and interpretation of data are enhanced, and a new approach is provided for revealing potential links between diseases and treatment opportunities.
Drawings
FIG. 1 is a schematic workflow diagram of the present invention;
FIG. 2 is a S1 refinement flowchart of the present invention;
FIG. 3 is a S2 refinement flowchart of the present invention;
FIG. 4 is a S3 refinement flowchart of the present invention;
FIG. 5 is a S4 refinement flowchart of the present invention;
FIG. 6 is a S5 refinement flowchart of the present invention;
FIG. 7 is a S6 refinement flowchart of the present invention;
FIG. 8 is a S7 refinement flowchart of the present invention;
fig. 9 is a system flow diagram of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
Embodiment one: referring to fig. 1, the present invention provides a technical solution: an intelligent remote diagnosis method, comprising the steps of:
s1: based on the medical image data, generating feature extraction image data by adopting a deep convolutional neural network comprising feature extraction and pattern recognition;
S2: based on the feature extraction image data, an association rule learning algorithm comprising an Apriori algorithm is used for analyzing the association between the image feature and the disease type, and an association analysis report is generated;
s3: based on the association analysis report, a time sequence prediction model comprising ARIMA and LSTM models is applied to predict future health trend of the patient, and a health risk prediction report is generated;
s4: based on the health risk prediction report, personalized adjustment is carried out on the model by using a transfer learning technology, and a personalized adaptation model is generated;
s5: based on the personalized adaptation model, carrying out data standardization processing comprising data standardization and format unification to generate a standardized medical data set;
s6: constructing a relationship network among symptoms by utilizing a graph database technology comprising Neo4j based on a standardized medical data set, and generating a disease relationship network graph;
s7: based on the disease relation network diagram, adopting a diagram mining technology to analyze potential relation and treatment opportunities among diseases and generating a comprehensive diagnosis and treatment scheme report;
the feature extraction image data comprises key image marks and abnormal region identification results, the association analysis report comprises a disease prediction rule and a feature association diagram, the health risk prediction report comprises short-term and long-term health trend prediction, the personalized adaptation model is specifically a diagnosis model adjusted for individual cases, the standardized medical data set comprises uniformly formatted health records, and the comprehensive diagnosis and treatment scheme report specifically comprises a treatment strategy of a target disease mode.
The key features can be accurately extracted from the medical images through the deep convolutional neural network, and powerful support is provided for early disease diagnosis. Then, by using association rule learning algorithms such as Apriori, the association between the image features and the disease types can be deeply analyzed, so that potential prediction rules of the disease can be revealed. In addition, short-term and long-term health trends of patients can be predicted by ARIMA and LSTM models, providing data support for prophylactic treatment. The personalized adaptation model of the model (1) is customized for each case by utilizing a transfer learning technology, so that the accuracy and the personalized level of the treatment scheme are improved. Meanwhile, the data standardization processing ensures the consistency and comparability of the data and enhances the reliability of data analysis. Finally, by constructing a disease relation network and applying a graph mining technology, complex connection among diseases and new treatment opportunities can be revealed, and a comprehensive treatment strategy is provided for doctors.
Referring to fig. 2, based on medical image data, the steps of generating feature extraction image data using a deep convolutional neural network including feature extraction and pattern recognition are specifically as follows:
s101: based on medical image data, performing image preprocessing including image normalization and denoising processing by adopting a deep learning frame TensorFlow, and generating preprocessed image data;
S102: based on the preprocessed image data, performing feature extraction by using a convolutional neural network to generate a feature mapping set;
s103: based on the feature mapping set, an activation function ReLU and batch normalization are applied to strengthen the nonlinear expression capacity and stability of the model, and optimized feature data are generated;
s104: based on the optimized feature data, mapping the features to an output layer through a full connection layer, and performing feature integration to generate feature extraction image data;
the deep learning framework comprises data flow graph construction, automatic differentiation and GPU acceleration, the convolutional neural network comprises a plurality of convolutional layers and a pooling layer, the convolutional neural network is used for extracting spatial features of images, the batch normalization specifically refers to standardized processing on each layer of output, the acceleration model training is carried out, and the full-connection layer specifically refers to a layer in which neurons in the network are connected with each neuron in the previous layer and is used for integrating the learned features.
In S101, data collection is performed, and medical image data is collected, including X-ray, MRI, or other medical image images. It is ensured that the dataset contains enough samples to cover different cases, conditions and disease types to ensure the generalization ability of the model. The source and collection method of data requires recording and documentation to ensure traceability and compliance of the data. And carrying out data preprocessing, including image normalization and denoising. Image normalization normalizes the pixel values of the image, typically scaling the pixel values to a range of 0 to 1, to reduce gradient explosion or vanishing problems in model training. The denoising process uses image processing technology such as Gaussian filtering, median filtering or wavelet denoising, so that noise in the image is removed, and the definition and the characteristic identifiability of the image are improved. And when needed, the data enhancement technology such as rotation, overturning, cutting and the like is applied, so that the diversity of training data is increased, and the generalization capability of the model is improved.
In S102, a deep learning model is constructed, and a Convolutional Neural Network (CNN) model is constructed using a deep learning framework such as TensorFlow, etc., to be used for feature extraction. The architecture of the model should be designed according to the requirements of the task, and generally includes a convolution layer, a pooling layer and a full connection layer. The convolution layer is used to detect various features in the image, including edges, texture, shape, and the like. The pooling layer is used for reducing the dimension of the feature map, reducing the computational complexity and increasing the translational invariance of the model. After the convolution and pooling layers, the model will generate a set of feature maps, each map corresponding to a particular image feature. These feature maps will be used for further processing and analysis to extract useful information in medical images.
In S103, an activation function ReLU (Rectified Linear Unit) is applied to enhance the nonlinear expression capabilities of the model. The ReLU function sets negative values to zero, preserving positive values, thereby introducing non-linear properties that help the model better capture complex features. Batch normalization (Batch Normalization) is applied to normalize each layer of output to increase the stability of the model and speed up the training process. Batch normalization is typically applied before activating the function. Enhanced feature data will be generated, the data comprising an abstract feature representation of the image.
In S104, a full connection layer is added, and feature data is mapped to an output layer. Neurons in the fully connected layer are connected to each neuron in the previous layer for integration of learned features. The architecture and size of the fully connected layer is designed according to the requirements of specific tasks. In the fully connected layer, feature extracted image data is generated, which is further analyzed according to the requirements of the task, e.g. for classification, segmentation, detection or other applications of medical images.
Referring to fig. 3, based on feature extraction image data, the association of image features and disease types is analyzed by using an association rule learning algorithm including Apriori algorithm, and the step of generating an association analysis report is specifically as follows:
s201: extracting image data based on the features, performing data format conversion and cleaning by adopting a data mining tool Python, and generating a cleaned feature data set;
s202: based on the cleaned characteristic data set, adopting an Apriori algorithm to learn association rules, and generating a frequent item set;
s203: based on the frequent item set, the Apriori algorithm is utilized again, and a strong association rule is found out through calculating confidence coefficient and lifting degree indexes, so that an association rule set is generated;
S204: based on the association rule set, displaying the relationship between the association rule and the disease type through the visualization tool, and generating an association analysis report;
the Apriori algorithm specifically sets a support threshold through the process of iteratively searching frequent item sets.
In S201, based on the feature extracted image data, cleaning and format conversion of the data are performed using Python or other data mining tools. This process aims to prepare the data for subsequent association rule analysis. Data cleansing includes detecting and processing missing, outliers, and duplicate values to ensure quality and consistency of the data. The data needs to be format converted to meet the input requirements of the Apriori algorithm. Typically, the data is converted to be represented in the form of a collection of transactions, each transaction containing one or more features for association rule learning.
In S202, association rule learning is performed by using Apriori algorithm based on the feature data set after cleaning. A support threshold needs to be set that is used to determine which item sets are considered frequent. The support represents the frequency of occurrence of the item set in the data, and the support threshold is usually set according to the requirements of the task and the characteristics of the data. An initial scan is performed to find frequent item sets that meet the support threshold, which are typically single features or a combination of features. By iteratively generating candidate item sets, the item sets are increased in size layer by layer until new frequent item sets cannot be generated any more. In each iteration, repeated scanning is carried out, the support degree of each candidate item set is calculated, and item sets which do not meet the support degree threshold are removed.
In S203, based on the generated frequent item set, the Apriori algorithm is again utilized to find a strong association rule. Confidence and promotion indicators between each pair of frequent item sets are calculated. The confidence level represents the credibility of the rule, and the promotion level represents the relevance of the rule. And filtering rules with confidence or lifting degree lower than the threshold according to the set confidence and lifting degree threshold, and only retaining the strong association rules. A set of association rules is generated that contains strong association rules that satisfy a threshold requirement, which describe relationships between features.
In S204, based on the generated set of association rules, a relationship between the association rules and the disease type is exposed using a visualization tool to generate an association analysis report. A list of strong association rules is listed, including the antecedents and postings of the rules, to provide detailed rule information. And the distribution and the relation of the association rules are displayed in a visual mode such as a chart or a graph, so that a user is helped to understand the data more intuitively. The generated association rules are interpreted and analyzed to provide conclusions regarding the association between the features and the disease type to aid decision making and further medical research.
Referring to fig. 4, based on the association analysis report, a time series prediction model including ARIMA and LSTM models is applied to predict future health trend of the patient, and the step of generating a health risk prediction report is specifically:
S301: based on the association analysis report, adopting an autoregressive integral moving average model to perform time sequence analysis, and generating a preliminary time sequence analysis result;
s302: based on the preliminary time sequence analysis result, carrying out deep learning time sequence prediction by adopting a long-term and short-term memory network, and generating a deep learning time sequence prediction result;
s303: based on the preliminary time sequence analysis result and the deep learning time sequence prediction result, generating a fusion prediction result by adopting a model fusion technology;
s304: based on the fusion prediction result, a performance evaluation method is adopted, wherein the performance evaluation method comprises MSE and R is a decision coefficient, and a health risk prediction report is generated;
autoregressive integral moving average models include autoregressive, differential and moving averages, long-short-term memory networks include forgetting gates, input gates and output gates, and model fusion techniques include weighted average and error correction.
In S301, based on the association analysis report, a time series analysis of an autoregressive integral moving average model (ARIMA) is performed. This process includes Autoregressive (AR) to establish a relationship with past observations, differencing (I, differential integration) to convert the time series to a stationary series, and running average (MA) to analyze the running average of the time series. The steps will generate preliminary time series analysis results, including model parameters and fitting degree information.
In S302, based on the preliminary time series analysis result, deep learning time series prediction is performed using a long short term memory network (LSTM). LSTM is a deep learning model suitable for sequence data, which includes key components such as forgetting gates, input gates, and output gates, capturing long-term dependencies and patterns in the sequence. The method comprises the steps of constructing an LSTM model, training the model to learn patterns and trends in sequence data, and carrying out time sequence prediction of future health trends by using the trained model to generate a deep learning time sequence prediction result.
In S303, a model fusion technique is employed to generate a fusion prediction result based on the preliminary time series analysis result and the deep learning time series prediction result. This step includes weighted averaging, wherein the preliminary time series analysis and the deep learning time series prediction result are weighted averaged according to a certain weight. And correcting the result according to the error condition of the two by using an error correction technology so as to improve the accuracy and the robustness of the prediction.
In S304, a performance evaluation method is employed, including a Mean Square Error (MSE) and a decision coefficient (R), based on the fusion prediction result, to generate a health risk prediction report. The report includes the predicted outcome, the calculated outcome of the performance assessment indicator, and the interpretation and suggestion of the predicted outcome. The report provides an assessment and advice regarding the patient's future health risk, facilitating the formulation of medical decisions and interventions.
Referring to fig. 5, based on the health risk prediction report, the model is personalized and adjusted by using the transfer learning technology, and the step of generating the personalized adaptive model specifically includes:
s401: based on the health risk prediction report, adopting a transfer learning technology to generate a pre-training model selection result;
s402: performing model fine adjustment based on a pre-training model selection result, including network level adjustment and parameter optimization, and generating a fine-tuned personalized model;
s403: based on the fine-tuned personalized model, a cross verification technology is adopted to generate a model performance verification result;
s404: based on the model performance verification result, model optimization and adjustment are carried out to generate a personalized adaptation model;
the transfer learning technology comprises pre-training model selection and model initialization, and the cross-validation technology comprises data segmentation and multi-round validation.
In S401, a pre-training model is selected.
The technology comprises the following steps: and (5) a migration learning technology.
The steps are as follows:
selecting a suitable pre-training model: and selecting a proper pre-training model according to the requirements of health risk prediction.
Model initialization: the selected model is initialized to fit a particular data set.
And generating a pre-training model selection result.
Code example (Python):
from tensorflow.keras.applications import VGG16；
# VGG16 was chosen as the pre-training model;
pretrained_model = VGG16(weights='imagenet', include_top=False)；
in S402, the model is fine-tuned.
The technology comprises the following steps: network hierarchy adjustment and parameter optimization.
The steps are as follows:
network hierarchy adjustment: specific layers are added or removed as desired.
Parameter optimization: the model parameters are fine-tuned to accommodate the new data.
And generating a fine-tuned personalized model.
Code example:
from tensorflow.keras import layers, models；
a# fine tuning model;
model = models.Sequential()；
model.add(pretrained_model)；
model.add(layers.Flatten())；
model.add(layers.Dense(256, activation='relu'))；
model. Add (layers. Dense (1, activation= 'signature'),) hypothesis is a classification problem;
compiling a model;
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])；
in S403, model performance is verified.
The technology comprises the following steps: cross-validation techniques.
The steps are as follows:
data segmentation: the dataset is partitioned into a plurality of subsets.
And (3) multiple rounds of verification: model performance is verified on different subsets of data.
And generating a model performance verification result.
Code example:
from sklearn.model_selection import cross_val_score；
cross-validation, # and (2);
score=cross_val_score (model, X, y, cv=5) # assuming that X is a feature and y is a tag;
in S404, the model is optimized and adjusted.
The technology comprises the following steps: and (5) model optimization.
The steps are as follows:
analysis and verification result: the behavior of the model in different aspects is determined.
Model optimization is carried out: and adjusting the model structure or parameters according to the verification result.
A personalized adaptation model is generated.
Code example:
def optimize_model(model, scores):
analyzing and optimizing the performance of the model;
# ...；
return optimized_model；
optimized_model = optimize_model(model, scores)；
Referring to fig. 6, based on the personalized adaptation model, the step of performing data normalization processing including data normalization and format unification to generate a normalized medical data set specifically includes:
s501: based on the personalized adaptation model, cleaning by adopting a data preprocessing algorithm, including abnormal value detection and processing and missing value filling, and generating a preprocessed original medical data set;
s502: based on the preprocessed original medical data set, carrying out standardization processing on the data by adopting a Z score standardization method to generate a Z score standardization medical data set;
s503: based on the Z-score standardized medical data set, carrying out format unification processing, verifying data format consistency, and generating a standardized medical data set with uniform format;
s504: based on the standardized medical data set with uniform format, executing data quality inspection, ensuring the accuracy and consistency of the data, and obtaining a final standardized medical data set;
the data preprocessing algorithm comprises the steps of detecting abnormal values by using a median filling and IQR method, and Z-score standardization is specifically to convert data points by using a mean value and a standard deviation of data, format unification comprises data type conversion and unification time format, and data quality inspection comprises integrity inspection and consistency inspection.
In S501, data cleansing and preprocessing are performed based on the personalized adaptation model. The process comprises the operations of abnormal value detection and processing, missing value filling and the like. The objective of outlier detection and processing is to identify and process outliers in the data to ensure the quality of the data. The median fill and IQR methods are typically employed to handle outliers. The missing value padding is used for processing missing values existing in the data set so as to ensure the integrity of the data. This step generates a preprocessed raw medical dataset.
In S502, based on the preprocessed raw medical dataset, the data is normalized using a Z-score normalization method. Z-score normalization is a common normalization method that uses the mean and standard deviation of data to transform the data points such that the mean of the data is 0 and the standard deviation is 1. This facilitates converting the data of different scales into relatively uniform scales for subsequent analysis and modeling. This step generates a Z-score normalized medical dataset.
In S503, the medical data set is normalized based on the Z score, and format unification processing is performed to ensure consistency of the data format. This step includes data type conversion and unifying time format operations to verify data format consistency. Data type conversion involves converting data fields from text to numeric values or unifying different data types to the same data type. The unified time format typically includes converting different time representations (e.g., dates, time stamps) into a unified time format. This helps to improve the comparability and operability of the data.
In S504, a data quality check is performed based on the standardized medical data set of uniform format to ensure accuracy and consistency of the data. Data quality checks include integrity checks and consistency checks. The integrity check is used to verify whether the data is complete, i.e. whether there is a missing value or a data loss situation. A consistency check is used to verify whether the data is consistent between different fields, e.g., whether the date and time fields match each other. Through these checks, it is ensured that the final standardized medical dataset is of high quality, accurate and consistent for further analysis and application.
Referring to fig. 7, based on the standardized medical data set, using graph database technology including Neo4j, a relationship network between diseases is constructed, and the steps of generating a disease relationship network graph are specifically as follows:
s601: based on the standardized medical data set, adopting a relation extraction algorithm to identify potential relations among symptoms and generating symptom relation metadata;
s602: based on the condition relation metadata, converting the metadata into a structure of a matching graph database by adopting a graph data modeling method, and generating a graph database applicable data model;
s603: constructing a relation network among symptoms by using Neo4j graph database technology based on a graph database applicable data model to obtain a symptom relation network database;
S604: based on the disease relation network database, a network visualization technology is applied to carry out graphical interface design, and a disease relation network diagram is generated;
the relation extraction algorithm comprises entity identification and relation mining, the graph data modeling specifically refers to using nodes of a graph to represent entities, edges to represent the relation among the entities, the Neo4j graph database technology comprises the steps of creating a graph structure, establishing an index and applying a Cypher query language, and the network visualization technology comprises the steps of using a graph layout algorithm and interactive design elements.
In S601, a relationship between potential medical conditions is identified based on the standardized medical dataset using a relationship extraction algorithm. This includes entity recognition, i.e., marking and extracting the entities of disorders from text data, and relational mining, which recognizes the relationships between disorders by analyzing the semantics and context of medical text. The generated condition relation metadata comprises condition entities and relation descriptions among the condition entities, and provides basic data for subsequent graph database modeling.
In S602, based on the condition relation metadata, the metadata is converted into a data structure suitable for a graph database using a graph data modeling method. This involves representing the symptom entity as nodes of a graph, each node having a unique identifier and attribute, such as a symptom name and description. Relationships between disorders are represented as edges of the graph, which contain relationship types and attributes, such as relationship strength and time information. The goal of this step is to create a data model that is suitable for graph database storage.
In S603, a relational network between disorders is constructed from the graph database using Neo4j graph database technology, and the data model is applied. This includes creating a graph structure, including nodes and edges, to store the disease entities and relationships to each other. And establishing an index to improve the data retrieval efficiency and ensure quick access and inquiry. Various query operations are performed using the cytoer query language, including relational lookup, screening, and analysis, to build a complete database of condition relational networks.
In S604, a graphical interface is designed to generate a disease relationship network graph based on the disease relationship network database by applying a network visualization technique. This includes selecting an appropriate graphical layout algorithm to visualize the condition relationship network data as a graphical structure that is easy to understand and view. Interactive design elements, such as zoom, filter, search, etc., are added to enhance the interactivity of the user with the graphical interface. The generated disease relationship network graph helps medical professionals and researchers to better understand and analyze disease relationships in medical data, and provides a useful tool for decision making and research.
Referring to fig. 8, based on a disease relationship network diagram, the potential links and treatment opportunities between diseases are analyzed by using a graph mining technique, and the steps for generating a comprehensive diagnosis and treatment plan report are specifically as follows:
S701: based on the disease relation network diagram, adopting a diagram centrality analysis algorithm to identify key disease nodes and generating a key disease node set;
s702: based on the key disease node set, a community detection algorithm is applied to reveal the group structures among diseases, and a disease group structure analysis result is generated;
s703: based on the disease group structure analysis result, analyzing the association mode among diseases by using an association rule mining technology, and generating an inter-disease association mode analysis result;
s704: based on the analysis result of the correlation mode between diseases, combining with a clinical knowledge base, adopting a decision support system technology to generate a comprehensive diagnosis and treatment scheme report;
the graph centrality analysis algorithm comprises centrality, medium centrality and near centrality, the community detection algorithm comprises modularized optimization and spectral clustering, the association rule mining technology comprises Apriori algorithm and confidence calculation, and the decision support system technology comprises an expert system and a machine learning prediction model.
In S701, based on the disease relationship network graph, a graph centrality analysis algorithm is employed to identify key disease nodes and generate a set of key disease nodes. The graph centrality analysis comprises indexes such as centrality, mid-number centrality, near centrality and the like, and is used for measuring importance of nodes in a network. By analyzing these metrics, critical disease nodes are determined that have high centrality and impact in the network.
In S702, a community detection algorithm is applied to reveal the population structure between diseases based on the set of key disease nodes, and a disease population structure analysis result is generated. The community detection algorithm comprises modularized optimization, spectral clustering and other technologies, and is used for dividing disease nodes into different groups or communities to reveal potential disease group structures. This helps to understand which diseases are associated with each other in a particular biology or clinic.
In S703, based on the disease population structure analysis result, the correlation pattern between diseases is analyzed by applying the correlation rule mining technique, and the inter-disease correlation pattern analysis result is generated. Association rule mining techniques include Apriori algorithms and confidence calculations for finding association rules and patterns between diseases. This reveals which diseases often occur simultaneously, as well as potential links to each other.
In S704, based on the results of the inter-disease correlation pattern analysis, a comprehensive diagnosis and treatment plan report is generated using decision support system techniques in combination with a clinical knowledge base. This includes providing personalized diagnostic advice and treatment regimens based on the association patterns and disease population structure using expert systems and machine learning predictive models. The report includes the patient's medical history, clinical data, and treatment options, providing decision support and treatment advice to the medical professional.
Referring to fig. 9, the intelligent remote diagnosis system is used for executing the intelligent remote diagnosis method, and the system comprises a feature extraction module, a relevance analysis module, a health trend prediction module, a personalized adjustment module, a data standardization module and a disease relationship network construction module.
The feature extraction module performs image preprocessing by adopting a TensorFlow deep learning framework based on medical image data, performs feature extraction by utilizing a convolutional neural network, and generates a feature mapping set by using a ReLU activation function and the nonlinear expression capability of a batch normalization strengthening model;
the relevance analysis module uses a Python data mining tool to perform data format conversion and cleaning based on the feature mapping set, and adopts an Apriori algorithm to learn the relevance rule of the image features and the disease types so as to generate a relevance analysis report;
the health trend prediction module performs deep learning time sequence prediction by using ARIMA and LSTM time sequence prediction models based on the association analysis report, and combines a model fusion technology to strengthen the prediction capability so as to generate a health risk prediction report;
the personalized adjustment module adopts a transfer learning technology to carry out model fine adjustment based on the health risk prediction report, and comprises network level adjustment and parameter optimization to generate a personalized adaptation model;
The data standardization module performs data cleaning and outlier processing based on the personalized adaptation model, performs standardization processing on the data by using a Z-score standardization method, and generates a standardized medical data set;
the disease relation network construction module constructs a relation network among diseases by using Neo4j graph database technology based on the standardized medical data set, analyzes potential relation among diseases by using graph mining technology, and generates a disease relation network graph.
The feature extraction module extracts features of medical image data by using a deep learning technology, and key features in complex medical images can be effectively identified, so that the accuracy of diagnosis is improved. By using the ReLU activation function and batch normalization, the model can better capture nonlinear characteristics, and further improve diagnosis efficiency and accuracy.
The relevance analysis module can discover relevance rules between image features and disease types by using a data mining technology and an Apriori algorithm. This helps reveal unobvious disease patterns and associations, providing more insight into clinical decisions, thereby helping physicians to better understand and predict disease progression.
The health trend prediction module is combined with a time sequence prediction model and a model fusion technology, so that the health trend of an individual can be predicted. The prediction can not only improve the effectiveness of preventive medical treatment, but also help doctors to adjust the treatment scheme in time, thereby improving the treatment success rate.
The personalized adjustment module can generate customized medical models for different individuals through transfer learning and model fine adjustment. The personalized method can adjust the diagnosis and treatment scheme according to the unique condition of each patient, and improves the pertinence and the effectiveness of treatment.
The data standardization module ensures the quality and consistency of all medical data through data cleaning and standardization processing. This is critical to ensure accuracy and reliability of data analysis, helping to avoid misdiagnosis and missed diagnosis due to data quality issues.
The disease relationship network construction module uses graph database technology and graph mining technology to construct a relationship network between diseases. This not only helps to understand the complex links between diseases, but also reveals underlying disease mechanisms and new therapeutic targets.
The present invention is not limited to the above embodiments, and any equivalent embodiments which can be changed or modified by the technical disclosure described above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matter of the present invention will still fall within the scope of the technical disclosure.
Claims (10)
1. The intelligent remote diagnosis method is characterized by comprising the following steps of:
based on the medical image data, generating feature extraction image data by adopting a deep convolutional neural network comprising feature extraction and pattern recognition;
extracting image data based on the features, analyzing the relevance of the image features and the disease types by using an association rule learning algorithm comprising an Apriori algorithm, and generating an association analysis report;
based on the association analysis report, a time sequence prediction model comprising ARIMA and LSTM models is applied to predict future health trend of the patient, and a health risk prediction report is generated;
based on the health risk prediction report, personalized adjustment is carried out on the model by using a transfer learning technology, and a personalized adaptation model is generated;
based on the personalized adaptation model, carrying out data standardization processing comprising data standardization and format unification to generate a standardized medical data set;
constructing a relationship network among symptoms by utilizing a graph database technology comprising Neo4j based on the standardized medical data set, and generating a disease relationship network graph;
based on the disease relation network diagram, analyzing potential links and treatment opportunities among diseases by adopting a diagram mining technology, and generating a comprehensive diagnosis and treatment scheme report;
The feature extraction image data comprises key image marks and abnormal region identification results, the association analysis report comprises disease prediction rules and feature association diagrams, the health risk prediction report comprises short-term and long-term health trend predictions, the personalized adaptation model is specifically a diagnosis model adjusted for individual cases, the standardized medical data set comprises uniformly formatted health records, and the comprehensive diagnosis and treatment scheme report specifically comprises a treatment strategy of a target disease mode.
2. The intelligent remote diagnosis method according to claim 1, wherein the step of generating feature extraction image data using a deep convolutional neural network including feature extraction and pattern recognition based on medical image data is specifically:
based on medical image data, performing image preprocessing including image normalization and denoising processing by adopting a deep learning frame TensorFlow, and generating preprocessed image data;
based on the preprocessed image data, performing feature extraction by using a convolutional neural network to generate a feature mapping set;
based on the feature mapping set, an activation function ReLU and batch normalization are applied to strengthen the nonlinear expression capacity and stability of the model, and optimized feature data are generated;
Based on the optimized feature data, mapping the features to an output layer through a full connection layer, and performing feature integration to generate feature extraction image data;
the deep learning framework comprises data flow graph construction, automatic differentiation and GPU acceleration, the convolutional neural network comprises a plurality of convolutional layers and a pooling layer, the convolutional neural network is used for extracting spatial features of images, the batch normalization specifically refers to standardized processing on output of each layer, model training is accelerated, and the full-connection layer specifically refers to a layer in which neurons in the network are connected with each neuron in the previous layer and is used for integrating the learned features.
3. The intelligent remote diagnosis method according to claim 1, wherein the step of generating a correlation analysis report by analyzing the correlation between the image feature and the disease type using a correlation rule learning algorithm including Apriori algorithm based on the feature extraction image data is specifically:
extracting image data based on the features, performing data format conversion and cleaning by adopting a data mining tool Python, and generating a cleaned feature data set;
based on the cleaned characteristic data set, adopting an Apriori algorithm to learn association rules, and generating a frequent item set;
Based on the frequent item set, the Apriori algorithm is utilized again, and a strong association rule is found out through calculating confidence coefficient and lifting degree indexes, so that an association rule set is generated;
based on the association rule set, displaying the relationship between the association rule and the disease type through a visualization tool, and generating an association analysis report;
the Apriori algorithm specifically sets a support threshold through the process of iteratively searching frequent item sets.
4. The intelligent remote diagnosis method according to claim 1, wherein based on the association analysis report, a time series prediction model including ARIMA and LSTM models is applied to predict future health trends of the patient, and the step of generating a health risk prediction report is specifically:
based on the association analysis report, adopting an autoregressive integral moving average model to perform time sequence analysis, and generating a preliminary time sequence analysis result;
based on the preliminary time sequence analysis result, carrying out deep learning time sequence prediction by adopting a long-term and short-term memory network, and generating a deep learning time sequence prediction result;
generating a fusion prediction result by adopting a model fusion technology based on the preliminary time sequence analysis result and the deep learning time sequence prediction result;
Based on the fusion prediction result, a performance evaluation method is adopted, wherein the performance evaluation method comprises MSE and R is a decision coefficient, and a health risk prediction report is generated;
the autoregressive integral moving average model comprises autoregressive, differential and moving average, the long-term memory network comprises a forgetting gate, an input gate and an output gate, and the model fusion technology comprises weighted average and error correction.
5. The intelligent remote diagnosis method according to claim 1, wherein based on the health risk prediction report, the model is personalized adjusted using a transfer learning technique, and the step of generating a personalized adaptive model is specifically:
based on the health risk prediction report, adopting a transfer learning technology to generate a pre-training model selection result;
performing model fine adjustment based on the pre-training model selection result, including network level adjustment and parameter optimization, and generating a fine-tuned personalized model;
based on the fine-tuned personalized model, a cross verification technology is adopted to generate a model performance verification result;
based on the model performance verification result, model optimization and adjustment are carried out to generate a personalized adaptation model;
the transfer learning technique includes pre-training model selection and model initialization, and the cross-validation technique includes data segmentation and multiple rounds of validation.
6. The intelligent remote diagnosis method according to claim 1, wherein the step of performing data normalization processing including data normalization and format unification based on the personalized adaptation model, to generate a normalized medical data set, is specifically:
based on the personalized adaptation model, cleaning by adopting a data preprocessing algorithm, including abnormal value detection and processing and missing value filling, and generating a preprocessed original medical data set;
based on the preprocessed original medical data set, carrying out standardization processing on the data by adopting a Z score standardization method to generate a Z score standardization medical data set;
based on the Z-score standardized medical data set, carrying out format unification processing, verifying data format consistency, and generating a standardized medical data set with uniform format;
based on the standardized medical data set with uniform format, executing data quality inspection, ensuring the accuracy and consistency of the data, and obtaining a final standardized medical data set;
the data preprocessing algorithm comprises the steps of detecting abnormal values by using a median filling and IQR method, the Z-score standardization is specifically used for converting data points by using the mean value and standard deviation of data, the format unification comprises data type conversion and unification time format, and the data quality inspection comprises integrity inspection and consistency inspection.
7. The intelligent remote diagnosis method according to claim 1, wherein the step of constructing a relationship network between disorders based on the standardized medical data set using a graph database technique including Neo4j, and generating a disease relationship network graph is specifically:
based on the standardized medical data set, a relationship extraction algorithm is adopted to identify potential relationships among symptoms, and symptom relationship metadata is generated;
based on the condition relation metadata, converting the metadata into a structure of a matching graph database by adopting a graph data modeling method, and generating a graph database applicable data model;
constructing a relationship network among symptoms by using Neo4j graph database technology based on the graph database applicable data model to obtain a symptom relationship network database;
based on the disease relation network database, a network visualization technology is applied to carry out graphical interface design, and a disease relation network diagram is generated;
the relation extraction algorithm comprises entity identification and relation mining, the graph data modeling specifically refers to using nodes of a graph to represent entities, edges to represent relations among the entities, the Neo4j graph database technology comprises creating a graph structure, establishing an index and applying a Cypher query language, and the network visualization technology comprises using a graph layout algorithm and interactive design elements.
8. The intelligent remote diagnosis method according to claim 1, wherein the step of generating a comprehensive diagnosis and treatment plan report based on the disease relationship network graph by analyzing potential links and treatment opportunities between diseases using graph mining technique comprises:
based on the disease relation network graph, identifying key disease nodes by adopting a graph centrality analysis algorithm, and generating a key disease node set;
based on the key disease node set, a community detection algorithm is applied to reveal the group structures among diseases, and a disease group structure analysis result is generated;
based on the disease group structure analysis result, analyzing the association mode among diseases by using an association rule mining technology, and generating an inter-disease association mode analysis result;
based on the analysis result of the inter-disease association mode, combining a clinical knowledge base, and adopting a decision support system technology to generate a comprehensive diagnosis and treatment scheme report;
the graph centrality analysis algorithm comprises degree centrality, medium number centrality and proximity centrality, the community detection algorithm comprises modularized optimization and spectral clustering, the association rule mining technology comprises an Apriori algorithm and confidence calculation, and the decision support system technology comprises an expert system and a machine learning prediction model.
9. An intelligent remote diagnosis system, characterized in that it comprises a feature extraction module, a relevance analysis module, a health trend prediction module, a personalized adjustment module, a data standardization module, and a disease relationship network construction module according to any one of claims 1-8.
10. The intelligent remote diagnosis system of claim 9, wherein the feature extraction module performs image preprocessing using a TensorFlow deep learning framework based on medical image data, performs feature extraction using a convolutional neural network, and generates a feature mapping set by a ReLU activation function and a nonlinear expression capability of a batch normalization reinforcement model;
the relevance analysis module uses a Python data mining tool to perform data format conversion and cleaning based on the feature mapping set, and adopts an Apriori algorithm to learn the relevance rule of the image features and the disease types so as to generate a relevance analysis report;
the health trend prediction module performs deep learning time sequence prediction by using ARIMA and LSTM time sequence prediction models based on the association analysis report, and combines a model fusion technology to strengthen the prediction capability so as to generate a health risk prediction report;
The personalized adjustment module adopts a transfer learning technology to carry out model fine adjustment based on a health risk prediction report, and comprises network level adjustment and parameter optimization to generate a personalized adaptation model;
the data standardization module performs data cleaning and outlier processing based on the personalized adaptation model, performs standardization processing on the data by using a Z score standardization method, and generates a standardized medical data set;
the disease relation network construction module is used for constructing a relation network among diseases by using Neo4j graph database technology based on a standardized medical data set, analyzing potential relations among diseases by using graph mining technology, and generating a disease relation network graph.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311774882.2A CN117457192A (en) | 2023-12-22 | 2023-12-22 | Intelligent remote diagnosis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311774882.2A CN117457192A (en) | 2023-12-22 | 2023-12-22 | Intelligent remote diagnosis method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117457192A true CN117457192A (en) | 2024-01-26 |
Family
ID=89589477
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311774882.2A Pending CN117457192A (en) | 2023-12-22 | 2023-12-22 | Intelligent remote diagnosis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117457192A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786373A (en) * | 2024-02-28 | 2024-03-29 | 山东鑫林纸制品有限公司 | Equipment operation diagnosis system based on big data corrugated paper processing |
-
2023
- 2023-12-22 CN CN202311774882.2A patent/CN117457192A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786373A (en) * | 2024-02-28 | 2024-03-29 | 山东鑫林纸制品有限公司 | Equipment operation diagnosis system based on big data corrugated paper processing |
| CN117786373B (en) * | 2024-02-28 | 2024-05-03 | 山东鑫林纸制品有限公司 | Equipment operation diagnosis system based on big data corrugated paper processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2016192612A1 (en) | Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof | |
| WO2021226132A2 (en) | Systems and methods for managing autoimmune conditions, disorders and diseases | |
| WO2022001123A1 (en) | Key point detection method and apparatus, and electronic device and storage medium | |
| CN117253614B (en) | Diabetes risk early warning method based on big data analysis | |
| CN114003734B (en) | Knowledge system and knowledge map system of breast cancer risk factors and construction method | |
| CN117457192A (en) | Intelligent remote diagnosis method and system | |
| CN115359873B (en) | Control method for operation quality | |
| CN117457217B (en) | Risk assessment method and system for diabetic nephropathy | |
| CN117393098A (en) | Medical image report generation method based on visual priori and cross-modal alignment network | |
| Mansouri | Application of Neural Networks in the Medical Field | |
| CN115036034B (en) | Similar patient identification method and system based on patient characterization map | |
| CN116883768A (en) | Lung nodule intelligent grading method and system based on multi-modal feature fusion | |
| Bhardwaj et al. | Improved healthcare monitoring of coronary heart disease patients in time-series fashion using deep learning model | |
| Chantamit-O-Pas et al. | A case-based reasoning framework for prediction of stroke | |
| US11816539B1 (en) | Selection system for machine learning module for determining target metrics for evaluation of health care procedures and providers | |
| Saraswat | Knowledge Discovery With Hybrid Data Mining Approach | |
| Kaur et al. | A review on Naive Baye’s (NB), J48 and K-means based mining algorithms for medical data mining | |
| CN117079821B (en) | Patient hospitalization event prediction method | |
| US20240028831A1 (en) | Apparatus and a method for detecting associations among datasets of different types | |
| CN110689964B (en) | Health data sample searching method and system | |
| Saha et al. | Predicting missing and noisy links via neighbourhood preserving graph embeddings in a clinical knowlegebase | |
| Aman | Disease Prediction using Deep Learning Algorithms in Healthcare Sector | |
| CN118016279A (en) | Analysis diagnosis and treatment platform based on artificial intelligence multi-mode technology in breast cancer field | |
| Zhu et al. | Naïve Bayes classifier based on reliability measurement for datasets with noisy labels | |
| Sasirekha et al. | Identification and Classification of Leukemia Using Machine Learning Approaches |
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 |