CN119296812B

CN119296812B - Construction method of myocardial rehabilitation decision support system based on case-based reasoning

Info

Publication number: CN119296812B
Application number: CN202411783091.0A
Authority: CN
Inventors: 谢妹伊; 解建强; 马琳; 朱男
Original assignee: First Hospital Of Qinhuangdao
Current assignee: First Hospital Of Qinhuangdao
Priority date: 2024-12-06
Filing date: 2024-12-06
Publication date: 2025-03-07
Anticipated expiration: 2044-12-06
Also published as: CN119296812A

Abstract

The present invention relates to the field of data mining technology, and in particular to a method for constructing a myocardial rehabilitation decision support system based on case reasoning. The method comprises the following steps: collecting and integrating historical myocardial case data, desensitizing sensitive features, and obtaining desensitized data; dividing myocardial rehabilitation cases and integrating abnormal electrocardiogram waveform features to obtain abnormal waveform data of rehabilitation cases; constructing an abnormal waveform detection model, and using it to detect abnormal waveforms to obtain abnormal electrocardiogram waveform myocardial case data; performing trend analysis on abnormal waveform data and calculating myocardial rehabilitation influencing factors; evaluating the doctor's treatment priority based on the influencing factors, and constructing a treatment doctor recommendation model; combining the recommendation model and the detection model, constructing and optimizing the myocardial rehabilitation decision system, thereby obtaining a myocardial rehabilitation decision support system. The present invention is based on data mining to achieve the accuracy and effectiveness of myocardial rehabilitation research on historical myocardial cases.

Description

Method for constructing myocardial rehabilitation decision support system based on case reasoning

Technical Field

The invention relates to the technical field of data mining, in particular to a method for constructing a myocardial rehabilitation decision support system based on case-based reasoning.

Background

Myocardial rehabilitation (Myocardial Rehabilitation) is a rehabilitation therapy process for patients with cardiovascular disease, particularly after myocardial infarction, with the aim of improving the quality of life and survival of the patient through systemic medical intervention. The myocardial rehabilitation process involves individualized treatment plan design, often requiring a combination of information from multiple aspects such as patient history, symptoms, and rehabilitation effect. In traditional myocardial rehabilitation practice, the formulation of a therapeutic regimen often depends on experience and intuition of a doctor, and lack of systematic and personalized support can lead to effects of the therapeutic regimen varying from individual to individual of the doctor. Meanwhile, in the traditional myocardial rehabilitation practice, it is difficult to effectively integrate data (such as medical records, image data and laboratory examination results) from different sources, so that information islands and decision basis are incomplete.

Disclosure of Invention

Based on the above, the present invention needs to provide a method for constructing a myocardial rehabilitation decision support system based on case-based reasoning, so as to solve at least one of the above technical problems.

In order to achieve the above purpose, a method for constructing a myocardial rehabilitation decision support system based on case-based reasoning comprises the following steps:

Step S1, acquiring historical myocardial case data, carrying out patient case data fusion on the historical myocardial case data so as to obtain patient myocardial case fusion data, and carrying out desensitization on the patient sensitive characteristics of the patient myocardial case fusion data so as to obtain historical myocardial case desensitization data;

Step S2, carrying out myocardial recovery case intersection division on the historical myocardial department case desensitization data so as to obtain historical myocardial department recovery case data, and carrying out abnormal electrocardiogram waveform characteristic integration according to the historical myocardial department recovery case data so as to obtain recovery case abnormal electrocardiogram waveform data;

step S3, constructing an electrocardiogram abnormal waveform detection model according to the abnormal electrocardiogram waveform data of the rehabilitation cases, and detecting abnormal waveforms of the historical myocardial rehabilitation case data through the electrocardiogram abnormal waveform detection model so as to obtain abnormal electrocardiogram waveform myocardial case data;

Step S4, carrying out electrocardiographic waveform trend analysis according to the abnormal electrocardiographic waveform myocardial case data so as to obtain electrocardiographic waveform trend data, and carrying out myocardial rehabilitation influence factor calculation on the electrocardiographic waveform trend data according to the historical myocardial rehabilitation case data so as to obtain myocardial rehabilitation influence factors;

Step S5, evaluating treatment priority of the myocardial doctor according to the myocardial rehabilitation influence factors to obtain treatment priority data of the myocardial doctor, and constructing a recommendation model of the myocardial doctor according to the treatment priority data of the myocardial doctor and the myocardial rehabilitation influence factors;

and S6, constructing a myocardial rehabilitation decision system according to the myocardial department doctor recommendation model and the electrocardiogram abnormal waveform detection model, and carrying out iterative optimization parameter adjustment on the myocardial rehabilitation decision system according to the historical myocardial department case data so as to obtain the myocardial rehabilitation decision support system.

By acquiring and fusing the historical myocardial case data, the invention can integrate medical records, image data and laboratory examination results from different sources, thereby forming a comprehensive patient data file. This integration helps to overcome the information islanding, allowing better utilization of the data in the treatment decision process. After the desensitization treatment of the patient case data, the privacy of the patient can be effectively protected, and the safety of the data is ensured. This is critical for the study and analysis of sensitive medical data. The fused data has more consistency and integrity, and is favorable for subsequent analysis and model training, so that the reliability of research results is improved. By carrying out myocardial rehabilitation case intersection division on desensitization data, different cases related to myocardial rehabilitation can be screened out more accurately, data which are not related to research purposes are eliminated, and pertinence of the data is enhanced. The waveform characteristics of the abnormal electrocardiogram are integrated, so that the abnormal modes in the electrocardiogram can be recognized and analyzed, and the diagnosis capability of the electrocardiogram abnormality is improved. The obtained abnormal electrocardiographic waveform data of the rehabilitation cases can provide a basis for subsequent model construction and trend analysis, and ensure the depth and breadth of analysis. By constructing an electrocardiogram abnormal waveform detection model, abnormal waveforms can be automatically detected and identified, the detection efficiency and accuracy are improved, and the burden of manual analysis is reduced. The model can be trained by using a large amount of historical data, so that the accuracy of abnormal waveform detection is improved, and doctors are helped to recognize potential electrocardiographic abnormalities earlier. By analyzing the trend of the electrocardiogram waveform, a long-term pattern of electrocardiographic changes in the patient can be found, thereby helping to assess the effects and progress of myocardial rehabilitation. Calculating myocardial rehabilitation influence factors is helpful for identifying factors which have great influence on myocardial rehabilitation effect, and provides basis for establishing personalized treatment schemes. Based on the analysis results of the electrocardiogram waveform trend and the influence factors, the rehabilitation treatment scheme can be optimized, and the treatment effect is improved. The treatment priority evaluation is carried out on the historical cases, so that the system can be helped to preferentially select doctors to process the cases needing the intervention according to the specific conditions of patients, and the treatment efficiency and effect are improved. By constructing a myocardial department treatment doctor recommendation model, the most suitable doctor and treatment scheme can be recommended according to the specific condition of the patient, and the capability of personalized medical service is enhanced. The data-driven decision support provided by the recommendation model helps the physician make more scientific decisions on treatment options. The myocardial rehabilitation decision system integrates various data and models, can provide comprehensive decision support for doctors, and improves the scientificity and rationality of treatment schemes. By iterative optimization and parameter adjustment of the decision system, the accuracy and reliability of the system can be continuously improved, and the effectiveness of the system in practical application is ensured. The decision making system can accelerate the making process of the treatment scheme, reduce human errors and improve the recovery efficiency and quality of patients. Overall, these steps can significantly improve the individuation and scientificity of the myocardial rehabilitation study through systematic data processing and model construction.

Optionally, step S1 specifically includes:

Step S11, acquiring historical myocardial case data, and classifying the structural data of the historical myocardial case data so as to obtain structural myocardial case data and unstructured myocardial case data;

Step S12, unstructured patient myocardial feature fusion is carried out on unstructured myocardial case data, so that unstructured patient myocardial fusion data are obtained;

S13, carrying out physiological characteristic fusion of a patient treatment course according to the structured myocardial case data so as to obtain structured patient physiological fusion data;

Step S14, carrying out patient information case combination on unstructured patient myocardial fusion data and structured patient physiological fusion data so as to obtain patient myocardial case fusion data;

and step S15, desensitizing the patient sensitive characteristics of the patient myocardial case fusion data, so as to obtain historical myocardial case desensitization data.

The method for acquiring and classifying the historical myocardial case data is helpful for classifying the data into two major types of structuring and unstructured, and ensures that the subsequent processing is more efficient and accurate. The structured data can be used directly for analysis and modeling, while the classification of unstructured data lays a foundation for further processing. Feature fusion is carried out on unstructured myocardial case data, myocardial feature information can be extracted, and comprehensive utilization efficiency of the data is improved, so that accuracy and comprehensiveness of myocardial research are improved. The physiological characteristics of the patient are fused according to the structured myocardial case data, so that the physiological data of the patient can be integrated, a more comprehensive physiological characteristic description is formed, and the effectiveness of a prediction model is enhanced. The unstructured patient myocardial fusion data and the structured patient physiological fusion data are combined, so that myocardial and physiological information of a patient can be comprehensively integrated, more comprehensive patient case data is provided, and the establishment condition of personalized treatment scheme research is improved. And (3) performing sensitive characteristic desensitization on the fused patient myocardial case data to ensure the privacy and data safety of the patient, thereby maintaining the privacy trust of the patient and promoting the safe sharing and use of the data.

Optionally, step S12 specifically includes:

Step S121, medical record abstract data, patient electrocardiogram record extraction and medical image record extraction are carried out on unstructured myocardial case data, so that myocardial medical record abstract data, myocardial patient electrocardiogram collection and myocardial patient medical image collection are obtained;

Step S122, performing myocardial description characteristic integration on the medical record abstract data of the department of cardiology so as to obtain myocardial description data;

Step 123, extracting CT scanning myocardial dynamic structural features according to the medical image set of the patient in the department of cardiomyopathy, so as to obtain CT scanning myocardial dynamic structural data, and constructing a patient myocardial three-dimensional dynamic structural model according to the CT scanning myocardial dynamic structural data;

step S124, filling myocardial structural details into the patient myocardial three-dimensional dynamic structural model according to myocardial description data, thereby obtaining the patient myocardial dynamic structural model;

Step S125, carrying out electrocardiogram dynamic fluctuation feature integration according to an electrocardiogram set of a patient in the department of cardiology so as to obtain electrocardiogram dynamic fluctuation data, and carrying out myocardial dynamic motion simulation on the electrocardiogram dynamic fluctuation data through a patient myocardial dynamic structure model so as to obtain patient myocardial dynamic simulation data;

And step S126, carrying out dynamic feature fusion of the patient cardiac muscle according to the dynamic simulation data of the patient cardiac muscle, thereby obtaining unstructured patient cardiac muscle fusion data.

The invention extracts the medical record abstract, the electrocardiogram record and the medical image record of the myocardial department, is beneficial to systematically collecting and integrating all relevant data of a patient and provides basic data support for subsequent analysis. The myocardial description features are integrated, so that the myocardial features in the medical record data can be generalized to form unified myocardial description data, and subsequent analysis and comparison are facilitated. The CT scanning myocardial dynamic structural features are extracted, a three-dimensional dynamic structural model is constructed, detailed structural information of the myocardium in space can be provided, and accurate three-dimensional analysis and simulation are facilitated. And carrying out detail filling on the three-dimensional dynamic structure model according to myocardial description data to ensure that the model is more real and accurate, thereby enhancing the reliability of simulation and analysis. And the electrocardiogram dynamic fluctuation characteristics are integrated, and the motion simulation is carried out through a myocardial dynamic structure model, so that the electrocardiogram data and the myocardial structure can be dynamically associated, and the real-time change of myocardial functions can be studied. The feature fusion is carried out according to the dynamic myocardial simulation data, so that the information of various data sources can be integrated into a comprehensive view, and the comprehensive understanding of the myocardial state of a patient and the accuracy of myocardial research are improved.

Optionally, step S13 specifically includes:

step S131, performing medicament dosage record extraction, disease course record extraction and physiological coefficient record extraction on the structured myocardial case data so as to obtain patient medicament dosage data, patient disease course data and patient physiological coefficient data;

Step S132, dividing treatment stages of the patient according to the patient course data so as to obtain treatment stage data of the patient;

step S133, analyzing the medicament dosage data to obtain medicament dosage characteristic data;

Step S134, carrying out physiological coefficient fluctuation association according to the medicament dosage characteristic data and the patient physiological coefficient data so as to obtain medicament dosage associated physiological fluctuation data;

And step S135, carrying out treatment stage physiological characteristic fusion on the patient treatment stage data according to the medicament dosage associated physiological fluctuation data so as to obtain structured patient physiological fusion data.

The extraction of agent dose records, course records, and physiological coefficient records of the present invention facilitates the disassembly of structured myocardial case data into specific, operational sub-data sets. The extraction provides basic data for subsequent detailed analysis, including the use condition of the medicine, the progress of the disease course and the physiological state, and lays a data foundation for accurate treatment analysis. Treatment staging based on patient course data helps to systemize the patient's treatment process. This process can identify different phases of treatment (e.g., acute phase, convalescence, etc.), thereby providing a support for staged data analysis of the relationship between dosage of agent and physiological response, helping to study the treatment regimen in the case. The dosage data is analyzed for the characteristics of the medicament, so that the dosage characteristics of different medicaments can be revealed, including the frequency of medicament administration, dosage change and the like. Such analysis helps to understand the actual application of the drug in different patients, thereby optimizing the study of drug treatment regimens and dose adjustment strategies. The relationship between the drug dosage change and the physiological parameter can be identified by carrying out physiological coefficient fluctuation correlation according to the drug dosage characteristic data and the physiological coefficient data of the patient. Such correlation analysis helps to understand the effects of the drug on physiological fluctuations, evaluate the drug effects and their side effects, and thus support research into personalized therapies. The physiological characteristic fusion is carried out on the patient treatment stage data according to the medicament dosage associated physiological fluctuation data, so that the relationship between the medicament dosage and physiological response in different treatment stages can be integrated. The fusion provides a comprehensive physiological data view, supports deeper treatment effect evaluation and research, and improves the research effect of the whole treatment scheme.

Optionally, step S2 specifically includes:

Step S21, performing treatment recovery stage feature extraction on the historical myocardial case desensitization data to obtain historical myocardial case treatment stage data, and performing treatment stage length classification on the historical myocardial case treatment stage data to obtain long treatment stage case data and short treatment stage case data;

step S22, carrying out treatment stage review classification on the historical myocardial case treatment stage data so as to obtain case treatment stage review data;

Step S23, carrying out review myocardial variable statistics according to the case treatment stage review data so as to obtain review myocardial variable data, and carrying out myocardial variable fluctuation amplitude classification on the review myocardial variable data so as to obtain high-amplitude fluctuation myocardial variable review data and low-amplitude fluctuation myocardial variable review data;

Step S24, classifying the high-amplitude fluctuation myocardial variable review data and the low-amplitude fluctuation myocardial variable review data according to the historical myocardial case treatment stage data, so as to obtain high-amplitude fluctuation treatment stage high-duty ratio case data and low-amplitude fluctuation treatment stage high-duty ratio case data;

Step S25, performing treatment case intersection operation on the long treatment stage case data and the low-amplitude fluctuation treatment stage high-duty ratio case data to obtain slow-effect treatment rehabilitation case data;

Step S26, performing treatment and rehabilitation case integration on the slow-effect treatment and rehabilitation case data and the quick-effect treatment and rehabilitation case data so as to obtain historical myocardial rehabilitation case data;

And step S27, carrying out abnormal electrocardiogram waveform characteristic integration according to the historical myocardial rehabilitation case data so as to obtain the rehabilitation case abnormal electrocardiogram waveform data.

The invention can structure the complex treatment process data in the case by extracting the treatment recovery stage characteristics, and provides a clear data base for the subsequent analysis. The treatment stage data are divided into a long treatment stage and a short treatment stage, so that the influence of different stages on the rehabilitation effect can be understood, and a reference is provided for personalized treatment. Further categorizing the treatment phase data into review data helps focus on post-treatment review results in the case, which is critical for assessing treatment efficacy and further improving treatment strategies. By classifying the review data, the change and trend in the review process can be clearly identified, and the tracking and management capacity of the recovery condition of the patient in the case is improved. The myocardial variables are subjected to statistics and fluctuation range classification, so that the change condition of the myocardial variables in the center of the case can be deeply known, and the treatment effect can be accurately estimated. By differentiating between high-amplitude and low-amplitude fluctuating myocardial variables, key indicators that have a significant impact on patient recovery can be identified and treatment regimens targeted. The myocardial variables with high amplitude and low amplitude fluctuation are classified according to the duty ratio, so that the influence of different fluctuation amplitudes on the treatment stage in the case can be known. By analyzing the ratio of myocardial variable fluctuation, the part possibly needing to be adjusted in treatment can be identified, and the effectiveness and accuracy of treatment scheme research are improved. The slow effect and the fast effect treatment case data are separated through intersection operation, so that the effects of different treatment methods can be clarified, and the optimal treatment scheme can be researched and optimized. The slow-acting and fast-acting treatment rehabilitation case data are integrated, comprehensive rehabilitation information in the case set can be summarized, and a foundation is provided for further analysis and research. By integrating various rehabilitation case data, the historical myocardial rehabilitation situation can be comprehensively evaluated, and common and individual problems in treatment can be found. By integrating the waveform characteristics of the abnormal electrocardiogram, potential abnormality or complications in the case can be found, and basis is provided for timely intervention.

Optionally, step S27 specifically includes:

step 271, carrying out electrocardiographic record extraction on the historical myocardial rehabilitation case data so as to obtain a slow-effect treatment rehabilitation case electrocardiograph set and a fast-effect treatment rehabilitation case electrocardiograph set;

step 272, performing Fourier transform on the slow-effect treatment recovery case electrocardiograph set and the fast-effect treatment recovery case electrocardiograph set respectively, so as to obtain a slow-effect recovery case electrocardiograph spectrum and a fast-effect recovery case electrocardiograph spectrum;

Step S273, respectively carrying out frequency spectrum amplitude threshold statistics on the slow-effect rehabilitation case electrocardiograph frequency spectrum and the fast-effect rehabilitation case electrocardiograph frequency spectrum, so as to obtain a slow-effect frequency spectrum amplitude threshold and a fast-effect frequency spectrum amplitude threshold;

Step S274, performing abnormal frequency spectrum amplitude classification on the slow-effect recovery case electrocardiograph frequency spectrum according to the slow-effect frequency spectrum amplitude threshold value so as to obtain an abnormal slow-effect recovery case electrocardiograph frequency spectrum;

Step 275, respectively performing time domain conversion on the abnormal slow effect recovery case electrocardiograph frequency spectrum and the abnormal fast effect recovery case electrocardiograph frequency spectrum, so as to obtain abnormal slow effect recovery case electrocardiograph data and abnormal fast effect recovery case electrocardiograph data;

Step S276, carrying out abnormal electrocardiogram waveform characteristic integration on the abnormal slow effect recovery case heart electric wave data and the abnormal fast effect recovery case heart electric wave data, thereby obtaining recovery case abnormal electrocardiogram waveform data.

By extracting the electrocardiogram records, the invention can know the electrocardiogram characteristics of slow effect and fast effect treatment in detail and provide basic data for subsequent analysis. The Fourier transform helps to convert the time domain signal into the frequency domain signal, so that the frequency spectrum analysis becomes visual, and the electrocardiogram frequency characteristics under different treatment effects in the case are revealed. The frequency spectrum amplitude threshold statistics can quantify abnormal fluctuation of the electrocardiogram, provide standardized indexes for identifying the abnormality, and help to find differences of treatment effects. The abnormal spectrum amplitude classification can accurately identify the abnormal characteristics of the electrocardiogram after treatment, and help to distinguish abnormal conditions under different treatment effects. The abnormal frequency spectrum data is converted into time domain data, so that the abnormal fluctuation characteristics are clearer, and the time domain change of the electrocardiogram is convenient to further analyze and understand. The integration of the abnormal waveform features can sum up specific abnormal patterns, help evaluate the treatment effect, and provide data support for formulating a more effective rehabilitation strategy.

Optionally, step S4 specifically includes:

Step S41, carrying out electrocardiographic waveform signal segmentation according to the abnormal electrocardiographic waveform myocardial case data, thereby obtaining abnormal electrocardiographic waveform signal data;

Step S42, performing time sequence trend line fitting on the abnormal electrocardiogram waveform signal data so as to obtain electrocardiogram waveform signal trend data;

Step S43, carrying out fluctuation measurement classification on the electrocardiographic waveform signal trend data so as to obtain excessive fluctuation waveform signal trend data and normal fluctuation waveform signal trend data;

Step S44, carrying out waveform signal trend correlation on the abnormal electrocardiogram waveform signal data according to the excessive fluctuation waveform signal trend data so as to obtain electrocardiogram waveform degradation trend data;

Step S45, carrying out data combination on the electrocardiogram waveform degradation trend data and the electrocardiogram waveform improvement trend data so as to obtain electrocardiogram waveform trend data;

And step S46, calculating myocardial rehabilitation influence factors according to the electrocardiogram waveform trend data according to the historical myocardial rehabilitation case data, so as to obtain the myocardial rehabilitation influence factors.

The invention can accurately extract abnormal signal data by dividing the electrocardiogram waveform signals, and provides clear abnormal signal basis for subsequent analysis. The time sequence trend line fitting helps to reveal the long-term trend of the electrocardiogram waveform signal, so that the change mode of the waveform is more visual and understandable. The fluctuation measurement classification divides the signal into excessive fluctuation and normal fluctuation, which is helpful for distinguishing the severity of abnormal waveform and normal variation, and improves the accuracy of analysis. The abnormal signals are subjected to degradation and good trend correlation, so that the influence of different fluctuation modes on the electrocardiogram waveform can be understood, and the deterioration or improvement of the illness state can be identified. The data combination provides comprehensive electrocardiogram waveform trend data, comprehensively considers degradation and good trend, and is convenient for overall evaluation of the development of myocardial rehabilitation. The myocardial rehabilitation influence factors can be calculated to evaluate the effects of different treatment schemes based on the historical case data, and scientific basis is provided for optimizing the rehabilitation strategy.

Optionally, step S46 specifically includes:

Step S461, extracting left ventricular ejection fraction characteristics and electrocardiogram waveform characteristics of the historical myocardial recovery case data so as to obtain recovery case left ventricular ejection fraction data and recovery case electrocardiogram waveform data;

step S462, constructing a myocardial recovery index evaluation system based on recovery case left ventricular ejection fraction data and recovery case electrocardiogram waveform data, and performing myocardial recovery index evaluation according to the myocardial recovery index evaluation system so as to obtain myocardial recovery index evaluation data;

step 463, extracting waveform amplitude characteristics of the electrocardiogram waveform trend data, thereby obtaining trend waveform amplitude data;

Step S464, constructing a myocardial rehabilitation linear regression model according to myocardial rehabilitation index evaluation data and trend waveform amplitude data;

And step S465, calculating myocardial rehabilitation influence factors according to the myocardial rehabilitation linear regression model, so as to obtain the myocardial rehabilitation influence factors.

The invention can comprehensively understand key physiological and electrophysiological indexes of myocardial rehabilitation by extracting left ventricular ejection fraction and electrocardiogram waveform characteristics, and provides basic data for analysis and evaluation. The myocardial rehabilitation index evaluation system is constructed and evaluated, so that the rehabilitation effect in the case can be systematically evaluated, and the key physiological and electrophysiological changes in the rehabilitation process can be determined. The waveform amplitude characteristics are extracted, so that the change characteristics of the electrocardiogram trend data are clearer, and the specific change condition of the electrocardiogram waveform can be evaluated. By constructing a myocardial rehabilitation linear regression model and combining the evaluation data with the trend waveform amplitude data, the specific action mechanism of myocardial rehabilitation influence factors can be revealed, and a quantitative analysis result is provided. The myocardial rehabilitation influence factors are calculated based on the linear regression model, so that the actual effect of different treatments on myocardial rehabilitation in the case can be quantified, and data support is provided for optimizing the treatment scheme.

Optionally, step S5 specifically includes:

step S51, carrying out electrocardiographic waveform feature extraction according to the myocardial rehabilitation influence factors so as to obtain myocardial rehabilitation electrocardiographic waveform influence factors;

Step S52, carrying out feature extraction of a cardiology therapist and feature extraction of a case electrocardiogram on the historical cardiology case data, thereby obtaining cardiology therapist data and case electrocardiogram data;

Step S53, carrying out electrocardiographic waveform similarity calculation on the case electrocardiographic data and the myocardial rehabilitation electrocardiographic waveform influencing factors so as to obtain case electrocardiographic similarity data, and carrying out high-similarity case statistics on the case electrocardiographic similarity data so as to obtain high-similarity case electrocardiographic data;

Step S54, carrying out treatment case association according to the high-similarity case electrocardiogram data and the myocardial department treatment doctor data so as to obtain treatment doctor-electrocardiogram associated data;

Step S55, carrying out treatment priority evaluation of the cardiologist according to the relevant data of the treating doctor and the electrocardiogram so as to obtain treatment priority data of the cardiologist;

And step S56, constructing a myocardial doctor recommendation model according to myocardial doctor treatment priority data and myocardial rehabilitation influence factors.

According to the invention, by extracting the electrocardiogram waveform characteristics according to the myocardial rehabilitation influence factors, the specific waveform influence factors aiming at the rehabilitation effect can be obtained. This process provides individualized electrocardiogram feature data that facilitates more accurate assessment and monitoring of electrocardiogram changes during case-centric muscle rehabilitation. The characteristics of different therapists and the specific conditions of the case electrocardiogram can be comprehensively known by carrying out the feature extraction of the cardiology therapists and the feature extraction of the case electrocardiogram on the historical cardiology case data. This process helps build a detailed profile between the physician and patient electrocardiographic data, supporting subsequent correlation analysis. And calculating the similarity between the case electrocardiogram data and the myocardial rehabilitation electrocardiogram waveform influence factor, and counting the cases with high similarity, so that the electrocardiogram case closest to the target waveform influence factor can be identified. This process can help to find similar cases, providing targeted therapeutic and rehabilitation regimens. The treatment case association is performed according to the high-similarity case electrocardiogram data and the myocardial department treatment doctor data, so that the treatment scheme and the treatment effect of which treatment doctors process similar electrocardiogram data can be clarified. This procedure helps to reveal the actual operation and effect of different doctors in handling similar cases, thus providing basis for personalized treatment. The treatment priority evaluation of the cardiologist is performed by the treatment doctor-electrocardiogram related data, and the doctors can be ordered according to the treatment effect of the doctors and the similarity of cases. This procedure can help hospitals or medical facilities optimize resource allocation, assigning efficient physicians to cases that are preferentially needed for treatment. According to the treatment doctor recommendation model constructed by the myocardial doctor treatment priority data and the myocardial rehabilitation influence factors, the most suitable myocardial doctor can be recommended for the patient. The process can ensure that the patient obtains the optimal medical service, improves the treatment effect, and improves the efficiency and quality of the whole myocardial rehabilitation.

Optionally, step S55 specifically includes:

Step S551, according to the doctor-electrocardiogram related data, carrying out doctor treatment times statistics and doctor-related electrocardiogram statistics so as to obtain doctor treatment times data and doctor-related electrocardiogram data;

Step S552, classifying the doctor treatment times data into doctor treatment times data, thereby obtaining doctor data with high treatment times and doctor data with low treatment times;

step S553, carrying out the associated electrocardiogram quantity classification on the doctor associated electrocardiogram data so as to obtain high electrocardiogram associated quantity doctor data and low electrocardiogram associated quantity doctor data;

Step 554, performing doctor intersection calculation on the doctor data with high treatment times and the doctor data with high electrocardiograph correlation number to obtain the doctor data with high treatment priority of the myocardial doctor;

Step S555, combining the data according to the high treatment priority myocardial doctor data and the low treatment priority myocardial doctor data, thereby obtaining the myocardial doctor treatment priority data.

The invention can determine the treatment frequency of each doctor and the associated electrocardiogram quantity through the association statistics of the treatment doctor and the electrocardiogram data. This step lays the foundation for the subsequent data classification and analysis, helping to identify the workload and professional practice conditions of different doctors. The doctor treatment times are classified, so that doctors with high treatment frequency and doctors with low treatment frequency can be distinguished. This allows the identification of the more therapeutically active physicians, providing basis for resource allocation and prioritization. The electrocardiogram related data of doctors are classified, so that doctors with high electrocardiogram related quantity and doctors with low electrocardiogram related quantity can be distinguished. This step reveals the physician's liveness in the case of an electrocardiogram, helping to assess his professional ability in the case of an electrocardiogram. Through intersection calculation, the doctors with high treatment times and high electrocardiogram association number are combined, so that the doctors with strong comprehensive treatment capacity can be identified, and the doctors with high treatment priority can be marked. Intersection calculations for a low number of treatment sessions and low electrocardiogram related numbers can be helpful in marking out low priority physicians. The classification can optimize resource allocation and improve treatment efficiency. Combining the high priority and low priority doctor data to obtain comprehensive treatment priority data. This allows for better physician management and scheduling, ensuring that high priority physicians get more resources and support, thereby improving overall treatment quality and efficiency.

Drawings

Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of a non-limiting implementation, made with reference to the accompanying drawings in which:

FIG. 1 is a flow chart of steps of a method for constructing a myocardial rehabilitation decision support system based on case-based reasoning;

FIG. 2 is a detailed step flow chart of step S1 of the present invention;

FIG. 3 is a detailed flowchart of step S12 of the present invention;

The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The following is a clear and complete description of the technical method of the present invention, taken in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

To achieve the above objective, referring to fig. 1 to 3, the present invention provides a method for constructing a myocardial rehabilitation decision support system based on case reasoning, the method comprising the following steps:

In this embodiment, historical patient case data of the cardiology department of the hospital is collected, which includes patient basic information, diagnostic records, treatment regimens, efficacy, and the like. These data are then integrated together by data fusion techniques to form patient cardiomyopathy case fusion data. This process involves unifying the formats of data from different sources (e.g., outpatient records, inpatient records, etc.), and removing duplicate data. Then, the fused data are subjected to desensitization processing of sensitive information, for example, sensitive fields such as names, identification card numbers, contact ways and the like are replaced by anonymous identifiers, so that privacy protection of patients is ensured. The resulting historical myocardial case desensitization data can be used for further analysis and model training.

in this embodiment, the intersection of myocardial rehabilitation cases is divided using the desensitized historical myocardial case data. Specifically, the patient cases subjected to the myocardial rehabilitation process are screened out, and the data of the patient cases are extracted and divided to form historical myocardial rehabilitation case data. Then, electrocardiographic waveform data are extracted from the data, waveform characteristics are integrated by combining expert knowledge and a machine learning algorithm, and abnormal electrocardiographic waveforms are identified. This includes analyzing peaks, bands, timing characteristics, etc. of the electrocardiogram waveform, labeling the abnormal waveform in combination with known abnormal patterns (e.g., ST elevation, T wave inversion), and generating recovery case abnormal electrocardiogram waveform data.

In this embodiment, an electrocardiographic abnormal waveform detection model is constructed based on the recovery case abnormal electrocardiographic waveform data. The electrocardiographic data may be feature extracted and classified using a deep learning method, such as Convolutional Neural Network (CNN) or Recurrent Neural Network (RNN). By training the model, abnormal waveforms in the electrocardiogram are detected and labeled. The training data includes electrocardiographic samples labeled with abnormal and normal waveforms. After training, the model can detect abnormal waveforms of the historical myocardial rehabilitation case data, identify potential abnormal electrocardiogram waveforms and generate abnormal electrocardiogram waveform myocardial case data.

In this embodiment, electrocardiographic waveform trend analysis is performed on abnormal electrocardiographic waveform myocardial case data. Time series analysis methods (such as moving average method and exponential smoothing method) are used to evaluate the variation trend of electrocardiogram waveform. For example, analysis of trends in ST segments in an electrocardiogram identifies potential signs of myocardial ischemia. And then, calculating the influence factor of the electrocardiogram waveform trend on myocardial rehabilitation by combining the historical myocardial rehabilitation case data. The influence degree of different waveform characteristics on the rehabilitation effect can be determined through a statistical analysis method, such as regression analysis, and myocardial rehabilitation influence factors are generated.

In this embodiment, doctor treatment priority evaluation is performed on the historical myocardial case data according to the calculated myocardial rehabilitation impact factor. The specific method includes combining the impact factors with data such as treatment history of the doctor, and using a weighted scoring system to determine treatment priority for each doctor. Next, a cardiologist recommendation model is constructed based on these priority data and myocardial rehabilitation impact factors. Algorithms such as decision trees, random forests, etc. can be used to predict and recommend the most appropriate doctor to improve the treatment outcome.

In this embodiment, a comprehensive myocardial rehabilitation decision system is constructed based on a cardiologist recommendation model and an electrocardiographic abnormal waveform detection model. The system may include portions of a user interface, a data input module, a model interface, and the like. First, the system functional requirements need to be determined, including user interfaces, data input modules, model interfaces, etc. And then designing a system architecture, and defining the functions and interfaces of the functional modules. The myocardial department therapist recommended model and the electrocardiogram abnormal waveform detection model are integrated in a system architecture, and then parameter adjustment, test and optimization of the system are carried out, so that a myocardial rehabilitation decision system is obtained. In the practical application process, the system performs iterative optimization according to new historical myocardial case data, and the accuracy and the practicability of the model are improved by continuously adjusting model parameters (such as learning rate, regularization coefficient and the like) and introducing new features, so that a high-efficiency myocardial rehabilitation decision support system is finally formed.

Optionally, step S1 specifically includes:

In this embodiment, acquiring historical myocardial case data includes extracting Electronic Health Record (EHR) and Electrocardiogram (ECG) data from a hospital database. The data is classified according to a format, for example, structured data (e.g., medical records, laboratory test results) and unstructured data (e.g., physician handwritten notes, image files) are stored separately. In particular implementations, electrocardiographic images and text annotations are converted into structured fields using a data annotation tool and stored in an SQL database, while images and text files are saved in a file system to form structured and unstructured myocardial case data.

In this embodiment unstructured myocardial case data (e.g., electrocardiographic images and doctor notes) are processed. Feature extraction is performed on electrocardiographic images using image processing techniques (e.g., convolutional neural networks, CNNs), while Natural Language Processing (NLP) techniques are applied to semantically analyze physician notes. And integrating the extracted characteristic data with a text analysis result through characteristic fusion to form unstructured patient myocardial fusion data, for example, integrating abnormal waveform characteristics in an image and illness state description in notes into a unified data set.

In this embodiment, the fusion of physiological characteristics of the treatment course is performed according to information (such as physiological parameters of the patient and treatment records) in the structured myocardial case data. The physiological parameters are subjected to dimensionality reduction and feature selection by using a statistical analysis and a machine learning method (such as principal component analysis and PCA) to form structured patient physiological fusion data. For example, the main features are extracted from heart rate and blood pressure data of multiple examinations, and a comprehensive physiological index data set is synthesized.

in this embodiment, unstructured patient myocardial fusion data is combined with structured patient physiological fusion data. Data from different sources (electrocardiogram features and physiological features) are integrated together using data fusion techniques, such as multi-modal data fusion methods. By constructing a joint database or data table, the two types of data are associated according to the patient ID to form comprehensive patient myocardial case fusion data, for example, the fused electrocardiogram features and physiological features are combined into a unified data set for further analysis.

In this embodiment, the patient myocardial case fusion data is subjected to a sensitive feature desensitization process. Personal identification information (e.g., name, identification card number) of the patient is removed from the dataset or replaced using data desensitization techniques, such as data masking or data encryption. For example, anonymizing tools are used to convert personal identity information to a universal identifier, thereby generating historical cardiomyopathy case desensitization data to preserve patient privacy.

Optionally, step S12 specifically includes:

in this embodiment, medical record summary data of the cardiomyopathy patient is extracted from unstructured myocardial case data, including complaints, diagnosis and treatment records. Next, an electrocardiogram record of the patient stored in DICOM format is extracted. Finally, medical image records of the patient, such as cardiac CT scan images, are extracted and format converted for processing.

In this embodiment, main description features (such as myocardial thickness, myocardial function, etc.) of the myocardium are extracted from the medical record summary data of the department of cardiology, and the features are integrated and generalized by using text mining technology to form structured myocardial description data. For example, myocardial related descriptions in medical records are identified by natural language processing techniques and converted to a standardized list of features.

In this embodiment, the CT scan image of the patient of the cardiology department is processed to extract dynamic structural features of the myocardium, such as the contraction and relaxation of the myocardium. The myocardial tissue is separated by utilizing an image segmentation technology, and a three-dimensional reconstruction technology is applied to generate a myocardial three-dimensional dynamic structure model, so that the structural change of the myocardial under different cardiac cycles can be accurately reflected.

in this embodiment, the myocardial three-dimensional dynamic structural model of the patient's myocardium is populated in detail using myocardial description data. By comparing the myocardial function description in the medical record with the existing data of the model, the details of myocardial thickness, morphology and the like in the model are adjusted so as to obtain a more accurate patient myocardial dynamic structure model.

In this embodiment, dynamic fluctuation feature analysis such as amplitude, frequency, and rhythm variation of an electrocardiogram is performed on the extracted electrocardiogram data. By combining with a myocardial dynamic structure model of a patient, the influence of electrocardiogram fluctuation on the myocardial dynamic motion is simulated, so that myocardial dynamic simulation data are generated, and the actual functional state of the cardiac muscle is estimated.

In this embodiment, feature fusion is performed by combining the dynamic myocardial simulation data and the myocardial description data of the patient. Dynamic characteristics of different sources are comprehensively analyzed through a data fusion technology to form a comprehensive unstructured myocardial fusion data set for further clinical analysis and personalized treatment scheme formulation.

Optionally, step S13 specifically includes:

In this embodiment, the dose records, course records, and physiological coefficient records are extracted from the structured myocardial case data. Specifically, the dose record of the medicament is extracted as the dose and the administration time of each administration, the course record is extracted as the disease progress and symptom change of each examination, and the physiological coefficient record is extracted as the physiological data of the patient at the time of examination, such as blood pressure, heart rate and the like. For example, the data stored in the electronic health record system is analyzed to obtain the dosage of the medicament, the disease description and the physiological parameter data of a certain patient at a certain time point, so as to form a structured patient data set.

In this embodiment, the treatment phases are divided according to patient course data. The specific method is to divide the whole treatment process into different stages such as early stage, middle stage and late stage according to the time stamp and the disease state description in the course record. For example, if the course of the disease records show that the patient had stable disease in the first three months, worsening in the middle, and improving in the late stages, the three periods may be marked as early, middle, and late stages, respectively, and the specific time frame and disease characteristics of each stage recorded.

In this example, the drug characteristic analysis was performed on the drug amount data. Specific analysis includes calculation of average dosage, frequency of administration, dosage range, etc. of the agent. For example, by statistical analysis of the data of use of drugs against heart failure, it was determined that the average daily dose of a certain drug was 50mg, the dose range was 30-70mg, and the frequency of use was once per day. These data help to understand the nature and variability of the agent in its actual application.

In this embodiment, the physiological coefficient fluctuation correlation analysis is performed based on the medicament dose characteristic data and the patient physiological coefficient data. In particular, the relationship between the change in dosage of the agent and the fluctuation of the physiological coefficient is analyzed statistically, for example, regression analysis or correlation analysis is used to determine whether the increase in dosage of the agent is related to a change in a certain physiological index such as blood pressure. For example, if a patient is found to have an increased fluctuation range of blood pressure after increasing a dose of a certain agent, correlation data between the dose of the agent and the fluctuation of blood pressure is recorded.

In this embodiment, the patient treatment phase data is subject to treatment phase physiological feature fusion using dose-related physiological fluctuation data. Specifically, the related data of the dosage and the physiological fluctuation of the medicament are combined with the data of the treatment stages, and the physiological characteristics of each treatment stage are extracted. For example, by integrating the dosage of the agent in the metaphase stage with the physiological fluctuation data, the average physiological index characteristic of the stage is obtained. Finally, a structured patient physiological fusion dataset is formed, including the integrated physiological characteristics of each stage, for further analysis and decision-making.

Optionally, step S2 specifically includes:

In this embodiment, raw data of historical myocardial cases including electrocardiography, medical records, and treatment protocols of the patient are obtained. By applying data preprocessing algorithms (e.g., denoising, normalization), treatment phase related features such as treatment duration, electrocardiogram parameters, drug dose, etc. are extracted. The treatment phase lengths are classified using a machine learning algorithm (e.g., support vector machine SVM or random forest) to divide the treatment phase into long phases (e.g., more than 6 months) and short phases (e.g., less than 6 months). The classification results obtained will be long treatment phase case data and short treatment phase case data.

in this embodiment, the treatment phase data is reviewed and classified. The review may be performed by examining the review date in the patient record, the electrocardiogram waveform, and the physician's review advice. And marking the re-checking data by adopting a classification algorithm (such as a K nearest neighbor algorithm KNN or a decision tree) and dividing the re-checking data into two types of 'checked' and 'not checked', thereby obtaining the re-checking data of the case treatment stage.

In this embodiment, statistical analysis is performed on myocardial variables (e.g., heart rate, ST segment changes, etc.) using the review data in the historical myocardial cases. The magnitude of the fluctuation (e.g., standard deviation or range) of the myocardial variable during the review is calculated. Based on the statistical results, the review myocardial variables are classified into high-amplitude fluctuations (e.g., standard deviation greater than 0.5 mV) and low-amplitude fluctuations (e.g., standard deviation less than 0.5 mV). The generated classification results will include high-amplitude fluctuating myocardial variable review data and low-amplitude fluctuating myocardial variable review data.

In this embodiment, the high-amplitude fluctuation and low-amplitude fluctuation myocardial variable review data is further classified according to the treatment phase data in the historic myocardial cases. The proportion of each fluctuation amplitude type that occurs during the treatment phase (e.g., the proportion of high amplitude fluctuation to the amplitude fluctuation throughout the treatment phase) is calculated. Cases are classified using a proportional threshold (e.g., 70%) to yield high-duty-cycle case data for high-amplitude fluctuation therapy phases and high-duty-cycle case data for low-amplitude fluctuation therapy phases.

in this embodiment, the intersection operation is performed on the case data of the long treatment stage and the high duty ratio case data of the low-amplitude fluctuation treatment stage, so as to obtain the slow-effect treatment rehabilitation case data. And performing intersection operation on the short-treatment-stage case data and the high-amplitude fluctuation-treatment-stage high-duty-ratio case data to obtain quick-effect treatment rehabilitation case data. Common features of these cases are determined using a collective operation method, such as a collective intersection operation.

In this embodiment, the slow-acting therapeutic rehabilitation case data and the fast-acting therapeutic rehabilitation case data are integrated. In particular, the integration process includes data cleansing, normalization, and feature matching. For example, if the slow-acting therapeutic recovery case data includes an electrocardiogram feature A, B, C and a treatment time X and the fast-acting therapeutic recovery case data includes an electrocardiogram feature D, E, F and a treatment time Y, then the representations of these features need to be unified and combined into a unified dataset during the integration process to generate the historical myocardial recovery case data. These integrated data will be used for further analysis and modeling.

In this embodiment, the integration of the abnormal electrocardiographic waveform features is performed according to the historical myocardial rehabilitation case data. Specifically, it is first necessary to extract all electrocardiographic waveform data from the integrated dataset, and then apply an abnormality detection algorithm (e.g., a machine-learning-based classification model or statistical analysis method) to identify and flag the abnormal waveform. For example, by comparing an electrocardiogram waveform in the rehabilitation case data with a normal waveform model, an abnormal waveform pattern can be detected. Finally, the abnormal electrocardiographic waveform data are organized into a data set, labeled as recovery case abnormal electrocardiographic waveform data. These data will be used for further analysis of abnormal patterns and clinical studies.

Optionally, step S27 specifically includes:

in this embodiment, electrocardiographic records of the slow-acting therapeutic recovery case and the fast-acting therapeutic recovery case are extracted from the historical myocardial recovery case data, so as to ensure that the data contains a plurality of electrocardiographic cycles. The records are classified to form a slow-effect treatment recovery case electrocardiograph set and a fast-effect treatment recovery case electrocardiograph set.

In this embodiment, fourier transform is applied to the two sets of electrocardiographic data. The fourier transform converts the time domain signal into a frequency domain signal for analysis of its spectral characteristics. For example, each electrocardiogram data is Fourier transformed using a scipy. Fft library in Python, generating a slow-effect rehabilitation case electrocardiograph spectrum and a fast-effect rehabilitation case electrocardiograph spectrum.

In this embodiment, the magnitude threshold for each set of spectra is counted. And respectively carrying out statistical analysis on the slow-effect rehabilitation case electrocardiograph frequency spectrum and the fast-effect rehabilitation case electrocardiograph frequency spectrum, calculating the amplitude distribution of each frequency spectrum, and setting a threshold value. For example, a 95% quantile of the spectral amplitude is used as the threshold. This will obtain a slow and a fast spectral amplitude threshold, respectively.

In this embodiment, the abnormal spectrum amplitude classification is performed according to the slow spectrum amplitude threshold and the fast spectrum amplitude threshold. For slow recovery cases, the spectrum with amplitude exceeding the threshold is marked as abnormal. Similarly, the same classification is made for quick recovery cases. Recording the abnormal frequency spectrums for subsequent analysis, and respectively generating an abnormal slow effect recovery case electrocardio frequency spectrum and an abnormal fast effect recovery case electrocardio frequency spectrum.

in this embodiment, the abnormal slow-effect recovery case electrocardiograph spectrum and the abnormal fast-effect recovery case electrocardiograph spectrum are respectively subjected to time domain conversion. An inverse fourier transform is applied to convert the abnormal spectrum back into time domain data. For example, inverse transformation using scipy. Fft. Ifft, obtaining the abnormal slow effect rehabilitation case heart electric wave data and the abnormal fast effect rehabilitation case heart electric wave data.

In this embodiment, the abnormal electrocardiographic waveform features are integrated. Waveform characteristic extraction is carried out on the obtained abnormal time domain data, for example, the characteristics such as R wave peak value, interval and the like of the waveform are detected, and abnormal electrocardiogram waveform data is formed through integration. These features are extracted and classified using an image processing tool, ultimately generating recovery case abnormal electrocardiographic waveform data.

Optionally, step S4 specifically includes:

In this embodiment, during electrocardiographic monitoring, an electrocardiographic instrument is first used to acquire electrocardiographic waveform signals. For continuous electrocardiogram signals in abnormal electrocardiogram waveform myocardial case data, an adaptive thresholding algorithm is used to segment the signals. These thresholds are determined by a series of predetermined criteria (e.g., R-wave peak, QRS complex). By detecting and locating the R wave in the electrocardiogram waveform, the electrocardiogram signal is divided into a plurality of heart cycle signal segments. Each signal segment represents a complete heart cycle, thereby extracting electrocardiographic waveform signal data containing anomalies. These anomaly signals may be characteristic of ST elevation, T wave inversion, etc., for subsequent processing.

In this embodiment, the obtained abnormal electrocardiographic waveform signal data is subjected to time-series trend line fitting by using a least square method. By plotting the signal points in each cycle on the time axis and fitting a trend line with a linear regression algorithm, the overall trend of the electrocardiogram signal over a certain time range can be described. Such trend lines may be linear or nonlinear, with particular choice being based on the fluctuating nature of the signal. The fitted trend data may help identify fundamental change patterns and long-term trends in the electrocardiogram signal.

In this embodiment, the obtained electrocardiographic waveform trend data is subjected to fluctuation measurement. Statistical indicators such as standard deviation, variance, etc. can be used to measure the volatility of the signal. The trend data is classified into excessive fluctuation and normal fluctuation according to the calculation result. For example, if the standard deviation exceeds a preset threshold value, it is classified as "excessive fluctuation waveform signal trend data", otherwise it is classified as "normal fluctuation waveform signal trend data". This classification helps to further analyze the stability and anomalies of the signal.

In this embodiment, waveform signal trend correlation analysis is performed on abnormal electrocardiographic waveform signal data according to excessive fluctuation waveform signal trend data and normal fluctuation waveform signal trend data. A correlation analysis algorithm (e.g., pearson correlation coefficient) is applied to each trend data to determine the relationship of the anomaly signal to excessive fluctuations or normal fluctuations. By this analysis, deterioration trend data (e.g., waveform distortion emphasis) and improvement trend data (e.g., waveform gradual recovery) of an electrocardiographic waveform can be obtained.

In this embodiment, the obtained electrocardiographic waveform degradation trend data and the obtained improvement trend data are combined. The merging operation may be performed by means of time series merging or data summarization. The specific method comprises the steps of splicing or reorganizing data of the degradation trend and the good trend according to time or waveform characteristics to form a comprehensive electrocardiogram waveform trend data set. This data set provides an overall view that reflects the overall change in the electrocardiogram waveform.

In this embodiment, the influence factor calculation is performed on the obtained electrocardiographic waveform trend data using the historical myocardial rehabilitation case data. By comparing the electrocardiogram waveform trend data with the rehabilitation data (such as the recovery speed and the rehabilitation effect) in the historical rehabilitation cases, the influence factors of myocardial rehabilitation are calculated by using regression analysis or other statistical methods. Specifically, a regression model may be constructed, using waveform trend data as an independent variable and rehabilitation effect as a dependent variable, to determine the weight and effect of each influence factor. This will help predict and evaluate the effects of myocardial rehabilitation and provide personalized rehabilitation advice.

Optionally, step S46 specifically includes:

In this embodiment, in processing the historical myocardial rehabilitation case data, cardiac ultrasound data is first extracted from the historical myocardial rehabilitation case data to calculate a Left Ventricular Ejection Fraction (LVEF). Waveform features, such as R-wave amplitude, P-R intervals, and Q-T intervals, are then extracted from the Electrocardiogram (ECG) data. Specialized cardiac ultrasound analysis software and ECG feature extraction algorithms can be used, and the extracted LVEF data should include detailed statistics for each rehabilitation case and convert the ECG waveform data to digitized features for subsequent analysis.

in this embodiment, a myocardial rehabilitation index assessment system is constructed using the extracted LVEF data and ECG waveform characteristics. This may be accomplished by determining the weights of the various metrics and normalizing the metrics. For example, a weighted scoring system is used to evaluate the effects of myocardial rehabilitation in combination with LVEF and ECG features. The assessment system may include indicators of recovery progress, electrocardiogram stability, and the like. Through statistical analysis, the generated assessment data should include a composite score and a itemized score for each case.

In this embodiment, when processing the electrocardiographic waveform trend data, it is first necessary to arrange the ECG waveform data in time series and then extract waveform amplitude features using fourier transform or the like. Features such as maximum amplitude, minimum amplitude, and average amplitude of the waveform can be extracted using a signal processing tool to generate trend waveform amplitude data for further analysis.

In the present embodiment, a myocardial rehabilitation linear regression model is constructed using the obtained myocardial rehabilitation index evaluation data and the trend waveform amplitude data obtained in step S463. In the model construction, linear regression analysis is performed with LVEF and waveform amplitude characteristics as independent variables and myocardial rehabilitation effect as dependent variables. The method specifically comprises the steps of selecting a regression algorithm, verifying the accuracy of a model, and adjusting model parameters to optimize a prediction result.

In this embodiment, the calculation of the myocardial rehabilitation influence factor is performed using a linear regression model. And (5) inputting the historical myocardial rehabilitation case data into a model to obtain the value of each influence factor. The calculation of the impact factors should include comparing the model predictions with the actual data and statistically determining the actual contribution of each factor to the rehabilitation effect to generate a final impact factor report.

Optionally, step S5 specifically includes:

In this embodiment, electrocardiographic waveform influence factor data extracted from myocardial rehabilitation influence factors is acquired, and these data may include various features of waveforms such as amplitudes, waveform intervals, shapes, and the like. The electrocardiogram waveform is feature extracted using a signal processing algorithm (e.g., fast fourier transform FFT, wavelet transform, etc.). The extracted features include the main frequency components, the amplitudes of the waveforms, the time intervals between the waveforms, the influence factors affecting the electrocardiogram waveform, etc. The extracted features are formatted into a data structure, such as a table or database record, that can be used for subsequent analysis for further processing.

In this embodiment, feature extraction is performed on the electrocardiogram of each case of the historic myocardial case data, including analysis of the frequency domain features and the time domain features of the electrocardiogram using fourier transforms. In addition, the physician's characteristics are extracted, including clinical experience years, the number of cases successfully treated, etc. The data are organized into a database, wherein the central electrogram data comprise P wave duration, QRS wave amplitude and the like, and the doctor data comprise personal data, seniority and the like.

In this embodiment, a dynamic time warping (DYNAMIC TIME WARPING, DTW) algorithm is used to calculate the similarity between the case electrocardiogram and the myocardial rehabilitation electrocardiogram waveform influencing factors. Specifically, the electrocardiogram of each case is compared with the standard waveform characteristics, and a similarity score is calculated. Then, sorting is performed based on the similarity scores, and the first 10 cases with high similarity are counted. For these high similarity cases, detailed statistical analysis is further performed, and the electrocardiogram data of these cases is extracted for subsequent analysis and correlation.

In this embodiment, high-similarity case electrocardiogram data is matched with therapist data. The specific operation is to correlate the electrocardiogram data of each high-similarity case with the processed doctor characteristic data to construct a treating doctor-electrocardiogram associated data table. For example, SQL database query statement "SELECT * FROM case_data JOIN doctor_data ON case_data.doctor_id = doctor_data.id WHERE case_data.similarity_score>0.9" is used to extract relevant data. Thus, the treatment record and the effect data of each doctor on the high-similarity electrocardiogram case can be obtained.

In this embodiment, the treatment priority of each doctor is calculated from the treating doctor-electrocardiogram related data. This may be achieved by a weighted scoring system, for example, using the success rate of each doctor in high similarity cases as a weighting factor. The specific operation is to calculate the treatment success rate and the number of cases of each doctor in the cases, and the priority score is obtained after the treatment is normalized. For example, the formula "Priority score= (Success Rate 0.7) + (Case Count 0.3)" is used for weighted scoring, thereby obtaining doctor's treatment Priority data.

In this embodiment, a doctor recommendation model is constructed using the cardiologist treatment priority data and the myocardial rehabilitation impact factors. A weighted linear regression model is used, with doctor priority scores as one of the input features, and myocardial rehabilitation impact factors as the other input feature. The model training process uses the historical data for cross-validation to optimize model parameters. After training, the model can predict the most suitable treating doctor according to the electrocardiogram waveform characteristics and the rehabilitation influence factors of the new case. For example, a linear regression model may be implemented using the scikit-learn library in Python. A recommendation model (e.g., weighted score based recommendation system, collaborative filtering model) may also be constructed using the priority data and the characteristics of the rehabilitation impact factors.

Optionally, step S55 specifically includes:

In the present embodiment, the number of treatments per doctor (for example, the total number of electrocardiographic cases processed by each doctor in the past year) is extracted from the treating doctor-electrocardiographic related data. At the same time, the number of electrocardiograms associated with each doctor (i.e., the number of electrocardiograms processed by each doctor) is counted.

In this embodiment, the doctor treatment frequency data is statistically analyzed, and doctors are classified according to the treatment frequency. Two thresholds are set, for example, a "high treatment times doctor" for a treatment times greater than or equal to 50 times, and a "low treatment times doctor" for a treatment times less than 50 times. Physicians can be divided into two groups using a data classification tool (e.g., pandas library in Python), high treatment times physician data containing physicians with treatment times greater than or equal to 50 times, and low treatment times physician data containing physicians with treatment times less than 50 times.

In this embodiment, the doctor-associated electrocardiogram data are classified. Two thresholds are set, for example, a number of associated electrocardiograms greater than or equal to 100 is "high electrocardiogram associated number doctor", and a number of associated electrocardiograms less than 100 is "low electrocardiogram associated number doctor". Doctors are divided into two groups using a data processing tool, high electrocardiogram-associated number of doctors data including a doctor having an associated electrocardiogram number of 100 or more, and low electrocardiogram-associated number of doctors data including a doctor having an associated electrocardiogram number of less than 100.

In this embodiment, the "high treatment times doctor data set" and the "high electrocardiogram-related number doctor data set" are subjected to intersection calculation to find out a doctor who satisfies both conditions. These doctors are considered "high treatment priority cardiologists". Similarly, the "low treatment times doctor data set" and the "low electrocardiogram associated number doctor data set" are subjected to intersection calculation to obtain the "low treatment priority cardiologist". The calculation is performed using a collective calculation tool (such as a collective calculation of Python), and doctors with high treatment times and high electrocardiographic association numbers are classified as high-treatment-priority cardiologists, and doctors with low treatment times and low electrocardiographic association numbers are classified as low-treatment-priority cardiologists.

In this embodiment, the obtained "high treatment priority cardiomyopathy data set" and "low treatment priority cardiomyopathy data set" are combined according to doctor ID to form a complete cardiomyopathy data set. After merging, the order is by priority (e.g., order by treatment priority). The final output data set contains all cardiologists and their treatment priorities.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The method for constructing the myocardial rehabilitation decision support system based on case reasoning is characterized by comprising the following steps of:

Step S4, carrying out electrocardiographic waveform trend analysis according to abnormal electrocardiographic waveform myocardial case data to obtain electrocardiographic waveform trend data, and carrying out myocardial rehabilitation influence factor calculation on the electrocardiographic waveform trend data according to historical myocardial rehabilitation case data to obtain myocardial rehabilitation influence factors, wherein the step S4 specifically comprises the following steps:

step S46, calculating myocardial rehabilitation influence factors according to the electrocardiogram waveform trend data according to the historical myocardial rehabilitation case data, so as to obtain myocardial rehabilitation influence factors;

Step S5, evaluating treatment priority of the myocardial doctor according to the myocardial rehabilitation influence factors to the historical myocardial case data so as to obtain treatment priority data of the myocardial doctor, and constructing a recommendation model of the myocardial doctor according to the treatment priority data of the myocardial doctor and the myocardial rehabilitation influence factors, wherein the step S5 specifically comprises the following steps:

step S56, constructing a myocardial doctor recommendation model according to myocardial doctor treatment priority data and myocardial rehabilitation influence factors;

2. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 1, wherein step S1 specifically comprises:

3. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 2, wherein step S12 specifically comprises:

4. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 2, wherein step S13 specifically comprises:

5. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 1, wherein step S2 specifically comprises:

6. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 5, wherein step S27 specifically comprises:

7. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 1, wherein step S46 specifically comprises:

8. The method for constructing a case-based myocardial rehabilitation decision support system according to claim 1, wherein step S55 specifically comprises: