A comprehensive prediction and assessment method for heart failure

Abstract

The present invention belongs to the technical field of auxiliary diagnosis of heart failure, and discloses a comprehensive prediction and evaluation method for heart failure, comprising: collecting patient indicator data and corresponding patient symptom labels; determining critical indicators of each type of data in the patient indicator data based on the patient symptom labels, constructing a scale standard function through the critical indicators, and constructing a heart failure risk index set based on the scale standard function; performing hierarchical analysis on each indicator in the patient indicator data using an improved hierarchical analysis method to obtain an indicator weight set; performing a comprehensive evaluation on the heart failure risk index set based on the indicator weight set to obtain a comprehensive heart failure risk index, performing risk grading on the comprehensive heart failure risk index using an improved clustering algorithm to obtain a heart failure risk rating table, and using the heart failure risk rating table to achieve a comprehensive evaluation of the patient's heart failure condition; thereby greatly improving the accuracy of the comprehensive evaluation of heart failure.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary diagnosis of heart failure, and more specifically, to a comprehensive prediction and evaluation method for heart failure. Background Art

[0002] The document with application publication number CN118737497A discloses a home rehabilitation management system for heart failure patients; it includes a data acquisition and storage module, and the data acquisition and storage module includes a human-computer communication unit and a self-monitoring unit, and the heart failure data is obtained through the data obtained by the human-computer communication unit and the self-monitoring unit; a trend analysis module, which analyzes the heart failure data through an improved ARIMA algorithm to obtain trend data; performs trend analysis on the trend data through an improved Adaboost algorithm; through the extraction of voice features, it can accurately capture the personal emotional data of the patient's answers, which is relatively accurate and avoids the patient from hiding his true personal emotions; it can monitor and evaluate the rehabilitation training process of heart failure patients undergoing home rehabilitation in real time; and can recommend guidance on rehabilitation training, drug adjustment, and diet adjustment to heart failure patients undergoing home rehabilitation.

[0003] Heart failure (HF) is a complex clinical syndrome involving functional imbalances of multiple systems, including the cardiovascular, respiratory, renal, and neuroendocrine systems. It poses a major challenge to the global healthcare system. Current HF prediction and assessment systems are mostly based on a single or a few biomarkers, which makes it difficult to capture the combined impact of multi-system decompensation in HF, resulting in incomplete risk prediction. The development of HF is a dynamic process, and existing assessment methods are mostly based on single measurements or short-term data, which makes it difficult to reflect the progressive or sudden changes in the patient's condition. Moreover, the assessment system relies on fixed parameters and lacks the ability to continuously optimize based on new clinical data, resulting in a decline in its assessment ability over time.

[0004] In view of this, the present invention proposes a comprehensive prediction and evaluation method for heart failure to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned objectives, the present invention provides the following technical solutions: a comprehensive prediction and evaluation method for heart failure, comprising:

[0006] S1. Collect patient index data and corresponding patient condition labels;

[0007] S2. Determine the critical index for each category of patient index data based on the patient's symptom label, construct an initial assimilation function based on the critical index, iteratively optimize the parameters of the initial assimilation function using an optimization algorithm to obtain a scale standard function, and construct a heart failure risk index set based on the scale standard function;

[0008] S3, using the improved hierarchical analysis method to perform hierarchical analysis on each indicator in the patient indicator data to obtain an indicator weight set;

[0009] S4. Comprehensively evaluate the heart failure risk index set based on the indicator weight set to obtain the comprehensive heart failure risk index. Use the improved clustering algorithm to perform risk grading on the comprehensive heart failure risk index to obtain a heart failure risk rating table. Use the heart failure risk rating table to achieve a comprehensive assessment of the patient's heart failure condition.

[0010] Furthermore, the patient indicator data include: physiological indicators, clinical symptom indicators and psychological state indicators; physiological indicators include: left ventricular ejection fraction, heart rate, respiratory rate and blood oxygen saturation; clinical symptom indicators include: cardiac function grade, weight fluctuation and dyspnea grade; psychological state indicators include: anxiety score, depression score and stress score.

[0011] Furthermore, the critical indicator is obtained by:

[0012] Each data of each indicator in the user indicator data is used as the data to be converted, and the statistical analysis method is used to count the patient condition label of each value in the converted data to obtain a label statistical table, which includes the value of the data to be converted, the number of patients with a condition label value of 1, and the number of patients with a condition label value of 0; the number of patients with a condition label value of 1 in the label statistical table is used as a positive label, and the number of patients with a condition label value of 0 in the label statistical table is used as a negative label; the value of each data to be converted in the label statistical table is used as a numerical cutoff point, and a performance evaluation is performed on each numerical cutoff point based on the label statistical table to obtain a performance index, and the numerical cutoff point with the largest value of the performance index is used as the critical index.

[0013] Furthermore, the formula of the initial assimilation function is: ;in, Represents risk indicators, Represents the index adjustment coefficient, the value range of the index adjustment coefficient is (0,1], Represents the critical indicator, Represents the data to be converted, Represents indicator attributes.

[0014] Furthermore, the method for obtaining the scale standard function includes:

[0015] Preset maximum number of iterations and Set the index adjustment coefficient, initialize the iteration coefficient and the optimal index adjustment coefficient. The initial value of the optimal index adjustment coefficient is null, and the initial fitness of the optimal adjustment coefficient is infinite.

[0016] The data to be converted is used as training data, and the adjustment coefficient of each group of indicators is substituted into the initial assimilation function. The risk index of each training data is calculated by the initial assimilation function. The skewness of the risk index is used as the fitness of each group of indicator adjustment coefficients. The indicator adjustment coefficient with the smallest fitness value is selected as the candidate indicator. When the fitness of the candidate indicator is less than the fitness of the optimal indicator adjustment coefficient, the candidate indicator is used as the new optimal indicator adjustment coefficient. A random number function is used to generate a perturbation random number, and the perturbation random number is used to perform a spiral update on the adjustment coefficient of each group of indicators. The formula for the spiral update of the adjustment coefficient of each group of indicators is as follows: ;in, Represents the index adjustment coefficient after The value after iterations, Represents the search coefficient, which is used to control the search behavior. represents the optimal indicator adjustment coefficient, represents the helical constant, Represents a random number for the direction, represents pi, Represents the index adjustment coefficient after The value after iterations, Represents the number of iterations; the calculation formula for the search coefficient is: ;in represents the iteration coefficient, represents the perturbed random number;

[0017] After each iteration, the iteration coefficient and the optimal index adjustment coefficient are updated. When the maximum number of iterations is reached, the iteration is stopped and the optimal index adjustment coefficient at this time is output. The optimal index adjustment coefficient is substituted into the initial assimilation function to obtain the scale standard function.

[0018] Furthermore, the heart failure risk index set is constructed in the following manner:

[0019] Each data to be converted is calculated using the scale standard function, and the calculation result is used as the standard risk indicator. All standard risk indicators constitute a standard risk set. For the standard risk set, indicator characteristic points are preset, and the standard risk indicators of the indicator characteristic points are calculated using the scale standard function. The standard risk indicators of the indicator characteristic points are used as the dividing points of the triangle membership algorithm. The triangle membership algorithm is used to fuzzify the risk of each data in the standard risk set to obtain the fuzzy membership. For the fuzzy membership corresponding to the data in each indicator in the same group of patient indicator data, the weighted average algorithm is used to calculate the comprehensive membership of each indicator. The comprehensive membership of all indicators in the same group of patient indicator data constitutes the heart failure risk index set.

[0020] Furthermore, the method for obtaining the indicator weight set includes:

[0021] Preset M groups of initial indicator scoring sets and corresponding initial weight matrices. Initialize the optimal indicator scoring set to be empty, and the scoring fitness of the optimal indicator scoring set to be infinitesimal. Construct an indicator judgment matrix based on the initial indicator scoring set. The dimension of the indicator judgment matrix is ​​consistent with the number of categories of indicators in the patient indicator data. The elements at each position in the indicator judgment matrix represent the importance between indicators.

[0022] For each set of initial indicator scoring sets, the initial weight matrix is ​​iteratively updated with the indicator judgment matrix to obtain the updated initial weight matrix. The vector normalization algorithm is used to normalize the initial weight matrix for each update to obtain the normalized weight matrix. The accuracy threshold is preset. When the output As the optimization weight matrix; where, Representative The normalized weight matrix of the initial weight matrix after the update, Representative The normalized weight matrix of the initial weight matrix after the update, represents the accuracy threshold; based on the optimized weight matrix, the adaptability of each initial indicator score set is evaluated to obtain the score fitness;

[0023] The initial indicator score set is selected by the tournament selection algorithm according to the value of the score fitness to obtain the preferred indicator score set; the preferred indicator score set is paired and combined using the permutation and combination algorithm to obtain the indicator score combination; for each group of indicator score combinations, an element is randomly selected as the exchange point, and the elements after the exchange point in the indicator score combination are exchanged to obtain the cross-indicator score set; a disturbance factor is preset, and a random mutation algorithm is used to perturb an element in the cross-indicator score set with the disturbance factor, and a disturbance factor is added to the element to obtain the variant indicator score set, and the score fitness of each group of variant indicator score sets is calculated. The variant indicator score set with a score fitness greater than the indicator score combination is used as the preferred variant, and the preferred variant with the highest score fitness is used as the candidate set. When the score fitness of the candidate set is greater than the score fitness of the optimal indicator score set, the candidate set is used as the new optimal indicator score set, and the preferred variant is used as the new initial indicator score set. This is repeated until the optimal indicator score set no longer changes, and the optimized weight matrix corresponding to the optimal indicator score set is used as the indicator weight set.

[0024] Furthermore, the formula for evaluating the adaptability of each initial indicator score set is:

[0025] ;in, represents the scoring fitness, Represents the number of categories of indicators in the patient indicator data, Represents the first elements, represents the indicator judgment matrix, represents the optimized weight matrix, The first one represents the product of the index judgment matrix and the optimization weight matrix. elements, Represents the consistency index.

[0026] Furthermore, the heart failure risk rating table is obtained in the following ways:

[0027] Preset fuzzy grouping value , based on fuzzy grouping values, selected from the comprehensive risk index of heart failure data as the initial risk center, and the data interval of the initial risk center is selected as , the data interval meets the interval condition: ;in, The total amount of data representing the comprehensive risk index of heart failure; the risk distance from the comprehensive risk index of heart failure to each initial risk center is calculated using the distance measurement formula, and the comprehensive risk index of heart failure is assigned to the initial risk center with the smallest risk distance, forming Risk clusters are formed. For each risk cluster, the mean of the comprehensive risk index of heart failure in each risk cluster is used as the new initial risk center. This process is repeated until the value of the initial risk center no longer changes, and the risk cluster at this time is output as the initial risk grouping. A differential threshold is preset, and the difference between the initial risk centers of different initial risk groups is used as the risk span. The initial risk groups with a risk span smaller than the differential threshold are merged to obtain the final risk grouping. The boundaries of the final risk grouping are optimized to obtain a grading threshold. The grading threshold of each final risk grouping is used as a grading indicator to divide the heart failure risk level and obtain a heart failure risk rating table.

[0028] Furthermore, the method of optimizing the boundaries of the final risk grouping includes:

[0029] The minimum and maximum values ​​of the comprehensive risk index of heart failure in the final risk group are used as the grouping interval, the minimum accuracy of the comprehensive risk index of heart failure is used as the exploration span, and the minimum value of the grouping interval is used as the initial boundary. The data with the comprehensive risk index of heart failure in the final risk group less than or equal to the initial boundary are used as the first group, and the data with the comprehensive risk index of heart failure in the final risk group greater than the initial boundary are used as the second group. The boundary evaluation of the initial boundary is performed based on the first and second groups. The formula for the boundary evaluation of the initial boundary is: ;in, represents the boundary quality index, Represents the group number, Representative The amount of data in the group, represents the mean of all heart failure comprehensive risk indices, Representative The mean of the group, Representative Grouping, Representative Data within the group; the initial boundary increases with the exploration span, and the boundary evaluation of the initial boundary is repeated, and the boundary quality index of each initial boundary is recorded until the initial boundary is equal to the maximum value of the grouping interval. The initial boundary with the largest boundary quality index is selected as the classification threshold.

[0030] The technical effects and advantages of the comprehensive prediction and evaluation method for heart failure of the present invention are as follows:

[0031] The present invention comprehensively considers physiological indicators, clinical symptom indicators and psychological state indicators to comprehensively evaluate the patient's health status. Compared with the traditional single physiological indicator evaluation method, it is more accurate and adaptable; through statistical analysis of patient data, the optimal cutoff point of the performance indicator is used as the critical indicator to ensure that the dividing point of each evaluation indicator is reasonable, which helps to improve the accuracy of the prediction; the parameters of the initial assimilation function are iteratively optimized through the optimization algorithm to improve the rationality of indicator normalization; an improved hierarchical analysis algorithm is used to analyze the weight of each indicator, avoiding the drawbacks of subjective human empowerment and further improving the objectivity and scientificity of the evaluation; an improved clustering algorithm is used to perform risk grading on the comprehensive risk index of heart failure, ensuring the accuracy of the grading standard, so that the risk level division is more in line with the actual clinical situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Schematic diagram of a comprehensive prediction and evaluation method for heart failure according to the present invention;

[0033] Figure 2 This is a schematic diagram of a comprehensive prediction and evaluation system for heart failure according to the present invention. DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] Example 1

[0036] See also Figure 1 As shown, the comprehensive prediction and evaluation method for heart failure described in this embodiment includes:

[0037] S1. Collect patient index data and corresponding patient condition labels;

[0038] S2. Determine the critical index for each category of patient index data based on the patient's symptom label, construct an initial assimilation function based on the critical index, iteratively optimize the parameters of the initial assimilation function using an optimization algorithm to obtain a scale standard function, and construct a heart failure risk index set based on the scale standard function;

[0039] S3, using the improved hierarchical analysis method to perform hierarchical analysis on each indicator in the patient indicator data to obtain an indicator weight set;

[0040] S4. Comprehensively evaluate the heart failure risk index set based on the indicator weight set to obtain the comprehensive heart failure risk index. Use the improved clustering algorithm to perform risk grading on the comprehensive heart failure risk index to obtain a heart failure risk rating table. Use the heart failure risk rating table to achieve a comprehensive assessment of the patient's heart failure condition.

[0041] Patient indicator data include: physiological indicators, clinical symptom indicators and psychological state indicators; the physiological indicators include: left ventricular ejection fraction, heart rate, respiratory rate and blood oxygen saturation; physiological indicators are obtained through medical equipment monitoring; clinical symptom indicators include: cardiac function grade, weight fluctuation and dyspnea grade; among them, weight fluctuation is the user's weekly weight fluctuation; psychological state indicators include: anxiety score, depression score and stress score; each indicator in the psychological state index is obtained through a questionnaire, such as the GAD-7 questionnaire for anxiety scoring and the PHQ-9 questionnaire for depression scoring; the patient's condition label is whether the patient has a worsening hospitalization condition, and the patient's condition label is a numerical label. The value of the patient's condition label is 0 and 1, 1 represents the presence of a worsening hospitalization condition, and 0 represents the absence of a worsening hospitalization condition.

[0042] In the traditional heart failure evaluation system, the critical values ​​of medical indicators have long relied on empirical judgment or guideline recommendations, lacking objective basis for specific populations. Setting the critical value too high leads to an increase in false positives, while setting it too low increases the missed diagnosis rate. It is difficult to determine the optimal balance in clinical practice. The test data obtained by different hospitals and equipment have systematic differences, and using a unified critical value may lead to judgment bias. Using a data-driven approach to objectively determine the optimal critical point for each medical indicator breaks away from the limitations of traditional reliance on expert subjective judgment. The critical point position is determined by the data itself. Data from different regions and populations can generate specific critical values, realizing regional precision medicine. Specifically:

[0043] Each data item of each indicator in the user indicator data is used as the data to be converted, and the statistical analysis method is used to count the patient symptom label of each value in the converted data to obtain a label statistical table. The label statistical table contains the value of the data to be converted, the number of patients with a symptom label value of 1, and the number of patients with a symptom label value of 0; the number of patients with a symptom label value of 1 in the label statistical table is used as a positive label, and the number of patients with a symptom label value of 0 in the label statistical table is used as a negative label; the value of each data to be converted in the label statistical table is used as a numerical cutoff point, and the performance evaluation of each numerical cutoff point is performed based on the label statistical table. The formula for performing performance evaluation on each numerical cutoff point is as follows: ;in, Represents the numerical cutoff point performance indicators, Represents less than or equal to the numerical cutoff point The sum of the positive labels, Represents greater than the numerical cutoff point The sum of the positive labels, Represents greater than the numerical cutoff point The sum of the negative labels, Represents less than or equal to the numerical cutoff point The sum of the negative labels of the performance index is taken as the critical index.

[0044] Heart failure assessment involves a variety of indicators with different properties. For example, the higher the LVEF, the better, while the lower the BNP, the better. Traditional models are difficult to handle uniformly. Medical indicators and disease risk usually have a nonlinear relationship. For example, both high and low blood pressure increase risk, which cannot be accurately represented by linear models. Traditional linear normalization methods are not sensitive enough near the critical point and cannot accurately reflect the risk changes brought about by small changes. By optimizing the initial assimilation function, the increase in the indicator after assimilation represents an increased risk. This is in line with clinical thinking, retains the actual relationship between the original indicator and risk, avoids information distortion, and appropriately suppresses changes in extreme areas. It adapts to the characteristics of different indicators and the convenience of unified standards, which simplifies the complexity of comprehensive multi-indicator evaluation. Specifically:

[0045] An indicator attribute library is constructed. Based on the indicator attribute library, the attributes of the data to be transformed are divided to obtain indicator attributes. The value of the indicator attribute is 0 or 1, where 1 represents a positive indicator, indicating that a higher value represents a worse state, and 0 represents a negative indicator, indicating that a lower value represents a worse state. The indicator attribute library is established by those skilled in the art. For example, a higher left ventricular ejection fraction indicates better cardiac pumping function, so the left ventricular ejection fraction is a negative indicator, and the value of the indicator attribute is 0; a higher respiratory rate represents a heavier respiratory burden, so the respiratory rate is a positive indicator, and the value of the indicator attribute is 1; an initial assimilation function is constructed based on the indicator attributes and critical indicators. The formula of the initial assimilation function is: ;in, Represents risk indicators, Represents the index adjustment coefficient, the value range of the index adjustment coefficient is (0,1], Represents the critical indicator, Represents the data to be converted, Represents indicator attributes;

[0046] Preset maximum number of iterations and Set the index adjustment coefficient, initialize the iteration coefficient and the optimal index adjustment coefficient, the initial value of the iteration coefficient is 2, the initial value of the optimal index adjustment coefficient is a null value, and the initial fitness of the optimal adjustment coefficient is infinite;

[0047] The data to be converted is used as training data. The adjustment coefficient of each group of indicators is substituted into the initial assimilation function. The risk index of each training data is calculated by the initial assimilation function. The skewness of the risk index is used as the fitness of each group of indicator adjustment coefficients. The smaller the fitness, the better the indicator adjustment coefficient. The calculation method of skewness is: ;in, represents skewness, represents the mean value of the risk indicator, represents the mode of the risk indicator, Represents the standard deviation of the risk indicator; select the indicator adjustment coefficient with the smallest fitness value as the candidate indicator. When the fitness of the candidate indicator is less than the fitness of the optimal indicator adjustment coefficient, the candidate indicator is used as the new optimal indicator adjustment coefficient; use the random number function to generate the perturbed random number. The value range of the perturbed random number is (0,1). Common random number functions include rand function and srand function. Based on the perturbed random number, the spiral update of each group of indicator adjustment coefficients is performed. The formula for spiral updating of each group of indicator adjustment coefficients is:

[0048] ;in, Represents the index adjustment coefficient after The value after iterations, Represents the search coefficient, which is used to control the search behavior. represents the optimal indicator adjustment coefficient, represents the helical constant. In this embodiment, the preferred helical constant is 0.5. Represents the direction random number, which is used to control the update direction of the index adjustment coefficient. The direction random number is generated by the random number function. The value of the direction random number is 0 or 1. represents pi, Represents the index adjustment coefficient after The value after iterations, represents the number of iterations, is an integer greater than zero; the search coefficient is calculated as follows: ;in represents the iteration coefficient, Represents the perturbation random number. The value range of the perturbation random number is (0,1). The perturbation random number is used to increase the randomness of the indicator adjustment coefficient exploration.

[0049] After each iteration, the iteration coefficient and the optimal index adjustment coefficient are updated. The formula for updating the iteration coefficient is: ;in, represents the updated value of the iteration coefficient, represents the iteration coefficient, represents the maximum number of iterations, Represents the number of iterations. When the maximum number of iterations is reached, the iteration is stopped and the optimal indicator adjustment coefficient at this time is output. The optimal indicator adjustment coefficient is substituted into the initial assimilation function to obtain the scale standard function; each data to be converted is calculated by the scale standard function, and the calculation result is used as the standard risk indicator. All standard risk indicators constitute a standard risk set; for the standard risk set, the indicator feature points are preset, and the standard risk indicators of the indicator feature points are calculated by the scale standard function. The standard risk indicators of the indicator feature points are used as the demarcation points of the triangle membership algorithm. The triangle membership algorithm is used to fuzzify the risk of each data in the standard risk set to obtain the fuzzy membership; for the fuzzy membership corresponding to the data in each indicator of the same group of patient indicator data, the weighted average algorithm is used to calculate to obtain the comprehensive membership of each indicator. The comprehensive membership of all indicators in the same group of patient indicator data constitutes the heart failure risk index set;

[0050] The characteristic points of indicators are determined by technicians in this field based on the characteristics of relevant indicators. Taking left ventricular ejection fraction as an example, the three characteristic points of clinical indicators are: 20%, 40% and 55%; 20% represents the starting point where the risk begins to increase, 40% represents the inflection point where the risk reaches the highest, and 55% represents the upper limit of risk (beyond this point, the risk no longer increases significantly); assuming that the fuzzy membership of each data in the physiological indicators in a set of patient indicator data are: 0.812, 0.6, 0.422 and 0.51, then the comprehensive membership of the physiological indicators is 0.586, which is the average of the fuzzy memberships; the elements in the heart failure risk index represent the degree of heart failure risk corresponding to the corresponding indicators.

[0051] The importance of different indicators in the medical field is difficult to directly quantify, such as the relative contribution of physiological and psychological indicators. Different experts have inconsistent criteria for judging the importance of indicators, resulting in inaccurate heart failure assessments. The improved analytic hierarchy process (AHP) improves the objectivity and accuracy of weight calculations while retaining medical knowledge. This enables the heart failure risk assessment system to more accurately reflect the actual contribution of each indicator, thereby providing more reliable risk assessment results. This has significant value for the clinical management and prognosis of heart failure patients. Specifically:

[0052] Preset M groups of initial indicator score sets and corresponding initial weight matrices, initialize the optimal indicator score set to empty, and the score fitness of the optimal indicator score set is infinitesimal; each element in the initial indicator score set represents the importance score between different indicators, and the value range of each element in the initial indicator score set is [1,10]; for example: the importance score of physiological indicators relative to psychological indicators is 9, which means that the contribution of physiological indicators to heart failure is much greater than that of psychological indicators, then the importance of psychological indicators relative to physiological indicators is ; Based on the initial indicator score set, an indicator judgment matrix is ​​constructed. The dimension of the indicator judgment matrix is ​​consistent with the number of categories of indicators in the patient indicator data. The elements at each position in the indicator judgment matrix represent the importance between indicators. Assuming that the element value of the first row and second column in the indicator judgment matrix is ​​2, it means that the importance of physiological indicators relative to clinical symptom indicators is 2, and the importance of each indicator relative to itself is 1; the initial weight matrix is The matrix of Represents the number of categories of indicators in the patient indicator data. The value of each element in the initial weight matrix is ​​1. Assuming that the initial indicator score set is: [2,5,7], then 2 represents the importance score of physiological indicators relative to psychological indicators is 2, 5 represents the importance score of physiological indicators relative to clinical symptom indicators is 5, and 7 represents the importance score of clinical symptom indicators relative to psychological indicators. The indicator judgment matrix corresponding to this score set is:

[0053] ;

[0054] For each set of initial indicator scoring sets, the initial weight matrix is ​​iteratively updated using the indicator judgment matrix. The formula for iteratively updating the initial weight matrix is: ;in, represents the indicator judgment matrix, Representative The initial weight matrix after the update, Representative The initial weight matrix after the update, is an integer greater than zero, and the vector normalization algorithm is used to normalize the initial weight matrix of each update to obtain the normalized weight matrix. The accuracy threshold is preset. When the output As the optimization weight matrix; where, Representative The normalized weight matrix of the initial weight matrix after the update, Representative The normalized weight matrix of the initial weight matrix after the update, represents the accuracy threshold. In this embodiment, the accuracy threshold is set to 0.001. Based on the optimized weight matrix, the adaptability evaluation of each initial indicator score set is performed. The formula for the adaptability evaluation of each initial indicator score set is: ;in, Represents the scoring fitness. The higher the scoring fitness, the better the initial indicator scoring set. Represents the number of categories of indicators in the patient indicator data, Represents the first elements, represents the indicator judgment matrix, represents the optimized weight matrix, The first one represents the product of the index judgment matrix and the optimization weight matrix. elements, represents the consistency index, which is obtained from the consistency index value table. The consistency index value of this embodiment is 0.58;

[0055] The initial indicator score set is selected using a tournament selection algorithm based on the score fitness value to obtain a preferred indicator score set; the preferred indicator score sets are paired and combined using a permutation and combination algorithm to obtain an indicator score combination; for each group of indicator score combinations, an element is randomly selected as an exchange point, and the elements after the exchange point in the indicator score combination are exchanged to obtain a cross-indicator score set; a perturbation factor is preset, and in this embodiment, the preferred perturbation factor is 0.5. An element in the cross-indicator score set is perturbed by the perturbation factor using a random mutation algorithm, and the element is perturbed by one perturbation factor to obtain a variant indicator score set. The score fitness of each variant indicator score set is calculated, and the variant indicator score set with a score fitness greater than the indicator score combination is used as the preferred variant, and the preferred variant with the highest score fitness is used as the candidate set. When the score fitness of the candidate set is greater than the score fitness of the optimal indicator score set, the candidate set is used as the new optimal indicator score set, and the preferred variant is used as the new initial indicator score set. This process is repeated until the optimal indicator score set no longer changes, and the optimized weight matrix corresponding to the optimal indicator score set is used as the indicator weight set.

[0056] Traditional risk stratification methods often rely on expert experience or simple statistical analysis, lacking data-driven, objective stratification methods. Heart failure risk distribution may vary across different populations, making it challenging to establish a risk rating system that adapts to the characteristics of diverse populations. By using an improved clustering algorithm to find the optimal cutoff point, the accuracy of risk stratification is improved, making the grading results more consistent with actual clinical needs. The risk grading criteria can be adaptively adjusted based on the data characteristics of different populations. Specifically:

[0057] For each heart failure risk index set, the weighted sum of each element in the heart failure risk index set is performed using the indicator weight set to obtain the heart failure comprehensive risk index; the preset fuzzy grouping value , based on fuzzy grouping values, selected from the comprehensive risk index of heart failure data as the initial risk center, and the data interval of the initial risk center is selected as , the data interval meets the interval condition: ;in, The total amount of data representing the comprehensive risk index of heart failure; the risk distance from the comprehensive risk index of heart failure to each initial risk center is calculated using the distance measurement formula, and the comprehensive risk index of heart failure is assigned to the initial risk center with the smallest risk distance, forming For each risk cluster, the mean of the heart failure comprehensive risk index in each risk cluster is used as the new initial risk center. This process is repeated until the value of the initial risk center no longer changes. The risk cluster at this point is output as the initial risk grouping. A differential threshold is preset, and the difference between the initial risk centers of different initial risk groups is used as the risk span. Initial risk groups with risk spans less than the differential threshold are merged to obtain the final risk grouping.

[0058] The boundaries of the final risk groups were optimized. The minimum and maximum values ​​of the comprehensive risk index of heart failure in the final risk groups were used as the grouping intervals. The minimum accuracy of the comprehensive risk index of heart failure was used as the exploration span. The minimum value of the grouping interval was used as the initial boundary. The data with a comprehensive risk index of heart failure in the final risk groups that was less than or equal to the initial boundary were grouped as the first group. The data with a comprehensive risk index of heart failure in the final risk groups that was greater than the initial boundary were grouped as the second group. The initial boundary was evaluated based on the first and second groups. The formula for evaluating the initial boundary was: ;in, represents the boundary quality index, Represents the group number, Representative The amount of data in the group, represents the mean of all heart failure comprehensive risk indices, Representative The mean of the group, Representative Grouping, Representative Data within the group; the initial boundary increases with the exploration span, and the boundary evaluation of the initial boundary is repeated, and the boundary quality index of each initial boundary is recorded until the initial boundary is equal to the maximum value of the grouping interval. The initial boundary with the largest boundary quality index is selected as the classification threshold; the classification threshold of each final risk group is used as the classification index to divide the heart failure risk level and obtain the heart failure risk rating table; taking the classification threshold: [0.30, 0.42, 0.58, 0.73] as an example, the heart failure risk rating table is:

[0059] ;

[0060] Use the heart failure risk rating table to achieve accurate assessment of heart failure status.

[0061] This embodiment comprehensively considers physiological indicators, clinical symptom indicators and psychological state indicators to comprehensively evaluate the patient's health status. Compared with the traditional single physiological indicator evaluation method, it is more accurate and adaptable; through statistical analysis of patient data, the optimal cutoff point of the performance indicator is used as the critical indicator to ensure that the dividing point of each evaluation indicator is reasonable, which helps to improve the accuracy of the prediction; the parameters of the initial assimilation function are iteratively optimized through the optimization algorithm to improve the rationality of indicator normalization; the improved hierarchical analysis algorithm is used to analyze the weight of each indicator, avoiding the drawbacks of subjective empowerment and further improving the objectivity and scientificity of the evaluation; the improved clustering algorithm is used to perform risk grading on the comprehensive risk index of heart failure to ensure the accuracy of the grading standard, so that the risk level division is more in line with the actual clinical situation.

[0062] Example 2;

[0063] See also Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description of Example 1. A comprehensive prediction and evaluation system for heart failure is provided, including:

[0064] Data collection module: collects patient index data and corresponding patient condition labels;

[0065] Data assimilation module: Based on the patient's disease label, the critical index of each type of data in the patient's index data is determined, and the initial assimilation function is constructed based on the critical index. The parameters of the initial assimilation function are iteratively optimized using an optimization algorithm to obtain a scale standard function. Based on the scale standard function, a heart failure risk index set is constructed;

[0066] Weight analysis module: uses the improved hierarchical analysis method to perform hierarchical analysis on each indicator in the patient indicator data to obtain the indicator weight set;

[0067] Risk classification module: Based on the indicator weight set, a comprehensive evaluation of the heart failure risk index set is performed to obtain a comprehensive heart failure risk index. An improved clustering algorithm is used to perform risk classification on the comprehensive heart failure risk index to obtain a heart failure risk rating table. The heart failure risk rating table is used to achieve a comprehensive assessment of the patient's heart failure condition.

[0068] The modules are connected via wired and / or wireless means to achieve data transmission between modules.

[0069] Example 3;

[0070] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the comprehensive prediction and evaluation method for heart failure provided above is implemented.

[0071] Since the electronic device described in this embodiment is an electronic device used to implement a comprehensive prediction and assessment method for heart failure in the embodiments of this application, based on the comprehensive prediction and assessment method for heart failure described in the embodiments of this application, those skilled in the art will be able to understand the specific implementation of the electronic device of this embodiment and its various variations. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. As long as those skilled in the art can implement the electronic device used in the comprehensive prediction and assessment method for heart failure in the embodiments of this application, it falls within the scope of protection of this application.

[0072] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0073] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A comprehensive prediction and evaluation method for heart failure, characterized in that: include: S1. Collect patient index data and corresponding patient condition labels; S2. Determine the critical index for each category of patient index data based on the patient's symptom label, construct an initial assimilation function based on the critical index, iteratively optimize the parameters of the initial assimilation function using an optimization algorithm to obtain a scale standard function, and construct a heart failure risk index set based on the scale standard function; S3, using the improved hierarchical analysis method to perform hierarchical analysis on each indicator in the patient indicator data to obtain an indicator weight set; S4. Comprehensively evaluate the heart failure risk index set based on the indicator weight set to obtain a comprehensive heart failure risk index, use an improved clustering algorithm to perform risk grading on the comprehensive heart failure risk index, obtain a heart failure risk rating table, and use the heart failure risk rating table to achieve a comprehensive assessment of the patient's heart failure condition; The formula of the initial assimilation function is: ;in, Represents risk indicators, Represents the index adjustment coefficient, the value range of the index adjustment coefficient is (0,1], Represents the critical indicator, Represents the data to be converted, Represents indicator attributes; The method for obtaining the scale standard function includes: Preset maximum number of iterations and Set the index adjustment coefficient, initialize the iteration coefficient and the optimal index adjustment coefficient. The initial value of the optimal index adjustment coefficient is null, and the initial fitness of the optimal adjustment coefficient is infinite. The data to be converted is used as training data, and the adjustment coefficient of each group of indicators is substituted into the initial assimilation function. The risk index of each training data is calculated by the initial assimilation function. The skewness of the risk index is used as the fitness of each group of indicator adjustment coefficients. The indicator adjustment coefficient with the smallest fitness value is selected as the candidate indicator. When the fitness of the candidate indicator is less than the fitness of the optimal indicator adjustment coefficient, the candidate indicator is used as the new optimal indicator adjustment coefficient. A random number function is used to generate a perturbation random number, and the perturbation random number is used to perform a spiral update on the adjustment coefficient of each group of indicators. The formula for the spiral update of the adjustment coefficient of each group of indicators is as follows: ;in, Represents the index adjustment coefficient after The value after iterations, Represents the search coefficient, which is used to control the search behavior. represents the optimal indicator adjustment coefficient, represents the helical constant, Represents a random number for the direction, represents pi, Represents the index adjustment coefficient after The value after iterations, Represents the number of iterations; the calculation formula of the search coefficient is: ;in represents the iteration coefficient, represents the perturbed random number; After each iteration, the iteration coefficient and the optimal index adjustment coefficient are updated. When the maximum number of iterations is reached, the iteration is stopped and the optimal index adjustment coefficient at this time is output. The optimal index adjustment coefficient is substituted into the initial assimilation function to obtain the scale standard function.

2. A comprehensive prediction and evaluation method for heart failure according to claim 1, characterized in that: The patient indicator data include: physiological indicators, clinical symptom indicators and psychological state indicators; physiological indicators include: left ventricular ejection fraction, heart rate, respiratory rate and blood oxygen saturation; clinical symptom indicators include: cardiac function grade, weight fluctuation and dyspnea grade; psychological state indicators include: anxiety score, depression score and stress score.

3. A comprehensive prediction and evaluation method for heart failure according to claim 2, characterized in that: The method for obtaining the critical indicator includes: Each data of each indicator in the user indicator data is used as the data to be converted, and the statistical analysis method is used to count the patient condition label of each value in the converted data to obtain a label statistical table, which includes the value of the data to be converted, the number of patients with a condition label value of 1, and the number of patients with a condition label value of 0; the number of patients with a condition label value of 1 in the label statistical table is used as a positive label, and the number of patients with a condition label value of 0 in the label statistical table is used as a negative label; the value of each data to be converted in the label statistical table is used as a numerical cutoff point, and a performance evaluation is performed on each numerical cutoff point based on the label statistical table to obtain a performance index, and the numerical cutoff point with the largest value of the performance index is used as the critical index.

4. A comprehensive prediction and evaluation method for heart failure according to claim 3, characterized in that: The heart failure risk index set is constructed in the following manner: Each data to be converted is calculated using the scale standard function, and the calculation result is used as the standard risk indicator. All standard risk indicators constitute a standard risk set. For the standard risk set, indicator characteristic points are preset, and the standard risk indicators of the indicator characteristic points are calculated using the scale standard function. The standard risk indicators of the indicator characteristic points are used as the dividing points of the triangle membership algorithm. The triangle membership algorithm is used to fuzzify the risk of each data in the standard risk set to obtain the fuzzy membership. For the fuzzy membership corresponding to the data in each indicator in the same group of patient indicator data, the weighted average algorithm is used to calculate the comprehensive membership of each indicator. The comprehensive membership of all indicators in the same group of patient indicator data constitutes the heart failure risk index set.

5. A comprehensive prediction and evaluation method for heart failure according to claim 4, characterized in that: The method of obtaining the indicator weight set includes: Preset M groups of initial indicator scoring sets and corresponding initial weight matrices. Initialize the optimal indicator scoring set to be empty, and the scoring fitness of the optimal indicator scoring set to be infinitesimal. Construct an indicator judgment matrix based on the initial indicator scoring set. The dimension of the indicator judgment matrix is ​​consistent with the number of categories of indicators in the patient indicator data. The elements at each position in the indicator judgment matrix represent the importance between indicators. For each set of initial indicator scoring sets, the initial weight matrix is ​​iteratively updated with the indicator judgment matrix to obtain the updated initial weight matrix. The vector normalization algorithm is used to normalize the initial weight matrix for each update to obtain the normalized weight matrix. The accuracy threshold is preset. When As the optimization weight matrix; where, Representative The normalized weight matrix of the initial weight matrix after the update, Representative The normalized weight matrix of the initial weight matrix after the update, represents the accuracy threshold; based on the optimized weight matrix, the adaptability of each initial indicator score set is evaluated to obtain the score fitness; The initial indicator score set is selected by the tournament selection algorithm according to the value of the score fitness to obtain the preferred indicator score set; the preferred indicator score set is paired and combined using the permutation and combination algorithm to obtain the indicator score combination; for each group of indicator score combinations, an element is randomly selected as the exchange point, and the elements after the exchange point in the indicator score combination are exchanged to obtain the cross-indicator score set; a disturbance factor is preset, and a random mutation algorithm is used to perturb an element in the cross-indicator score set with the disturbance factor, and a disturbance factor is added to the element to obtain the variant indicator score set, and the score fitness of each group of variant indicator score sets is calculated. The variant indicator score set with a score fitness greater than the indicator score combination is used as the preferred variant, and the preferred variant with the highest score fitness is used as the candidate set. When the score fitness of the candidate set is greater than the score fitness of the optimal indicator score set, the candidate set is used as the new optimal indicator score set, and the preferred variant is used as the new initial indicator score set. This is repeated until the optimal indicator score set no longer changes, and the optimized weight matrix corresponding to the optimal indicator score set is used as the indicator weight set.

6. A comprehensive prediction and evaluation method for heart failure according to claim 5, characterized in that: The formula for evaluating the adaptability of each initial indicator score set is: ;in, represents the scoring fitness, Represents the number of categories of indicators in the patient indicator data, Represents the first elements, represents the indicator judgment matrix, represents the optimized weight matrix, The first one represents the product of the index judgment matrix and the optimization weight matrix. elements, Represents the consistency index.

7. A comprehensive prediction and evaluation method for heart failure according to claim 6, characterized in that: Methods for obtaining the heart failure risk rating table include: Preset fuzzy grouping value , based on fuzzy grouping values, selected from the comprehensive risk index of heart failure data as the initial risk center, and the data interval of the initial risk center is selected as , the data interval meets the interval condition: ;in, The total amount of data representing the comprehensive risk index of heart failure; the risk distance from the comprehensive risk index of heart failure to each initial risk center is calculated using the distance measurement formula, and the comprehensive risk index of heart failure is assigned to the initial risk center with the smallest risk distance, forming Risk clusters are formed. For each risk cluster, the mean of the comprehensive risk index of heart failure in each risk cluster is used as the new initial risk center. This process is repeated until the value of the initial risk center no longer changes, and the risk cluster at this time is output as the initial risk grouping. A differential threshold is preset, and the difference between the initial risk centers of different initial risk groups is used as the risk span. The initial risk groups with a risk span smaller than the differential threshold are merged to obtain the final risk grouping. The boundaries of the final risk grouping are optimized to obtain a grading threshold. The grading threshold of each final risk grouping is used as a grading indicator to divide the heart failure risk level and obtain a heart failure risk rating table.

8. A comprehensive prediction and evaluation method for heart failure according to claim 7, characterized in that: The method of optimizing the boundaries of the final risk grouping includes: The minimum and maximum values ​​of the comprehensive risk index of heart failure in the final risk group are used as the grouping interval, the minimum accuracy of the comprehensive risk index of heart failure is used as the exploration span, and the minimum value of the grouping interval is used as the initial boundary. The data with the comprehensive risk index of heart failure in the final risk group less than or equal to the initial boundary are used as the first group, and the data with the comprehensive risk index of heart failure in the final risk group greater than the initial boundary are used as the second group. The boundary evaluation of the initial boundary is performed based on the first and second groups. The formula for the boundary evaluation of the initial boundary is: ;in, represents the boundary quality index, Represents the group number, Representative The amount of data in the group, represents the mean of all heart failure comprehensive risk indices, Representative The mean of the group, Representative Grouping, Representative Data within the group; the initial boundary increases with the exploration span, and the boundary evaluation of the initial boundary is repeated, and the boundary quality index of each initial boundary is recorded until the initial boundary is equal to the maximum value of the grouping interval. The initial boundary with the largest boundary quality index is selected as the classification threshold.

Images