Method and device for determining decision rule, equipment and storage medium

Abstract

Embodiments of the present application disclose a decision rule determination method and device, equipment and a storage medium, the method comprising: obtaining sample data comprising a sub-feature combination label of a plurality of sample electrocardio signals, the sub-feature combination label comprising at least a sub-feature threshold of each sub-feature of a confirmed QT interval in the sample electrocardio signals; performing evolution processing on the sub-feature threshold by using the sample data and a preset evolution algorithm, determining a target sub-feature threshold of each sub-feature of an optimal sub-feature combination label; and performing rule construction processing according to the target sub-feature threshold, and determining a target decision rule for feature recognition of a QT interval in a to-be-recognized electrocardio signal. In the foregoing manner, the evolution algorithm can be used to perform evolution iteration on the sub-feature threshold in the sample data, to obtain a target sub-feature threshold with the highest recognition accuracy, so as to formulate and use the target decision rule to recognize the QT interval, thereby facilitating improvement of the accuracy of recognizing the QT interval.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, device, and storage medium for determining decision rules. Background Technology

[0002] The QT algorithm is an algorithm used in patient monitors to calculate the QT interval of electrocardiogram (ECG) signals. The QT interval begins at the onset of the QRS complex in the ECG signal and ends at the end of the T wave. Variations in the QT interval contain a wealth of information about cardiovascular regulation. Extracting and analyzing this information can be used for the quantitative assessment of ventricular repolarization and autonomic nervous system activity, and is of great significance for the diagnosis and risk prediction of diseases such as arrhythmias, myocardial ischemia, and myocardial infarction.

[0003] However, due to the complexity and variability of ECG waveforms and the presence of various interferences, it is difficult to calculate the QT interval or the QT interval may be calculated incorrectly under special circumstances. To solve this problem, a more robust QT algorithm needs to be developed. In the QT algorithm, formulating decision rules is a crucial step. However, the selection of thresholds in sub-features is often based on experience, without a specific, scientific standard for selecting sub-feature thresholds. This results in insufficient accuracy in identifying QT intervals through decision rules.

[0004] Currently, there is still a lack of methods to improve the accuracy of QT interval identification. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for determining decision rules, which can solve the problem of low accuracy in identifying QT intervals in the prior art.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for determining decision rules, the method comprising:

[0007] Acquire sample data, which includes sub-feature combination labels of several sample electrocardiogram signals, and the sub-feature combination labels include at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signals.

[0008] Using the sample data and a preset evolutionary algorithm, the sub-feature thresholds are processed to determine the target sub-feature thresholds of each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label.

[0009] Based on the target sub-feature threshold, rule construction processing is performed to determine the target decision rule, which is used to identify the QT interval in the ECG signal to be identified.

[0010] In one feasible implementation, the step of using the sample data and a preset evolutionary algorithm to perform evolutionary processing on the sub-feature thresholds to determine the target sub-feature thresholds for each sub-feature of the optimal sub-feature combination label includes:

[0011] Using a preset fitness function and the individual's encoding value, the target fitness of the individual is determined. The individual corresponds one-to-one with the sub-feature combination label. The encoding value includes the encoding value of the sub-feature threshold of the sub-feature combination label. The fitness is used to reflect the recognition accuracy of the QT interval.

[0012] Based on the target fitness and the preset subpopulation division rules, the subpopulation to which each individual belongs is determined;

[0013] The evolutionary coding value of each individual is determined using the coding value, the type of the subpopulation, and the evolutionary algorithm.

[0014] The fitness of an individual is determined by comparing the encoded value and the evolutionary encoded value. The target encoded value is used as the encoded value, and the number of iterations is increased by a first preset threshold. The process of determining the target fitness of an individual using a preset fitness function and the encoded value of the individual is repeated until the number of iterations is not less than the maximum number of iterations. Then, the target encoded value of each individual is output.

[0015] The optimal individual is determined using the target encoding value. The sub-feature combination label corresponding to the optimal individual is the optimal sub-feature combination label. The target sub-feature threshold is the sub-feature threshold of each sub-feature corresponding to the optimal sub-feature combination label.

[0016] In one feasible implementation, determining the target fitness of an individual using a preset fitness function and the individual's encoded value further includes:

[0017] The algorithm parameters of the evolutionary algorithm are initialized based on the sample data. The algorithm parameters include at least the individual dimension of the individual, the maximum number of iterations, and the search space of the individual encoding.

[0018] Encode the sub-feature thresholds of the sub-feature combination label within the search space to obtain the encoded value of each individual. The encoding rule includes that each dimension of each individual represents a sub-feature threshold.

[0019] In one feasible implementation, the subpopulation includes at least a sterile line and a maintainer line, and the evolutionary algorithm includes at least a hybridization algorithm. Then, determining the evolutionary code value of each individual using the type of the subpopulation and the preset evolutionary algorithm includes:

[0020] The first individual of the sterile line and the second individual of the maintainer line were randomly selected;

[0021] The coding values ​​of the first individual and the second individual are input into the hybridization algorithm to determine the evolutionary coding value of the first individual of the sterile line.

[0022] In one feasible implementation, the subpopulation further includes a restorer line, and the evolutionary algorithm includes at least a first self-crossing algorithm. Then, determining the evolutionary code value of each individual using the type of the subpopulation and the preset evolutionary algorithm further includes:

[0023] Randomly select a third body from the recovery system;

[0024] Determine the number of historical self-intersections of the third individual;

[0025] If the number of historical self-crosses is not higher than a preset self-crossing threshold, the encoding value of the third individual is used as input to the first self-crossing algorithm to determine the evolutionary encoding value of the third individual of the restorer line, and the number of historical self-crosses is increased by a second preset threshold.

[0026] In one feasible implementation, the evolutionary algorithm further includes a second self-crossing algorithm, then the method further includes:

[0027] When the number of historical self-crosses exceeds a preset self-crossing threshold, the evolutionary encoding value of the third individual of the restorer line is determined using the third individual, the upper and lower bounds of the search space, and the second self-crossing algorithm, and the number of historical self-crosses is increased by the second preset threshold.

[0028] In one feasible implementation, the subpopulation includes sterile lines, maintainer lines, and restorer lines. The step of determining the subpopulation to which each individual belongs based on the target fitness and a preset subpopulation division rule includes:

[0029] The target fitness of each individual is arranged in descending order of value to determine the fitness decreasing sequence of the population. The fitness decreasing sequence includes the correspondence between individuals and target fitness.

[0030] All individuals in the top proportion of the fitness-decreasing sequence are identified as the maintainer line;

[0031] All individuals in the second-to-last proportion of the fitness-decreasing sequence are identified as sterile lines;

[0032] The individuals in the fitness decreasing sequence, excluding all individuals in the first proportion of the top ranking and all individuals in the second proportion of the bottom ranking, are identified as the restorer line.

[0033] To achieve the above objectives, a second aspect of the present invention provides a decision rule determination apparatus, the apparatus comprising:

[0034] Sample determination module: used to acquire sample data, the sample data including sub-feature combination labels of several sample electrocardiogram signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signal;

[0035] Threshold determination module: used to perform evolutionary processing of the sub-feature threshold using the sample data and a preset evolutionary algorithm, and determine the target sub-feature threshold of each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature threshold of the sub-feature combination label.

[0036] Rule determination module: used to perform rule construction processing based on the target sub-feature threshold to determine the target decision rule, which is used to identify the QT interval in the ECG signal to be identified.

[0037] To achieve the above objectives, a third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0038] To achieve the above objectives, a fourth aspect of the present invention provides a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, causes the processor to perform the steps shown in the first aspect and any feasible implementation.

[0039] The embodiments of the present invention have the following beneficial effects:

[0040] This invention provides a method for determining decision rules. The method includes: acquiring sample data, which includes sub-feature combination labels of several sample electrocardiogram (ECG) signals. Each sub-feature combination label includes at least a sub-feature threshold of each sub-feature of the confirmed QT interval in the sample ECG signals. The method further includes: performing evolutionary processing on the sub-feature thresholds using the sample data and a preset evolutionary algorithm to determine target sub-feature thresholds for each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label that achieves the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label. Finally, the method involves rule construction processing based on the target sub-feature thresholds to determine a target decision rule. This target decision rule is used to perform feature recognition of the QT interval in the ECG signal to be identified. Through this method, the sub-feature thresholds in the sample data can be iteratively evolved using an evolutionary algorithm to obtain target sub-feature thresholds with recognition accuracy not lower than a preset accuracy threshold. Target decision rules are then formulated based on these target sub-feature thresholds to identify QT intervals, which helps improve the accuracy of QT interval recognition and reduces errors caused by manually setting thresholds. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] in:

[0043] Figure 1 This is a flowchart of a method for determining decision rules in an embodiment of the present invention;

[0044] Figure 2 This is another flowchart of a method for determining a decision rule in an embodiment of the present invention;

[0045] Figure 3 This is a structural block diagram of a decision rule determination device according to an embodiment of the present invention;

[0046] Figure 4 This is a structural block diagram of a computer device in an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Please see Figure 1 , Figure 1 The following is a flowchart of a method for determining decision rules in an embodiment of the present invention, such as... Figure 1 The method shown can be applied to both terminals and servers. The terminal can be a desktop terminal or a mobile terminal; a mobile terminal can be at least one of a mobile phone, tablet, or laptop. The server can be a standalone server or a server cluster consisting of multiple servers. This embodiment uses a terminal application as an example. Figure 1 The method shown includes the following steps:

[0049] 101. Obtain sample data, the sample data including sub-feature combination labels of several sample electrocardiogram signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signals;

[0050] It should be noted that the sample data is pre-collected and can be signal data from several pre-collected electrocardiogram (ECG) signals. The sample data includes the correspondence between sample ECG signals and sub-feature combination labels. The ECG signal must include at least the QT interval, which includes the Q wave position and the T wave position. An ECG signal can have multiple QT intervals, and the types of sub-features for each QT interval can be different or the same. Therefore, there are N possible combinations of QT interval sub-features. Sub-feature labels are annotated onto the ECG signal through manual annotation or other annotation methods to obtain the sub-feature combination labels. The combined label includes at least the sub-feature thresholds for each sub-feature of the confirmed QT interval in the sample ECG signal. Sub-features represent the characteristics set for identifying the Q wave and T wave positions. For example, determining the T wave requires finding several slopes in the ECG signal; these slopes can be considered sub-features. Furthermore, the slope must meet a slope threshold to be considered a T wave; this slope threshold can be considered a sub-feature threshold. The identification of the Q wave position is similar. Therefore, sub-features are the signal features needed to identify the QT interval, and sub-feature thresholds are the thresholds for these signal features. These thresholds can be numerical values ​​or numerical ranges, etc., without limitation here. Furthermore, sample ECG signals are pre-collected and labeled to obtain sample data with combined labels for sub-features related to the QT interval. Alternatively, labeled sample ECG signals can be collected directly. The sub-feature combined label contains multiple sub-features, and consequently, multiple sub-feature thresholds. The labeling method can be manual or automatic, without limitation here, to statistically analyze the distribution of each sub-feature in labeled signals with different characteristics.

[0051] 102. Using the sample data and a preset evolutionary algorithm, perform evolutionary processing on the sub-feature thresholds to determine the target sub-feature thresholds of each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label.

[0052] Furthermore, after obtaining the sample data, the sub-feature thresholds can be optimized using the sample data to obtain the optimal sub-feature combination label. Specifically, the sample data and a preset evolutionary algorithm are used to perform evolutionary processing on the sub-feature thresholds to determine the target sub-feature thresholds for each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label.

[0053] It's important to note that evolutionary algorithms, or "evolutionary algorithms," constitute a family of algorithms. While they exhibit numerous variations, including different genetic expression methods, crossover and mutation operators, special operator references, and different regeneration and selection methods, their inspiration all stems from biological evolution in nature. Compared to traditional optimization algorithms based on calculus and exhaustive search, evolutionary computation is a mature, robust, and widely applicable global optimization method. It possesses self-organizing, adaptive, and self-learning characteristics, effectively handling complex problems that traditional optimization algorithms struggle with, regardless of the nature of the problem. Evolutionary algorithms include, but are not limited to, hybrid rice algorithms. The Hybrid Rice Optimization (HRO) algorithm is a metaheuristic algorithm that simulates the breeding process of three-line hybrid rice. This algorithm consists of three stages: hybridization, self-crossing, and renewal. It can optimize sample data through different breeding methods to obtain optimal data. Therefore, the aforementioned sample data can be input into an evolutionary algorithm to evolve the sub-feature thresholds. Through a process of natural selection, continuous updates and reproduction are achieved to obtain the final optimal sub-feature combination label, thereby determining the target sub-feature thresholds for each sub-feature of the optimal sub-feature combination label. By using target sub-feature thresholds to identify the QT interval of ECG signals, the accuracy of identification can be improved.

[0054] 103. Based on the target sub-feature threshold, perform rule construction processing to determine the target decision rule, which is used to identify the QT interval in the ECG signal to be identified.

[0055] Furthermore, after obtaining the target sub-feature threshold, rule construction can be performed based on the target sub-feature threshold to determine the target decision rule, thereby further improving the accuracy of QT interval identification.

[0056] This invention provides a method for determining decision rules. The method includes: acquiring sample data, which includes sub-feature combination labels of several sample electrocardiogram (ECG) signals. Each sub-feature combination label includes at least a sub-feature threshold of each sub-feature of the confirmed QT interval in the sample ECG signals. The method further includes: performing evolutionary processing on the sub-feature thresholds using the sample data and a preset evolutionary algorithm to determine target sub-feature thresholds for each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label that achieves the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label. Finally, the method involves rule construction processing based on the target sub-feature thresholds to determine a target decision rule. This target decision rule is used to perform feature recognition of the QT interval in the ECG signal to be identified. Through this method, the sub-feature thresholds in the sample data can be iteratively evolved using an evolutionary algorithm to obtain target sub-feature thresholds with recognition accuracy not lower than a preset accuracy threshold. Target decision rules are then formulated based on these target sub-feature thresholds to identify QT intervals, which helps improve the accuracy of QT interval recognition and reduces errors caused by manually setting thresholds.

[0057] Please see Figure 2 , Figure 2 This is another flowchart of a method for determining decision rules in an embodiment of the present invention, as shown below. Figure 2 The method shown includes the following steps:

[0058] 201. Obtain sample data, the sample data including sub-feature combination labels of several sample electrocardiogram signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signals;

[0059] It should be noted that step 201 and Figure 1 The content of step 101 shown is similar, and will not be repeated here to avoid repetition. Please refer to [link / reference needed] for details. Figure 1 The content of step 101 shown.

[0060] 202. Using a preset fitness function and the individual's encoding value, determine the target fitness of the individual. The individual corresponds one-to-one with the sub-feature combination label. The encoding value includes the encoding value of the sub-feature threshold of the sub-feature combination label. The fitness is used to reflect the recognition accuracy of the QT interval.

[0061] It should be noted that, in order to optimize the sub-feature threshold and determine the optimal sub-feature threshold, firstly, the sub-feature combination label can be regarded as an individual to be evolved, that is, an individual and a sub-feature combination label are mapped one-to-one. The sub-feature threshold in the sub-feature combination label can be used to encode the individual, so that the encoded value of the individual includes the encoded value of the sub-feature threshold of the sub-feature combination label. In this way, the individual to be evolved is evolved through an evolutionary algorithm to obtain the optimal individual, which is the optimal sub-feature combination label.

[0062] Taking the hybrid rice algorithm as an example of an evolutionary algorithm, the hybrid rice algorithm needs to be initialized before executing step 202. Therefore, steps L01-L02 are also included before step 202:

[0063] L01. Initialize the algorithm parameters of the evolutionary algorithm based on the sample data. The algorithm parameters include at least the individual dimension of the individual, the maximum number of iterations, and the search space of the individual encoding.

[0064] L02. Encode the sub-feature thresholds of the sub-feature combination label within the search space to obtain the encoded value of each individual. The encoding rule includes that each dimension in each individual represents a sub-feature threshold.

[0065] That is, firstly, the algorithm parameters of the hybrid rice algorithm are initialized in step L01. These parameters include, but are not limited to, population size, individual dimension, maximum number of iterations, and upper and lower bounds of individual encoding. Population size can be understood as the population scale, reflecting the number of individuals in the population, and can be any N combinations of the aforementioned sub-feature combination labels. Individual dimension refers to the number of sub-feature thresholds in the sub-feature combination label corresponding to the individual; the number of sub-feature thresholds is used as the individual dimension. The maximum number of iterations is the number of iterations in the evolutionary process. The upper and lower bounds of individual encoding indicate the range of sub-feature threshold values, and the space formed by these upper and lower bounds is the search space. Further, the sub-feature thresholds are initialized in step L02. The population is randomly initialized within the search space, and each individual is encoded (i.e., the sub-feature threshold is initialized). The encoding rule includes each dimension of each individual representing a sub-feature threshold, thus mapping each individual to a sub-feature combination label. The encoded value can also be considered as the real value of the sub-feature threshold.

[0066] Furthermore, fitness is used as an individual evaluation criterion in the evolutionary process to determine the survival of the fittest. Therefore, after initialization, it is necessary to calculate the fitness of each individual. In step 202, the target fitness of the individual is determined using a preset fitness function and the individual's encoded value. The fitness function is the total length of the intersection of the threshold ranges of each sub-feature (e.g., sub-features A1, A2, A3, A4, A5, A6). Fitness reflects the accuracy of QT interval recognition, and fitness is directly proportional to recognition accuracy.

[0067] 203. Based on the target fitness and the preset subpopulation division rules, determine the subpopulation to which each individual belongs;

[0068] After obtaining the target fitness for each individual, subpopulations can be formed based on the target fitness. These subpopulations are used for subsequent reproductive evolution. For example, after calculating the fitness function for each individual, the individuals are ranked according to the quality of their fitness functions, thus dividing the population into subpopulations. Subpopulations include sterile lines, maintainer lines, and restorer lines, and different subpopulations have different reproductive evolutionary pathways.

[0069] In one feasible implementation, to improve the accuracy of subpopulation division, this embodiment divides the subpopulations to which individuals belong based on fitness ranking results. That is, step 203 may include steps P01 to P02:

[0070] P01. Arrange the target fitness of each individual in descending order of value to determine the fitness decreasing sequence of the population. The fitness decreasing sequence includes the correspondence between individuals and target fitness.

[0071] P02. All individuals in the first proportion of the fitness decreasing sequence are identified as maintainer lines; all individuals in the second proportion of the fitness decreasing sequence are identified as sterile lines; and the remaining individuals in the fitness decreasing sequence, excluding all individuals in the first proportion and all individuals in the second proportion, are identified as restorer lines.

[0072] It should be noted that after obtaining the target fitness of individuals, the individuals can be sorted according to the magnitude of the target fitness, and the resulting individual sequence can be used to divide the subpopulations. Specifically, the target fitness of each individual is arranged in descending order to determine the fitness decreasing sequence of the population. The fitness decreasing sequence includes the correspondence between individuals and target fitness. All individuals in the first proportion of the fitness decreasing sequence are identified as maintainer lines; all individuals in the second proportion of the fitness decreasing sequence are identified as sterile lines; and all individuals in the fitness decreasing sequence other than those in the first proportion and the second proportion are identified as restorer lines. The individuals in the sequence can be equally divided into the subpopulations, i.e., the first proportion and the second proportion are equal, or they can be unequal, in which case the first proportion and the second proportion are not equal.

[0073] Taking hybrid rice algorithms as an example, rice seeds are the individuals mentioned above. The Hybrid Rice Optimization (HRO) algorithm is a metaheuristic algorithm that simulates the three-line hybrid rice breeding process. In each iteration, the rice seed population is sorted according to fitness from best to worst, divided into three subpopulations. Based on self-balancing and symmetry, each subpopulation is designed with the same number of individuals. The top third of the individuals in terms of fitness are selected as maintainer lines, the bottom third as sterile lines, and the rest as restorer lines.

[0074] Among them, the sterile line is a rice male-sterile line, in which the male organs in its own flower are underdeveloped and cannot produce normal pollen, while its female organs are normally developed. Therefore, it cannot reproduce on its own and needs to rely on foreign rice pollen to produce seeds. The restorer line is a male-sterile restorer line, which refers to a line that can restore male fertility in its offspring after being crossed with a sterile line. It is achieved by using a variety with the genotype N (RR) as the male parent and crossing it with a male-sterile line with the genotype S (rr). The maintainer line is a male-sterile line, because the pollen of the sterile line itself is sterile and it does not produce seeds through self-pollination. It cannot reproduce offspring with sterile characteristics through self-pollination. It must be pollinated by a specific variety with normal fertility and produce seeds so that the offspring of the sterile line can still retain its male sterility. This specific male parent variety that can maintain the sterile line's characteristics from generation to generation is called a male-sterile maintainer line.

[0075] 204. Using the encoded value, the type of the subpopulation, and the evolutionary algorithm, determine the evolutionary encoded value of each individual;

[0076] Furthermore, after dividing into subpopulations, the evolution of individuals within each subpopulation can be performed. Specifically, using coding values, subpopulation types, and evolutionary algorithms, the evolutionary coding value of each individual is determined. Continuing with the hybrid rice algorithm as an example, subpopulations include at least sterile lines, maintainer lines, and restorer lines. Different subpopulations correspond to different reproductive evolutionary methods. The evolutionary process involves using the coding values ​​of individuals to reproduce and obtain new coding values.

[0077] Wherein, for the subpopulations of the sterile line and the subpopulations of the maintainer line, the evolutionary algorithm is a hybridization algorithm, then step 204 may include: randomly selecting the first individual of the sterile line and the second individual of the maintainer line; inputting the coding value of the first individual and the coding value of the second individual into the hybridization algorithm to determine the evolutionary coding value of the first individual of the sterile line.

[0078] It should be noted that for the subpopulation of the sterile line, its individual evolution is achieved through the hybridization stage. Specifically, the hybridization stage is to use hybridization technology to update the individuals of the sterile line. The hybridization technology includes the hybridization algorithm. Specifically, two kinds of rice seeds are randomly selected from the maintainer line and the sterile line to construct a new individual, that is, the first individual of the sterile line and the second individual of the maintainer line are randomly selected; the coding values ​​of the first individual and the second individual are input into the hybridization algorithm to determine the evolutionary coding value of the first individual of the sterile line. This evolutionary coding value is the coding value of the new individual. And through the following step 205, it is determined whether the new individual can replace the original individual. If the new rice seed is superior to the current rice seed, the current rice seed will be replaced by the new seed, that is, the new individual is used to replace the subpopulation of the sterile line. Therefore, the evolutionary algorithm includes at least the hybridization algorithm (1):

[0079]

[0080] In the formula, The k-th gene of the i-th individual in the sterile line, where the k-th gene is also the k-th sub-characteristic; That is, the evolutionary code value of the new individual obtained by crossing the sterile line and the maintainer line, which is also the first individual of the sterile line;

[0081] The kth gene corresponding to an individual randomly selected from the sterile line, which is also the coding value of the first individual in the sterile line;

[0082] The k-th gene corresponding to an individual randomly selected from the maintainer line, which is also the coding value of the second individual in the maintainer line;

[0083] r1 and r2: random numbers between -1 and 1.

[0084] Furthermore, for the subpopulations of the restorer line, evolution is carried out through self-pollination. The evolutionary algorithm includes at least a first self-pollination algorithm and a second algorithm. Step 204 can also be used to evolve individuals of the restorer line. Step 204 can also include: randomly selecting a third individual of the restorer line; determining the historical self-pollination count of the third individual; when the historical self-pollination count is not higher than a preset self-pollination count threshold, the encoding value of the third individual is input into the first self-pollination algorithm to determine the evolutionary encoding value of the third individual of the restorer line, and the historical self-pollination count is increased by a second preset count threshold.

[0085] It should be noted that for individuals in the subpopulation of the restorer line, individual reproduction and renewal are carried out through the self-pollination stage. Self-pollination is the behavior of optimizing the gene sequence of rice seeds of the restorer line, so that the rice seeds gradually approach the best seeds. The individual evolution of the restorer line is achieved by having the best individuals participate in the self-pollination process. The first self-pollination algorithm is shown in Equation (2):

[0086] X new(i) = rand(0,1)·(X best -X j,r )+X i (2)

[0087] In the formula, X new(i) This represents the evolutionary code value of the new individual produced by self-pollination of the i-th individual of the restorer line, which is also the third individual of the restorer line;

[0088] X best This represents the current optimal solution, or the optimal individual. The optimal individual can be the individual with the highest fitness obtained by comparing the fitness of individuals, which is also the encoded value of the optimal individual.

[0089] X j,r This represents the encoded value of the j-th individual randomly selected from the restorer line, which is also the third individual in the restorer line;

[0090] X i This represents the encoded value of the i-th individual in the restorer system, that is, the i-th individual in the restorer system.

[0091] Then, in subsequent step 205, it is determined whether the new individual can replace the old individual, and a process of natural selection is carried out. If the new individual after self-pollination is superior to the old individual, then the new individual replaces the old individual, and the current historical self-pollination number t is set. i Set to 0, otherwise the number of historical self-intersections t i =t i +1.

[0092] In one feasible implementation, the evolutionary algorithm further includes a second self-crossing algorithm. The method further includes: when the historical self-crossing count of the third individual is higher than a preset self-crossing count threshold, the evolutionary encoding value of the third individual of the restorer line is determined using the third individual, the upper and lower bounds of the search space, and the second self-crossing algorithm, and the historical self-crossing count is increased by the second preset count threshold.

[0093] It should be noted that when the historical self-pollination count of a third individual exceeds a preset self-pollination count threshold, the individual is updated through an update phase. This update phase resets the rice seeds in the restorer line that have not been updated consecutively (i.e., have reached the maximum self-pollination count). The preset self-pollination count threshold can be the maximum self-pollination count t. max The second self-intersection algorithm in the update phase is shown in equation (3):

[0094] X new(i) =X i +rand(0,1)·(R max -R min )+R min (3)

[0095] In the formula, X new(i) The evolutionary code value of the new individual generated by updating the i-th individual of the restorer line, which is also the third individual;

[0096] Xi is the encoded value of the i-th individual in the restorer line, which is also the third individual in the restorer line;

[0097] R max This is the upper bound of the search space;

[0098] R min This is the lower bound of the search space.

[0099] 205. Based on the encoding value and the evolutionary encoding value, compare the fitness to determine the target encoding value of the individual; use the target encoding value as the encoding value, increase the number of iterations by a first preset threshold, and return to execute the step of determining the target fitness of the individual using the preset fitness function and the individual's encoding value, until the number of iterations is not less than the maximum number of iterations, then output the target encoding value of each individual;

[0100] After obtaining the evolutionary coding value of an individual in step 204, the best and worst individuals can be selected through a process of natural selection. The fitness of each individual is calculated using a fitness function. By comparing the fitness values ​​with the evolutionary coding values, the result of the natural selection is determined. If the fitness of a new individual is higher than that of the corresponding old individual, then the new individual can replace the old individual for subsequent iterations; otherwise, no replacement is needed, and the old individual continues to be used for subsequent iterations. This process determines the target coding value for each individual. Furthermore, the iteration count needs to be updated at the end of each iteration. When the maximum number of iterations is reached, iteration stops, and the optimal individual is obtained.

[0101] 206. The optimal individual is determined using the target encoding value. The sub-feature combination label corresponding to the optimal individual is the optimal sub-feature combination label. The target sub-feature threshold is the sub-feature threshold of each sub-feature corresponding to the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature threshold of the sub-feature combination label.

[0102] For example, the simplified process of steps 204 to 206 can be referred to as steps H21 to H27 as follows: Step H21: Initialize parameters; Step H22: Initialize the population; Step H23: Calculate the fitness function and divide the subpopulations; Step H24: Perform self-crossing, renewal, and hybridization on the subpopulations; Step H25: Obtain the optimal individual; Step H26: Determine whether the maximum number of iterations has been reached. If yes, proceed to step H27; otherwise, repeat steps H23, H24, and H25; Step H27: Output the optimal threshold for the sub-feature. At this point, the optimal threshold is the target sub-feature threshold. That is, after completing the above evolutionary iteration process, the optimal individual can be obtained through step 206 using the individual's target encoding value. For example, by comparing the fitness of individuals, the individual with the highest hardness is considered the optimal individual. This yields the sub-feature combination label with the highest accuracy in QT interval identification. It is understandable that the above-mentioned evolutionary process of hybridization or self-pollination optimizes and updates the sub-feature thresholds of each individual through reproductive evolution, so that the sub-feature thresholds of the individuals obtained in each iteration can be optimal, and then the final optimal individual is obtained through the selection of the best among the best, thereby improving the accuracy of the sub-feature thresholds.

[0103] 207. Based on the target sub-feature threshold, perform rule construction processing to determine the target decision rule, which is used to identify the QT interval in the ECG signal to be identified.

[0104] It should be noted that step 207 and Figure 1 Step 103 in the method shown is similar, and will not be repeated here to avoid repetition. Please refer to [link to relevant documentation] for details. Figure 1 The content of step 103 in the method shown.

[0105] This invention provides a method for determining decision rules, the method comprising: acquiring sample data, the sample data including sub-feature combination labels of several sample electrocardiogram signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signals; determining the sub-population to which each individual belongs based on the target fitness and a preset sub-population division rule; determining the evolutionary coding value of each individual using the coding value, the type of the sub-population, and an evolutionary algorithm; comparing the fitness of the individual based on the coding value and the evolutionary coding value to determine the target coding value of the individual; using the target coding value as the coding value, increasing the iteration number by a first preset number threshold, and returning to execute the algorithm using a preset fitness function to determine the sub-population ... The process involves determining the target fitness of an individual by analyzing its encoded value, iterating until the number of iterations is not less than the maximum number of iterations, and then outputting the target encoded value for each individual. The optimal individual is determined using the target encoded value, and the sub-feature combination label corresponding to the optimal individual is called the optimal sub-feature combination label. The target sub-feature threshold is the sub-feature threshold of each sub-feature corresponding to the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature threshold of the sub-feature combination label. Rule construction is performed based on the target sub-feature threshold to determine the target decision rule, which is used to perform feature recognition of the QT interval in the ECG signal to be identified. Through this method, an evolutionary algorithm, such as the hybrid rice algorithm, can be used to iterate the sub-feature threshold in the sample data to obtain a scientific and standard optimal sub-feature threshold in the QT algorithm. This yields the target sub-feature threshold with recognition accuracy not less than a preset accuracy threshold. The target decision rule is then formulated based on the target sub-feature threshold to identify the QT interval, which helps improve the accuracy of QT interval recognition and reduces errors caused by manually setting thresholds. Furthermore, by selecting the optimal individual based on fitness, which is proportional to the recognition accuracy, the accuracy and reliability of the results are further improved. The hybrid rice optimization algorithm used in this application has a simple structure and strong optimization ability, and has good application conditions for the QT algorithm based on a set threshold. Moreover, the QT algorithm decision rule formulated by the optimal threshold of the sub-features obtained by the algorithm can have a good effect on the calculation results of the QT interval.

[0106] Please see Figure 3 , Figure 3 This is a structural block diagram of a decision rule determination device according to an embodiment of the present invention, such as... Figure 3 The apparatus shown includes:

[0107] Sample determination module 301: used to acquire sample data, the sample data including sub-feature combination labels of several sample electrocardiogram signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample electrocardiogram signal;

[0108] Threshold determination module 302: used to perform evolutionary processing of the sub-feature threshold using the sample data and a preset evolutionary algorithm, and determine the target sub-feature threshold of each sub-feature of the optimal sub-feature combination label. The optimal sub-feature combination label is the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature threshold of the sub-feature combination label.

[0109] Rule determination module 303: is used to perform rule construction processing based on the target sub-feature threshold to determine the target decision rule, which is used to identify the QT interval in the ECG signal to be identified.

[0110] It should be noted that, Figure 3 The function of each module in the device shown is as follows: Figure 1 The steps in the method shown are similar, and will not be repeated here to avoid repetition. Please refer to [link / reference needed] for details. Figure 1 The content of each step in the method shown.

[0111] This invention provides a device for determining decision rules. The device includes: a sample determination module for acquiring sample data, the sample data including sub-feature combination labels of several sample electrocardiogram (ECG) signals, the sub-feature combination labels including at least the sub-feature thresholds of each sub-feature of the confirmed QT interval in the sample ECG signals; a threshold determination module for using the sample data and a preset evolutionary algorithm to perform evolutionary processing on the sub-feature thresholds, determining the target sub-feature thresholds of each sub-feature of the optimal sub-feature combination label, the optimal sub-feature combination label being the sub-feature combination label with the highest recognition accuracy when performing QT interval feature recognition based on the sub-feature thresholds of the sub-feature combination label; and a rule determination module for performing rule construction processing based on the target sub-feature thresholds to determine the target decision rule, the target decision rule being used for feature recognition of the QT interval in the ECG signal to be recognized. Through the above method, the sub-feature thresholds in the sample data can be iteratively evolved using an evolutionary algorithm to obtain target sub-feature thresholds with recognition accuracy not lower than a preset accuracy threshold. Target decision rules are formulated based on the target sub-feature thresholds to identify QT intervals, which helps improve the accuracy of QT interval recognition and reduces errors caused by manually setting thresholds.

[0112] Figure 4 An internal structural diagram of a computer device in one embodiment is shown. This computer device can specifically be a terminal or a server. Figure 4As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. The internal memory may also store a computer program, which, when executed by the processor, causes the processor to perform the aforementioned methods. Those skilled in the art will understand that… Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0113] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform... Figure 1 or Figure 2 The steps of the method shown.

[0114] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, causes the processor to perform... Figure 1 or Figure 2 The steps of the method shown.

[0115] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0116] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method of determining a decision rule, characterized by, The method comprises: acquiring sample data, wherein the sample data comprises a sub-feature combination label of a plurality of sample electrocardio signals, and the sub-feature combination label at least comprises a sub-feature threshold value of each sub-feature of a confirmed QT interval in the sample electrocardio signal; evolving the sub-feature threshold value by using the sample data and a preset evolution algorithm to determine a target sub-feature threshold value of each sub-feature of an optimal sub-feature combination label, wherein the optimal sub-feature combination label is a sub-feature combination label with the highest recognition accuracy when performing feature recognition on the QT interval based on the sub-feature threshold value of the sub-feature combination label; performing rule construction processing according to the target sub-feature threshold value to determine a target decision rule, wherein the target decision rule is used for performing feature recognition on the QT interval in a to-be-recognized electrocardio signal; wherein the evolving the sub-feature threshold value by using the sample data and the preset evolution algorithm to determine the target sub-feature threshold value of each sub-feature of the optimal sub-feature combination label comprises: determining a target fitness of an individual by using a preset fitness function and a coding value of the individual, wherein the individual corresponds to the sub-feature combination label one by one, the coding value comprises a coding value of the sub-feature threshold value of the sub-feature combination label, and the fitness is used for reflecting the recognition accuracy of the QT interval; determining a sub-population to which each individual belongs according to the target fitness and a preset sub-population division rule; determining an evolution coding value of each individual by using the coding value, the type of the sub-population and the evolution algorithm; performing fitness comparison according to the coding value and the evolution coding value to determine a target coding value of the individual; taking the target coding value as the coding value, increasing an iteration number by a first preset number threshold value, and returning to perform the step of determining the target fitness of the individual by using the preset fitness function and the coding value of the individual until the iteration number is not less than a maximum iteration number, and then outputting the target coding value of each individual; determining an optimal individual by using the target coding value, wherein the sub-feature combination label corresponding to the optimal individual is an optimal sub-feature combination label, and the target sub-feature threshold value is a sub-feature threshold value of each sub-feature corresponding to the optimal sub-feature combination label.

2. The method of claim 1, wherein, Before the determining the target fitness of the individual by using the preset fitness function and the coding value of the individual, the method further comprises: initializing an algorithm parameter of the evolution algorithm based on the sample data, wherein the algorithm parameter at least comprises an individual dimension of the individual, the maximum iteration number and a search space of individual coding; coding the sub-feature threshold value of the sub-feature combination label within the search space to obtain the coding value of each individual, and a coding rule of the coding comprises that each dimension in each individual represents a sub-feature threshold value.

3. The method of claim 2, wherein, The sub-population at least comprises a sterile line and a maintenance line, and the evolution algorithm at least comprises a hybridization algorithm, and the determining the evolution coding value of each individual by using the type of the sub-population and the preset evolution algorithm comprises: randomly selecting a first individual of the sterile line and a second individual of the maintenance line; Inputting the coding value of the first individual and the coding value of the second individual into the hybrid algorithm to determine the evolutionary coding value of the first individual of the sterile line.

4. The method of claim 2, wherein, The sub-population further includes a restoration line, and the evolutionary algorithm at least includes a first self-crossing algorithm, so that the evolutionary coding value of each individual is determined by using the type of the sub-population and the preset evolutionary algorithm, and further includes: Randomly selecting a third individual of the restoration line; Determining the historical self-crossing times of the third individual; When the historical self-crossing times are not higher than a preset self-crossing times threshold, inputting the coding value of the third individual into the first self-crossing algorithm to determine the evolutionary coding value of the third individual of the restoration line, and increasing the historical self-crossing times by a second preset times threshold.

5. The method of claim 4, wherein, The evolutionary algorithm further includes a second self-crossing algorithm, so that the method further includes: When the historical self-crossing times are higher than the preset self-crossing times threshold, inputting the third individual, the upper and lower bounds of the search space and the second self-crossing algorithm to determine the evolutionary coding value of the third individual of the restoration line, and increasing the historical self-crossing times by the second preset times threshold.

6. The method of claim 1, wherein, The sub-population includes a sterile line, a maintainer line and a restoration line, so that the sub-population to which each individual belongs is determined according to the target fitness and a preset sub-population division rule, and the method includes: Arranging the target fitness of each individual in a descending order of value to determine a fitness descending sequence of the population, and the fitness descending sequence includes the corresponding relationship between the individual and the target fitness; All individuals ranked in a first proportion in the fitness descending sequence are determined as the maintainer line; All individuals ranked in a second proportion after the first proportion in the fitness descending sequence are determined as the sterile line; All individuals in the fitness descending sequence except the individuals ranked in the first proportion and the individuals ranked in the second proportion after the first proportion are determined as the restoration line.

7. An apparatus for determining a decision rule, characterized by The device includes: A sample determination module is configured to obtain sample data, wherein the sample data includes a sub-feature combination label of a plurality of sample electrocardio signals, and the sub-feature combination label at least includes a sub-feature threshold value of each sub-feature of a confirmed QT interval in the sample electrocardio signal. A threshold value determination module is configured to perform evolutionary processing on the sub-feature threshold value by using the sample data and a preset evolutionary algorithm to determine a target sub-feature threshold value of each sub-feature of an optimal sub-feature combination label, wherein the optimal sub-feature combination label is a sub-feature combination label with the highest recognition accuracy when performing feature recognition on a QT interval based on the sub-feature threshold value of the sub-feature combination label. A rule determination module is configured to perform rule construction processing according to the target sub-feature threshold value to determine a target decision rule, wherein the target decision rule is used for performing feature recognition on a QT interval in a to-be-recognized electrocardio signal. The threshold determination module is specifically configured to: determine a target fitness of the individual by using a preset fitness function and a coding value of the individual, the individual corresponding to the sub-feature combination label one-to-one, the coding value including a coding value of a sub-feature threshold of the sub-feature combination label, the fitness being used to reflect an identification accuracy of the QT interval; determine a sub-population to which each individual belongs according to the target fitness and a preset sub-population division rule; determine an evolution coding value of each individual by using the coding value, a type of the sub-population and an evolution algorithm; determine a target coding value of the individual by comparing the fitness according to the coding value and the evolution coding value; take the target coding value as the coding value, increase an iteration number by a first preset number threshold, and return to execute the step of determining the target fitness of the individual by using the preset fitness function and the coding value of the individual until the iteration number is not less than a maximum iteration number, and then output the target coding value of each individual; determine an optimal individual by using the target coding value, the sub-feature combination label corresponding to the optimal individual being an optimal sub-feature combination label, and the target sub-feature threshold being a sub-feature threshold of each sub-feature corresponding to the optimal sub-feature combination label.

8. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor, so that the processor executes the steps of the method according to any one of claims 1 to 6. 9.A computer device, comprising a memory and a processor, and characterized in that, The memory stores a computer program, and the computer program is executed by the processor, so that the processor executes the steps of the method according to any one of claims 1 to 6.

Images