Classification method and related equipment for incremental medical data based on rule classifier

CN114255830BActive Publication Date: 2024-11-26NAT UNIV OF DEFENSE TECH
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202111576268.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-21
Publication Date
2024-11-26
Estimated Expiration
2041-12-21

AI Technical Summary

Technical Problem

The existing medical data classification methods are expensive to obtain and analyze, especially the PSG evaluation method, which requires a large number of expensive monitoring signals and special instruments, and the data acquisition is in sequence, resulting in waste of resources and high costs.

Method used

The incremental medical data classification method based on the rule classifier is adopted, and the attribute reliability and classification accuracy are calculated by preset categories and attributes, the initial feature subset is constructed, and the feature subset is gradually adjusted to achieve effective classification of medical data.

Benefits of technology

Under the circumstances of limited and incremental data, effective classification of medical data is achieved, monitoring and analysis costs are reduced, resources are saved, and classification accuracy is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114255830B_ABST
    Figure CN114255830B_ABST
Patent Text Reader

Abstract

The present application provides a classification method and related equipment for incremental medical data based on a rule classifier; the method comprises: presetting multiple categories for a medical data set, and determining the attributes and categories possessed by each sample in the medical data set; determining the classification order when performing classification on each category; in each classification, calculating the attribute reliability of each attribute, and obtaining a reliability ranking; selecting a predetermined number of attributes as an initial feature subset according to the reliability ranking, and constructing a rule classifier, and classifying the samples using the rule classifier to obtain a classification accuracy; adjusting the attributes in the initial feature subset according to the reliability ranking and the classification accuracy to obtain a target feature subset; classifying all categories according to the classification order to obtain a target feature subset and a rule classifier for each classification operation; integrating all target feature subsets and rule classifiers to classify new samples in the medical data set.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present application relate to the technical field of data processing, and in particular, to a classification method for incremental medical data based on a rule classifier and related equipment. Background Art

[0002] In existing medical data classification, such as the PSG (polysomnography) evaluation method, a large number of different monitoring signals are used as multiple features or attributes of the data, and the sleep state is evaluated based on the numerical value generated by each feature. However, in actual applications, data acquisition often has a sequence, and the conditions for obtaining certain attributes are harsh or expensive. For example, in the PSG evaluation method, a large number of monitoring signals come from monitoring different parts of the body and using different special instruments, which often means that patients need to pay a lot of costs, and many signal data monitoring processes need to be carried out in special places, which brings very high monitoring costs and data analysis costs.

[0003] In some cases, such as the classification of medical data in PSG assessment, it is possible to use some attributes to make a rough classification, and then determine whether to obtain subsequent attributes based on the results. This greatly saves time and resources. Therefore, it is necessary to first use the existing limited data for classification.

[0004] Based on this, a solution is needed that can effectively classify medical data in the case of limited data and incremental data. Summary of the invention

[0005] In view of this, the purpose of the present application is to propose a classification method and related equipment for incremental medical data based on a rule classifier.

[0006] Based on the above purpose, the present application provides a classification method for incremental medical data based on a rule classifier, which is applied to a database storing a medical data set, including:

[0007] Presetting a plurality of categories for the medical data set, and determining a plurality of attributes possessed by each sample in the medical data set and a category to which each sample initially belongs;

[0008] determining, among all of the categories, a sorting order in which the sorting is performed on each of the categories;

[0009] In each classification for a single category, for each attribute, the attribute reliability of the attribute is calculated according to the number of samples in each category when the attribute is used alone for classification, and the reliability ranking is obtained according to the reliabilities of all the attributes;

[0010] Selecting a predetermined number of the attributes from all the attributes as an initial feature subset according to the reliability ranking, constructing the rule classifier using the initial feature subset, and classifying all the samples using the rule classifier to obtain a classification accuracy rate for the initial feature subset;

[0011] Adjust the attributes in the initial feature subset according to the reliability ranking and the classification accuracy to obtain a target feature subset;

[0012] According to the classification order, all the categories are classified to obtain the target feature subset and the rule classifier for each classification operation; all the target feature subsets and the rule classifiers are integrated to classify new samples in the medical data set.

[0013] Furthermore, in each classification for a single category, for each attribute, the attribute reliability of the attribute is calculated according to the number of samples in each category when the attribute is used alone for classification, and the reliability ranking is obtained according to the reliabilities of all the attributes, including:

[0014] In each classification for a single category, a binary classification operation is performed on the category and non-category;

[0015] For each of the attributes, when all the samples are classified by the attribute alone, determining in each of the categories the number of samples that are determined to belong to the category by the attribute;

[0016] determining the attribute reliability of the attribute according to the number of determinations of the attribute;

[0017] According to the attribute reliability of each of the attributes, all the attributes are arranged in order of the attribute reliability from high to low.

[0018] Further, constructing the rule classifier using the initial feature subset includes:

[0019] For each attribute of the initial feature subset, multiple reference values ​​are matched for the attribute, and corresponding reference evidence is assigned to each reference value, and corresponding reference evidence weight is assigned to each reference evidence;

[0020] For each attribute of the initial feature subset, construct a similarity distribution between the attribute and the two categories in the binary classification operation according to the attribute and the reference value corresponding to the attribute, and construct a reference evidence matrix for the attribute according to the similarity distribution;

[0021] For each of the attributes, the reference evidence of the attribute is weighted according to the similarity distribution on the attribute and the reference evidence matrix to obtain the attribute evidence of the attribute, and the reference evidence weight of the reference evidence of the attribute is weighted to obtain the attribute evidence weight of the attribute, and all the attribute evidence is fused using the reliability and the attribute evidence weight to obtain the rule classifier.

[0022] Furthermore, the reliability ranking and the classification accuracy adjust the attributes in the initial feature subset to obtain a target feature subset, including:

[0023] Among the attributes that are not added to the initial feature subset, select the attribute with the highest reliability ranking and add it to the initial feature subset;

[0024] Based on the constructed rule classifier, classify all the samples using all the attributes in the initial feature subset;

[0025] Based on the classification result, calculating the classification accuracy rate of the initial feature subset;

[0026] According to the numerical change of the classification accuracy, adjusting the attribute in the initial feature subset;

[0027] The adjusted initial feature subset is used as the target feature subset.

[0028] Further, adjusting the attributes in the initial feature subset according to the numerical change of the classification accuracy includes:

[0029] In response to determining that the classification accuracy rate increases, retaining the attribute added last, and continuing to select the attribute with the highest reliability ranking from the attributes not added to the initial feature subset, and adding it to the initial feature subset;

[0030] In response to determining that the classification accuracy has not increased, the last added attribute is removed, and the attributes included in the initial feature subset are kept unchanged.

[0031] Further, selecting one of the attributes with the highest reliability ranking and adding it to the initial feature subset includes:

[0032] setting reliability thresholds for all of said attributes;

[0033] Selecting one of the attributes with the highest reliability ranking, and determining whether the attribute reliability of the attribute is greater than the reliability threshold;

[0034] In response to determining that the attribute reliability of the attribute is greater than or equal to a reliability threshold, adding the attribute to the initial feature subset;

[0035] In response to determining that the attribute reliability of the attribute is less than a reliability threshold, the attribute is not added to the initial feature subset, and the attribute included in the initial feature subset remains unchanged.

[0036] Further, according to the attribute and the reference value corresponding to the attribute, a similarity distribution between the attribute and the two categories in the binary classification operation is formed, and a reference evidence matrix of the attribute is constructed according to the similarity distribution, including:

[0037] On each of the samples, for each of the reference values ​​in each of the attributes, a first difference between the reference value and another adjacent reference value is calculated; and a second difference between the attribute and another adjacent reference value is calculated; and a matching degree between the attribute and the reference value is determined according to a ratio of the first difference to the second difference;

[0038] Based on the attribute, the similarity distribution is formed using the matching degrees of all the reference values ​​and the two categories in the binary classification operation;

[0039] For each of the categories, in all the samples, the sum of the matching degrees belonging to the same attribute and the same reference value is calculated as the first parameter of the similarity distribution;

[0040] For each of the categories, in all the samples, the sum of the first parameters of all the attributes is calculated as the second parameter of the similarity distribution;

[0041] Using the first parameter, the second parameter and the matching degree to form the similarity distribution;

[0042] In the case where the category is determined, for each reference value matched by each attribute, constructing the likelihood of taking the attribute as the reference value according to the ratio of the first parameter to the second parameter in the similarity distribution;

[0043] By respectively normalizing the likelihood of each category, the reliability of the reference evidence when each of the reference evidence judges the sample is obtained;

[0044] When the attribute is determined, the reference evidence matrix is ​​constructed according to the reference evidence reliability and the category of each reference evidence of the attribute.

[0045] Further, the reference evidence of the attribute is weighted to obtain the attribute evidence of the attribute, and the reference evidence weight of the reference evidence of the attribute is weighted to obtain the attribute evidence weight of the attribute, including:

[0046] Based on the distribution of the reference evidence matrix, in response to determining that the attribute value is between two adjacent reference values, obtaining the evidence credibility of the attribute according to the matching degree of the two reference values ​​and the reference evidence credibility; obtaining the attribute evidence weight of the attribute according to the matching degree of the two reference values ​​and the reference evidence weight;

[0047] The attribute evidence weight, the evidence credibility and the reliability are combined to obtain the attribute evidence of the attribute.

[0048] Furthermore, all the attribute evidences are fused using the reliability and the attribute evidence weight to obtain the rule classifier, including:

[0049] For each of the attribute evidence, according to the fusion rule in the evidence reasoning rule, all the attribute evidence of the sample is fused using the reliability factor and the attribute evidence weight to obtain the discrimination reliability of the attribute evidence in judging the sample as each of the categories;

[0050] The rule classifier is constructed using each of the categories and the discrimination confidence of the category.

[0051] Based on the same inventive concept, the present application also provides a classification device for incremental medical data based on a rule classifier, connected to a database storing medical data sets, comprising: a preprocessing module, a single classification module and a global classification module;

[0052] The preprocessing module is configured to preset multiple categories for the medical data set, and determine multiple attributes possessed by each sample in the medical data set and the category to which each sample initially belongs.

[0053] Furthermore, a classification order when performing classification on each of the categories is determined among all the categories.

[0054] The single classification module is configured to, in each classification for a single category, calculate the attribute reliability of each attribute according to the number of samples in each category when the attribute is used alone for classification, and obtain a reliability ranking according to all the attribute reliabilities.

[0055] Furthermore, a predetermined number of the attributes are selected from all the attributes as an initial feature subset according to the reliability ranking, the rule classifier is constructed using the initial feature subset, and all the samples are classified using the rule classifier to obtain the classification accuracy of the initial feature subset.

[0056] Furthermore, the attributes in the initial feature subset are adjusted according to the reliability ranking and the classification accuracy to obtain a target feature subset.

[0057] The global classification module is configured to classify all the categories according to the classification order to obtain the target feature subset and the rule classifier for each classification operation; integrate all the target feature subsets and the rule classifiers to classify new samples in the medical data set.

[0058] Based on the same inventive concept, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the classification method for incremental medical data based on a rule classifier as described in any one of the above items is implemented.

[0059] Based on the same inventive concept, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the classification method of incremental medical data based on the rule classifier as mentioned above.

[0060] From the above, it can be seen that the classification method and related equipment for incremental medical data based on rule classifiers provided by the present application classify the classification order of each category according to actual needs, and only classify one category in each classification. In each classification, all attributes are sorted based on attribute reliability, and only attributes that can improve the accuracy of this classification are used to establish an initial feature subset. By comprehensively considering the attribute reliability sorting and classification accuracy, the attributes in the initial feature subset are adjusted, and the rule classifier is trained using the first data attributes for rough classification, and then other classifiers are continuously trained according to the attributes obtained subsequently, and more detailed classification is gradually performed. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1 A flow chart of a method for classifying incremental medical data based on a rule classifier according to an embodiment of the present application;

[0063] Figure 2 This is a schematic diagram of a module of a classification device for incremental medical data based on a rule classifier according to an embodiment of the present application;

[0064] Figure 3a Schematic diagram a of the classification process of incremental medical data based on a rule classifier in an embodiment of the present application;

[0065] Figure 3b Schematic diagram b of the classification process of incremental medical data based on a rule classifier in an embodiment of the present application;

[0066] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0067] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0068] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0069] As described in the background technology section, the related classification methods for incremental medical data based on rule classifiers are still difficult to meet the needs of practical applications.

[0070] In the process of implementing the present application, the applicant discovered that in existing medical data classification, such as PSG (polysomnography) evaluation methods, a large number of different monitoring signals are used as multiple features or attributes of the data, and the sleep state is evaluated based on the numerical value generated by each feature. However, in actual applications, the acquisition of data often has a sequence, and the conditions for obtaining certain attributes are harsh or expensive.

[0071] For example, in the PSG evaluation method related to medical data classification, a large number of monitoring signals come from monitoring different parts of the body and using different special instruments, which often means that patients need to pay a lot of costs. In addition, the monitoring process of many signal data needs to be carried out at different times. For example, when the patient is just about to fall asleep, the data attributes related to the rapid eye movement sleep state are useless. Therefore, there is a situation where some attributes can be used to roughly classify the patient's sleep state. According to the classification results, it is determined whether subsequent attributes need to be obtained, which greatly saves time and resources.

[0072] Taking polysomnography as an example, in order to save the cost of monitoring and data processing, the medical data samples can be roughly classified first, only distinguishing whether the patient is in a sleeping state. When the patient is in a sleeping state, other data attributes obtained subsequently can be used to classify whether the patient is in deep sleep or light sleep. After that, data attributes related to rapid eye movement sleep and slow wave sleep can be obtained to divide the medical data into more detailed categories.

[0073] It can be understood that the method can be executed by any device, equipment, platform, or device cluster having computing and processing capabilities.

[0074] The technical method of the present application is described in detail below through specific embodiments.

[0075] refer to Figure 1 A classification method for incremental medical data based on a rule classifier according to an embodiment of the present application comprises the following steps:

[0076] Step S101, presetting multiple categories for the medical data set, and determining multiple attributes possessed by each sample in the medical data set and the category to which each sample initially belongs; and further determining the classification order when performing classification on each category among all the categories.

[0077] In the embodiments of the present application, human sleep health monitoring is taken as a specific example. In the monitoring technology of human sleep health, PSG is often used as an evaluation method, wherein PSG needs to monitor multiple vital signs of the patient through multiple sensors to obtain multiple signals about the sleep state of the monitored patient. For example, in a data sample obtained from each monitoring, more than a dozen vital signs such as the patient's respiratory rhythm, heart rate, electroencephalogram, electrocardiogram, electrooculogram, electromyogram, SPO2 (pulse oxygen saturation) and HR (heart rate) may be included. In the data obtained each time, the above-mentioned multiple signals are used as multiple attributes of the data sample obtained this time. Based on the above-mentioned more than ten attributes, it is inferred that the patient is in a deep sleep, shallow sleep or awake sleep state.

[0078] In this embodiment, the collected patient data can be used as a medical data set, and the patient data collected each time can be used as a data sample in the medical data set, also referred to as a sample in this application, and different sleep states can be used as multiple categories for classifying samples.

[0079] In this embodiment, if Figure 3a and Figure 3b As shown, the medical data samples will be classified multiple times, and each classification will only divide one of the above categories, that is, in each classification operation, a binary classification operation is performed.

[0080] Among them, the binary classification operation can be performed in two ways. In one way, Figure 3a As shown, classification is performed for a selected specific feature category, and the classification results are two categories: the specific category and non-specific category. Based on the attributes continuously acquired incrementally, the samples in non-specific category are again subjected to another specific category binary classification operation until all detailed categories are distinguished at once. In this embodiment, in each classification, the selection of a specific category can be selected according to actual needs and circumstances.

[0081] In another way, Figure 3b As shown, based on the existing data attributes, a coarse classification binary classification operation can be first performed on the medical data set, and then the two categories obtained by the coarse classification can be continuously refined according to the continuously acquired data attributes, that is, they can be gradually divided into more detailed categories.

[0082] In this embodiment, Figure 3a The first method shown is used as an example. It should be noted that the two methods only have slight differences in the classification order, and can be selected according to the specific scenario. There is no difference in the construction of the rule classifier and the operation of the method.

[0083] Furthermore, there are 6 categories marked by human sleep health monitoring data, namely a wakefulness state w, and five sleep states: S1-S5. Sleep states are divided into rapid eye movement sleep and non-rapid eye movement sleep, and non-rapid eye movement is divided into stages 1, 2, 3, and 4, of which stage 1 is the sleep latency period, stage 2 is the shallow sleep period, and stages 3 and 4 are deep sleep periods. Sample categories S1 to S4 correspond to stages 1, 2, 3, and 4 of the non-rapid eye movement period, and sample category S5 corresponds to the rapid eye movement period. The applicant's research found that the number of samples in sleep states S1 and S5 is small, because non-rapid eye movement stages 1 and 4 are closer to stages 2 and 3, respectively. In order to make the sample size of each category close and the categories easy to distinguish, non-rapid eye movement stages 1 and 2 are collectively referred to as shallow sleep states, and non-rapid eye movement stages 3 and 4 are collectively referred to as deep sleep states. After redefining the sleep categories, sleep samples are divided into 4 categories, namely, awake state y1, light sleep state y2, deep sleep state y3, and rapid eye movement sleep state y4.

[0084] In this embodiment, PSG evaluation may involve 12 monitoring signals, and the acquisition of which requires multiple sensors. For most users who need daily sleep monitoring, the acquisition of PSG is considered expensive and inconvenient. However, the acquisition of SPO2 and HR, which are commonly used to distinguish between sleep and wakefulness, is relatively easy, and when classifying sleep states, if the person is determined to be in a wakeful state, there is no need to determine which sleep state it belongs to, which can simplify the acquisition and classification process of attributes related to other sleep states; and after determining that the patient has entered a sleep state, further related attributes can be obtained to determine whether it is a light sleep or other sleep state, and finally determine whether it is a deep sleep or a REM sleep state.

[0085] In this embodiment, it is determined that a total of three classifications are required. First, the awake state y1 and the awake state, that is, all other sleep states: y2, y3 and y4, are distinguished; then the light sleep state y2 and the non-light sleep state, that is, the other two sleep states: y3 and y4, are distinguished; finally, the light sleep state y3 and the rapid eye movement sleep state y4 are distinguished, and a rule classifier needs to be constructed in each of the three classifications.

[0086] It should be noted that this embodiment only takes a limited number of categories as an example. In actual operation, the number of categories can be any number, and when the number of samples is large enough, the categories can be divided in an infinitely detailed manner.

[0087] In this embodiment, the medical data set can be set to have K samples, and for the convenience of description, the above 12 attributes of each sample are set to M attributes: x = {x1, x2, ..., x M}, the 4 categories of the medical data set are represented as N categories: Θ = {y1,…y n ,…,y N}, where N = 4, y n is the nth category, then a sample vector can be expressed as: x = {(x1, x2, ..., x M ),y n}, and can be decomposed into M sample pairs {(x i ,y n ),i=1,2,...,M}.

[0088] Furthermore, it is necessary to set multiple reference values ​​for each attribute. In this embodiment, the number of reference values ​​for each attribute is set to 6 and expressed as: Among them, i represents the i-th attribute, J i =6,i=1,2,...,M.

[0089] Furthermore, a corresponding reference evidence is set for each reference value, the attribute reliability calculated as follows is equivalent to the reliability factor of the reference evidence about the attribute, and a reference evidence weight is set for each reference evidence.

[0090] Among them, the value of the reference evidence weight can be determined subjectively after considering multiple factors, with the purpose of reflecting the relative importance of the reference evidence corresponding to the reference evidence weight compared with other reference evidence.

[0091] Step S102: In each classification for a single category, for each attribute, calculate the attribute reliability of the attribute according to the number of samples in each category when the attribute is used alone for classification, and obtain a reliability ranking according to all the attribute reliabilities.

[0092] In the embodiment of the present application, since the operations in each classification are the same operations, only the first classification is used as an example for description in this embodiment.

[0093] As mentioned above, although each classification is for a single specific category, it is a binary classification operation, so two categories will inevitably be obtained.

[0094] In each classification, the attribute reliability of all the attributes held is first calculated, and the attribute reliability is equivalent to the reliability factor of each reference evidence of the attribute.

[0095] Specifically, for each attribute, it is necessary to determine the number of samples that can be classified independently based on the two categories of this classification, and determine the attribute reliability r of attribute i based on the definition formula of attribute reliability as shown below: i :

[0096]

[0097] Among them, Q i It can be done through the attribute x i Directly determine the number of samples belonging to each category, if That is, the attribute reliability r of attribute i i =1, then attribute x is considered i is the most reliable attribute, through attribute x i The value of can directly classify the most samples.

[0098] After calculating the attribute reliability of all attributes, all attributes are sorted in descending order of attribute reliability.

[0099] Step S103, selecting a predetermined number of the attributes from all the attributes as an initial feature subset according to the reliability ranking, constructing the rule classifier using the initial feature subset, and classifying all the samples using the rule classifier to obtain the classification accuracy of the initial feature subset.

[0100] In the embodiment of the present application, in each classification, based on the attribute reliability ranking determined above, a predetermined number of attributes are selected from all attributes, and the selected attributes are used as the initial feature subset.

[0101] In this embodiment, when each classification is performed, two attributes are first selected from all attributes as an initial feature subset, and a reliability threshold is set for the initial feature subset.

[0102] Furthermore, a rule classifier is constructed based on the attributes in the initial feature subset, as well as the reference value, reference evidence, reference evidence weight and reliability factor of each attribute, and the classification accuracy of the initial feature subset based on the current attribute is calculated.

[0103] Specifically, for each attribute in the initial feature subset in this classification, the similarity between the attribute and each of its reference values ​​can be obtained by using the matching degree calculation formula shown below:

[0104]

[0105]

[0106] α i,j' =0,j'=1,...J i ,j'≠j,j+1

[0107] Among them, α i,j Represents the i-th attribute among the M attributes of the k-th sample The value of the attribute is the jth reference value of The matching degree describes the attribute With reference value The degree of similarity between them.

[0108] Specifically, as shown in the above formula, on each sample, for each reference value in each attribute, the first difference between the reference value and another adjacent reference value is calculated; and the second difference between the attribute and another adjacent reference value is calculated; and the matching degree between the attribute and the reference value is determined according to the ratio of the first difference to the second difference.

[0109] Due to the intervention of the reference value, the attribute x i The relationship with category y is approximately transformed into attribute x i Specific reference value The relationship between the sample pairs (x i ,y) is transformed and uniquely represented as a similarity distribution of category y as shown in Table 1 below:

[0110] Table 1 Sample pairs (x i ,y) in attribute x i Similarity distribution on

[0111]

[0112] Among them, a n,j Is the attribute value With reference value The matching degree can be calculated and belongs to y n The sum of the matching degrees of all sample pairs of the class is taken as the first parameter.

[0113] belongs to category y n The sum of the matching degrees of all reference values ​​of the sample pairs is taken as the second parameter.

[0114] is the reference value In the case of determination, for the attribute value In all categories of sample pairs, the The sum of the matching degrees is taken as the third parameter, and there is

[0115] The first parameter, the second parameter and the third parameter together constitute the similarity distribution shown above.

[0116] According to Table 1, when the category is determined, we can construct the attribute x i The value is a reference value The likelihood of n,j :

[0117]

[0118] Furthermore, the reference evidence for each reference value can be Defined as: when attribute x i The value is taken as reference value When , the sample is judged as category y n The reliability of the reference evidence is Furthermore, the reliability of the above attributes is equivalent to the reliability factor of the reference evidence

[0119] Further, reference evidence Reliability of reference evidence It can be obtained by normalizing the likelihood using the following formula:

[0120]

[0121] Furthermore, according to the above calculation, for each attribute x i , we can get the reference evidence matrix for this attribute as shown in Table 2:

[0122] Table 2 Attribute x i Reference Evidence Matrix

[0123]

[0124]

[0125] Furthermore, according to the distribution of the reference evidence matrix, for the kth sample with M attributes in this embodiment: If its i-th attribute Values ​​in the range If the reference value is within and The corresponding two adjacent reference evidences in the reference evidence matrix and And weighted to calculate the attribute evidence of the attribute.

[0126] Specifically, as shown in the following formula, the matching degree between the attribute and the two reference values ​​and the reference evidence reliability of the two reference evidences are used to complete the reference evidence and Weighting of:

[0127]

[0128] Among them, the result p n,iIndicates: The value of is in the interval Internal, that is, activation and In the case of n The credibility of the evidence, that is, the credibility of the attribute Evidence of the property i credibility of evidence.

[0129] Furthermore, for reference evidence and Weight of evidence and Take the formula shown below and weight it in the same way:

[0130]

[0131] Among them, the result w i Represents: About attributes Evidence of the property i The attribute weight of evidence.

[0132] The attribute reliability of the attribute calculated above is equivalent to the reliability factor of the attribute evidence. Combining the attribute evidence weight, reliability factor and evidence credibility of the attribute evidence, we can get the complete attribute evidence e i .

[0133] Furthermore, according to the above method, all M attribute evidences of M attributes can be obtained: e1, e2, ..., e M .

[0134] Furthermore, based on the fusion rule in the evidence reasoning rule, the M pieces of attribute evidence are fused in combination with the attribute evidence weight and the reliability factor.

[0135] Among them, the two independent attribute evidences e1 and e2 can be fused through the evidence reasoning fusion rule, and the confidence function p of e1 and e2 jointly supporting the proposition θ is obtained: θ,e(2) , the calculation method is shown in the formula:

[0136]

[0137]

[0138] When M pieces of evidence are fused, the confidence function m that jointly supports the proposition θ θ,e(i) It can also be expressed as:

[0139]

[0140]

[0141]

[0142] After fusing M pieces of evidence by combining attribute evidence weights and reliability factors, we get the following fusion function for sample categories, which is used as a rule classifier to classify medical data sets:

[0143] P(x k )={(y n ,p n,e(M) ),n=1,...,N}

[0144] Among them, y n represents the nth category, p n,e(M) Indicates that the sample belongs to category y n The reliability of the fusion result P(x k ), judge the sample x k The category is the maximum confidence p n,e(m) The corresponding category y n .

[0145] Based on the classification results obtained above, the accuracy calculation formula describing the classification performance of this rule classifier is constructed and expressed as:

[0146]

[0147] Among them, totalNum represents the total number of samples involved in the classification operation, rightNum represents the number of samples classified correctly, and the accuracy describes the classification ability of samples based on the existing attributes of the initial feature subset, that is, the classification performance of the rule classifier constructed based on the existing attributes of the initial feature subset.

[0148] Step S104: adjusting the attributes in the initial feature subset according to the reliability ranking and the classification accuracy to obtain a target feature subset.

[0149] In the embodiment of the present application, among the other remaining attributes that are not added to the initial feature subset, an attribute with the highest reliability ranking is selected and compared with the reliability threshold set above.

[0150] Further, when the selected attribute is smaller than the reliability threshold, it is not added to the initial feature subset, and the existing attributes of the initial feature subset are maintained.

[0151] When the attribute is selected to be greater than or equal to the reliability threshold, the attribute is added to the initial feature subset, and in the same manner as above, the rule classifier is constructed using the current initial feature subset, and the classification accuracy of the current initial feature subset is calculated.

[0152] Based on the obtained classification accuracy, if the classification accuracy does not increase, the attributes last added to the initial feature subset are removed, the attributes in the initial feature subset are kept, and the classification is terminated.

[0153] If the classification accuracy rate increases, the attribute that was last added to the initial feature subset is retained, and the attribute with the highest reliability ranking is selected again from the other remaining attributes that have not been added to the initial feature subset. The comparison of the reliability thresholds and the comparison of the classification accuracy rates in step S104 are repeated again to determine whether to retain the attribute that was added again in the initial feature subset.

[0154] Furthermore, if the classification accuracy continues to increase after all attributes are added to the initial feature subset, but no redundant attributes can be added, the classification operation can be terminated.

[0155] Based on the above classification operation, the samples in the medical data set can be divided into two categories: awake state y1, and another category including light sleep state y2, deep sleep state y3, and rapid eye movement sleep state y4.

[0156] Step S105, classify all the categories according to the classification order to obtain the target feature subset and the rule classifier for each classification operation; integrate all the target feature subsets and the rule classifiers to classify new samples in the medical data set.

[0157] In the embodiments of the present application, Figure 3a As shown, the remaining three categories of all four categories are classified according to the same operations as the above steps S103 and S104.

[0158] Furthermore, in the second and third classifications, since all attributes need to be sorted by attribute reliability when the initial feature subset is built and attributes are added in each classification, in addition to classifying samples based on existing attributes, in some cases, if there are newly added attributes, the newly added attributes can also be sorted by attribute reliability, thereby achieving the screening of incremental attributes.

[0159] After completing all three classification operations, we can get Figure 3a The three feature subsets and three rule classifiers shown.

[0160] Furthermore, all feature subsets and all rule classifiers can be integrated to classify new samples in the medical dataset.

[0161] It can be seen that the classification method and related equipment for incremental medical data based on rule classifiers provided by the present application classify the classification order of each category according to actual needs, and only classify one category in each classification. In each classification, all attributes are sorted based on attribute reliability, and only attributes that can improve the accuracy of this classification are used to establish an initial feature subset. By comprehensively considering the attribute reliability sorting and classification accuracy, the attributes in the initial feature subset are adjusted, and the rule classifier is trained using the first-arrived data attributes for rough classification, and then other classifiers are continuously trained according to the attributes obtained subsequently, and more detailed classification is gradually performed.

[0162] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.

[0163] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0164] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, an embodiment of the present application further provides a classification device for incremental medical data based on a rule classifier.

[0165] refer to Figure 2 The incremental medical data classification device based on the rule classifier is connected to a database storing medical data sets, and includes: a preprocessing module 201, a single classification module 202 and a global classification module 203;

[0166] The preprocessing module 201 is configured to preset multiple categories for the medical data set, and determine multiple attributes possessed by each sample in the medical data set and the category to which each sample initially belongs.

[0167] Furthermore, a classification order when performing classification on each of the categories is determined among all the categories.

[0168] The single classification module 202 is configured to, in each classification for a single category, calculate the attribute reliability of each attribute according to the number of samples in each category when the attribute is used alone for classification, and obtain a reliability ranking according to all the attribute reliabilities.

[0169] Furthermore, a predetermined number of the attributes are selected from all the attributes as an initial feature subset according to the reliability ranking, the rule classifier is constructed using the initial feature subset, and all the samples are classified using the rule classifier to obtain the classification accuracy of the initial feature subset.

[0170] Furthermore, the attributes in the initial feature subset are adjusted according to the reliability ranking and the classification accuracy to obtain a target feature subset.

[0171] The global classification module 203 is configured to classify all the categories according to the classification order to obtain the target feature subset and the rule classifier for each classification operation; integrate all the target feature subsets and the rule classifiers to classify new samples in the medical data set.

[0172] For the convenience of description, the above devices are described in terms of functions and are divided into various modules. Of course, when implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0173] The device of the above embodiment is used to implement the corresponding incremental medical data classification method based on the rule classifier in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0174] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, an embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the classification method for incremental medical data based on a rule classifier as described in any of the above embodiments is implemented.

[0175] Figure 4 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0176] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0177] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solution provided in the embodiment of the present application is implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0178] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0179] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0180] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0181] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiment of the present application, and does not necessarily include all the components shown in the figure.

[0182] The device of the above embodiment is used to implement the corresponding incremental medical data classification method based on the rule classifier in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0183] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the classification method of incremental medical data based on a rule classifier as described in any of the above embodiments.

[0184] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0185] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the classification method of incremental medical data based on the rule classifier as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0186] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0187] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present application will be implemented (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0188] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0189] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A classification method for incremental medical data based on a rule classifier, characterized in that: Applicable to databases storing medical data sets, including: Presetting a plurality of categories for the medical data set, and determining a plurality of attributes possessed by each sample in the medical data set and a category to which each sample initially belongs; determining, among all of the categories, a sorting order in which the sorting is performed on each of the categories; In each classification for a single category, for each attribute, the attribute reliability of the attribute is calculated according to the number of samples in each category when the attribute is used alone for classification, and the reliability ranking is obtained according to the reliabilities of all the attributes; Selecting a predetermined number of the attributes from all the attributes as an initial feature subset according to the reliability ranking, constructing the rule classifier using the initial feature subset, and classifying all the samples using the rule classifier to obtain a classification accuracy rate for the initial feature subset; Adjust the attributes in the initial feature subset according to the reliability ranking and the classification accuracy to obtain a target feature subset; According to the classification order, all the categories are classified to obtain the target feature subset and the rule classifier for each classification operation; all the target feature subsets and the rule classifiers are integrated to classify new samples in the medical data set.

2. The method according to claim 1, characterized in that In each classification for a single category, for each attribute, the attribute reliability of the attribute is calculated according to the number of samples in each category when the attribute is used alone for classification, and the reliability ranking is obtained according to the reliabilities of all the attributes, including: In each classification for a single category, a binary classification operation is performed on the category and non-category; For each of the attributes, when all the samples are classified by the attribute alone, determining in each of the categories the number of samples that are determined to belong to the category by the attribute; determining the attribute reliability of the attribute according to the number of determinations of the attribute; According to the attribute reliability of each of the attributes, all the attributes are arranged in order of the attribute reliability from high to low.

3. The method according to claim 2, characterized in that The step of constructing the rule classifier using the initial feature subset includes: For each attribute of the initial feature subset, multiple reference values ​​are matched for the attribute, and corresponding reference evidence is assigned to each reference value, and corresponding reference evidence weight is assigned to each reference evidence; For each attribute of the initial feature subset, construct a similarity distribution between the attribute and the two categories in the binary classification operation according to the attribute and the reference value corresponding to the attribute, and construct a reference evidence matrix for the attribute according to the similarity distribution; For each of the attributes, the reference evidence of the attribute is weighted according to the similarity distribution on the attribute and the reference evidence matrix to obtain the attribute evidence of the attribute, and the reference evidence weight of the reference evidence of the attribute is weighted to obtain the attribute evidence weight of the attribute, and all the attribute evidence is fused using the reliability and the attribute evidence weight to obtain the rule classifier.

4. The method according to any one of claims 1 to 3, characterized in that: The attributes in the initial feature subset are adjusted according to the reliability ranking and the classification accuracy to obtain a target feature subset, including: Among the attributes that are not added to the initial feature subset, select the attribute with the highest reliability ranking and add it to the initial feature subset; Based on the constructed rule classifier, classify all the samples using all the attributes in the initial feature subset; Based on the classification result, calculating the classification accuracy rate of the initial feature subset; According to the numerical change of the classification accuracy, adjusting the attribute in the initial feature subset; The adjusted initial feature subset is used as the target feature subset.

5. The method according to claim 4, characterized in that The step of adjusting the attribute in the initial feature subset according to the numerical change of the classification accuracy rate includes: In response to determining that the classification accuracy rate increases, retaining the attribute added last, and continuing to select the attribute with the highest reliability ranking from the attributes not added to the initial feature subset, and adding it to the initial feature subset; In response to determining that the classification accuracy has not increased, the last added attribute is removed, and the attributes included in the initial feature subset are kept unchanged.

6. The method according to claim 4, characterized in that The selecting the attribute with the highest reliability ranking and adding it to the initial feature subset includes: setting reliability thresholds for all of said attributes; Selecting one of the attributes with the highest reliability ranking, and determining whether the attribute reliability of the attribute is greater than the reliability threshold; In response to determining that the attribute reliability of the attribute is greater than or equal to a reliability threshold, adding the attribute to the initial feature subset; In response to determining that the attribute reliability of the attribute is less than a reliability threshold, the attribute is not added to the initial feature subset, and the attribute included in the initial feature subset remains unchanged.

7. The method according to claim 3, characterized in that The step of forming a similarity distribution between the attribute and the two categories in the binary classification operation according to the attribute and the reference value corresponding to the attribute, and constructing a reference evidence matrix for the attribute according to the similarity distribution includes: On each of the samples, for each of the reference values ​​in each of the attributes, a first difference between the reference value and another adjacent reference value is calculated; and a second difference between the attribute and another adjacent reference value is calculated; and a matching degree between the attribute and the reference value is determined according to a ratio of the first difference to the second difference; Based on the attribute, the similarity distribution is formed using the matching degrees of all the reference values ​​and the two categories in the binary classification operation; For each of the categories, in all the samples, the sum of the matching degrees belonging to the same attribute and the same reference value is calculated as the first parameter of the similarity distribution; For each of the categories, in all the samples, the sum of the first parameters of all the attributes is calculated as the second parameter of the similarity distribution; Using the first parameter, the second parameter and the matching degree to form the similarity distribution; In the case where the category is determined, for each reference value matched by each attribute, constructing the likelihood of taking the attribute as the reference value according to the ratio of the first parameter to the second parameter in the similarity distribution; By respectively normalizing the likelihood of each category, the reliability of the reference evidence when each of the reference evidence judges the sample is obtained; When the attribute is determined, the reference evidence matrix is ​​constructed according to the reference evidence reliability and the category of each reference evidence of the attribute.

8. The method according to claim 7, characterized in that The weighting of the reference evidence of the attribute to obtain the attribute evidence of the attribute, and weighting of the reference evidence weight of the reference evidence of the attribute to obtain the attribute evidence weight of the attribute, include: Based on the distribution of the reference evidence matrix, in response to determining that the attribute value is between two adjacent reference values, obtaining the evidence credibility of the attribute according to the matching degree of the two reference values ​​and the reference evidence credibility; obtaining the attribute evidence weight of the attribute according to the matching degree of the two reference values ​​and the reference evidence weight; The attribute evidence weight, the evidence credibility and the reliability are combined to obtain the attribute evidence of the attribute.

9. The method according to claim 8, characterized in that The step of fusing all the attribute evidence using the reliability and the attribute evidence weight to obtain the rule classifier includes: For each of the attribute evidence, according to the fusion rule in the evidence reasoning rule, all the attribute evidence of the sample is fused using the reliability factor and the attribute evidence weight to obtain the discrimination reliability of the attribute evidence in judging the sample as each of the categories; The rule classifier is constructed using each of the categories and the discrimination confidence of the category.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable by the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.

Citation Information

Patent Citations

  • Rule classifier-based medical data classification method and related equipment

    CN114239742A

  • Identification and management system for log entries

    US20180357299A1

  • Method, Apparatus, and Device for Classifying LiDAR Point Cloud Data, and Storage Medium

    US20210224613A1

  • Automatic detection and association of new attributes with entities in knowledge bases

    US20210279606A1