Intelligent management method for pulmonary interstitial fibrosis diagnosis data

Abstract

The invention relates to the technical field of electrical digital data processing, in particular to a pulmonary interstitial fibrosis diagnostic data intelligent management method, which comprises the following steps: acquiring pulmonary interstitial fibrosis diagnostic data; according to the size condition of each item of data of all the data points, the possibility that the diagnosis data represented by each data point corresponds to abnormal diagnosis data is obtained, and a neighborhood minimum sample number threshold value is obtained in a self-adaptive mode; judging whether each data point is a core object or not according to a neighborhood minimum sample number threshold value of each data point and the number of data points contained in a neighborhood; and completing a DBScan clustering algorithm according to the core object to obtain a clustering result, and judging whether the data points in each cluster are abnormal data or not according to the clustering result. According to the method, the abnormal degree of the pulmonary interstitial fibrosis data is calculated, the neighborhood minimum sample number threshold value of each data point is obtained in a self-adaptive mode, clustering operation is carried out, abnormal data are clustered into one class more accurately, and the intelligence of pulmonary interstitial fibrosis data management is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular to an intelligent management method for pulmonary interstitial fibrosis diagnosis data. Background Art

[0002] Pulmonary fibrosis is a chronic progressive lung disease characterized by progressive pulmonary fibrosis and lung function deterioration. The diagnosis and treatment of pulmonary fibrosis rely on a large amount of medical imaging data, clinical data and biomarker data, and the management and analysis of these data are crucial for the diagnosis, treatment and research of the disease. Therefore, an intelligent management method for pulmonary fibrosis diagnostic data is needed. The existing technology uses DBScan clustering to manage pulmonary fibrosis diagnostic data. However, in the process of selecting the minimum number of samples in the neighborhood, incorrect clustering occurs due to the selection of too large or too small number thresholds. Therefore, by optimizing the DBScan clustering algorithm, the misjudgment in the clustering process is reduced, and the intelligent management of diagnostic data is achieved. Summary of the invention

[0003] The present invention provides an intelligent management method for pulmonary interstitial fibrosis diagnostic data to solve the problem that the existing management of pulmonary interstitial fibrosis diagnostic data causes incorrect clustering due to the selection of too large or too small threshold of the minimum number of samples in the neighborhood, and the data management fails to achieve intelligentization. The technical solution adopted is as follows:

[0004] An embodiment of the present invention provides a method for intelligent management of pulmonary fibrosis diagnosis data, the method comprising the following steps:

[0005] Collect respiratory data of pulmonary interstitial fibrosis from all patients, the respiratory data includes several items of data, and all items of data for each patient constitute a data point; obtain a neighborhood size parameter according to the range of each item of data of all data points; obtain the number of data points in the neighborhood of each data point according to the neighborhood size parameter; obtain an adjustment factor for each item of data according to all data points; obtain a baseline value for each item of data according to the adjustment factor for each item of data; obtain the abnormal possibility of each item of data at each data point according to the baseline value of each item of data; obtain a minimum sample number threshold corresponding to the data point according to the abnormal possibility of the data point, obtain a core object according to the minimum sample number threshold corresponding to the data point and the number of data points in the neighborhood of the data point; complete a DBScan clustering algorithm according to the core object, obtain a clustering result, and determine whether the data point in each cluster cluster is abnormal data according to the clustering result.

[0006] Furthermore, the step of obtaining the neighborhood size parameter according to the range of each data item of all data points includes the following specific steps:

[0007] Get the range of each data point of all data points, and record the range of the j-th data as R j, obtain the neighborhood size parameter based on the range of all item data and the number of collected diagnostic data:

[0008]

[0009] Among them, ∈ represents the neighborhood size parameter of the DBScan algorithm, n represents the total number of data points, and N represents the number of collected data items. The number of data points in the neighborhood of each data point is obtained according to the neighborhood size parameter.

[0010] Furthermore, the step of obtaining the number of data points in the neighborhood of each data point according to the neighborhood size parameter includes the following specific steps:

[0011] For the i-th data point, obtain all data points whose Euclidean distance to it is less than ∈ as the data points in the neighborhood of the i-th data point ∈, and count the number of data points in the neighborhood of the i-th data point ∈.

[0012] Furthermore, the step of obtaining the adjustment factor of each item of data according to each item of data of all data points includes the following specific steps:

[0013] Get the average value of each data item of all data points, and record the average value of the j-th data as Get the abnormal index of the j-th data of the i-th data point based on the difference between the j-th data of the i-th data point and the average value of the data:

[0014]

[0015] P j (i) = 0 P j (i)<0

[0016] Among them, P j (i) represents the abnormal index of the jth data point of the i-th data point, D j (i) represents the jth data item of the i-th data point;

[0017] The adjustment factor of each data item is obtained according to the abnormal index of each data item of all data points.

[0018] Furthermore, the step of obtaining the adjustment factor of each data item according to the abnormal index of each data item of all data points includes the following specific steps:

[0019] Preset an abnormal threshold T1, and select the data points with an abnormal index greater than T1 from all data points corresponding to the j-th data as the abnormal data of the j-th data, and record the number of abnormal data in the j-th data as C j , get the adjustment factor based on the ratio of abnormal data to all data:

[0020]

[0021] Among them, A j Represents the adjustment factor of each data item, and n represents the total number of data points.

[0022] Furthermore, the base value of each data item is obtained according to the adjustment factor of each data item, including the specific formula as follows:

[0023]

[0024] Among them, B j represents the reference value of the j-th data, A j represents the adjustment factor for the j-th data, Represents the average value of the j-th data.

[0025] Furthermore, the abnormal possibility of each data item at each data point is obtained according to the reference value of each data item, and the specific formula included is as follows:

[0026]

[0027] Q j (i)=0 Q j (i)<0

[0028] Among them, Q j (i) represents the abnormality of the jth data point in the i-th data point, B j represents the reference value of the j-th data, D j (i) represents the jth data item of the i-th data point, σ j Represents the standard deviation of the j-th data; the abnormal possibility of each data point is obtained according to the abnormal degree of each data point.

[0029] Furthermore, the abnormal possibility of each data point is obtained according to the abnormal degree of each data item of each data point, and the specific formula included is as follows:

[0030]

[0031] Among them, S i represents the abnormal possibility of the ith data point, and N represents the number of data items collected for each data point.

[0032] Furthermore, the minimum sample number threshold corresponding to the data point is obtained according to the abnormal possibility of the data point, and the core object is obtained according to the minimum sample number threshold corresponding to the data point and the number of data points in the neighborhood of the data point, including the specific formula as follows:

[0033]

[0034] Among them, Ui represents the minimum sample number threshold corresponding to the i-th data point, N represents the number of data items collected for each data point, S i Indicates the abnormal possibility of the i-th data point;

[0035] When the number of data points in the neighborhood of the ith data point is greater than the minimum sample number threshold corresponding to the ith data point, the ith data point is recorded as a core object.

[0036] Furthermore, judging whether the data point in each cluster is abnormal data according to the clustering result includes the following specific steps:

[0037] The average abnormal probability of all data points in each cluster is obtained, all clusters are sorted according to the average abnormal probability, and the respiratory data in the clusters whose average abnormal probability is greater than the preset threshold T0 is recorded as abnormal data according to the cluster sorting result.

[0038] The beneficial effect of the technical solution of the present invention is: in order to address the problem that the conventional DBScan algorithm is difficult to cluster abnormal data into one category or mistakenly clusters abnormal data into normal data during the clustering process of density-uneven pulmonary interstitial fibrosis diagnostic data, the present invention calculates the abnormal possibility of each data point and adaptively obtains the minimum neighborhood sample number threshold of each data point according to the abnormal possibility for clustering, so as to achieve the beneficial effect of more accurately clustering abnormal data into one category without affecting the clustering of normal data. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 The present invention is a flowchart of the steps of an intelligent management method for pulmonary interstitial fibrosis diagnostic data. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] The specific scheme of the intelligent management method for pulmonary interstitial fibrosis diagnosis data provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a flow chart of a method for intelligent management of pulmonary interstitial fibrosis diagnosis data provided by an embodiment of the present invention, the method comprising the following steps:

[0044] Step S001: Collect some pulmonary interstitial fibrosis diagnosis data.

[0045] It should be noted that the diagnostic data of pulmonary interstitial fibrosis is collected. The main manifestations of pulmonary interstitial fibrosis are decreased vital capacity, forced expiratory volume in the first second (FEV1) and peak expiratory flow (PEF), while the residual volume / total lung volume ratio (RV / TLC) increases. Therefore, the condition of pulmonary interstitial fibrosis in patients can be reflected by collecting the decrease of vital capacity, forced expiratory volume in the first second (FEV1) and peak expiratory flow (PEF) and the residual volume / total lung volume ratio (RV / TLC).

[0046] The four required data obtained by performing a vital capacity test using a breathing device include vital capacity, forced expiratory volume in the first second (FEV1) and peak expiratory flow rate (PEF) and other indicators are reduced, while the residual volume / total lung volume ratio (RV / TLC); the collected respiratory diagnostic data is recorded as D, abbreviated as respiratory data D, vital capacity, forced expiratory volume in the first second (FEV1) and peak expiratory flow rate (PEF) and other indicators are reduced, while the residual volume / total lung volume ratio (RV / TLC) and other four data are recorded as D1, D2, D3, D4 respectively. When these data are abnormal, for example, the abnormal data deviates from the normal data, it means that the patient's lung function is abnormal and there may be pulmonary interstitial fibrosis. The number of patients is recorded as n, and the number of data items collected from each patient is recorded as N.

[0047] Obtain the data of each patient, and use the vital capacity, forced expiratory volume in the first second (FEV1) and peak expiratory flow (PEF) as four coordinate axes, and construct a four-dimensional rectangular coordinate system. All the data of each patient correspond to a data point in the four-dimensional rectangular coordinate system, that is, each patient corresponds to a data point, so each data point contains each data of each patient. Obtain the data points corresponding to all patients, and record the j-th data of the i-th data point as D j (i).

[0048] So far, some respiratory data of pulmonary interstitial fibrosis have been collected.

[0049] Step S002: preset neighborhood parameters to obtain the number of data points contained in the neighborhood of each data point.

[0050] It should be noted that the DBScan algorithm is used to cluster the diagnostic data of pulmonary interstitial fibrosis, and the neighborhood parameters are set by fully considering the data dimension and the fluctuation range of each dimensional data; the square root of the sum of the squares of the ranges of all items of data can represent the maximum Euclidean distance that all range values ​​can form. The maximum Euclidean distance is divided by the number of collected diagnostic data to obtain the average Euclidean distance between two data points. Since the average Euclidean distance obtained by the range takes all unread data into consideration, it is easily affected by the curse of dimensionality, so it is multiplied by a data dimension N to reduce the impact of the curse of dimensionality.

[0051] Specifically, obtain the range of each data point of all data points, and record the range of the j-th data of all patients as R j , obtain the neighborhood size parameter based on the range of all item data and the number of collected diagnostic data:

[0052]

[0053] Among them, ∈ represents the neighborhood size parameter of the DBScan algorithm, n represents the number of collected data points, and N represents the number of collected data items.

[0054] Furthermore, according to the preset neighborhood size parameter ∈, for the ith data point, all data points whose Euclidean distance to it is less than ∈ are obtained as the data points in the neighborhood of the ith data point, and the number of data points in its neighborhood is recorded as F i .

[0055] At this point, the neighborhood size parameter of the DBScan algorithm is obtained, and the number of data points in the neighborhood of each data point ∈ is obtained.

[0056] Step S003: Obtain the abnormality level of each data item of all data points according to the size of each data item of all data points.

[0057] It should be noted that for the diagnosis data of pulmonary interstitial fibrosis, the abnormal possibility of each data item is obtained according to its size level in all data points, and the average value of each data item is used as the average level of the data item to determine whether each data item of each data point is too high or too low; in the data of pulmonary interstitial fibrosis, when there is a difference between the level of data and most of the same data items, it is believed that the patient corresponding to the data may have pulmonary interstitial fibrosis, so only the abnormal degree of the smaller data value needs to be obtained. Compare these three data items with their corresponding average values ​​to quantify the abnormal degree of data less than the average value. Only the abnormal degree corresponding to the larger data value needs to be obtained; in the diagnosis data of pulmonary interstitial fibrosis, a certain proportion of abnormal data will be included. The existence of abnormal data will cause the data to be different from the overall data level. Therefore, the calculation of the abnormal index based only on the average value will cause the obtained abnormal index to be biased because the average value is affected by the abnormal data. Therefore, it is necessary to adjust the average value according to the proportion of data with a large degree of abnormality calculated for the first time; according to the number of abnormal data, the more abnormal data there are, the more the average value calculation of the data will be affected by the abnormal data, and the corresponding adjustment factor will be larger. Different data items are also affected differently by abnormal data. The average value of respiratory data in pulmonary interstitial fibrosis is affected by abnormal data and is smaller, so it should be increased by the adjustment factor. Therefore, the initial average value can be adjusted according to the adjustment factor. Pulmonary interstitial fibrosis will have different effects on the patient's body, such as reducing indicators such as forced vital capacity (FVC) or increasing other indicator data. Different adjustment methods should be used for abnormal data in different indicator data; because different data items are affected differently by abnormal data, the adjustment strategies adopted are also different. After adjustment, the degree of abnormality of each data is obtained according to the degree of deviation of each data from the baseline value: each data value is subtracted from the baseline value of the corresponding data item, and the difference is compared with the standard deviation of the data item. If the difference is larger than the standard deviation, it is considered that the data item is outside the normal data range and the degree of abnormality of the data item is large. In the diagnostic data of pulmonary interstitial fibrosis, the smaller the data items are, the more likely they are abnormal data. Therefore, only the abnormality of a smaller part of these data items is considered. For those that are greater than the baseline value, they are not considered to be abnormal. Similarly, for some data items, only the abnormality of the data with larger values ​​is considered.

[0058] Specifically, get the average value of the jth data point in all data points, denoted as The standard deviation is denoted as σ j , according to the difference between the j-th data item of the i-th data point and the average value of the data item, the abnormal index of the j-th data item of the i-th data point is obtained:

[0059]

[0060] P j (i) = 0 Pj (i)<0

[0061] Among them, P j (i) represents the abnormal index of the jth data point of the i-th data point, D j (i) represents the jth data item of the i-th data point.

[0062] Furthermore, an abnormal threshold T1 is preset, and the empirical value is T1=0.3. This embodiment does not limit the abnormal threshold, and the implementer can set the abnormal threshold according to the actual implementation situation. For the abnormal index of the j-th data of all data points, all data with abnormal index greater than the abnormal threshold are obtained and recorded as abnormal data, and the number of abnormal data in the j-th data is recorded as C j , get the adjustment factor based on the ratio of abnormal data to all data:

[0063]

[0064] Among them, A j represents the adjustment factor of the j-th data, and n represents the total number of data points.

[0065] Furthermore, the average value of each data item is adjusted according to the adjustment factors of different data items to obtain the benchmark value of each data item:

[0066]

[0067] Among them, B j A represents the base value of the j-th data average after adjustment factor adjustment, j represents the adjustment factor for the j-th data, Represents the average value of the j-th data.

[0068] Furthermore, the abnormality degree of each data item at each data point is obtained according to the baseline value of each data item and the data value corresponding to each data item:

[0069]

[0070] Q j (i)=0 Q j (i)<0

[0071] Among them, Q j (i) represents the abnormality of the jth data point of the i-th data point, B j represents the reference value of the j-th data, D j (i) represents the jth data item of the i-th data point, σ j Represents the standard deviation of the j-th data.

[0072] At this point, the abnormality level of each data item of all data points is obtained.

[0073] Step S004: Obtain the abnormal possibility of each data point according to the abnormal degree of all items of data of the data point.

[0074] It should be noted that in order to adaptively obtain the minimum number of samples in the neighborhood according to the degree of abnormality of each data point, it is necessary to comprehensively consider the degree of abnormality of each data point, so it is necessary to obtain the abnormal possibility of the data point according to the degree of abnormality of all items of data of each data point; add the degree of abnormality of all items of data of each data point and multiply them together to obtain the cumulative value of the degree of abnormality, and then raise the cumulative value of the degree of abnormality to the power of N to obtain the abnormal possibility value, which can better consider the contribution of each degree of abnormality to the abnormal possibility; according to the abnormal possibility of each data point, the minimum sample threshold of each data point is adaptively obtained. The data point with a greater abnormal possibility should have a smaller minimum sample threshold to ensure that more abnormal data points can be identified as a cluster.

[0075] Specifically, the abnormal possibility of each data point is obtained according to the abnormal degree of all items of data of each data point:

[0076]

[0077] Among them, S i represents the abnormal possibility of the i-th data point, N represents the number of data items collected for each data point, and Q j (i) represents the jth data item of the i-th data point.

[0078] Furthermore, according to the abnormal possibility of each data point, the corresponding minimum sample number threshold is obtained:

[0079] U i =N×S i

[0080] Among them, U i represents the minimum sample number threshold corresponding to the i-th data point, N represents the number of data items collected for each data point, that is, there are N dimensions of data (each data item is regarded as one dimension), S i represents the abnormal possibility of the i-th data point.

[0081] It should be noted that in the process of neighborhood size calculation, the curse of dimensionality is reduced through the data dimension, and the number of data points in the neighborhood of each data point is also made close to the data dimension N. Therefore, the data dimension is used as the benchmark value here to adaptively calculate the minimum number of samples threshold in the neighborhood of each data point; when the abnormal probability of each data point is greater than 1, it is considered that the abnormal probability of the data point is large, so its minimum number of samples threshold is greater than N, and it is adjusted to N.

[0082] At this point, the minimum sample number threshold for all data points is obtained.

[0083] Step S005: adaptively obtain the minimum number of neighborhood samples threshold according to the possibility that each data point corresponds to abnormal data, determine whether each data point is a core object according to the minimum number of neighborhood samples threshold of each data point and the number of data points contained in the neighborhood, complete the DBScan clustering algorithm according to the core object, obtain the clustering result, and determine whether the data point in each cluster cluster is abnormal data according to the clustering result.

[0084] Specifically, the number of data points in the neighborhood of each data point is compared with the adaptive minimum sample number threshold, and the number of data points in the neighborhood of ∈ that is greater than the adaptive minimum sample number threshold is recorded as a core object, and all core objects are obtained. It is a known technology to perform clustering of all core objects using the DBScan algorithm. Clustering is performed based on all core objects using the DBScan algorithm to obtain the final division result, and the average abnormal probability of all data points in each cluster is obtained based on the division result. All clusters are sorted based on the average abnormal probability, and the respiratory data in the clusters with an average abnormal probability greater than the threshold T0=0.6 is recorded as abnormal data based on the cluster sorting result.

[0085] At this point, abnormal data in the respiratory data is obtained.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent management of pulmonary fibrosis diagnostic data, characterized in that: The method comprises the following steps: Collect respiratory data of pulmonary interstitial fibrosis from all patients, the respiratory data includes several items of data, and all items of data for each patient constitute a data point; obtain a neighborhood size parameter according to the range of each item of data of all data points; obtain the number of data points in the neighborhood of each data point according to the neighborhood size parameter; obtain an adjustment factor for each item of data according to all data points; obtain a baseline value for each item of data according to the adjustment factor for each item of data; obtain the abnormal possibility of each item of data at each data point according to the baseline value of each item of data; obtain a minimum sample number threshold corresponding to the data point according to the abnormal possibility of the data point, obtain a core object according to the minimum sample number threshold corresponding to the data point and the number of data points in the neighborhood of the data point; complete a DBScan clustering algorithm according to the core object, obtain a clustering result, and determine whether the data point in each cluster cluster is abnormal data according to the clustering result.

2. According to claim 1, a method for intelligent management of pulmonary fibrosis diagnostic data, characterized in that: The specific steps of obtaining the neighborhood size parameter according to the range of each data item of all data points are as follows: Get the range of each data point of all data points, and record the range of the j-th data as R j , obtain the neighborhood size parameter based on the range of all item data and the number of collected diagnostic data: Among them, ∈ represents the neighborhood size parameter of the DBScan algorithm, n represents the total number of data points, and N represents the number of collected data items. The number of data points in the neighborhood of each data point is obtained according to the neighborhood size parameter.

3. According to claim 2, a method for intelligent management of pulmonary fibrosis diagnostic data, characterized in that: The specific steps of obtaining the number of data points in the neighborhood of each data point according to the neighborhood size parameter are as follows: For the i-th data point, obtain all data points whose Euclidean distance to it is less than ∈ as the data points in the neighborhood of the i-th data point ∈, and count the number of data points in the neighborhood of the i-th data point ∈.

4. According to claim 1, a method for intelligent management of pulmonary fibrosis diagnostic data, characterized in that: The specific steps of obtaining the adjustment factor of each data item according to each data item of all data points are as follows: Get the average value of each data item of all data points, and record the average value of the j-th data as Get the abnormal index of the j-th data of the i-th data point based on the difference between the j-th data of the i-th data point and the average value of the data: P j (i)=0 P j (i)<0 Among them, P j (i) represents the abnormal index of the jth data point of the i-th data point, D j (i) represents the jth data item of the i-th data point; The adjustment factor of each data item is obtained according to the abnormal index of each data item of all data points.

5. According to claim 4, a method for intelligent management of pulmonary fibrosis diagnosis data is characterized in that: The specific steps of obtaining the adjustment factor of each data item according to the abnormal index of each data item of all data points are as follows: Preset an abnormal threshold T1, and select the data points with an abnormal index greater than T1 from all data points corresponding to the j-th data as the abnormal data of the j-th data, and record the number of abnormal data in the j-th data as C j , get the adjustment factor based on the ratio of abnormal data to all data: Among them, A j Represents the adjustment factor of each data item, and n represents the total number of data points.

6. According to claim 1, a method for intelligent management of pulmonary fibrosis diagnostic data, characterized in that: The base value of each data item is obtained according to the adjustment factor of each data item, including the specific formula as follows: Among them, B j represents the reference value of the j-th data, A j represents the adjustment factor for the j-th data, Represents the average value of the j-th data.

7. The method for intelligent management of pulmonary fibrosis diagnostic data according to claim 1, characterized in that: The specific formula for obtaining the abnormal probability of each data item at each data point according to the benchmark value of each data item is as follows: Q j (i)=0 Q j (i)<0 Among them, Q j (i) represents the abnormality of the jth data point in the i-th data point, B j represents the reference value of the j-th data, D j (i) represents the jth data item of the i-th data point, σ j Represents the standard deviation of the j-th data; the abnormal possibility of each data point is obtained according to the abnormal degree of each data point.

8. The method for intelligent management of pulmonary fibrosis diagnosis data according to claim 7, characterized in that: The specific formula for obtaining the abnormal possibility of each data point according to the abnormal degree of each data point is as follows: Among them, S i represents the abnormal possibility of the ith data point, and N represents the number of data items collected for each data point.

9. The method for intelligent management of pulmonary fibrosis diagnostic data according to claim 1, characterized in that: The method of obtaining the minimum sample number threshold corresponding to the data point according to the abnormal possibility of the data point, and obtaining the core object according to the minimum sample number threshold corresponding to the data point and the number of data points in the neighborhood of the data point, includes the following specific formula: Among them, U i represents the minimum sample number threshold corresponding to the i-th data point, N represents the number of data items collected for each data point, S i Indicates the abnormal possibility of the i-th data point; When the number of data points in the neighborhood of the ith data point is greater than the minimum sample number threshold corresponding to the ith data point, the ith data point is recorded as a core object.

10. The method for intelligent management of pulmonary fibrosis diagnostic data according to claim 1, characterized in that: The specific steps of judging whether the data points in each cluster are abnormal data according to the clustering results are as follows: The average abnormal probability of all data points in each cluster is obtained, all clusters are sorted according to the average abnormal probability, and the respiratory data in the clusters whose average abnormal probability is greater than the preset threshold T0 is recorded as abnormal data according to the cluster sorting result.

Images