Cardiovascular intervention postoperative care risk prediction system based on physiological feedback

Abstract

The invention relates to the technical field of postoperative risk prediction, in particular to a cardiovascular intervention postoperative nursing risk prediction system based on physiological feedback. The method comprises the following steps: firstly, acquiring physiological data of historical patients and risk values of postoperative care, and grouping; the aggregation of the risk values of the patients is further analyzed, and the initial typical degree is obtained in combination with the risk values; further analyzing the reference weight of other groups for each physiological data of the target group, analyzing the correlation between the physiological data of other groups and the maximum initial typical degree in other groups, and obtaining the prediction importance of each physiological data of the target group; the final typical degree is further obtained; and finally, based on the final canonical degree fusion risk value of the patients in each group, obtaining a risk optimization value and constructing a prediction decision tree, thereby improving the representativeness and individualized prediction capability of the decision tree.

Description

Technical Field

[0001] This invention relates to the field of postoperative risk prediction technology, specifically to a cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback. Background Technology

[0002] Cardiovascular interventional procedures can lead to various complications, such as bleeding, infection, thrombosis, and arrhythmia. Therefore, predicting the postoperative care risk of cardiovascular interventional procedures is crucial. Current methods for constructing decision trees for predicting postoperative care risk of cardiovascular interventional procedures based on historical patient physiological feedback data typically use the average risk value of multiple patients corresponding to each set of physiological feedback data as the unique postoperative care risk value for that set of data.

[0003] However, different patients have varying physiological systems that adapt to cardiovascular interventional surgery. Some patients have robust cardiovascular systems and recover quickly after surgery; while others may have weaker cardiac function or other systemic abnormalities, leading to slower recovery and even a higher risk of complications. Simply using the average risk value of multiple patients corresponding to each group of physiological feedback data as the sole risk value for that group cannot accurately represent typical postoperative care risks. This results in a lack of representativeness in the construction of the decision tree, consequently affecting the accuracy of risk prediction and the effectiveness of individualized treatment. Summary of the Invention

[0004] To address the problem in existing technologies that use the average risk value of multiple patients corresponding to each set of physiological feedback data as the sole risk value for that set of data, which fails to accurately represent typical postoperative care risks and leads to a lack of representativeness in the construction of decision trees, the present invention aims to provide a cardiovascular interventional postoperative care risk prediction system based on physiological feedback. The specific technical solution adopted is as follows:

[0005] Data acquisition module: Acquires physiological data and postoperative care risk values ​​of patients undergoing historical cardiovascular interventional surgery; groups patients with identical physiological data into groups, and selects each group as the target group;

[0006] Initial analysis module: Based on the clustering of each patient's risk value within its respective group, and in conjunction with the risk value, an initial typicality is obtained; based on the differences in the same physiological data among different groups, and in conjunction with the initial typicality, reference weights for various physiological data of other groups relative to the target group are obtained; for each physiological data, the correlation between the physiological data of other groups and the maximum initial typicality within other groups is analyzed based on the reference weights to obtain the predictive importance of various physiological data of the target group.

[0007] Prediction optimization module: Based on the similarity of changes in various physiological data and risk values ​​of each patient in the target group to all patients in other groups, and combined with the prediction importance and the initial typicality, the final typicality is obtained; based on the final typicality of patients in each group, the risk value is fused to obtain the risk optimization value and construct a prediction decision tree.

[0008] Furthermore, the method for obtaining the initial typicality includes:

[0009] The clustering coefficient of each risk value in its respective group is obtained based on the LOF algorithm; the risk value and the corresponding clustering coefficient are fused to obtain the initial typicality of the risk value for each patient.

[0010] Furthermore, the method for obtaining the clustering coefficient includes:

[0011] Each risk value in a group is used as its own horizontal and vertical coordinates and mapped onto a two-dimensional plane to obtain the LOF value of each risk value. The LOF value is then negatively correlated, normalized, and used as the clustering coefficient of each risk value.

[0012] Furthermore, the method for obtaining the reference weights includes:

[0013] For each physiological data item, the reference weights of the other groups relative to the target group are obtained based on the absolute value of the difference between the physiological data of the other groups and the target group, combined with the largest initial typicality within the other groups.

[0014] Furthermore, the method for obtaining the predicted importance includes:

[0015] Select each of the aforementioned physiological data as the target physiological data;

[0016] The reference weights of each other group relative to the target physiological data of the target group and the target physiological data of each other group are fused to construct a first sequence with a fusion value; the reference weights of each other group relative to the target physiological data of the target group and the maximum initial typicality within each other group are fused to construct a second sequence with a fusion value.

[0017] Analyze the correlation between the first sequence and the second sequence to obtain the predictive importance of the target physiological data for the target group.

[0018] Furthermore, the correlation between the first sequence and the second sequence was analyzed using the Pearson correlation coefficient.

[0019] Furthermore, the method for obtaining the final typicality includes:

[0020] A relative typicality factor is obtained based on the similarity of the differences in various physiological data between each patient in the target group and each patient in other groups and the differences in risk values.

[0021] By fusing the predictive importance of each physiological data of each patient in the target group with all the corresponding relative typical factors, a typicality correction factor is obtained; by fusing the typicality correction factor and the initial typicality, the final typicality of the risk value of each patient in the target group is obtained.

[0022] Furthermore, the method for obtaining the risk optimization value includes:

[0023] Using the proportion of each patient's final typicality to the total final typicality of all patients in the same group as the weight, the risk values ​​of each patient in each group are weighted and summed to obtain the risk optimization value corresponding to each physiological data feedback of each group.

[0024] Furthermore, the prediction decision tree is a CART regression tree.

[0025] Furthermore, the risk value for the postoperative care is the DAPT score.

[0026] The present invention has the following beneficial effects:

[0027] This invention first acquires and groups historical patients' physiological data and postoperative care risk values ​​to provide a basis for subsequent analysis. It then obtains initial typicality to preliminarily quantify the representativeness of patients' postoperative care risk values. Further analysis of the reference weights of other groups for various physiological data of the target group helps to introduce highly relevant external data in subsequent analyses, improving the model's predictive accuracy by leveraging other groups to analyze the target group. Based on the reference weights, the correlation between the physiological data of other groups and the maximum initial typicality within other groups is analyzed to obtain the predictive importance of various physiological data of the target group, quantifying the contribution of physiological data and facilitating subsequent optimization of initial typicality. Furthermore, based on the similarity of changes in various physiological data and risk values ​​of each patient in the target group with all patients in other groups, the relative typicality of each patient among all patients in different groups is determined. Combining predictive importance and initial typicality, the final typicality is obtained, characterizing the representativeness of each patient's risk value within the group, providing a basis for final optimization of the decision tree. Finally, based on the final typicality of patients in each group, the risk value is fused to obtain the optimized risk value and construct a predictive decision tree, improving the representativeness and individualized predictive ability of the decision tree. Attached Figure Description

[0028] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a system block diagram of a cardiovascular interventional postoperative care risk prediction system based on physiological feedback, provided as an embodiment of the present invention.

[0030] Figure 2 This is a flowchart of a method for obtaining predicted importance provided in one embodiment of the present invention. Detailed Implementation

[0031] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a cardiovascular interventional postoperative care risk prediction system based on physiological feedback proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0033] The following description, in conjunction with the accompanying drawings, details the specific scheme of the cardiovascular interventional postoperative care risk prediction system provided by this invention.

[0034] Please see Figure 1 The diagram illustrates a system block diagram of a cardiovascular interventional postoperative care risk prediction system based on physiological feedback, according to an embodiment of the present invention. The system includes: a data acquisition module 101, an initial analysis module 102, and a prediction optimization module 103.

[0035] Data acquisition module 101: Acquires physiological data and postoperative care risk values ​​of patients undergoing historical cardiovascular interventional surgery; groups patients with the same physiological data into groups, and selects each group as the target group.

[0036] In one embodiment of the present invention, the physiological data reported by patients undergoing historical cardiovascular interventional surgery includes at least heart rate, blood pressure, blood oxygen saturation, respiratory rate, body temperature, and blood glucose levels. Each data point undergoes preprocessing such as linear normalization in its respective dimension. Considering that even with similar physiological data, individual recovery abilities and complication risks still differ, and that a patient's underlying health condition affects postoperative recovery—for example, two patients may have similar heart rate and blood pressure, but one patient may have a history of underlying conditions such as diabetes, hypertension, or renal insufficiency, which increase the risk of postoperative complications—the risk values ​​for postoperative care of patients undergoing historical cardiovascular interventional surgery are also obtained.

[0037] In one embodiment of the present invention, the risk value for postoperative care is the DAPT score. The scoring factors involved in the DAPT score may include: age {age < 60 years: -2 points; 60 ≤ age < 70 years: 0 points; age ≥ 70 years: +2 points}; smoking history {no smoking history: 0 points; smoking history: +1 point}; diabetes {no diabetes: 0 points; diabetes: +1 point}; myocardial infarction at the time of initial visit {no myocardial infarction: 0 points; myocardial infarction: +1 point}; paclitaxel-eluting stent {non-paclitaxel-eluting stent: 0 points; paclitaxel-eluting stent: +1 point}; coronary artery stent diameter < 3 mm {stent diameter ≥ 3 mm: 0 points; stent diameter < 3 mm: +1 point}. The scoring factors and scoring methods can be adjusted by the implementer and will not be elaborated further.

[0038] To analyze the postoperative risk of patients with the same physiological data and thus optimize the decision tree for more accurate risk prediction, patients with the same physiological data were grouped together, and each group was selected as the target group. Each group was analyzed in turn.

[0039] This method involves data processing in the initial analysis module 102 and the prediction optimization module 103 after data transmission, and execution in the corresponding processing chip.

[0040] Initial Analysis Module 102: Based on the clustering of each patient's risk value within its respective group, and combined with the risk value, initial typicality is obtained; based on the differences in the same physiological data among different groups, and combined with the initial typicality, reference weights of other groups for various physiological data of the target group are obtained; for each physiological data, the correlation between the physiological data of other groups and the maximum initial typicality within other groups is analyzed based on the reference weights to obtain the predictive importance of various physiological data of the target group.

[0041] Considering that patients with higher clustering of postoperative care risk values ​​can better represent the postoperative care risk values ​​of this group of physiological feedback data, the initial typicality is obtained by combining the risk values ​​with the clustering of each patient's risk values ​​in their respective groups, thus initially quantifying the typical representativeness of the patients' postoperative care risk values.

[0042] Preferably, in one embodiment of the present invention, considering that the LOF algorithm can analyze the local density of data points, the clustering coefficient of each risk value in its respective group is obtained based on the LOF algorithm;

[0043] As an example, each risk value in a group is used as its own x and y coordinates (the x and y coordinates are the same), and mapped onto a two-dimensional plane to obtain the LOF value of each risk value;

[0044] Considering that patients with lower LOF values ​​indicate that their postoperative care risk is similar to that of most patients and there are no obvious abnormalities, and the higher the degree of typical representativeness, the LOF values ​​are negatively correlated and normalized, and then used as the clustering coefficient for each risk value, representing the clustering of the patient's risk value in its respective group.

[0045] Patients with high post-cardiovascular surgery risk values ​​typically face a greater risk of post-operative complications such as bleeding, infection, and heart failure. These patients often require more attention and intervention because their physical condition is more vulnerable and the recovery process is riskier. Applying extra, typical care to these patients helps to facilitate timely intervention in the post-operative stage and prevent further deterioration of their condition.

[0046] Furthermore, the risk profiles of these patients help the medical team make more accurate decisions, thereby improving the quality of care and ensuring the patient's postoperative recovery.

[0047] Based on this, the initial typicality of the risk value for each patient is obtained by fusing the risk value and the corresponding clustering coefficient.

[0048] As an example, the LOF value of each patient's hazard value is used as the independent variable. After negative correlation mapping and normalization by the exponential function exp(-x) with the natural constant e as the base, it is used as the clustering coefficient of each patient's hazard value; x is the independent variable.

[0049] The hazard value of each patient is linearly normalized across all hazard values ​​in its group. The product of the normalization result and the corresponding clustering coefficient is used as the initial typicality of the hazard value for each patient.

[0050] It should be noted that the LOF algorithm is an existing technology, and the initial typicality of the risk value of postoperative care for each patient in each group is obtained in the same way, so it will not be described again. In other embodiments of the present invention, the LOF value can also be linearly normalized in the corresponding data dimension first, and then negative correlation mapping and normalization can be performed by subtracting the normalization result from the constant 1.

[0051] Considering that the initial typicality initially characterizes the available weights of risk values ​​within the group, and that the differences in the same physiological data across different groups reflect the similar characteristics of the physiological states of patients in different groups, it is necessary to obtain the reference weights of other groups for each physiological data point of the target group. These reference weights can measure the reference value of data from other groups in the analysis of the target group, facilitating the introduction of highly relevant external data in subsequent analyses. By leveraging other groups to analyze the target group, the predictive accuracy of the model can be improved. Therefore, based on the differences in the same physiological data across different groups, and in conjunction with the initial typicality, the reference weights of other groups for each physiological data point of the target group are obtained.

[0052] Preferably, in one embodiment of the present invention, considering that the maximum initial typicality among all patients in other groups needs to be large so that other groups can serve as the benchmark for data analysis and provide more stable and reliable reference data, the larger the maximum initial typicality in other groups, the greater the reference weight of other groups.

[0053] Furthermore, considering that the smaller the absolute value of the difference between the physiological data of other groups and the target group, the smaller the difference in the same physiological data, the more similar the physiological state, and the greater the reference weight, the difference in the same physiological data is expressed in the form of the absolute value of the difference.

[0054] Based on this, for each physiological data point, the reference weights of other groups relative to the target group are obtained by combining the absolute value of the difference between the physiological data of other groups and the target group with the maximum initial typicality within other groups.

[0055] As an example, for each physiological data point, the absolute value of the difference between the physiological data of other groups and the target group is used as the independent variable. After negative correlation mapping using the function exp(-x), the product of the mapped value and the maximum initial typicality within the other groups is used as the reference weight of the physiological data of other groups relative to the target group.

[0056] It should be noted that "other groups" refers to groups other than the target group. The analysis process for each other group is the same as that for the target group, and will not be repeated here. This gives us the reference weight of each other group for each physiological data point of the target group.

[0057] Considering that different physiological indicators have different impacts on postoperative risk values ​​and thus different importance for risk prediction, it is necessary to analyze the predictive importance. Considering that the correlation between the physiological data of each other group and the maximum initial typicality within each other group represents the association characteristics between physiological data and initial typicality, and the reference weight represents how much reference value other groups have in the target group analysis, for each physiological data, the correlation between the physiological data of other groups and the maximum initial typicality within other groups is analyzed based on the reference weight, to obtain the predictive importance of each physiological data of the target group, quantify the contribution of physiological data, and facilitate subsequent optimization of initial typicality.

[0058] Preferably, in one embodiment of the present invention, please refer to Figure 2 The diagram illustrates a flowchart of a method for obtaining predictive importance according to an embodiment of the present invention, specifically including:

[0059] Step S201: Select each physiological data point as the target physiological data point one by one.

[0060] Select target physiological data one by one to analyze each physiological data point individually.

[0061] Step S202: Fuse the reference weights of each other group to the target physiological data of the target group, and the target physiological data of each other group, to construct a first sequence with the fusion value; fuse the reference weights of each other group to the target physiological data of the target group, and the maximum initial typicality within each other group, to construct a second sequence with the fusion value.

[0062] Considering that the reference weights represent how much reference value other groups have in the analysis of the target group, the reference weights are merged before analyzing the correlation.

[0063] As an example, using reference weights as weights, the data is fused through multiplication. The first sequence is constructed by multiplying the reference weights of each other group relative to the target physiological data of the target group with the product of the target physiological data of each other group; the first sequence is {K}. j,i,g ×S j,g ,j=1,2,3…J};K j,i,g S represents the reference weight of the j-th other group relative to the g-th target physiological data of the i-th target group; j,g This represents the data value of the g-th target physiological data of the patient in the j-th other group; j is the index of the other group excluding the target group, and J is the number of other groups excluding the target group.

[0064] The second sequence is constructed by multiplying the reference weight of each other group relative to the target physiological data of the target group with the product of the maximum initial typicality within each other group. The second sequence is {K}. j,i,g ×Y j,max ,j=1,2,3…J};Y j,max This represents the maximum initial typicality of the risk value of patients in the j-th other group.

[0065] Step S203: Analyze the correlation between the first and second sequences to obtain the predictive importance of the target physiological data of the target group.

[0066] The Pearson correlation coefficient, by quantifying the strength and direction of the relationship between variables, can directly reflect the importance of physiological feedback components in risk prediction. Therefore, in one embodiment of the present invention, the correlation between the first sequence and the second sequence is analyzed using the Pearson correlation coefficient.

[0067] When the Pearson correlation coefficient is close to 1 or -1, it indicates a strong positive or negative correlation between the feedback physiological data and the initial typicality of the risk value, suggesting that this component plays an important role in prediction; when the correlation coefficient is close to 0, it indicates a weak correlation and low importance to prediction.

[0068] Therefore, the absolute value of the Pearson correlation coefficient between the first and second sequences is used as the predictive importance of the target physiological data for the target group, representing the correlation between the physiological data of other groups and the maximum initial typicality within other groups after weighting based on the reference weight.

[0069] The method for obtaining the predicted importance of each physiological data point in each group is the same, and will not be repeated here.

[0070] Prediction optimization module 103: Based on the similarity of changes in various physiological data and risk values ​​of each patient in the target group to all patients in other groups, and combining the prediction importance and initial typicality, the final typicality is obtained; based on the final typicality of patients in each group, the risk value is fused to obtain the risk optimization value and construct the prediction decision tree.

[0071] Based on the importance of physiological data in predicting post-cardiovascular interventional nursing risks, when the relationship between changes in physiological data and nursing risk values ​​is consistent between each patient and patients in other groups, it indicates that the role of the physiological data is relatively stable relative to patients in different groups and is not easily affected by other factors. This consistency helps to determine the relative typicality of the patient among all patients in different groups, that is, whether the patient's risk value can represent the commonalities of this type of patient.

[0072] Therefore, based on the similarity of changes in various physiological data and risk values ​​of each patient in the target group to all patients in other groups, combined with the predictive importance and initial typicality, the final typicality is obtained, which ultimately characterizes the typical representativeness of the risk value of each patient in the group, and provides a basis for the final optimization of the decision tree.

[0073] Preferably, in one embodiment of the present invention, the changes in physiological data and risk values ​​are represented by the differences between data; a relative typicality factor is obtained based on the similarity of the differences in various physiological data and risk values ​​between each patient in the target group and each patient in other groups.

[0074] As an example, to represent the differences between data in terms of absolute values ​​of the difference, select a patient in the target group and a patient in another group, select a physiological data point, and use the linearly normalized result of the absolute difference of this physiological data point between the two patients in the corresponding data dimension as the first difference factor; use the linearly normalized result of the absolute difference of the risk values ​​between the two patients in the corresponding data dimension as the second difference factor.

[0075] The absolute value of the difference between the first and second differential factors is used as the independent variable. After negative correlation mapping using the function exp(-x), the mapped value is used as the relative canonical factor of the two selected patients with respect to the selected physiological data, representing the similarity of the changes of the two selected patients with respect to the selected physiological data and risk value.

[0076] In this regard, considering that the magnitude of change of data in different dimensions is different, linear normalization is first performed on the respective data dimensions before comparing differences. For the first difference factor, linear normalization is performed on the data dimension consisting of the absolute value of the difference in physiological data among all patients for a certain physiological data. For the second difference factor, linear normalization is performed on the data dimension consisting of the absolute value of the difference in risk values ​​among all patients.

[0077] Further integrate the predictive importance of various physiological data of each patient in the target group with all corresponding relatively typical factors to obtain typical correction factors.

[0078] As an example, a typical formula for calculating the correction factor includes:

[0079]

[0080] Among them, DX i,h Let f[] represent the typical correction factor for the h-th patient in the i-th target group; f[] represents the linear normalization function; j is the index of the other groups excluding the target group, J is the number of other groups excluding the target group; l is the index of the patient in the other groups; n jZ represents the number of patients in the j-th other group; g represents the index of the physiological indicator type; G represents the total number of all physiological indicator types; Z i,g D represents the predictive importance of the g-th physiological data in the i-th target group; [i,h],[j,l]g Let represent the relative canonical factor of the h-th patient in the i-th target group and the l-th patient in the j-th other group regarding the g-th physiological indicator.

[0081] The formula integrates predictive importance and relative canonicity factors through multiplication, and includes all patients within each group. The data dimensions are linearly normalized to characterize the degree to which a patient's risk value represents the commonality of all patients in the group, providing a basis for revising the initial typicality.

[0082] Finally, the typicality correction factor and the initial typicality are combined to obtain the final typicality of the risk value for each patient in the target group.

[0083] As an example, the product of the typicality correction factor and the initial typicality for each patient is used as the final typicality of each patient's risk value.

[0084] The method for obtaining the final typicality of the risk value for each patient within each group is the same and will not be repeated here.

[0085] Because decision tree algorithms classify data step by step by splitting nodes, thus outputting a clear prediction, all patients in each data set should be assigned to a unique postoperative care risk value.

[0086] Having determined the final typicality of postoperative care risk values ​​for patients in each group of physiological feedback data, it's crucial to ensure that patients with high typicality contribute more to the decision tree model, providing more effective information for the classification process and helping the model make more accurate predictions. Simultaneously, this approach reduces interference from outliers or extreme data, improving the model's accuracy and robustness. Therefore, the final risk values ​​are fused based on the final typicality of patients within each group to obtain optimized risk values ​​and construct predictive decision trees.

[0087] Preferably, in one embodiment of the present invention, considering that the higher the proportion of the final typicality of a patient in a certain group, the higher the reference value of the risk value relative to other patients in the same group, and the greater the weight should be assigned, the proportion of the final typicality of each patient to all final typicalities in the same group is used as the weight, and the risk values ​​of each patient in each group are weighted and summed, and the weighted sum is used as the risk optimization value corresponding to the physiological data of each feedback in each group.

[0088] This allows for the determination of a unique risk value for post-cardiovascular interventional care corresponding to each set of physiological data.

[0089] By weighting the final typicality of the postoperative care risk for each cardiovascular surgery patient in each set of physiological data feedback, the problem of inconsistent physiological systems adaptability to cardiovascular interventional surgery among different patients, which could not accurately represent typical postoperative care risks and led to a lack of representativeness in the construction of the decision tree, was eliminated. This resulted in a unique optimized value for typical postoperative care risk corresponding to each set of physiological feedback data.

[0090] Next, this step automatically learns the optimal decision rule based on the unique postoperative care risk optimization value corresponding to all physiological feedback groups, constructs a typical and representative decision tree, and predicts the postoperative care risk of new patients. Moreover, this prediction result is more typical and can provide data support for clinical nursing.

[0091] In one embodiment of the present invention, the prediction decision tree is a CART regression tree, and the specific construction process is existing technology and will not be described in detail here.

[0092] In summary, addressing the technical problem that the single risk value used in existing technologies cannot accurately represent typical postoperative care risks, leading to a lack of representativeness in decision tree construction, this invention proposes a cardiovascular interventional postoperative care risk prediction system based on physiological feedback. This invention first acquires historical patients' physiological data and postoperative care risk values ​​and groups them; further, it analyzes the clustering of patients' risk values ​​and combines these risk values ​​to obtain initial typicality; it further analyzes the reference weights of other groups for various physiological data of the target group, and uses this to analyze the correlation between the physiological data of other groups and the maximum initial typicality within other groups, obtaining the predictive importance of various physiological data of the target group; it further obtains the final typicality; finally, it integrates the risk values ​​based on the final typicality of patients within each group to obtain optimized risk values ​​and construct a predictive decision tree. Addressing the problem that average risk values ​​cannot represent typical postoperative care risks and affect prediction accuracy, this invention obtains initial typicality by analyzing risk value clustering, and combines group differences to obtain predictive importance and final typicality, optimizing risk values ​​and improving the representativeness and individualized prediction ability of the decision tree.

[0093] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0094] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A cardiovascular interventional postoperative care risk prediction system based on physiological feedback, characterized in that, The system includes: Data acquisition module: Acquires physiological data and postoperative care risk values ​​of patients undergoing historical cardiovascular interventional surgery; groups patients with identical physiological data into groups, and selects each group as the target group; Initial analysis module: Based on the clustering of each patient's risk value within its respective group, and in conjunction with the risk value, an initial typicality is obtained; based on the differences in the same physiological data among different groups, and in conjunction with the initial typicality, reference weights for various physiological data of other groups relative to the target group are obtained; for each physiological data, the correlation between the physiological data of other groups and the maximum initial typicality within other groups is analyzed based on the reference weights to obtain the predictive importance of various physiological data of the target group. Prediction optimization module: Based on the similarity of changes in various physiological data and risk values ​​of each patient in the target group to all patients in other groups, and combined with the prediction importance and the initial typicality, the final typicality is obtained; based on the final typicality of patients in each group, the risk value is fused to obtain the risk optimization value and construct a prediction decision tree.

2. The cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The method for obtaining the initial typicality includes: The clustering coefficient of each risk value in its respective group is obtained based on the LOF algorithm; the risk value and the corresponding clustering coefficient are fused to obtain the initial typicality of the risk value for each patient.

3. The cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 2, characterized in that, The method for obtaining the aggregation coefficient includes: Each risk value in a group is used as its own horizontal and vertical coordinates and mapped onto a two-dimensional plane to obtain the LOF value of each risk value. The LOF value is then negatively correlated, normalized, and used as the clustering coefficient of each risk value.

4. The cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The method for obtaining the reference weights includes: For each physiological data item, the reference weights of the other groups relative to the target group are obtained based on the absolute value of the difference between the physiological data of the other groups and the target group, combined with the largest initial typicality within the other groups.

5. The cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The method for obtaining the predicted importance includes: Select each of the aforementioned physiological data as the target physiological data; The reference weights of each other group relative to the target physiological data of the target group and the target physiological data of each other group are fused to construct a first sequence with a fusion value; the reference weights of each other group relative to the target physiological data of the target group and the maximum initial typicality within each other group are fused to construct a second sequence with a fusion value. Analyze the correlation between the first sequence and the second sequence to obtain the predictive importance of the target physiological data for the target group.

6. The cardiovascular interventional postoperative care risk prediction system based on physiological feedback according to claim 5, characterized in that, The correlation between the first sequence and the second sequence was analyzed using the Pearson correlation coefficient.

7. The cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The method for obtaining the final typicality includes: A relative typicality factor is obtained based on the similarity of the differences in various physiological data between each patient in the target group and each patient in other groups and the differences in risk values. By fusing the predictive importance of each physiological data of each patient in the target group with all the corresponding relative typical factors, a typicality correction factor is obtained; by fusing the typicality correction factor and the initial typicality, the final typicality of the risk value of each patient in the target group is obtained.

8. The cardiovascular interventional postoperative care risk prediction system based on physiological feedback according to claim 1, characterized in that, The method for obtaining the risk optimization value includes: Using the proportion of each patient's final typicality to the total final typicality of all patients in the same group as the weight, the risk values ​​of each patient in each group are weighted and summed to obtain the risk optimization value corresponding to each physiological data feedback of each group.

9. A cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The prediction decision tree is a CART regression tree.

10. A cardiovascular interventional postoperative nursing risk prediction system based on physiological feedback according to claim 1, characterized in that, The risk value for postoperative care is the DAPT score.

Images