Intelligent warning system for postoperative complications based on liver transplant patients

Abstract

The present application relates to the technical field of medical auxiliary, in particular to an intelligent early warning system for postoperative complications of liver transplantation patients, which clusters postoperative patients without complications and postoperative patients with each type of complications, obtains the stage influence degree according to the distribution difference of the corresponding two types of patients in the cluster, obtains the contribution weight according to the difference of the feature vectors of the to-be-evaluated patient and the postoperative patient with early warning complications in each stage and the stage influence degree, clusters the postoperative patient with early warning complications and the to-be-evaluated patient, obtains the risk early warning value according to the position distribution of the to-be-evaluated patient in the cluster to which it belongs and the mainstream degree of the patient's complications, adjusts the risk early warning value to obtain the optimized early warning value according to the contribution weight, and early warns the postoperative complications of the to-be-evaluated patient. The present application adjusts the risk early warning by using the driving effect of different stages of operation on complication early warning, and improves the accuracy of complication early warning.

Description

Technical Field

[0001] This invention relates to the field of medical auxiliary technology, specifically to an intelligent early warning system for postoperative complications in liver transplant patients. Background Technology

[0002] Liver transplantation is the fundamental treatment for patients with end-stage liver disease. However, postoperative complications not only increase the risk of death but also lead to a surge in the consumption of medical resources, prolonged hospital stays, and significantly higher treatment costs. Therefore, achieving early and accurate warning of complications, and providing clinicians with a critical intervention window, is crucial for improving the survival rate of liver transplant patients.

[0003] The prior art patent document with publication number CN118609801A discloses a surgical anesthesia information early warning method and system based on clinical auxiliary decision-making. Specifically, it collects historical patient data and current patient data, uses knowledge graph to analyze and process the anesthesia data of historical patients in the three stages of preoperative, intraoperative and postoperative periods to form an early warning event database, and uses artificial intelligence technology to realize intelligent assessment of anesthesia risk level.

[0004] However, the occurrence of postoperative complications in liver transplant patients is a dynamic cumulative process driven by multiple stages and dimensions of factors. The risk depends not only on the patient's baseline condition and the severity of the primary liver disease, but also on the dynamic evolution of the physiological state throughout the perioperative period from preoperative to postoperative. Furthermore, the impact of different surgical stages on postoperative complications varies. For example, in the early postoperative stage, patients are in an immunosuppressed state, and changes in routine indicators are often misinterpreted as ordinary inflammatory reactions, masking early signs of complications and leading to a high false alarm rate for postoperative complications in liver transplant patients. Summary of the Invention

[0005] To address the technical problem of high false alarm rates in early warning systems caused by the varying impacts of different surgical stages during the perioperative period on complications, this invention aims to provide an intelligent early warning system for postoperative complications in liver transplant patients. The specific technical solution adopted is as follows:

[0006] This invention proposes an intelligent early warning system for postoperative complications in liver transplant patients, the system comprising:

[0007] The data acquisition module is used to acquire feature vectors of postoperative patients with complications, postoperative patients without complications, and patients to be evaluated at each surgical stage. Postoperative patients with complications have different types of complications.

[0008] The stage impact analysis module is used to cluster patients without postoperative complications and patients with postoperative complications of each type. Based on the distribution differences between the two types of patients within the obtained clusters, the stage impact of each surgical stage on each type of complication is obtained.

[0009] The stage contribution analysis module is used to select early warning complications for patients to be evaluated; based on the difference in feature vectors between the patient to be evaluated and postoperative patients with early warning complications at each surgical stage and the stage influence degree, the contribution weight of each surgical stage of the patient to be evaluated to early warning complications is obtained.

[0010] The risk warning analysis module is used to cluster postoperative patients with warning complications and patients to be evaluated. Based on the distribution of patients to be evaluated within their respective clusters and the prevalence of complications, the risk warning value of patients to be evaluated is obtained at each surgical stage.

[0011] The complication warning module is used to adjust the risk warning value according to the contribution weight of all surgical stages, obtain an optimized warning value, and provide postoperative complication warnings for patients to be evaluated.

[0012] Furthermore, the acquisition of the stage impact of each surgical stage on each type of complication includes:

[0013] Choose any type of complication and record it as an example complication. Record postoperative patients who develop the example complication and postoperative patients who do not develop the example complication as target patients for the example complication.

[0014] Based on the feature distance between the feature vectors of different target patients in each surgical stage of the example complication, all target patients of the example complication are clustered to obtain the first cluster of each surgical stage under the example complication; the first cluster containing patients who did not develop complications after surgery is denoted as the target cluster of each surgical stage under the example complication.

[0015] The proportion of patients within the target cluster who did not develop postoperative complications is used as the first probability distribution; the proportion of patients within the target cluster who developed postoperative complications is used as the second probability distribution.

[0016] Obtain the JS divergence between the first probability distribution and the second probability distribution of all target clusters for each surgical stage under the example complication, and denote it as the stage influence degree of each surgical stage on the example complication.

[0017] Furthermore, the selected early warning complications for patients to be evaluated include:

[0018] Choose any surgical stage as the example stage, calculate the mean of the feature distances between the patient to be evaluated in the example stage and all postoperative patients with each type of complication in the example stage, and perform negative correlation mapping on the mean to obtain the stage similarity between the patient to be evaluated in the example stage and each type of complication.

[0019] The sum of the similarities between the patient to be evaluated at all surgical stages and the stages of each type of complication is taken as the overall similarity of the patient to be evaluated for each type of complication.

[0020] The complication corresponding to the highest value among the risk levels of all types of complications for the patient to be evaluated is selected and recorded as the early warning complication for the patient to be evaluated.

[0021] Furthermore, the contribution weight of each surgical stage under the early warning complications of the patient to be evaluated includes:

[0022] The product of the similarity of the patient to be evaluated at each surgical stage to the stage of the early warning complication and the influence of each surgical stage on the stage of the early warning complication is taken as the early warning drive degree of the patient to be evaluated for each surgical stage of the early warning complication.

[0023] Based on the warning drive degree of all surgical stages, the warning drive degree of each surgical stage is normalized to obtain the contribution weight of each surgical stage to the warning of complications for the patient to be evaluated.

[0024] Furthermore, obtaining the risk warning value for the patient to be evaluated at each surgical stage includes:

[0025] The patients to be evaluated and the postoperative patients who developed the aforementioned early warning complications are referred to as the analysis patients;

[0026] Based on the feature distance between the feature vectors of different patients at each surgical stage, all analyzed patients are clustered to obtain a second cluster for each surgical stage;

[0027] The proportion of patients with postoperative complications within the second cluster to which the patient to be evaluated belongs in each surgical stage, among all postoperative patients who have the aforementioned warning complications, is used as the mainstream of each surgical stage;

[0028] The ratio of the distance between the patient to be evaluated and the cluster center of the second cluster to which the patient belongs in each surgical stage is used as the numerator, and the maximum distance between the patient to be evaluated and the cluster center of all postoperative complications in the second cluster to which the patient belongs in each surgical stage is used as the denominator, is used as the deviation degree of each surgical stage.

[0029] Based on the mainstream degree and the deviation degree, obtain the risk warning value of the patient to be evaluated at each surgical stage.

[0030] Furthermore, obtaining the optimized early warning value includes:

[0031] The risk warning values ​​for all surgical stages of the patient to be evaluated are weighted and summed according to the contribution weights, and the weighted summation result is normalized to obtain the optimized warning value.

[0032] Furthermore, the mainstream degree is positively correlated with the warning value, and the deviation degree is negatively correlated with the warning value.

[0033] Furthermore, the feature distance is the Euclidean distance between any two feature vectors.

[0034] Furthermore, the method for clustering all target patients with example complications is the K-means clustering algorithm.

[0035] Furthermore, the elements at the same position in different feature vectors have the same dimension. This invention has the following beneficial effects:

[0036] In this embodiment of the invention, the stage influence degree is obtained by the distribution differences of postoperative patients with each type of complication within clusters, effectively revealing the dynamic correlation between different surgical stages and complications, and evaluating the predictive value of different surgical stages for complication early warning. Furthermore, by combining the differences in feature vectors between the patient to be evaluated and postoperative patients with early warning complications at different surgical stages, and by integrating individual performance with group patterns, the driving effect of different surgical stages on early warning complications is accurately quantified. This allows the system to focus on key risk stages and filter out interference from non-key stages, fundamentally solving the problem of false alarms caused by the differentiated impact of different surgical stages on complications during the perioperative period. The potential risk signals of patients to be evaluated exhibiting early warning complications at each surgical stage are accurately captured from two dimensions: group representativeness and individual typicality, resulting in risk warning values. Optimized warning values ​​are generated by adjusting the risk warning values ​​of each surgical stage using contribution weights, capturing the accumulation and evolution of early warning complication risks at different stages, thereby achieving earlier and more accurate complication early warning. Attached Figure Description

[0037] 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.

[0038] Figure 1 This is a system structure diagram of an intelligent early warning system for postoperative complications in liver transplant patients, provided in one embodiment of the present invention.

[0039] Figure 2 This is a schematic diagram of a computer device for an intelligent early warning system for postoperative complications in liver transplant patients, provided as an embodiment of the present invention. Detailed Implementation

[0040] 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 an intelligent early warning system for postoperative complications in liver transplant patients 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.

[0041] 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.

[0042] The following description, in conjunction with the accompanying drawings, details a specific scheme for an intelligent early warning system for postoperative complications in liver transplant patients provided by the present invention.

[0043] Example 1

[0044] Please see Figure 1 The diagram illustrates a system block diagram of an intelligent early warning system for postoperative complications in liver transplant patients according to an embodiment of the present invention. The system includes: a data acquisition module 110, a stage impact analysis module 120, a stage contribution analysis module 130, a risk early warning analysis module 140, and a complication early warning module 150.

[0045] The data acquisition module 110 is used to acquire feature vectors of postoperative patients with complications, postoperative patients without complications, and patients to be evaluated at each surgical stage. Postoperative patients with complications have different types of complications.

[0046] The surgical process is divided into different consecutive stages: the preoperative stage, the early postoperative stage, and the late postoperative stage. Data from the preoperative stage focuses on assessing the patient's baseline condition; data from the intraoperative stage focuses on recording the surgical procedure; data from the early postoperative stage focuses on monitoring the acute phase; and data from the late postoperative stage focuses on assessing long-term recovery. Postoperative complications refer to patients who develop complications after liver transplantation; postoperative complications-free patients refer to patients who do not develop complications after liver transplantation; and patients requiring evaluation refer to patients who need to be monitored for postoperative complications after liver transplantation. These three types of patients are collectively referred to as liver transplant patients. Postoperative complications can manifest in various forms.

[0047] Using the HL7 V2.5 protocol interface of the hospital information system, the medical data of all liver transplant patients at each surgical stage can be queried and extracted from the hospital information system by their unique identifiers. The specific data includes: Preoperative medical data such as patient basic information (age, gender, body mass index), liver function indicators (Child-Pugh classification and end-stage liver disease model score), degree of cirrhosis assessed by FibroScan or liver biopsy, imaging results (liver volume and vascular structure in computed tomography), and medical history (hepatitis type, history of abdominal surgery, etc.); Intraoperative medical data such as operation duration, blood loss, transfusion volume, and surgical record summary (donor type and vascular anastomosis method, etc.); Early postoperative medical data such as vital sign monitoring data (heart rate, blood pressure, respiratory rate, body temperature, and blood oxygen saturation), and laboratory test data (white blood cell count, C-reactive protein, and liver function retest indicators); and Late postoperative medical data such as follow-up time points, liver function retest data (bilirubin and international normalized ratio (INR), and medication records (immunosuppressant dosage, etc.).

[0048] First, the k-anonymization algorithm is used to remove direct identifiers from liver transplant patients, generalizing indirect identifiers to age ranges. Direct identifiers include the patient's name and ID number, while concise identifiers include the patient's date of birth and age. Then, the medical data for each surgical stage is organized by surgical stage and timestamp, and linked to data from the hospital information system, laboratory information system, and medical imaging archive and communication system through patient identifiers to generate a staged medical data set. The medical data for each liver transplant patient at each surgical stage constitutes a stage subset. Based on the medical data for each stage subset, several feature parameters are obtained. Specifically, the medical data is divided into numerical data, categorical data, and time-series data. For numerical data, Z-scores are used for standardization. For categorical data, one-hot encoding is used to convert it into a binary vector. For time-series data, statistical features are extracted, including mean, standard deviation, and slope. For example, the patient's age, end-stage liver disease model score, and blood loss are numerical data; Child-Pugh classification and donor type are categorical data; and postoperative vital sign monitoring data are time-series data. Different types of feature parameters are standardized to obtain standard feature parameters. A feature vector is constructed by mapping the medical data of each liver transplant patient at each surgical stage to the standard feature parameters. It is important to note that elements at the same position in different feature vectors have the same dimension, and all elements in the feature vectors are numerical data.

[0049] In this embodiment of the invention, the range standardization method is selected for standardization. However, Z-score standardization and decimal scaling standardization methods can also be used for standardization, and this is not limited here.

[0050] In this embodiment of the invention, the specific method for dividing the surgical stage is as follows: the 30 days before the operation are defined as the preoperative stage, the period from the induction of anesthesia to the end of the operation is defined as the preoperative stage, the first 8 days after the operation are defined as the early postoperative stage, and the period from the 8th day after the operation to the end of the follow-up is defined as the early postoperative stage. The implementer can set the duration of the preoperative stage, the early postoperative stage and the early postoperative stage according to the specific circumstances, which is not limited here.

[0051] In this embodiment of the invention, the Euclidean distance between any two feature vectors is used as the feature distance, or the result of negatively correlated mapping of the cosine similarity between any two feature vectors can be used as the feature distance. Since the cosine similarity ranges from -1 to 1, the negative correlation mapping is achieved by subtracting the cosine similarity from the constant 1.

[0052] The stage impact analysis module 120 is used to cluster patients without postoperative complications and patients with postoperative complications of each type. Based on the distribution differences between the two types of patients within the obtained clusters, the stage impact of each surgical stage on each type of complication is obtained.

[0053] By clustering, postoperative patients with each type of complication were divided into several patient groups with similar physiological states, including those without postoperative complications. Based on the differences in the distribution patterns of postoperative patients with and without complications within the clusters, the degree of deviation of the physiological pattern of complications from the healthy pattern during the surgical stage was characterized. This effectively revealed the dynamic correlation between different surgical stages and complications, assessed the predictive value of different surgical stages for complication early warning, and obtained the stage impact degree.

[0054] The stage contribution analysis module 130 is used to select the early warning complications of the patients to be evaluated; based on the difference in feature vectors and stage influence between the patients to be evaluated and the postoperative complications that have early warning complications at each surgical stage, the contribution weight of each surgical stage of the patients to be evaluated to the early warning complications is obtained.

[0055] Early warning of complications predicts the most likely types of complications for patients under evaluation. The differences in feature vectors between patients under evaluation and those who developed early warning complications at each surgical stage reflect the confidence that patients under evaluation will develop early warning complications in the future. Combined with the stage influence of the surgical stage on the predictive value of complication early warning, and by integrating individual performance and group patterns, the driving role of the surgical stage on early warning of complications is precisely quantified to obtain the contribution weight, thereby improving the adaptability and accuracy of complication early warning.

[0056] The risk warning analysis module 140 is used to cluster postoperative patients with warning complications and patients to be evaluated. Based on the distribution of patients to be evaluated within their respective clusters and the prevalence of complications, the risk warning value of patients to be evaluated at each surgical stage is obtained.

[0057] By clustering postoperative patients with pre-existing complications with those under evaluation, a patient group with similar complication patterns to the patients under evaluation is dynamically identified. Due to the heterogeneity of physiological patterns among postoperative patients with pre-existing complications, the representativeness of the physiological patterns of the patients under evaluation can be assessed by the prevalence of complications within their respective clusters; the individual typicality of the physiological patterns of the patients under evaluation can be assessed by their location distribution within their respective clusters. Combining these two methods, the potential risk of future pre-existing complications in the patients under evaluation is analyzed, resulting in a risk warning value. This allows for the precise capture of potential risk signals for pre-existing complications in the patients under evaluation at each surgical stage.

[0058] The complication warning module 150 is used to adjust the risk warning value according to the contribution weight of all surgical stages, obtain the optimized warning value, and provide postoperative complication warnings for patients to be evaluated.

[0059] Since the impact of different surgical stages on complication warning varies during the perioperative period, and the occurrence of complications is the result of the accumulation and evolution of risks at multiple stages, the risk warning value is adjusted by using the contribution weights of all surgical stages to obtain the overall effective risk intensity of the perioperative period, i.e., the optimized warning value. By combining spatiotemporal dynamic information and individualized characteristics, the final output warning result is guaranteed to be robust and reliable to the greatest extent.

[0060] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the stage impact degree includes: arbitrarily selecting a type of complication as an example complication, and arbitrarily selecting postoperative patients with the example complication and postoperative patients without the example complication as target patients of the example complication; clustering all target patients of the example complication based on the feature distance between the feature vectors of different target patients of the example complication at each surgical stage to obtain a first cluster for each surgical stage under the example complication; arbitrarily selecting the first cluster containing postoperative patients without the example complication as the target cluster for each surgical stage under the example complication; using the proportion of postoperative patients without the example complication within the target cluster as a first probability distribution; using the proportion of postoperative patients with the example complication within the target cluster as a second probability distribution; and obtaining the JS divergence between the first probability distribution and the second probability distribution of all target clusters at each surgical stage under the example complication, which is recorded as the stage impact degree of each surgical stage on the example complication.

[0061] It should be noted that the healthy pathway and the complication pathway intersect within the target cluster. Based solely on the characteristics of the current surgical stage, it is difficult to distinguish future complication patients from healthy patients. While valuable for complication early warning, analyzing the differences among different patients within the target cluster can make the stage impact more sensitive. If the JS divergence is larger, the distribution of postoperative patients without complications within the target cluster is less similar to the distribution of postoperative patients with example complications. This means that the physiological pattern of postoperative patients with example complications deviates more significantly from the healthy pattern, and the information from that surgical stage is more likely to induce example complications in patients. In this embodiment of the invention, the formula for calculating the stage impact of each surgical stage on example complications is:

[0062]

[0063] In the formula, the example complication is defined as the c-th type of complication; The stage impact of each surgical stage on the c-th type of complication; m is the total number of target clusters for each surgical stage under the c-th type of complication; This represents the percentage of patients within the i-th target cluster at each surgical stage who did not experience postoperative complications under the c-th type of complication. Let be the percentage of postoperative complications of type c within the i-th target cluster at each surgical stage under type c complication; ln is a logarithmic function with the natural constant e as the base.

[0064] In one implementation of this invention, the K-means clustering algorithm is used to cluster all target patients with example complications, where the K value can be determined using the elbow method. Both the K-means clustering algorithm and the elbow method are well-known techniques to those skilled in the art and will not be described in detail here.

[0065] It should be noted that the method for obtaining the stage impact of each surgical stage on each complication type is the same as that for obtaining the stage impact of each surgical stage on the example complication.

[0066] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining early warning complications includes: selecting any surgical stage as the example stage; calculating the mean of the feature distances between the patient to be evaluated and all postoperative patients with each type of complication in the example stage; performing a negative correlation mapping on the mean to obtain the stage similarity between the patient to be evaluated and each type of complication in the example stage; summing the stage similarities between the patient to be evaluated and each type of complication in all surgical stages as the overall similarity of the patient to be evaluated for each type of complication; and selecting the complication with the highest value from the risk of the patient to be evaluated for all types of complications, and recording it as the early warning complication of the patient to be evaluated.

[0067] It should be noted that the smaller the characteristic distance between the patient to be evaluated at the example stage and the postoperative complications of the example complication at the example stage, the more similar the physiological patterns of the patient to be evaluated and the postoperative complication group are at the example stage. Therefore, a negative correlation mapping of the mean is needed to obtain the stage similarity. Considering that complications are a multi-stage dynamic cumulative process, the stage similarity between all surgical stages and the example complication needs to be summed. A larger sum indicates a closer match between the physiological patterns of the patient to be evaluated and the example complication during the perioperative period, and thus a higher risk of the patient to develop the example complication. The complication with the highest risk, i.e., the warning complication, is the complication that the patient to be evaluated is most likely to experience in the future.

[0068] In this embodiment of the invention, the data to be processed is used as an exponential function with the natural constant as the base to achieve a negative correlation mapping of the data to be processed. Alternatively, the negative correlation mapping can be achieved by taking the reciprocal or other methods, which are not limited here.

[0069] It should be noted that the methods for obtaining the similarity between the patient to be evaluated at each surgical stage and the stage of each type of complication, as well as the similarity between the example stage and the stage of each type of complication, are the same.

[0070] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the contribution weight includes: multiplying the similarity between the patient to be evaluated and the stage of the early warning complication at each surgical stage and the stage influence of each surgical stage on the early warning complication as the early warning driving degree of each surgical stage of the patient to be evaluated on the early warning complication; and normalizing the early warning driving degree of each surgical stage based on the early warning driving degree of all surgical stages to obtain the contribution weight of each surgical stage of the patient to be evaluated on the early warning complication.

[0071] It should be noted that the greater the stage similarity, the more similar the physiological patterns of the patient under evaluation and the postoperative complications with early warning of complications are during the surgical stage, and the greater the confidence that the patient under evaluation will develop early warning complications in the future. Conversely, the greater the stage influence, the more easily the information from the surgical stage can induce early warning complications in the patient under evaluation, and the higher the predictive value of that surgical stage for complication early warning. When both stage similarity and stage influence are greater, the information from the surgical stage of the patient under evaluation has higher confidence and early warning value for assessing early warning complications, and the driving effect of the surgical stage on early warning complications is more significant, resulting in a greater early warning drive degree. Therefore, both stage similarity and stage influence are positively correlated with the early warning drive degree. In this embodiment of the invention, the product of the stage similarity of the patient under evaluation with the example complication at each surgical stage and the stage influence of each surgical stage on early warning complications is used as the early warning drive degree of the patient under evaluation for each surgical stage on early warning complications. By calculating the ratio of the early warning drive degree corresponding to each surgical stage to the sum of the early warning drive degrees corresponding to all surgical stages, the early warning drive degree corresponding to each surgical stage is normalized to obtain the contribution weight. The surgical stage with the greater contribution weight is more likely to be a critical period for predicting early warning complications in patients to be evaluated.

[0072] It should be noted that the methods for obtaining the warning drive degree of patients to be assessed for early warning complications are the same for all surgical stages and the example stage, and the methods for obtaining the contribution weights are the same for all surgical stages and the example stage.

[0073] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the risk warning value includes: designating the patient to be evaluated and postoperative patients with warning complications as analyzed patients; clustering all analyzed patients based on the feature distance between the feature vectors of different analyzed patients at each surgical stage to obtain a second cluster for each surgical stage; using the proportion of postoperative patients with warning complications within the second cluster to which the patient to be evaluated belongs in each surgical stage as the mainstream of each surgical stage; using the distance between the patient to be evaluated and the cluster center of the second cluster to which it belongs in each surgical stage as the numerator, and the maximum distance between all postoperative patients with warning complications and the cluster center within the second cluster to which the patient to be evaluated belongs in each surgical stage as the denominator, the ratio obtained is used as the deviation of each surgical stage; and obtaining the risk warning value of the patient to be evaluated in each surgical stage based on the mainstream and the deviation. In this embodiment, distance refers to feature distance.

[0074] It should be noted that the second cluster to which the patient under evaluation belongs at each surgical stage defines its risk environment. A higher mainstream prevalence indicates a greater likelihood that the physiological pattern exhibited by the patient under evaluation is a common or typical complication pattern of a historically predicted complication, thus increasing the potential risk of the patient developing the predicted complication in the future, and consequently, the higher the risk warning value. Patients located at the cluster center are known to be typical representatives of the complication pattern; a smaller deviation indicates a closer patient to the cluster center, meaning the patient's complication pattern highly matches the typical complication pattern of its cluster. In other words, the more typical the patient's clinical presentation is of the complication, the higher the potential risk of the patient developing the predicted complication in the future, and consequently, the higher the risk warning value. Therefore, mainstream prevalence is positively correlated with the warning value, while deviation is negatively correlated. In this embodiment, the difference between the constant 1 and the deviation at each surgical stage is calculated, and the product of this difference and the mainstream prevalence is used as the warning value. In this embodiment, the deviation is negatively correlated by subtracting the constant 1 from the deviation. Negative correlation can also be achieved by taking the reciprocal or by function transformation, which is not limited here.

[0075] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the optimized warning value includes: weighting and summing the risk warning values ​​of all surgical stages of the patient to be evaluated according to the contribution weight, and normalizing the weighted summation result to obtain the optimized warning value.

[0076] In one specific implementation of this invention, the optimized warning value is expressed by the formula:

[0077]

[0078] in, To optimize early warning values; The contribution weight for the nth surgical stage under the early warning complications of the patient to be evaluated; is the warning value for the patient to be evaluated at the nth surgical stage; N is the number of surgical stages; is the Sigmoid normalization function.

[0079] It should be noted that the risk warning value measures the potential risk of the patient under evaluation developing a predicted complication in the future, while the contribution weight measures the probability that the surgical stage is a critical period for predicting the occurrence of predicted complications. The risk warning value is weighted using the contribution weight to obtain the overall effective risk intensity in the perioperative period. Since the occurrence of complications is the result of the accumulation and evolution of risks at multiple stages, it is necessary to sum the risk warning values ​​of all surgical stages to capture the continuity and cumulative effect of risk, obtaining an optimized warning value representing the overall global risk.

[0080] In this embodiment of the invention, when the optimized warning value is greater than the preset warning threshold, the risk of the patient to be assessed developing a warning complication is higher, and a warning message is issued; conversely, when the optimized warning value is less than the preset warning threshold, the risk of the patient to be assessed developing a warning complication is higher, and no warning is issued.

[0081] In one implementation of this invention, the preset warning threshold is set to 0.8, which can be set by the implementer according to specific circumstances.

[0082] This invention is now complete.

[0083] Example 2

[0084] Figure 2 This is a schematic diagram of a computer device for an intelligent early warning system for postoperative complications in liver transplant patients, provided as an embodiment of the present invention. For example,... Figure 2 As shown, the computer device includes: a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202, wherein when the processor 202 executes the computer program 203, the computer device can execute any of the aforementioned intelligent early warning systems for postoperative complications of liver transplant patients.

[0085] Furthermore, embodiments of this application also protect an apparatus that may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to execute the intelligent early warning system for postoperative complications of liver transplant patients provided in embodiments of this application.

[0086] This embodiment can divide the device into functional modules based on the above method example. For example, each module can correspond to a separate function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0087] It should be understood that the device provided in this embodiment is used to execute the above-described intelligent early warning system for postoperative complications in liver transplant patients, and therefore can achieve the same effect as the above-described implementation method.

[0088] When using integrated units, the device may include a processing module and a storage module. When applied to a workpiece, the processing module can be used to control and manage the workpiece's operations. The storage module can be used to support the execution of program code by the workpiece.

[0089] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits as disclosed in this application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and microprocessors, etc., and the storage module may be a memory.

[0090] Example 3

[0091] This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-described related method steps to implement the intelligent early warning system for postoperative complications of liver transplant patients provided in the above embodiment.

[0092] In this embodiment, the device and computer-readable storage medium are used to execute the corresponding system provided above. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding system provided above, and will not be repeated here.

[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.

[0095] 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 within the protection scope of the present invention.

Claims

1. An intelligent early warning system for postoperative complications in liver transplant patients, characterized in that, The system includes: The data acquisition module is used to acquire feature vectors of postoperative patients with complications, postoperative patients without complications, and patients to be evaluated at each surgical stage. Postoperative patients with complications have different types of complications. The stage impact analysis module is used to cluster patients without postoperative complications and patients with postoperative complications of each type. Based on the distribution differences between the two types of patients within the obtained clusters, the stage impact of each surgical stage on each type of complication is obtained. The stage contribution analysis module is used to select early warning complications for patients to be evaluated; based on the difference in feature vectors between the patient to be evaluated and postoperative patients with early warning complications at each surgical stage and the stage influence degree, the contribution weight of each surgical stage of the patient to be evaluated to early warning complications is obtained. The risk warning analysis module is used to cluster postoperative patients with warning complications and patients to be evaluated. Based on the distribution of patients to be evaluated within their respective clusters and the prevalence of complications, the risk warning value of patients to be evaluated is obtained at each surgical stage. The complication warning module is used to adjust the risk warning value according to the contribution weight of all surgical stages, obtain an optimized warning value, and provide postoperative complication warnings for patients to be evaluated. The acquisition of the stage impact of each surgical stage on each type of complication includes: Choose any type of complication and record it as an example complication. Record postoperative patients who develop the example complication and postoperative patients who do not develop the example complication as target patients for the example complication. Based on the feature distance between the feature vectors of different target patients in each surgical stage of the example complication, all target patients of the example complication are clustered to obtain the first cluster of each surgical stage under the example complication; the first cluster containing patients who did not develop complications after surgery is denoted as the target cluster of each surgical stage under the example complication. The proportion of patients within the target cluster who did not develop postoperative complications is used as the first probability distribution; the proportion of patients within the target cluster who developed postoperative complications is used as the second probability distribution. Obtain the JS divergence between the first probability distribution and the second probability distribution of all target clusters at each surgical stage under the example complication, and denote it as the stage influence degree of each surgical stage on the example complication; The selected early warning complications for patients to be evaluated include: Choose any surgical stage as the example stage, calculate the mean of the feature distances between the patient to be evaluated in the example stage and all postoperative patients with each type of complication in the example stage, and perform negative correlation mapping on the mean to obtain the stage similarity between the patient to be evaluated in the example stage and each type of complication. The sum of the similarities between the patient to be evaluated at all surgical stages and the stages of each type of complication is taken as the overall similarity of the patient to be evaluated for each type of complication. The complication with the highest value among the overall similarities of all types of complications in the patients to be evaluated is recorded as the early warning complication of the patients to be evaluated. The method of obtaining the contribution weight of each surgical stage of the patient to the early warning of complications includes: The product of the similarity of the patient to be evaluated at each surgical stage to the stage of the early warning complication and the influence of each surgical stage on the stage of the early warning complication is taken as the early warning drive degree of the patient to be evaluated for each surgical stage of the early warning complication. Based on the warning drive degree of all surgical stages, the warning drive degree of each surgical stage is normalized to obtain the contribution weight of each surgical stage to the warning of complications for the patient to be evaluated.

2. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 1, characterized in that, The acquisition of risk warning values ​​for the patient to be evaluated at each surgical stage includes: The patients to be evaluated and the postoperative patients who developed the aforementioned early warning complications are referred to as the analysis patients; Based on the feature distance between the feature vectors of different patients at each surgical stage, all analyzed patients are clustered to obtain a second cluster for each surgical stage; The proportion of patients with postoperative complications within the second cluster to which the patient to be evaluated belongs in each surgical stage, among all postoperative patients with the aforementioned warning complications, is used as the degree of complication prevalence in each surgical stage. The ratio of the distance between the patient to be evaluated and the cluster center of the second cluster to which the patient belongs in each surgical stage is used as the numerator, and the maximum distance between the patient to be evaluated and the cluster center of all postoperative complications in the second cluster to which the patient belongs in each surgical stage is used as the denominator, is used as the deviation degree of each surgical stage. Based on the prevalence and deviation of the aforementioned complications, risk warning values ​​are obtained for the patients to be evaluated at each surgical stage.

3. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 1, characterized in that, The process of obtaining the optimized early warning value includes: The risk warning values ​​for all surgical stages of the patient to be evaluated are weighted and summed according to the contribution weights, and the weighted summation result is normalized to obtain the optimized warning value.

4. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 2, characterized in that, The prevalence of the complications is positively correlated with the warning value, while the deviation is negatively correlated with the warning value.

5. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 1 or 2, characterized in that, The feature distance is the Euclidean distance between any two feature vectors.

6. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 1, characterized in that, The method used to cluster all target patients with the example complication is the K-means clustering algorithm.

7. The intelligent early warning system for postoperative complications in liver transplant patients according to claim 1, characterized in that, Elements at the same position in different feature vectors have the same dimension.

Images