ICU medical equipment alarm management system based on SEIPS model

Abstract

The present invention relates to the field of medical equipment alarm technology, and specifically, to an ICU medical equipment alarm management system based on the SEIPS model. It includes: an alarm generation unit monitors the patient's physiological parameters, and generates an alarm signal based on a distribution threshold algorithm, and optimizes the distribution threshold algorithm considering the different physiological parameters of different patients and different postoperative recovery periods; the alarm processing unit receives the alarm signals from different devices, and classifies and filters the alarm signals; the alarm communication unit receives the screened alarm signals, and transmits the alarm information to relevant medical staff through wireless transmission. The design of the present invention emphasizes the interaction between different medical work elements through the SEIPS model, aiming to improve the overall performance of medical services by improving the coordination and integration between the various parts within the system; by monitoring the patient's physiological parameters, and dynamically adjusting the alarm threshold using an optimized distribution threshold algorithm.

Description

Technical Field

[0001] The invention relates to the technical field of medical equipment alarms, and in particular to an ICU medical equipment alarm management system based on a SEIPS model. Background Art

[0002] The SEIPS model is not a specific unit module, but a methodological framework for understanding human behavior in work systems. The SEIPS model emphasizes that when designing and optimizing work systems, it is necessary to comprehensively consider multiple dimensions such as people, tools and technologies, organizational structure, work environment, and tasks. When designing an ICU medical equipment alarm management system based on the SEIPS model, the SEIPS model is not directly integrated into the above-mentioned unit modules, but serves as a principle and guide for designing and evaluating these modules.

[0003] Medical equipment in the ICU usually generates a large number of alarm signals. These alarms are sometimes too frequent or not accurate enough, causing alarm fatigue among medical staff, reducing their sensitivity and trust in alarms, and ignoring or even turning off some non-emergency alarms. This may cause alarms that really require emergency treatment to be ignored, thus affecting patient safety. Due to the large individual differences among patients, especially those in the postoperative recovery period, their physiological parameters may fluctuate within the normal range. Traditional fixed threshold alarm systems may not be able to adapt to such fluctuations, resulting in false alarms or missed alarms. In the busy ICU environment, medical staff are faced with the interference of different sounds and visual signals, as well as spatial layout limitations in the physical environment. How to effectively transmit alarm information to the appropriate medical staff is a challenge. Therefore, an ICU medical equipment alarm management system based on the SEIPS model is designed. Summary of the invention

[0004] The purpose of the present invention is to provide an ICU medical equipment alarm management system based on the SEIPS model to solve the problems of alarm fatigue, false alarms and missed alarms, and poor alarm information transmission proposed in the above background technology.

[0005] To achieve the above object, the present invention aims to provide an ICU medical equipment alarm management system based on the SEIPS model, comprising:

[0006] An alarm generating unit, wherein the alarm generating unit monitors the physiological parameters of the patient and generates an alarm signal based on a distribution threshold algorithm, and optimizes the distribution threshold algorithm taking into account the different physiological parameters and postoperative recovery periods of different patients;

[0007] An alarm processing unit, which receives alarm signals from different devices and classifies and filters the alarm signals;

[0008] An alarm communication unit receives the screened alarm signal and transmits the alarm information to medical staff via wireless transmission.

[0009] As a further improvement of the present technical solution, the physiological parameters monitored by the alarm generation unit include heart rate, blood pressure and blood oxygen saturation.

[0010] As a further improvement of the technical solution, the alarm generation unit generates an alarm signal based on a distribution threshold algorithm, comprising the following steps:

[0011] S1.1, preprocessing of physiological parameters;

[0012] S1.2. Use the distribution threshold algorithm to discover the distribution pattern of normal physiological parameters, and set the threshold according to this distribution pattern. Considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized. Since different patients in the postoperative recovery period have physiological differences and their physiological states change over time, the distribution threshold algorithm is further optimized;

[0013] S1.3. Compare the physiological parameters monitored in real time with the preset thresholds. When the monitored physiological parameters exceed the preset thresholds, an alarm signal is immediately generated.

[0014] As a further improvement of the technical solution, in S1.2, the distribution threshold algorithm is:

[0015] ;

[0016] Considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized as follows:

[0017] ;

[0018] Since different patients have physiological differences during the postoperative recovery period and their physiological states change over time, the distribution threshold algorithm is further optimized as follows:

[0019] ;

[0020] in, represents the threshold set point; It represents the set points of different optimized thresholds for different patients’ physiological parameters and postoperative recovery period; It represents the threshold setting point that is further optimized according to the physiological differences among different patients and the changes of their physiological status over time during the postoperative recovery period; Indicates physiological parameters; Indicates the central value of normal physiological parameters; timestamp indicating physiological parameters; represents the attenuation coefficient, which is used to determine the rate at which physiological parameters decay over time during the postoperative recovery period; represents the parameters that adjust for the effects of physiological differences among patients; Indicates physiological differences among patients; represents a time-based weight function, which is used to adjust the weights that change over time; represents the trend factor; represents the total number of physiological parameter measurements included in the dataset; An index variable representing the physiological parameter measurements contained in the dataset.

[0021] As a further improvement of the technical solution, the alarm processing unit classifies and filters the alarm signal, including the following steps:

[0022] S2.1, receiving an alarm signal, and taking a set of the alarm signals as a data set, and extracting key features from the alarm signal;

[0023] S2.2, using the alarm classification model to classify the alarm signals into different categories according to the key features of the alarm;

[0024] S2.3. Use the relationship evaluation model to evaluate the relationship between multiple alarms and determine the priority. Use the repeated search algorithm to find the correlation between different alarms and eliminate redundant alarms caused by the same reason.

[0025] S2.4. Summarize the classified and filtered alarm signals and send them to the alarm communication unit.

[0026] As a further improvement of the technical solution, in S2.2, the alarm signal is classified using an alarm classification model, and the alarm signal is classified into different categories according to the key features of the alarm, including the following steps:

[0027] S2.21. For each key feature, use the alarm index as the feature selection criterion, consider the impact of common alarm and rare alarm features on the alarm index, and then consider the impact of the dynamic nature of heart rate, blood oxygen saturation and blood pressure over time on the urgency of the alarm index, and optimize the alarm index;

[0028] S2.22, select the feature with the lowest alarm index as the best split feature for the current node;

[0029] S2.23, dividing the data set into several subsets according to different values ​​of the selected splitting feature;

[0030] S2.24, continue to split the data in the subset until a predetermined stop condition is reached, and identify abnormal alarms through an alarm anomaly algorithm;

[0031] S2.25. Use the constructed alarm classification model to classify the alarm signal. Input the key features of the alarm. The classification tree makes judgments along the path from the root node to a leaf node. The leaf node finally reached determines the category of the alarm signal.

[0032] As a further improvement of the technical solution, in S2.21, the alarm index is:

[0033] ;

[0034] Consider the impact of common and rare alarms on the alarm index. Common alarms include abnormal heart rate, low blood oxygen saturation, and abnormal blood pressure alarms. Rare alarms include acute myocardial infarction, acute cerebrovascular event, and severe respiratory failure alarms.

[0035] The alarm index is:

[0036] ;

[0037] Since the dynamic changes of heart rate, blood oxygen saturation and blood pressure over time affect the urgency of the alarm index, the alarm index is optimized as follows:

[0038] ;

[0039] in, Represents a dataset; Indicates the alarm index; represents the alarm index after considering the characteristics of common alarms and rare alarms; An alarm index that takes into account the dynamics of common and rare alarm characteristics over time; Represents the total number of categories in the data set; Represents the category index; Indicates the total number of alarm records in the data set; Indicates the index of the alarm record; Indicates the current time; Indicates The time when each alarm record occurs; Indicates the decay coefficient that controls the decay speed of the alarm record weight over time; Indicates The characteristic parameters of the class are used to adjust the weights of common and rare alarms; Indicates The alarm category to which the alarm record belongs; represents the base of natural logarithms; Represents the indicator function.

[0040] As a further improvement of the technical solution, in S2.3, a relationship evaluation model is used to evaluate the relationship between multiple alarms and determine the priority, including the following steps:

[0041] S2.31, receiving the classified and filtered alarm signal, and preprocessing the alarm signal data;

[0042] S2.32. Determine the key attributes of the alarm signal and establish an alarm signal relationship matrix;

[0043] S2.33, collect historical alarm signal data, and use the historical alarm signal data to estimate the probability distribution of each node in the alarm relationship matrix;

[0044] S2.34. When a new alarm signal appears, the probability between the alarm signals is updated, the priority of the alarm is determined based on the updated probability, and the priority rules are dynamically adjusted based on the current ICU load and the patient's urgency factors.

[0045] As a further improvement of the technical solution, in S2.3, a repeated search algorithm is used to find the correlation between different alarms, including the following steps:

[0046] S2.35. Record the time, type, and level of all alarm signals, determine the criteria for alarm signals to be considered duplicates, and build an environmental alarm database to record the alarm conditions of all patients within the same time window;

[0047] S2.36. Set a time window within which alarm signals appearing are considered potential duplicate alarms, and identify alarm signals appearing within this time window;

[0048] S2.37, record the occurrence time and type of each alarm signal as an alarm sequence, use a sequence pattern mining algorithm to discover frequent alarm sequences, and construct a multi-patient alarm sequence to integrate the alarm sequences of all patients in the same time window;

[0049] S2.38. For the mined frequent alarm sequences, evaluate the support and confidence of the frequent alarm sequence association rules, and select the rules with higher support and confidence as the alarm association rules;

[0050] S2.39. Apply correlation rules to alarm signal data to identify redundant alarms triggered by the same cause and filter out redundant alarms, and adjust the alarm sequence for individual patients based on environmental alarm data to exclude false alarms caused by environmental factors.

[0051] As a further improvement of the technical solution, the alarm communication unit transmits the alarm information to relevant medical personnel by wireless transmission, including the following steps:

[0052] S3.1, extracting key information of the classified and filtered alarm signals, and formatting the alarm information into data packets;

[0053] S3.2, encrypting the data before transmission, packaging the alarm information into a data frame for wireless transmission, and sending the alarm information through wireless transmission;

[0054] S3.3, the device carried by the medical staff receives the data frame from the wireless network, parses the received data packet, extracts the alarm information, and displays the alarm information on the receiving device;

[0055] S3.4. The medical staff confirms the alarm, the system records the confirmation information, and closes the alarm according to the medical staff's needs.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. In the ICU medical equipment alarm management system based on the SEIPS model, the SEIPS model emphasizes the interaction between different medical work elements, aiming to improve the overall performance of medical services by improving the coordination and integration between the various parts of the system; by monitoring the patient's physiological parameters and using the optimized distribution threshold algorithm to dynamically adjust the alarm threshold, the individual differences of patients in the postoperative recovery period and changes in physiological status are taken into account, thereby reducing the occurrence of false positive alarms. In addition, by classifying and filtering the alarm signals, not only the accuracy of the alarm is improved, but also the correlation between alarms is identified through the relationship evaluation model and repeated search algorithm, which avoids redundant alarms and ensures that medical staff can pay attention to the situations that really need emergency treatment in a timely manner.

[0058] 2. The ICU medical equipment alarm management system based on the SEIPS model realizes efficient transmission of alarm information, ensuring that medical staff can quickly obtain key alarm information and respond. The system quickly and safely sends alarm information to medical staff through wireless transmission. Combined with the application of smart devices, medical staff can understand the patient's condition at the first time, which is crucial in the critical care environment where every second counts. In addition, the intelligence of the system is also reflected in the fact that it can dynamically adjust the alarm priority according to the load of the ICU and the urgency of the patient, which helps to reasonably allocate medical resources and improve treatment efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is the overall flow chart of the present invention;

[0060] The meaning of each number in the figure is:

[0061] 1. Alarm generation unit; 2. Alarm processing unit; 3. Alarm communication unit. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Example: See Figure 1 As shown, an ICU medical equipment alarm management system based on the SEIPS model is provided, including:

[0064] The alarm generating unit 1 monitors the physiological parameters of the patient and generates an alarm signal based on a distribution threshold algorithm, and optimizes the distribution threshold algorithm considering the different physiological parameters and postoperative recovery periods of different patients;

[0065] In this embodiment, the monitored physiological parameters include heart rate, blood pressure, and blood oxygen saturation;

[0066] The generation of an alarm signal based on a distribution threshold algorithm includes the following steps:

[0067] S1.1. Preprocess physiological parameters, including data cleaning;

[0068] S1.2. Use the distribution threshold algorithm to find the distribution pattern of normal physiological parameters, and set the threshold according to the distribution pattern. Considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized. Since different patients in the postoperative recovery period have physiological differences and their physiological states change over time, the distribution threshold algorithm is further optimized, wherein the physiological differences include the age, gender and weight of the patients;

[0069] Furthermore, the distributed threshold algorithm dynamically monitors the patient's physiological parameters and sets the alarm threshold based on the historical data of these parameters. This method can more accurately reflect the patient's current status. Traditional alarm systems often generate a large number of unnecessary alarms due to the use of fixed alarm thresholds, causing alarm fatigue for medical staff. The alarm system based on the distributed threshold algorithm reduces the generation of unnecessary alarms by optimizing the alarm generation logic, allowing medical staff to focus on those situations that really require emergency response. By reducing false alarms and missed alarms, the alarm system based on the distributed threshold algorithm improves the reliability of alarms, allowing medical staff to promptly notice situations that really require intervention, thereby improving patient safety. The distributed threshold algorithm is:

[0070] ;

[0071] The physiological parameters of patients in the postoperative recovery period may change with the progress of rehabilitation, and fixed thresholds are difficult to adapt to such changes. Therefore, the optimization algorithm aims to provide each patient with a personalized alarm threshold to make it more suitable for the individual's actual condition; by dynamically adjusting the threshold, it can reduce false alarms (i.e., situations where an alarm is issued but does not actually require immediate treatment) and missed alarms (i.e., situations where no alarm is issued but actually require emergency treatment) caused by improper threshold settings; the distribution threshold algorithm is optimized to ensure that the alarm signal is highly correlated with the patient's actual health status, so that medical staff can more quickly identify situations that really require emergency treatment; considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized as follows:

[0072] ;

[0073] Factors such as age, gender, and weight directly affect the normal range of the patient's physiological parameters. For example, the normal range of heart rate and blood pressure is different between children and adults, and the physiological indicators of men and women are also different. Therefore, incorporating these factors into the threshold setting of the alarm system can make the alarm more personalized and more in line with actual clinical needs; because of individual differences, the alarm system can more accurately determine when the alarm should be triggered. For example, a patient with a larger weight may have higher blood pressure in the early postoperative period, which may be a normal physiological reaction rather than an abnormal situation that requires immediate intervention. Therefore, personalized alarm settings can reduce the occurrence of false alarms; postoperative recovery is a dynamic process, and the patient's physical condition will gradually improve over time or deteriorate in some cases. Therefore, the threshold of the alarm system needs to be able to be dynamically adjusted according to the patient's status changes to ensure the timeliness and accuracy of the alarm; the optimized alarm system can more effectively identify which alarms really need the attention of medical staff, thereby reducing those false alarms or false positive alarms caused by not taking into account the patient's specific situation; because different patients in the postoperative recovery period have physiological differences and their physiological states change over time, the distribution threshold algorithm is further optimized as follows:

[0074] ;

[0075] ;

[0076] in, represents the threshold set point; It represents the set points of different optimized thresholds for different patients’ physiological parameters and postoperative recovery period; It represents the threshold setting point that is further optimized according to the physiological differences among different patients and the changes of their physiological status over time during the postoperative recovery period; Indicates physiological parameters; It represents the central value of normal physiological parameters. It is a representative value calculated by statistical methods based on the patient's historical data. timestamp indicating physiological parameters; represents the attenuation coefficient, which is used to determine the rate at which physiological parameters decay over time during the postoperative recovery period; represents the parameters that adjust for the effects of physiological differences among patients; to represent physiological differences (patient age, sex, weight); represents a time-based weight function, which is used to adjust the weights that change over time; represents the total number of physiological parameter measurements included in the dataset; index variables representing the physiological parameter measurements included in the dataset; represents the trend factor, i.e., the weight function, which depends on the rate of change And affects the final threshold setting, where ;

[0077] Furthermore, if there is a specific patient, his personal characteristics are: age is 40 years old, gender is male, and weight is 70 kg; can be represented as a vector, such as ; Then convert the gender into binary code (male = 0, female = 1), then ;like ; then the formula ; Therefore, the formula Some will consider the impact of these features and adjust the threshold setting;

[0078] Furthermore, if there is heart rate data of an ICU patient, and it is known that the patient is in the postoperative recovery period, the parameters used are: attenuation coefficient , patient attribute adjustment coefficient , change rate threshold , the weight when the rate of change threshold is exceeded ; If at a certain moment , the heart rate changes at a rate of 2 beats / minute, then ,at this time , therefore, the optimized distribution threshold algorithm can be expressed as: ;

[0079] S1.3. Compare the physiological parameters monitored in real time with the preset thresholds. When the monitored physiological parameters exceed the preset thresholds, an alarm signal is immediately generated (the alarm signal includes visual prompts (such as warning lights on the display), auditory prompts (such as beeps or voice prompts), and other forms of notifications (such as text messages, emails, etc.)).

[0080] After receiving the alarm signals from different devices, the alarm processing unit 2 classifies and filters the alarm signals;

[0081] In this embodiment, classifying and filtering the alarm signal includes the following steps:

[0082] S2.1. Receive alarm signals, and use the alarm signal set as a data set to extract key features from the alarm signals. The key features include alarm type (abnormal heart rate, abnormal blood pressure, etc.) and alarm level (minor, severe, etc.);

[0083] S2.2, using the alarm classification model to classify the alarm signals into different categories according to the key features of the alarm;

[0084] Among them, the role of the alarm classification model is to quickly and accurately identify which alarms are the most urgent after receiving a series of alarm signals, and classify them so that medical staff can give priority to the most important situations; automatically classify complex alarm information according to certain standards so that medical staff can quickly understand the importance and urgency of the alarm; by classifying the alarms, they can be sorted according to the severity and urgency of the alarms to ensure that the most critical situations are given priority; the classified alarms can be immediately sent to the corresponding medical staff, shortening the time difference from alarm generation to processing; using the alarm classification model to classify the alarm signals, and classifying the alarm signals into different categories according to the key features of the alarm, including the following steps:

[0085] S2.21. For each key feature, use the alarm index as the feature selection criterion, consider the impact of common alarm and rare alarm features on the alarm index, and then consider the impact of the dynamic nature of heart rate, blood oxygen saturation and blood pressure over time on the urgency of the alarm index, and optimize the alarm index;

[0086] Furthermore, the alarm index is:

[0087] ;

[0088] By optimizing the alarm index, the system can better identify those alarms that are rare but may lead to serious consequences, such as acute myocardial infarction, acute cerebrovascular events, etc.; for common alarms, such as abnormal heart rate and low blood oxygen saturation, although they need attention, they are usually not as immediately life-threatening as rare alarms. Optimizing the alarm index can help the system distinguish between these common alarms and rare alarms; by distinguishing between common alarms and rare alarms and assigning different weights to them, false alarms caused by common alarms can be reduced; the optimized alarm index can more accurately capture those alarms that really need to be handled immediately, thereby increasing the proportion of true positive alarms; consider the impact of common alarms (abnormal heart rate, low blood oxygen saturation and abnormal blood pressure alarms) and rare alarms (acute myocardial infarction, acute cerebrovascular events and severe respiratory failure alarms) on the alarm index;

[0089] The alarm index is:

[0090] ;

[0091] By analyzing the changing trends of alarm characteristics over time, it is possible to predict possible problems in advance and take preventive measures in time; short-term fluctuations in physiological parameters may not require immediate intervention, but if these parameters show a trend of continued deterioration, they need to be taken seriously. Optimizing the alarm index can better identify these trends; by considering the changes in alarm characteristics over time, it is possible to more accurately identify which alarms are caused by actual changes in physiological status, thereby reducing false alarms; the optimized alarm system can more accurately identify those situations that require immediate attention, especially when the patient's condition deteriorates significantly; the optimized alarm system can ensure that those alarms indicating that the patient's condition is deteriorating are handled first; because the dynamic nature of heart rate, blood oxygen saturation, and blood pressure over time affects the urgency of the alarm index, the alarm index is optimized as follows:

[0092] ;

[0093] in, represents a data set, i.e., a collection of alarm signals; Represents the alarm index. The smaller the value, the higher the purity of the data set. represents the alarm index after considering the characteristics of common alarms and rare alarms; An alarm index that takes into account the dynamics of common and rare alarm characteristics over time; Represents the total number of categories in the data set; Represents the category index; Indicates the total number of alarm records in the data set; Indicates the index of the alarm record; Indicates the current time; Indicates The time when each alarm record occurs; represents the base of natural logarithms; Indicates an indicator function. The function of this function is to return a binary value (binary value is 0 or 1) based on whether a certain condition is met. If equal ,but ,if Not equal to ,but ; Indicates the decay coefficient that controls the decay speed of the alarm record weight over time; Indicates The characteristic parameters of the class are used to adjust the weights of common and rare alarms. Rare alarms (such as acute myocardial infarction) can be weighted by setting a larger A value (such as 2) is used to ensure its importance in the alarm index. Since common alarms (such as abnormal heart rate) occur more frequently but may not be as urgent as rare alarms (such as acute myocardial infarction), a smaller value can be set. value (such as 0.5) to balance their impact on the alarm index; Indicates The alarm category to which the alarm record belongs (such as abnormal heart rate, abnormal blood pressure, etc.); is a way to express how weights change over time. and As the time difference between and When approaching (i.e. the time when the most recent alarm record occurred), is close to 1, which means that the most recent data has a greater impact on the results. and When the gap between the alarms becomes larger (i.e. the occurrence time of the earlier alarm records), will become very small, which means that older data has less impact on the result. This function measures the importance of observations in different time periods by introducing the time dimension, thus reflecting the dynamics of changes in physiological parameters such as heart rate, blood oxygen saturation and blood pressure over time. In simple terms, the closer the physiological parameter changes are to the current time, the greater their impact on the alarm index; conversely, the farther the physiological parameter changes are from the current time, the smaller their impact on the alarm index.

[0094] S2.22, select the feature with the lowest alarm index as the best split feature for the current node;

[0095] S2.23. Divide the data set into several subsets according to the different values ​​of the selected split feature (each time a split occurs, an optimal feature is selected to divide the data set. The optimal feature is selected based on the ability of the feature to distinguish different types of alarms to the greatest extent. The criterion used here is the alarm index. The purpose is to find a feature that maximizes information gain or minimizes the alarm index. Once the optimal split feature is selected, the entire data set is divided into several subsets according to the different values ​​of the feature (for example, heart rate is above or below a certain threshold). Each subset contains data that have the same value of the feature).

[0096] S2.24, continue to split the data in the subset until a predetermined stopping condition is reached, the predetermined stopping condition includes reaching a maximum depth, a minimum number of leaf nodes, or all instances belong to the same category, and identifying anomaly alarms through an alarm anomaly algorithm;

[0097] S2.25. Use the alarm classification model that has been built to classify the alarm signal. Input the key features of the alarm. The classification tree makes judgments along the path from the root node to a leaf node. The leaf node finally reached determines the category of the alarm signal (the classification tree is a tree structure composed of a series of nodes. The root node is at the top of the tree, indicating the initial feature selection point; the middle node represents the subsequent split point; the leaf node represents the final classification result; when a new alarm signal enters the system, it will start from the root node and move down step by step along the classification tree according to the key features of the alarm (such as heart rate, blood oxygen saturation, etc.). Each time it encounters a split node, it will decide which child node to move to according to the feature value corresponding to the node; eventually, the alarm signal will reach a leaf node in the classification tree, which represents the category of the alarm. For example, an alarm of abnormal heart rate may eventually be classified as "medium urgency" or "high urgency");

[0098] S2.3. Use the relationship evaluation model to evaluate the relationship between multiple alarms and determine the priority. Use the repeated search algorithm to find the correlation between different alarms and eliminate redundant alarms caused by the same reason.

[0099] Among them, the relationship evaluation model evaluates and determines the correlation between multiple alarms, and determines the priority of processing alarms accordingly; by evaluating the relationship between multiple alarms, it can be determined which alarms should be processed first, thereby optimizing the workflow of medical staff; through priority sorting, the system can ensure that urgent and important alarms are noticed and processed by medical staff first, thereby improving the response speed, giving priority to the most urgent alarms, ensuring that patients can receive medical intervention in time when needed, and reducing the risks caused by delays; through priority sorting, medical staff can allocate their workload more reasonably and avoid wasting time on non-urgent alarms. The relationship evaluation model can dynamically adjust the alarm priority based on historical data, making the system more intelligent and adaptive; using the relationship evaluation model to evaluate the relationship between multiple alarms and determine the priority includes the following steps:

[0100] S2.31. Receive classified and filtered alarm signals and pre-process the alarm signal data (clean and standardize the alarm data to remove duplicate records, missing values ​​or erroneous information);

[0101] S2.32. Determine the key attributes of the alarm signal, including the alarm type, alarm level, timestamp of the alarm occurrence, etc., and establish the alarm signal relationship matrix (create a matrix or graph in which each alarm is a node, the relationship between alarms is an edge, and the weight can represent the strength of the correlation between alarms);

[0102] S2.33. Collect historical alarm signal data, including the frequency, sequence, and simultaneous occurrence of alarms, and use the historical alarm signal data to estimate the probability distribution of each node in the alarm relationship matrix (through maximum likelihood estimation or other probability estimation techniques);

[0103] S2.34. When a new alarm signal appears, update the probability between alarm signals, determine the priority of the alarm according to the updated probability, and dynamically adjust the priority rules according to the current ICU load and the urgency of the patient. For example, during peak hours, the system may need to identify the highest priority alarms more quickly.

[0104] Furthermore, by using a repeated search algorithm, multiple alarms triggered by the same reason can be identified to avoid wasting resources. Potential correlations between different alarms can be found to help medical staff understand the patient's overall condition and provide a basis for diagnosis and treatment. By identifying the relationship between alarms, the alarm processing flow can be optimized and the overall performance of the alarm system can be improved. By identifying and filtering repeated alarms, unnecessary interference can be reduced, allowing medical staff to focus on dealing with situations that really need attention, avoiding multiple unnecessary treatments for the same problem, and saving medical resources. By identifying the correlation between alarms, it is possible to more accurately determine which alarms are false alarms caused by system failures or other non-emergency situations, and by excluding repeated alarms, ensure that alarms that really need to be processed can be discovered and processed in a timely manner. Using a repeated search algorithm to find the correlation between different alarms includes the following steps:

[0105] S2.35. Record the time, type, and level of all alarm signals, and determine the criteria for alarm signals to be considered repeated, such as multiple occurrences of the same type of alarm within a short period of time, and build an environmental alarm database to record the alarm conditions of all patients within the same time window;

[0106] S2.36. Set a time window within which alarm signals appearing are considered potential duplicate alarms, and identify alarm signals appearing within this time window;

[0107] S2.37, record the occurrence time and type of each alarm signal as the alarm sequence, use the sequence pattern mining algorithm (PrefixSpan) to discover frequent alarm sequences, and construct a multi-patient alarm sequence to integrate the alarm sequences of all patients in the same time window;

[0108] The data processed by the PrefixSpan algorithm is usually a sequence composed of item sets, each sequence represents a sequential list of events. In the ICU environment, each alarm sequence is an item set sequence, in which each item represents an alarm event. A frequent pattern refers to an item set sequence that appears at least a specified number of times (minimum support threshold) in one or more sequences. Support refers to the frequency of a pattern appearing in all sequences, and confidence refers to the probability of a conclusion appearing when the premise is known.

[0109] S2.38. For the mined frequent alarm sequences, evaluate the support and confidence of the frequent alarm sequence association rules, and select the rules with higher support and confidence as the alarm association rules;

[0110] S2.39, apply correlation rules to alarm signal data, identify redundant alarms triggered by the same cause, filter out redundant alarms, and adjust the alarm sequence of a single patient based on environmental alarm data to exclude false alarms caused by environmental factors;

[0111] S2.4. Summarize the classified and filtered alarm signals and send them to the alarm communication unit 3.

[0112] The alarm communication unit 3 receives the screened alarm signal and transmits the alarm information to the relevant medical staff through wireless transmission;

[0113] In this embodiment, the alarm information can be sent to various mobile devices, such as smart phones, PDAs or dedicated alarm receivers, through wireless transmission, so that medical staff can receive information at any time. Wireless transmission is not limited by physical distance and can cover a wider area, including medical staff inside and outside the hospital. The wireless transmission method allows the alarm information to be transmitted to the medical staff's equipment in almost real time, ensuring that they can respond quickly. Compared with traditional wired transmission or manual notification, wireless transmission greatly reduces the time delay of information transmission. Encrypting data before wireless transmission can protect the security of alarm information and prevent sensitive information from being leaked. The alarm information is transmitted to relevant medical staff through wireless transmission, including the following steps:

[0114] S3.1. Extract the key information of the classified and filtered alarm signals, including alarm type, level, patient ID, and alarm time, and format the alarm information into data packets to ensure that the information is complete and easy to parse;

[0115] S3.2, encrypt the data before transmission to protect the security of the alarm information, package the alarm information into data frames for wireless transmission, and send the alarm information through wireless transmission (LoRaWAN);

[0116] S3.3. The device carried by the medical staff (such as a smart phone, PDA, dedicated alarm receiver, etc.) receives data frames from the wireless network. If the data is encrypted during transmission, the receiving end needs to decrypt the data, parse the received data packets, extract the alarm information, and display the alarm information on the receiving device in the form of text messages, sound prompts, vibration reminders, or other forms of notifications;

[0117] S3.4. The medical staff confirms the alarm, the system records the confirmation information, and closes the alarm or takes the next step according to the needs of the medical staff.

[0118] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. ICU medical equipment alarm management system based on SEIPS model, characterized by: include: An alarm generating unit (1), wherein the alarm generating unit (1) monitors the physiological parameters of the patient and generates an alarm signal based on a distribution threshold algorithm, and optimizes the distribution threshold algorithm taking into account the physiological parameters of different patients and the different recovery periods after surgery; An alarm processing unit (2), wherein the alarm processing unit (2) receives alarm signals from different devices and classifies and filters the alarm signals; An alarm communication unit (3), the alarm communication unit (3) receiving the screened alarm signal and transmitting the alarm information to medical personnel via wireless transmission; The physiological parameters monitored by the alarm generation unit (1) include heart rate, blood pressure and blood oxygen saturation; The alarm generating unit (1) generates an alarm signal based on a distribution threshold algorithm, comprising the following steps: S1.1, preprocessing of physiological parameters; S1.

2. Use the distribution threshold algorithm to discover the distribution pattern of normal physiological parameters, and set the threshold according to this distribution pattern. Considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized. Since different patients in the postoperative recovery period have physiological differences and their physiological states change over time, the distribution threshold algorithm is further optimized; S1.3, comparing the physiological parameters monitored in real time with the preset thresholds, and immediately generating an alarm signal when the monitored physiological parameters exceed the preset thresholds; In S1.2, the distribution threshold algorithm is: ； Considering the different physiological parameters and postoperative recovery periods of different patients, the distribution threshold algorithm is optimized as follows: ； Since different patients have physiological differences during the postoperative recovery period and their physiological states change over time, the distribution threshold algorithm is further optimized as follows: ； in, represents the threshold set point; It represents the set points of different optimized thresholds for different patients’ physiological parameters and postoperative recovery period; It represents the threshold setting point that is further optimized according to the physiological differences among different patients and the changes of their physiological status over time during the postoperative recovery period; Indicates physiological parameters; Indicates the central value of normal physiological parameters; timestamp indicating physiological parameters; represents the attenuation coefficient, which is used to determine the rate at which physiological parameters decay over time during the postoperative recovery period; represents the parameters that adjust for the effects of physiological differences among patients; Indicates physiological differences among patients; represents a time-based weight function, which is used to adjust the weights that change over time; represents the trend factor; represents the total number of physiological parameter measurements included in the dataset; An index variable representing the physiological parameter measurements contained in the dataset.

2. The ICU medical equipment alarm management system based on the SEIPS model according to claim 1, characterized in that: The alarm processing unit (2) classifies and filters the alarm signal, including the following steps: S2.1, receiving an alarm signal, and taking a set of the alarm signals as a data set, and extracting key features from the alarm signal; S2.2, using the alarm classification model to classify the alarm signals into different categories according to the key features of the alarm; S2.

3. Use the relationship evaluation model to evaluate the relationship between multiple alarms and determine the priority. Use the repeated search algorithm to find the correlation between different alarms and eliminate redundant alarms caused by the same reason. S2.

4. Summarize the classified and filtered alarm signals and send them to the alarm communication unit (3).

3. The ICU medical equipment alarm management system based on the SEIPS model according to claim 2 is characterized in that: In S2.2, the alarm classification model is used to classify the alarm signal, and the alarm signal is classified into different categories according to the key features of the alarm, including the following steps: S2.

21. For each key feature, use the alarm index as the feature selection criterion, consider the impact of common alarm and rare alarm features on the alarm index, and then consider the impact of the dynamic nature of heart rate, blood oxygen saturation and blood pressure over time on the urgency of the alarm index, and optimize the alarm index; S2.22, select the feature with the lowest alarm index as the best split feature for the current node; S2.23, dividing the data set into several subsets according to different values ​​of the selected splitting feature; S2.24, continue to split the data in the subset until a predetermined stop condition is reached, and identify abnormal alarms through an alarm anomaly algorithm; S2.

25. Use the constructed alarm classification model to classify the alarm signal. Input the key features of the alarm. The classification tree makes judgments along the path from the root node to a leaf node. The leaf node finally reached determines the category of the alarm signal.

4. The ICU medical equipment alarm management system based on the SEIPS model according to claim 3 is characterized in that: In S2.21, the alarm index is: ； Consider the impact of common and rare alarms on the alarm index. Common alarms include abnormal heart rate, low blood oxygen saturation, and abnormal blood pressure alarms. Rare alarms include acute myocardial infarction, acute cerebrovascular event, and severe respiratory failure alarms. The alarm index is: ； Since the dynamic changes of heart rate, blood oxygen saturation and blood pressure over time affect the urgency of the alarm index, the alarm index is optimized as follows: ； in, Represents a dataset; Indicates the alarm index; represents the alarm index after considering the characteristics of common alarms and rare alarms; An alarm index that takes into account the dynamics of common and rare alarm characteristics over time; Represents the total number of categories in the data set; Represents the category index; Indicates the total number of alarm records in the data set; Indicates the index of the alarm record; Indicates the current time; Indicates The occurrence time of each alarm record; Indicates the decay coefficient that controls the decay speed of the alarm record weight over time; Indicates The characteristic parameters of the class are used to adjust the weights of common and rare alarms; Indicates The alarm category to which the alarm record belongs; represents the base of natural logarithms; Represents the indicator function.

5. The ICU medical equipment alarm management system based on the SEIPS model according to claim 4 is characterized in that: In S2.3, a relationship assessment model is used to assess the relationship between multiple alarms and determine priorities, including the following steps: S2.31, receiving the classified and filtered alarm signal, and preprocessing the alarm signal data; S2.

32. Determine the key attributes of the alarm signal and establish an alarm signal relationship matrix; S2.33, collect historical alarm signal data, and use the historical alarm signal data to estimate the probability distribution of each node in the alarm relationship matrix; S2.

34. When a new alarm signal appears, the probability between the alarm signals is updated, the priority of the alarm is determined based on the updated probability, and the priority rules are dynamically adjusted based on the current ICU load and the patient's urgency factors.

6. The ICU medical equipment alarm management system based on the SEIPS model according to claim 5 is characterized in that: In S2.3, a repeated search algorithm is used to find the correlation between different alarms, including the following steps: S2.

35. Record the time, type, and level of all alarm signals, determine the criteria for alarm signals to be considered duplicates, and build an environmental alarm database to record the alarm conditions of all patients within the same time window; S2.

36. Set a time window within which alarm signals appearing are considered potential duplicate alarms, and identify alarm signals appearing within this time window; S2.37, record the occurrence time and type of each alarm signal as an alarm sequence, use a sequence pattern mining algorithm to discover frequent alarm sequences, and construct a multi-patient alarm sequence to integrate the alarm sequences of all patients in the same time window; S2.

38. For the mined frequent alarm sequences, evaluate the support and confidence of the frequent alarm sequence association rules, and select the rules with higher support and confidence as the alarm association rules; S2.

39. Apply correlation rules to alarm signal data to identify redundant alarms triggered by the same cause and filter out redundant alarms, and adjust the alarm sequence for individual patients based on environmental alarm data to exclude false alarms caused by environmental factors.

7. The ICU medical equipment alarm management system based on the SEIPS model according to claim 6 is characterized in that: The alarm communication unit (3) transmits the alarm information to relevant medical personnel via wireless transmission, comprising the following steps: S3.1, extracting key information of the classified and filtered alarm signals, and formatting the alarm information into data packets; S3.2, encrypting the data before transmission, packaging the alarm information into a data frame for wireless transmission, and sending the alarm information through wireless transmission; S3.3, the device carried by the medical staff receives the data frame from the wireless network, parses the received data packet, extracts the alarm information, and displays the alarm information on the receiving device; S3.

4. The medical staff confirms the alarm, the system records the confirmation information, and closes the alarm according to the medical staff's needs.

Images