Catheter-related complication risk early warning information system based on intelligent AI

Abstract

The invention discloses a catheter-related complication risk early warning information system based on intelligent AI, and belongs to the technical field of catheter-related complication, and the system comprises a data collection module which is used for collecting multi-modal data of a patient in real time; the data processing module is used for processing the multi-modal data of the patient and determining multi-modal characteristic data of the patient; the risk identification module is used for analyzing the multi-modal characteristic data of the patient and determining a risk identification result of catheter-related complications; and the early warning and intervention module is used for carrying out early warning and intervention on the patient in time according to the catheter related complication risk identification result. The problems that in the prior art, catheter related complication risks cannot be recognized in advance, early warning intervention cannot be conducted in time, catheter use safety is low, and the complication occurrence rate cannot be effectively reduced are solved. The catheter related complication risk can be recognized in advance, early warning intervention can be conducted in time, the use safety of the catheter can be improved, and the complication occurrence rate can be effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of catheter-related complications, in particular to an intelligent AI-based catheter-related complication risk early warning information system. BACKGROUND

[0002] Catheter-related complications are various complications caused by catheter puncture placement and long-term indwelling in the body. For example, puncture injury of blood vessels, pleura, nerves, catheter-related infection, thrombus, phlebitis, and catheter obstruction and shedding, etc. Not only affect the health of patients, but also increase the medical burden.

[0003] Traditional catheter-related complication monitoring methods rely on the experience of medical staff, and have hysteresis and subjectivity, and cannot identify catheter-related complication risks in advance and timely early warning intervention, resulting in low catheter use safety and ineffective reduction of complication incidence. SUMMARY

[0004] The purpose of the present application is to provide an intelligent AI-based catheter-related complication risk early warning information system, which can identify catheter-related complication risks in advance and timely early warning intervention, improve catheter use safety and effectively reduce complication incidence, and solve the problems raised in the above background technology.

[0005] To achieve the above purpose, the present application provides the following technical solutions: The intelligent AI-based catheter-related complication risk early warning information system comprises: A data acquisition module for monitoring patient physiological indicators, patient catheter status and patient laboratory test results, and acquiring patient multi-modal data in real time; A data processing module for processing patient multi-modal data and extracting key features valuable for catheter-related complication risk early warning to determine patient multi-modal feature data; A risk identification module for analyzing patient multi-modal feature data, identifying patient catheter-related complication risks, and determining catheter-related complication risk identification results; An early warning intervention module for timely early warning and intervention of patients according to catheter-related complication risk identification results.

[0006] Preferably, the patient multi-modal data is processed by performing the following operations: Integrate the patient multi-modal data, integrate the patient multi-modal data from different sources into a unified data view, and verify the integrity of the integrated patient multi-modal data; Feature extraction is performed on the patient multi-modal data to extract features valuable for catheter-related complication risk early warning from the patient multi-modal data to determine patient multi-modal feature data.

[0007] Preferably, the patient multi-modal feature data is analyzed to identify the patient catheter-related complication risk, and the following operations are performed: According to the catheter-related complication risk early warning demand based on intelligent AI, a catheter-related complication risk identification model is established, and the catheter-related complication risk identification model is deployed in an actual catheter-related complication risk identification environment; The patient multi-modal feature data is input into the catheter-related complication risk identification model, the patient multi-modal feature data is analyzed according to the catheter-related complication risk identification model, and the catheter-related complication risk is effectively identified, so as to determine the catheter-related complication risk identification result.

[0008] Preferably, the catheter-related complication risk identification model is established, and the following operations are performed: The patient multi-modal historical data is collected, and the collected patient multi-modal historical data is divided to determine the training set and the test set; Based on the deep learning technology, the training set is used to train the deep learning model, so that the deep learning model learns the catheter-related complication risk identification behavior from the training set, effectively identifies the catheter-related complication risk, and determines the deep learning-based catheter-related complication risk identification model; The test set is input into the deep learning-based catheter-related complication risk identification model, the deep learning-based catheter-related complication risk identification model is tested according to the test set, and the performance of the deep learning-based catheter-related complication risk identification model is evaluated to determine the model test evaluation result; According to the model test evaluation result, the parameters of the deep learning-based catheter-related complication risk identification model are adjusted and optimized until the deep learning-based catheter-related complication risk identification model can achieve the expected effect of effectively identifying the catheter-related complication risk, and then the optimal catheter-related complication risk identification model is determined.

[0009] Preferably, according to the catheter-related complication risk identification result, timely warning and intervention management scheme are provided, medical staff are prompted to timely intervene and manage the patients with catheter-related complication risk, and the patients after intervention management are monitored in real time, the intervention management scheme is adjusted according to the monitoring feedback, and closed-loop management of the patients is formed.

[0010] Preferably, the patient physiological indicators, patient catheter state and patient laboratory examination results are monitored, the patient multi-modal data is collected in real time, and the following operations are performed: The heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, blood glucose, electrolyte and electroencephalogram of the patient are monitored in real time based on sensors and Internet of Things devices, and the physiological indicators of the patient are collected. Based on the sensors and Internet of Things devices, the catheter position, catheter flow rate and catheter pressure of the patient are monitored in real time, and the patient catheter state is collected; Based on the sensors and Internet of Things devices, the blood indicators, creatinine levels, liver function indicators, urine indicators and pathogenic microorganism detection of the patient are monitored in real time, and the patient laboratory examination results are collected; Among them, according to the patient physiological indicators, patient catheter state and patient laboratory examination results, the patient multi-modal data is determined.

[0011] Preferably, the patient multi-modal data is processed to perform the following operations: The patient multi-modal data is cleaned to remove noise in the patient multi-modal data, and abnormal values in the patient multi-modal data that are not valuable for catheter-related complication risk warning are deleted; The patient multi-modal data is normalized to convert the patient multi-modal data into a unified data format, remove dimensional differences in the patient multi-modal data, and form standardized patient multi-modal data.

[0012] Preferably, the integrated patient multi-modal data is verified for integrity, and the following operations are performed: The integrated patient multi-modal data is compared with the patient multi-modal data before integration one by one, the similarity between the integrated patient multi-modal data and the patient multi-modal data before integration is analyzed, and whether the integrated patient multi-modal data has data integrity is evaluated; When the integrated patient multi-modal data is the same as the patient multi-modal data before integration, the integrated patient multi-modal data has data integrity, and the integrated patient multi-modal data is stored safely at this time; When the integrated patient multi-modal data is different from the patient multi-modal data before integration, the integrated patient multi-modal data does not have data integrity, and the missing data in the integrated patient multi-modal data is searched and recorded at this time, so that the integrated patient multi-modal data has data integrity.

[0013] Preferably, the patient multi-modal feature data is determined by extracting features valuable for catheter-related complication risk warning from the patient multi-modal data, including: The patient multi-modal data is analyzed to obtain a plurality of to-be-extracted features; A universal feature set and a customized feature set are constructed respectively; wherein the universal feature set includes a first feature which is valuable for catheter-related complication risk warning; and a second feature which is valuable for catheter-related complication risk warning and is customized for the individual situation of the patient; The extraction indication value of each to-be-extracted feature is calculated by the following algorithm: wherein, is an extraction indication value of the feature to be extracted, is a first intermediate variable, is a first preset weight corresponding to the first intermediate variable, is a second intermediate variable, is a second preset weight corresponding to the second intermediate variable, is a first maximum similarity between the feature to be extracted and a first feature in the universal feature set, is a first preset similarity threshold, is a second maximum similarity between the feature to be extracted and a second feature in the customized feature set, is a second preset similarity threshold. When the extraction indication value exceeds the indication value threshold, the corresponding feature to be extracted is considered as a feature valuable for early warning of catheter-related complication risk, and is extracted as patient multi-modal feature data.

[0014] Preferably, the intelligent AI-based catheter-related complication risk early warning information system further comprises: a tracking management module configured to: track a follow-up situation of early warning and intervention on the patient according to the catheter-related complication risk identification result; when the follow-up situation indicates early warning abnormalities or intervention abnormalities, configure an explainability capability of the catheter-related complication risk identification model; input the follow-up situation and the early warning abnormalities or intervention abnormalities indicated by the follow-up situation into the catheter-related complication risk identification model after the explainability capability configuration, and obtain an explanatory content output by the catheter-related complication risk identification model after the explainability capability configuration; extract valuable content from the patient multi-modal data based on the explanatory content; based on the valuable content, control the catheter-related complication risk identification model to re-identify the catheter-related complication risk, and based on a new identification result, re-early warn and intervene on the patient.

[0015] Compared with the prior art, the present application has the following advantages: The application monitors physiological indicators of patients, states of patient catheters and laboratory test results of patients through sensors and Internet of Things devices, thereby collecting patient multi-modal data in real time, processes the patient multi-modal data, extracts key features valuable for early warning of catheter-related complication risks, determines patient multi-modal feature data, analyzes the patient multi-modal feature data through the establishment of a catheter-related complication risk identification model, identifies the catheter-related complication risks of patients, thereby determines the catheter-related complication risk identification results, timely early warns and provides intervention management schemes according to the catheter-related complication risk identification results, prompts medical staff to timely intervene and manage patients with catheter-related complication risks, and real-time monitors the patients after intervention management, adjusts the intervention management schemes according to the monitoring conditions, thereby forms a closed-loop management of patients, can early identify and timely early warn and intervene the catheter-related complication risks, can improve the safety of catheter use and effectively reduce the incidence of complications.

[0016] Through the double-feature set cooperative screening mechanism, the group common risk and the individual specific index are considered, the universal feature set covers extensive complication data, the customized feature set captures the unique risk of patients, and the false negative rate is significantly reduced. Secondly, the first preset weight and the second preset weight can be adjusted according to the patient type (such as focusing on universal features for ordinary patients and increasing the weight of customized features for chronic disease patients), so as to avoid interference of invalid features. In addition, only high-value features are retained, the calculation load of the subsequent risk identification module is reduced, and the real-time early warning efficiency is ensured.

[0017] The model is reversely optimized through the actual intervention effect, the false positives and false negatives are reduced, the early warning accuracy is improved, the explainable content directly shows the AI decision basis, assists medical staff in verifying the logic, promotes human-machine cooperation, the scheme is dynamically adjusted according to the patient response, a cycle of early warning, intervention and re-evaluation is formed, the value content is continuously expanded to train the data set, the model can adapt to new risk patterns, and the system life cycle is prolonged. BRIEF DESCRIPTION OF DRAWINGS

[0018] Fig. 1 A module diagram of the catheter-related complication risk early warning information system of the application; Fig. 2 A flowchart of the catheter-related complication risk early warning information system of the application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0020] In order to solve the problem that the existing catheter-related complications risk cannot be identified in advance and timely warning intervention is not performed, leading to low catheter use safety and being unable to effectively reduce the incidence of complications, please refer to Figs. 1-2 The embodiment provides the following technical scheme: The catheter-related complication risk warning information system based on intelligent AI comprises a data acquisition module, a data processing module, a risk identification module and a warning intervention module.

[0021] Specifically, through the interactive communication between the data acquisition module, the data processing module, the risk identification module and the warning intervention module, the catheter-related complication risk can be identified in advance and timely warning intervention can be performed, the catheter use safety can be improved, and the incidence of complications can be effectively reduced.

[0022] The data acquisition module is used for monitoring physiological indexes of a patient, a catheter state of the patient and laboratory examination results of the patient, and collecting multi-modal data of the patient in real time.

[0023] In the embodiment, the physiological indexes of the patient, the catheter state of the patient and the laboratory examination results of the patient are monitored, and the multi-modal data of the patient is collected in real time, and the following operations are performed: The heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, blood glucose, electrolyte and electroencephalogram of the patient are monitored in real time based on sensors and Internet of Things devices, and the physiological indexes of the patient are collected. It should be noted that the heart rate refers to the frequency of heartbeats, which is used to evaluate the function of the heart; the blood pressure refers to the systolic pressure and diastolic pressure, which reflects the state of the cardiovascular system; the blood oxygen saturation refers to the evaluation of the oxygen-carrying capacity of blood, which reflects the respiratory and circulatory functions; the body temperature refers to the temperature in the body, which is convenient for timely discovery of infection or inflammation; the respiratory rate refers to the evaluation of the state of the respiratory system, which reflects whether the respiratory function is normal; the blood glucose refers to the blood glucose level, which is used for the risk assessment of diabetes patients or catheter-related infections; the electrolyte such as sodium, potassium and calcium reflects the body fluid balance and organ function; the electroencephalogram is used to evaluate the brain function and monitor the nervous system complications.

[0024] The catheter position, catheter flow rate and catheter pressure of the patient are monitored in real time based on sensors and Internet of Things devices, and the catheter state of the patient is collected. It should be noted that by monitoring the catheter position, the position change can be discovered in time to prevent the catheter from being displaced or pulled out; by monitoring the catheter flow rate, the patency of the catheter can be ensured; and by monitoring the catheter pressure, the hemodynamic state can be evaluated.

[0025] The blood index, creatinine level, liver function index, urine index and pathogenic microorganism detection situation of the patient are monitored in real time based on sensors and Internet of Things devices, and the laboratory examination results of the patient are collected. It should be noted that the blood indicators include white blood cell count and C-reactive protein, wherein the white blood cell count reflects the risk of infection, and the C-reactive protein is used to assess the inflammatory response; the creatinine level reflects the kidney function; the liver function indicators such as bilirubin and transaminase are used to assess the liver function status; the urine indicators include urine routine and urine microalbumin, wherein the urine routine is used to detect urinary tract infection or metabolic disorder, and the urine microalbumin is used to assess kidney damage; pathogenic microorganism detection such as bacterial culture, by detecting pathogenic bacteria in blood, catheter tip and other samples, for infection diagnosis.

[0026] The patient multi-modal data is determined according to the patient physiological indicators, the patient catheter state and the patient laboratory examination results.

[0027] Therefore, the patient physiological indicators, the patient catheter state and the patient laboratory examination results are monitored by the sensors and the Internet of Things devices, so as to collect the patient multi-modal data in real time, and provide data support for subsequent identification of the patient catheter-related complication risk.

[0028] The data processing module is used to process the patient multi-modal data, and extract key features valuable for catheter-related complication risk warning to determine the patient multi-modal feature data.

[0029] In this embodiment, the patient multi-modal data is processed to perform the following operations: The patient multi-modal data is cleaned to remove noise in the patient multi-modal data, and delete abnormal values in the patient multi-modal data which are not valuable for catheter-related complication risk warning, so as to improve the data quality of the patient multi-modal data; The patient multi-modal data is normalized to convert the patient multi-modal data into a unified data format, remove dimensional differences in the patient multi-modal data, and form standardized patient multi-modal data, so as to facilitate subsequent analysis of the patient multi-modal data; The patient multi-modal data is integrated to integrate patient multi-modal data from different sources into a unified data view, and the integrated patient multi-modal data is verified for integrity, so as to fully guarantee the data integrity of the patient multi-modal data; The patient multi-modal data is processed to extract features valuable for catheter-related complication risk warning from the patient multi-modal data to determine the patient multi-modal feature data.

[0030] It should be noted that the patient multi-modal data is processed, and key features valuable for catheter-related complication risk warning are extracted to determine the patient multi-modal feature data, so as to facilitate subsequent effective identification of the patient catheter-related complication risk.

[0031] In this embodiment, the integrated patient multi-modal data is verified for integrity, and the following operations are performed: The integrated patient multi-modal data is compared with the pre-integrated patient multi-modal data one by one, the similarity between the integrated patient multi-modal data and the pre-integrated patient multi-modal data is analyzed, and whether the integrated patient multi-modal data has data integrity is evaluated; When the integrated patient multi-modal data is the same as the pre-integrated patient multi-modal data, the integrated patient multi-modal data has data integrity, and the integrated patient multi-modal data is stored safely at this time; When the integrated patient multi-modal data is different from the pre-integrated patient multi-modal data, the integrated patient multi-modal data does not have data integrity, and the missing data in the integrated patient multi-modal data is searched and recorded at this time, so that the integrated patient multi-modal data has data integrity.

[0032] The risk identification module is configured to analyze the patient multi-modal feature data, identify the catheter-related complication risk of the patient, and determine the catheter-related complication risk identification result.

[0033] In this embodiment, the patient multi-modal feature data is analyzed to identify the catheter-related complication risk of the patient, and the following operations are performed: According to the catheter-related complication risk early warning requirement based on intelligent AI, a catheter-related complication risk identification model is established; The patient multi-modal historical data is collected, and the collected patient multi-modal historical data is divided to determine the training set and the test set; Based on deep learning technology, the training set is used to train the deep learning model, so that the deep learning model learns the catheter-related complication risk identification behavior from the training set, effectively identifies the catheter-related complication risk, and thus determines the deep learning-based catheter-related complication risk identification model; The test set is input into the deep learning-based catheter-related complication risk identification model, the deep learning-based catheter-related complication risk identification model is tested according to the test set, and the performance of the deep learning-based catheter-related complication risk identification model is evaluated to determine the model test evaluation result; According to the model test evaluation result, the parameters of the deep learning-based catheter-related complication risk identification model are adjusted and optimized until the deep learning-based catheter-related complication risk identification model can achieve the expected effect of effectively identifying the catheter-related complication risk, and thus the optimal catheter-related complication risk identification model is determined; The optimal catheter-related complication risk identification model is deployed, and the catheter-related complication risk identification model is deployed in the actual catheter-related complication risk identification environment; The patient multi-modal feature data is input into the catheter-related complication risk identification model, the patient multi-modal feature data is analyzed according to the catheter-related complication risk identification model, and the catheter-related complication risk is effectively identified, so as to determine the catheter-related complication risk identification result.

[0034] The early warning intervention module is used for timely early warning and intervention of the patient according to the catheter-related complication risk identification result.

[0035] In this embodiment, the timely early warning and intervention management scheme are provided according to the catheter-related complication risk identification result, the medical staff is prompted to timely intervene and manage the patient with the catheter-related complication risk, and the patient after the intervention management is monitored in real time, the intervention management scheme is adjusted according to the monitoring feedback, and the closed-loop management of the patient is formed.

[0036] In summary, the physiological indicators of the patient, the catheter state of the patient and the laboratory examination results of the patient are monitored through the sensors and the Internet of Things devices, so as to collect the multi-modal data of the patient in real time, the key features valuable for the early warning of the catheter-related complication risk are extracted through the processing of the multi-modal data of the patient, the multi-modal feature data of the patient is determined, the catheter-related complication risk of the patient is identified through the establishment of the catheter-related complication risk identification model and the analysis of the multi-modal feature data of the patient, and the catheter-related complication risk identification result is determined. The timely early warning and intervention management scheme are provided according to the catheter-related complication risk identification result, the medical staff is prompted to timely intervene and manage the patient with the catheter-related complication risk, and the patient after the intervention management is monitored in real time, the intervention management scheme is adjusted according to the monitoring feedback, and the closed-loop management of the patient is formed. The catheter-related complication risk can be identified in advance and timely early warning intervention can be performed, the safety of catheter use can be improved, and the incidence of complications can be effectively reduced.

[0037] In this embodiment, the features valuable for the early warning of the catheter-related complication risk are extracted from the multi-modal data of the patient, and the multi-modal feature data of the patient is determined, including: The multi-modal data of the patient is analyzed, and a plurality of to-be-extracted features are obtained; A universal feature set and a customized feature set are respectively constructed; the universal feature set includes a first feature which is valuable for the early warning of the catheter-related complication risk; and the customized feature set includes a second feature which is valuable for the early warning of the catheter-related complication risk and is customized for the individual situation of the patient; The extraction indication value of each to-be-extracted feature is calculated by the following algorithm: wherein, is an extraction indication value of a feature to be extracted, is a first intermediate variable, is a first preset weight corresponding to the first intermediate variable, is a second intermediate variable, is a second preset weight corresponding to the second intermediate variable, is a first maximum similarity between the feature to be extracted and a first feature in the universal feature set, is a first preset similarity threshold, is a second maximum similarity between the feature to be extracted and a second feature in the customized feature set, is a second preset similarity threshold; When the extraction indication value exceeds the indication value threshold, the corresponding feature to be extracted is considered as a feature valuable for the catheter-related complication risk warning, and is extracted as the patient multi-modal feature data.

[0038] The working principle and beneficial effects of the above technical solution are: The first feature with universal value for the catheter-related complication risk warning refers to a feature with value for the catheter-related complication risk warning of different patients. For example, such first feature can be catheter pressure sudden rise, abnormal white blood cell count, etc. The second feature with value for the catheter-related complication risk warning customized for the individual condition of the patient refers to a feature with value for the catheter-related complication risk warning of the patient itself. For example, such second feature can be a blood glucose fluctuation pattern specific to diabetic patients, etc.

[0039] The first maximum similarity between the feature to be extracted and the first feature is calculated. If it is greater than the first similarity threshold, it means that there is a certain universal value, and the first maximum similarity is assigned to the first intermediate variable for the calculation of the extraction indication value. Otherwise, it means that there is no universal value, and 0 is assigned to the first intermediate variable. Similarly, the second maximum similarity between the feature to be extracted and the second feature is calculated. If it is greater than the second similarity threshold, it means that there is a certain customized value, and the second maximum similarity is assigned to the second intermediate variable for the calculation of the extraction indication value. Otherwise, it means that there is no customized value, and 0 is assigned to the second intermediate variable.

[0040] Then, the first intermediate variable and the second intermediate variable are weighted and summed. Only the feature to be extracted with the extraction indication value exceeding the indication value threshold is retained, and is extracted as the patient multi-modal feature data.

[0041] The embodiment of the present application considers the group common risk and individual specific indicators through the double feature set cooperative screening mechanism, covers a wide range of complications data through the universal feature set, captures the unique risk of the patient through the customized feature set, and significantly reduces the false negative rate. Secondly, the first preset weight and the second preset weight can be adjusted according to the patient type (such as focusing on the universal features for ordinary patients and increasing the weight of the customized features for chronic disease patients), so as to avoid the interference of invalid features. In addition, only high-value features are retained, the calculation load of the subsequent risk identification module is reduced, and the real-time early warning efficiency is ensured.

[0042] In the embodiment, the intelligent AI-based catheter-related complication risk early warning information system further includes: The tracking management module is configured to: Track the subsequent situation of early warning and intervention on the patient according to the catheter-related complication risk identification result; When the subsequent situation represents early warning abnormalities or intervention abnormalities, configure the explainability of the catheter-related complication risk identification model; Input the subsequent situation and the early warning abnormalities or intervention abnormalities represented by the subsequent situation into the catheter-related complication risk identification model after the explainability configuration, and obtain the explanatory content output by the catheter-related complication risk identification model after the explainability configuration; Based on the explanatory content, extract the valuable content from the patient multi-modal data; Based on the valuable content, control the catheter-related complication risk identification model to re-identify the catheter-related complication risk, and based on the new identification result, re-warn and intervene the patient.

[0043] The working principle and beneficial effects of the above technical solution are: The system continuously tracks the intervention effect of the patient after early warning, and when early warning abnormalities (such as false positives) or intervention abnormalities (such as unrelieved complications) are found, an explanation algorithm such as SHAP (SHapley Additive exPlanations, a machine learning model explanation tool based on game theory Shapley value) is injected into the risk identification model to analyze the decision logic and realize the explainability configuration of the catheter-related complication risk identification model. The subsequent situation and the early warning abnormalities or intervention abnormalities represented by the subsequent situation are input into the catheter-related complication risk identification model after the explainability configuration, and the explanatory content output by the catheter-related complication risk identification model after the explainability configuration is obtained. For example: the thrombosis risk early warning is triggered by the sudden rise of catheter pressure and the sudden drop of platelets, but the patient did not use anticoagulant drugs that day. Based on this, the valuable content is extracted from the original multi-modal data, such as coagulation indicators and medication records in the relevant period. The valuable content is used to retrain the model to generate a new early warning result and dynamically adjust the intervention scheme.

[0044] The embodiments of the present application reduce false positives and false negatives by actually intervening to optimize the model, improve the accuracy of early warning, and intuitively display the AI decision basis to assist medical staff in verifying the logic and promoting human-machine collaboration. The scheme is dynamically adjusted according to the patient response, forming a cycle of early warning, intervention, and re-evaluation, and the value content continuously expands the training data set, which can adapt the model to new risk patterns and extend the system life cycle.

[0045] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0046] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. The AI-based catheter-related complication risk warning information system is characterized by: include: The data acquisition module is used to monitor the patient's physiological indicators, the patient's catheter status and the patient's laboratory test results, and collect the patient's multimodal data in real time; A data processing module is used to process the patient's multimodal data, extract key features that are valuable for warning of catheter-related complication risks, and determine the patient's multimodal feature data; A risk identification module is used to analyze the patient's multimodal feature data, identify the patient's catheter-related complication risk, and determine the catheter-related complication risk identification results; The early warning and intervention module is used to provide timely early warning and intervention to patients based on the results of catheter-related complication risk identification.

2. The intelligent AI-based catheter-related complication risk warning information system according to claim 1, characterized in that: Process the patient's multimodal data and perform the following operations: Integrate multimodal patient data from different sources into a unified data view and verify the integrity of the integrated multimodal patient data; Feature extraction is performed on the patient's multimodal data, and features valuable for warning of catheter-related complication risks are extracted from the patient's multimodal data to determine the patient's multimodal feature data.

3. The intelligent AI-based catheter-related complication risk warning information system according to claim 2, characterized in that: Analyze the patient's multimodal feature data to identify the patient's risk of catheter-related complications and perform the following operations: Based on the need for early warning of catheter-related complications based on intelligent AI, a catheter-related complication risk identification model was established and deployed in the actual catheter-related complication risk identification environment; The patient's multimodal feature data is input into the catheter-related complication risk identification model, the patient's multimodal feature data is analyzed according to the catheter-related complication risk identification model, and the catheter-related complication risk is effectively identified, thereby determining the catheter-related complication risk identification result.

4. The intelligent AI-based catheter-related complication risk warning information system according to claim 3, characterized in that: A catheter-related complication risk identification model was developed to perform the following: Collect multimodal historical data of patients, divide the collected multimodal historical data of patients, and determine the training set and test set; Based on deep learning technology, a training set is used to train the deep learning model, so that the deep learning model can autonomously learn the catheter-related complication risk identification behavior from the training set and effectively identify the catheter-related complication risk, thereby determining the catheter-related complication risk identification model based on deep learning; Input the test set into the deep learning-based catheter-related complication risk identification model, test the deep learning-based catheter-related complication risk identification model based on the test set, evaluate the performance of the deep learning-based catheter-related complication risk identification model, and determine the model test evaluation results; According to the model test evaluation results, the parameters of the deep learning-based catheter-related complication risk identification model are adjusted and optimized until the deep learning-based catheter-related complication risk identification model can achieve the expected effect of effectively identifying the risk of catheter-related complications, and then the optimal catheter-related complication risk identification model is determined.

5. The intelligent AI-based catheter-related complication risk warning information system according to claim 4, characterized in that: Based on the results of catheter-related complication risk identification, timely warnings and intervention management plans are provided to prompt medical staff to conduct timely intervention management on patients at risk of catheter-related complications, and real-time monitoring of patients after intervention management is carried out. The intervention management plan is adjusted according to the monitoring feedback to form a closed-loop management of patients.

6. The intelligent AI-based catheter-related complication risk warning information system according to claim 5, characterized in that: Monitor patient physiological indicators, catheter status, and laboratory test results, collect multimodal patient data in real time, and perform the following operations: Based on sensors and IoT devices, patients' heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, blood sugar, electrolytes and electroencephalogram are monitored in real time to collect physiological indicators; Real-time monitoring of the patient's catheter position, catheter flow rate, and catheter pressure based on sensors and IoT devices to collect the patient's catheter status; Using sensors and IoT devices, patients' blood indicators, creatinine levels, liver function indicators, urine indicators, and pathogenic microorganism detection are monitored in real time, and laboratory test results are collected; Among them, the patient's multimodal data is determined based on the patient's physiological indicators, the patient's catheter status and the patient's laboratory test results.

7. The intelligent AI-based catheter-related complication risk warning information system according to claim 6, characterized in that: Process the patient's multimodal data and perform the following operations: Clean the patient's multimodal data to remove noise and delete outliers that are not valuable for warning of catheter-related complication risks; Normalize the patient multimodal data to convert it into a unified data format, remove the dimensional differences in the patient multimodal data, and form standardized patient multimodal data.

8. The intelligent AI-based catheter-related complication risk warning information system according to claim 7, characterized in that: Verify the integrity of the integrated multimodal patient data by performing the following operations: Comparing the integrated patient multimodal data with the patient multimodal data before the integration one by one, analyzing the similarity between the integrated patient multimodal data and the patient multimodal data before the integration, and evaluating whether the integrated patient multimodal data has data integrity; When the integrated patient multimodal data is identical to the patient multimodal data before the integration, the integrated patient multimodal data has data integrity, and the integrated patient multimodal data is securely stored. When the integrated patient multimodal data is different from the patient multimodal data before the integration, the integrated patient multimodal data does not have data integrity. At this time, the missing data in the integrated patient multimodal data is searched and supplemented to ensure that the integrated patient multimodal data has data integrity.

9. The intelligent AI-based catheter-related complication risk warning information system according to claim 2, characterized in that: The method of extracting features valuable for early warning of catheter-related complication risks from the patient's multimodal data and determining the patient's multimodal feature data includes: Perform feature analysis on the patient's multimodal data to obtain multiple features to be extracted; A universal feature set and a customized feature set are constructed separately; the universal feature set includes a first feature with universal value for early warning of catheter-related complication risks; and the second feature includes a second feature customized for individual patient conditions and valuable for early warning of catheter-related complication risks. The extraction indicator value of each feature to be extracted is calculated by the following algorithm: in, is the extraction indicator value of the feature to be extracted, is the first intermediate variable, is the first preset weight corresponding to the first intermediate variable, is the second intermediate variable, is the second preset weight corresponding to the second intermediate variable, is the first maximum similarity between the feature to be extracted and the first feature in the universal feature set, is the first preset similarity threshold, is the second maximum similarity between the feature to be extracted and the second feature in the custom feature set, is a second preset similarity threshold; When the extracted indicator value exceeds the indicator value threshold, the corresponding feature to be extracted is regarded as a valuable feature for warning of the risk of catheter-related complications and is extracted as the patient's multimodal feature data.

10. The intelligent AI-based catheter-related complication risk warning information system according to claim 3, characterized in that: Also includes: Tracking management module, including: Follow up on the follow-up of early warning and intervention for patients based on the results of catheter-related complication risk identification; When subsequent situations indicate anomalies in early warning or intervention, the catheter-related complication risk identification model is configured with interpretability capabilities. Inputting the subsequent situation and the warning anomalies or intervention anomalies represented by the subsequent situation into the catheter-related complication risk identification model after the interpretability capability is configured, and obtaining the explanatory content output by the catheter-related complication risk identification model after the interpretability capability is configured; Extracting valuable content from multimodal patient data based on explanatory content; Based on the value content, the catheter-related complication risk identification model is controlled to re-identify the risk of catheter-related complications, and based on the new identification results, patients are re-warned and intervened.

Images