Intelligent infusion management method, system and application

Abstract

The invention belongs to the technical field of infusion management, and provides an intelligent infusion management method and system and application, and the method comprises the steps: obtaining static feature data and chemotherapy treatment data of a patient; obtaining time sequence monitoring data of the puncture area, and performing preprocessing and feature extraction on the time sequence monitoring data to obtain a dynamic feature vector; inputting the static characteristic data and the chemotherapy treatment data into a first machine learning model for processing, and outputting a baseline risk probability; inputting the dynamic feature vector into a second machine learning model for processing and outputting a real-time abnormal probability; fusing the baseline risk probability and the real-time abnormal probability to obtain a comprehensive risk score; dynamically adjusting an early warning threshold based on the baseline risk probability; when the comprehensive risk score exceeds the adjusted early warning threshold value, generating and pushing an early warning signal; and optimizing and updating the first machine learning model and the second machine learning model based on the early warning signal and the corresponding clinical verification result. The chemotherapeutic drug extravasation early warning system can realize early and accurate early warning of chemotherapeutic drug extravasation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of infusion management, and particularly relates to an intelligent infusion management method and system and application. BACKGROUND

[0002] Intelligent infusion pumps have become an indispensable device in modern medicine, significantly improving the safety and efficiency of infusion therapy through automated control, precise drug delivery, and safety monitoring. The basic functions of intelligent infusion pumps have matured, including high-precision infusion (up to μL / h level) and standard safety mechanisms such as air detection and occlusion alarm. The core of its intelligence lies in the Dose Error Reduction Software (DERS), which consists of a drug library, pump-end control software, and a user interface. The drug library serves as a pre-programmed database with soft and hard infusion limits for different drugs and clinical scenarios. In the workflow, nurses select drugs from the drug library and input parameters, and the DERS compares the set values with the safety range in real time: if within the soft limit, normal infusion; if beyond the soft limit but within the hard limit, an alarm is triggered, and after double-checking, it can be forced to continue; if beyond the hard limit, the pump is locked to prohibit infusion. All operations are recorded and can be used for quality analysis. DERS can effectively intercept dose errors, standardize clinical practice, improve safety culture, and provide data support for continuous improvement.

[0003] Currently, the complexity of tumor treatment, the high risk of drugs, and the demands of patients for quality of life have driven the development of intelligent infusion pumps towards an integrated platform of precise drug delivery, safety monitoring, comfortable experience, and data interconnection. However, current intelligent infusion has the following problems:

[0004] 1. Insufficient matching of infusion precision and drug characteristics: For some sensitive drugs, long-term flow stability and transient pulsation may still affect drug efficacy and toxicity.

[0005] 2. Delayed warning of chemotherapy drug extravasation: Chemotherapy drug extravasation is a serious complication that can cause tissue necrosis. Current intelligent infusion pumps can only monitor abnormal pipe pressure, but this is usually an after-the-fact indicator of extravasation. Existing technologies lack prospective techniques that can warn of extravasation in its early stages or impending occurrence. Once extravasation occurs, the consequences are serious, and current technologies cannot achieve early prevention, mainly relying on frequent patrols by nurses and subjective feelings of patients, which is inefficient and risky.

[0006] 3. Limited patient activity freedom and comfort: Traditional portable pumps have problems such as large size, long pipe length, and alarm disturbance, affecting patient quality of life and treatment compliance.

[0007] 4. Weak data island and individualized support: Infusion data is not associated with patient individual information, lacking intelligent decision support; and not interfaced with the daytime chemotherapy call-in system, resulting in inaccurate queuing times.

[0008] 5. Break in management in and out of hospital: Patients with home continuous infusion lack real-time monitoring, resulting in management vacuum and safety risks.

[0009] 6. High complexity of operation and labor cost: Increased functions lead to complex operation, increasing the workload of nurses and training costs.

[0010] 7. Shortcomings of DERS: including alarm fatigue, difficulty in maintaining drug library, and complexity of initial configuration.

[0011] Currently, intelligent infusion pumps for tumor patients are transforming from safe automation tools to intelligent diagnosis and treatment partners. Although the basic safety and precision problems have been solved, there is still a lot of room for improvement in terms of forward-looking early warning, personalized adaptation, patient experience coherence, and data-driven decision-making. SUMMARY

[0012] In view of the above-mentioned shortcomings of the prior art, the present application provides an intelligent infusion management method, system and application, which can realize early and accurate early warning of chemotherapy drug extravasation, reduce the risk of medication errors, and improve nursing efficiency.

[0013] To achieve the above and related purposes, the present application adopts the following technical solutions:

[0014] The first aspect of the present application provides an intelligent infusion management method, comprising the following steps:

[0015] Step S100, acquiring static feature data and chemotherapy treatment data of the patient; acquiring time sequence monitoring data of the puncture area of the patient;

[0016] Step S200, pre-processing and feature extraction are performed on the time sequence monitoring data to obtain a dynamic feature vector for abnormality identification;

[0017] Step S300, inputting the static feature data and the chemotherapy treatment data into a trained first machine learning model for processing, and outputting a baseline risk probability representing the basic risk of extravasation of the patient;

[0018] Step S400, inputting the dynamic feature vector into a trained second machine learning model for processing, and outputting a real-time abnormality probability representing the occurrence of extravasation at the current time;

[0019] Step S500, fusing the baseline risk probability and the real-time abnormality probability to obtain a comprehensive risk score; dynamically adjusting the early warning threshold based on the baseline risk probability, wherein the higher the baseline risk probability, the lower the early warning threshold used; when the comprehensive risk score exceeds the adjusted early warning threshold, an early warning signal is generated and pushed;

[0020] Step S600, based on the early warning signal and its corresponding clinical verification result, the first machine learning model and the second machine learning model are optimized and updated.

[0021] Further, in step S100, the static feature data and the chemotherapy treatment data include the age, gender, BMI, medical history, chemotherapy drug attributes and infusion parameters of the patient; the time sequence monitoring data includes infrared thermal imaging data, tissue bioimpedance data, spectral data and image data of the puncture region of the patient.

[0022] Further, in step S300, the first machine learning model is a logistic regression model.

[0023] Further, in step S400, the second machine learning model is an auto-encoding model based on LSTM, and the second machine learning model is trained through time sequence data of a normal infusion process to learn a normal infusion mode.

[0024] Further, step S400 further comprises: organizing the dynamic feature vector into a time sequence according to a time window, calculating the reconstruction error of the time sequence and the normal infusion mode through the second machine learning model, and normalizing the reconstruction error into a probability value to obtain a real-time abnormal probability.

[0025] Further, the method further comprises: based on the real-time infusion speed, the nurse operation interval historical data and the patient pause record, dynamically updating the total remaining time length of infusion through a prediction model, and synchronizing the length information to a hospital management information system and a patient mobile terminal application.

[0026] Further, the method executes a preset digital chemotherapy scheme through an intelligent infusion pump, and the digital chemotherapy scheme is automatically synchronized from a hospital management information system and includes infusion sequence, speed and time length parameters of multiple drugs.

[0027] The second aspect of the present application provides an intelligent infusion management system, comprising:

[0028] The acquisition module is used for acquiring the static feature data and the chemotherapy treatment data of the patient; and acquiring the time sequence monitoring data of the puncture region of the patient.

[0029] The preprocessing and extraction module is used for preprocessing and feature extraction of the time sequence monitoring data to obtain a dynamic feature vector for abnormal identification.

[0030] The prospective risk assessment module is used for inputting the static feature data and the chemotherapy treatment data into the trained first machine learning model for processing, and outputting a baseline risk probability representing the individual extravasation risk of the patient.

[0031] a dynamic anomaly detection module, configured to input the dynamic feature vector into the trained second machine learning model for processing, and output a real-time anomaly probability representing the occurrence of extravasation at the current time;

[0032] a fusion decision and early warning module, configured to fuse the baseline risk probability and the real-time anomaly probability to obtain a comprehensive risk score, and dynamically adjust the early warning threshold based on the baseline risk probability, wherein the higher the baseline risk probability, the lower the early warning threshold to be adopted; and when the comprehensive risk score exceeds the adjusted early warning threshold, generate and push an early warning signal;

[0033] an optimization module, configured to optimize and update the first machine learning model and the second machine learning model based on the early warning signal and the corresponding clinical verification result.

[0034] The third aspect of the present application provides a computer readable storage medium having computer readable instructions stored thereon, which, when executed by a processor of a computer, cause the computer to perform the intelligent infusion management method.

[0035] The fourth aspect of the present application provides a computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the intelligent infusion management method when executing the computer program.

[0036] The beneficial technical effects of the present application are as follows:

[0037] The present application captures the changes of multiple biophysical signals in the early stage of extravasation of chemotherapy drugs, i.e., the timing monitoring data, to improve the sensitivity and specificity of early extravasation warning; and the present application classifies the risk of patients before infusion through the first machine learning model, and performs real-time monitoring through the second machine learning model, to improve the foresight and initiative of the warning; in addition, the present application dynamically adjusts the early warning threshold by taking the static baseline risk probability as a regulating factor, and adopts a lower threshold for patients with high basic risk, so as to be more vigilant, so as to realize personalized adaptive warning, which can alarm high-risk patients as early as possible without significantly increasing the overall false positive rate, thereby realizing the application of the concept of precision medicine in infusion safety.

[0038] The present application can continuously correct model bias through feedback optimization, so that the warning ability of the model is continuously enhanced with the passage of time and the accumulation of data.

[0039] The present application is deeply integrated with a hospital management information system (HIS), which can reduce the risk of medication errors; and the present application can improve nursing efficiency, promote the standardization of nursing services, and improve the patient experience.

[0040] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application, as claimed. BRIEF DESCRIPTION OF DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and serve to explain the principles of the application. It is readily apparent to one skilled in the art that the following figures are merely illustrative of some embodiments of the application and that other figures can be obtained from these figures without resorting to inventive skill. In the drawings:

[0042] Figure 1 Flow chart of the intelligent infusion management method of the present application;

[0043] Figure 2 Framework diagram of the intelligent infusion management system of the present application;

[0044] Figure 3 Structure diagram of a computer system of a computer device suitable for the embodiments of the present application is shown. DETAILED DESCRIPTION

[0045] Unless otherwise defined, all technical and / or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application pertains. It should be understood that some of the features of the present application, which are described in the context of separate embodiments, can also be provided in combination in a single embodiment. Conversely, many of the features of the present application, which are described in the context of a single embodiment, can also be provided separately or in any appropriate combination. Certain features described in the context of various embodiments are not to be considered essential features of those embodiments, unless the embodiment is inoperative without those elements. The present application is further defined by the following Examples, but it should be understood that these Examples, while indicating the preferred embodiment of the application, are given by way of illustration only. Any equivalent variation of the specific procedure and results shown in the Examples is to be considered as falling within the scope of the present application.

[0046] Reference will now be made to Figure 1 Flow chart of the intelligent infusion management method of the present application, which is described in detail as follows:

[0047] In step S100, static feature data and chemotherapy treatment data of a patient are acquired, and time-series monitoring data of a puncture region of the patient is acquired.

[0048] Specifically, the static feature data of the application includes the age, gender, BMI, medical history (such as diabetes, vascular disease), cancer type and stage, and previous chemotherapy history of the patient, which are used to establish the individual risk characteristics of the patient for prospective risk assessment, for example, an elderly diabetic patient with poor vascular conditions has a higher baseline extravasation risk. The static feature data also includes chemotherapy drug properties and infusion parameters, such as drug name, concentration, pH value, osmotic pressure, infusion speed, planned infusion time, and vascular access type (such as PICC, CVC, peripheral vein), which are used to quantify the irritant risk of the treatment regimen itself. High osmotic pressure, extreme pH value, and highly irritating drugs (such as anthracyclines and vinca alkaloids) or high-speed infusion through a peripheral vein can significantly increase the probability of extravasation and the potential severity of injury.

[0049] Specifically, the time series monitoring data of the application is the key to early and real-time warning. Through continuous and dynamic physiological and physical signal monitoring, subtle changes in the early stages of extravasation or even early warning can be captured. The time series monitoring data of the application includes infrared thermal imaging data, which is the temperature change and thermal radiation distribution of the skin surface of the puncture area. It can be used as a core early warning indicator. Inflammation caused by early extravasation often leads to local temperature rise. When certain drugs (such as gemcitabine) cause vasospasm, they can cause local temperature to drop. Continuous temperature maps can sensitively reflect these abnormalities. It also includes tissue bioimpedance data, which is the electrical impedance value of the tissue around the puncture point. It can also be a core early warning indicator. When the drug solution (with a specific conductivity) seeps into the tissue gap from the blood vessel, it will change the electrical properties (impedance) of the local tissue. This change can be detected before swelling, enabling very early warning. It also includes optical / spectral data, which uses specific wavelengths of light to detect local tissue blood oxygen saturation and edema. It serves as an auxiliary judgment indicator. Extravasation can affect local microcirculation, causing changes in blood oxygen levels. Tissue edema also changes the scattering properties of light in tissue. This data can be verified with other indicators. It also includes image data, which is visual information of the puncture site, used to analyze slight swelling and color changes. It serves as an auxiliary verification means. Through AI image analysis, subtle visual changes that the human eye cannot detect can be identified, providing visual evidence for other sensor data.

[0050] Step S200, pre-process and feature extract the time series monitoring data to obtain a dynamic feature vector for anomaly recognition.

[0051] Specifically, the present application time aligns the time series monitoring data from different sensors with different sampling frequencies, for example, infrared thermal imaging collects at a frequency of 1 frame / second, while bioelectrical impedance collects at a frequency of 10 sampling points / second, the present application establishes a unified timestamp for all data streams, ensuring that different modal data collected at the same time can be correctly associated. Then, the present application applies digital filtering algorithms, such as low-pass filters, median filters, to remove random noise and high-frequency interference introduced by sensor electronic interference, patient slight movement, environmental light changes, etc.; and normalizes the data of different dimensions and orders of magnitude to ensure that data from different sources have equal weights in model training and avoid a certain dimension being dominated by large factor values.

[0052] Specifically, the present application extracts features from the time series monitoring data after the above preprocessing that can represent the differences between normal and abnormal states and have clear physical and statistical significance to form a dynamic feature vector. For example, the present application extracts temperature change rate, temperature difference gradient between local and peripheral regions, etc. from infrared thermal imaging data; extracts impedance value curve slope over time, phase angle change of specific frequency band impedance, etc. from tissue bioimpedance data; extracts image texture features of the puncture region using computer vision technology from spectral / image data, etc.

[0053] Specifically, the present application time series monitoring data completes the above preprocessing and feature extraction in the edge computing module (on a bedside computing device), and outputs a refined dynamic feature vector.

[0054] Step S300, input the static feature data and chemotherapy treatment data into the trained first machine learning model for processing, and output a baseline risk probability representing the individual extravasation risk of the patient.

[0055] Specifically, the first machine learning model of the present application is a logistic regression model.

[0056] Specifically, the first machine learning model of the present application is run before the start of infusion, for evaluating the individual baseline risk of the patient. Wherein, the first machine learning model of the present application receives static feature data and chemotherapy treatment data, and constructs them into a structured input vector , wherein the features specifically include:

[0057] x1 is age; x2 is BMI; x3 is whether there is diabetes, x3=1 for yes and x3=0 for no; x4 is vascular quality score, evaluated by a nurse; x5 is drug osmotic pressure; x6 is drug pH value; x7 is infusion speed; x8 is whether it is a central static catheter, x8=1 for yes and x8=0 for no; x n are other features.

[0058] More specifically, the trained logistic regression model has learned a set of optimal weight coefficients internally through historical data The calculation formula of the logistic regression model is:

[0059] (Formula 1),

[0060] In formula 1, β0 is the intercept, representing the baseline risk; is the weight coefficient corresponding to each feature, whose sign and size directly quantify the influence direction and contribution degree of the feature on the extravasation risk. For example, β3 (corresponding to whether there is diabetes) has a positive and large value, indicating that diabetes is a strong risk factor for extravasation. The output result P pre is the baseline risk probability of extravasation occurrence, which does not depend on any real-time monitoring during the infusion process, but is purely based on the patient's own conditions and the established treatment plan.

[0061] Step S400, input the dynamic feature vector into the trained second machine learning model for processing, and output the real-time abnormal probability representing the occurrence of extravasation at the current time.

[0062] Specifically, the second machine learning model of the present application is an auto-encoding model based on LSTM, and the second machine learning model is trained through time series data of normal infusion processes to learn the normal infusion mode. The present application organizes the dynamic feature vector into a time series according to a time window, calculates the reconstruction error of the time series and the normal infusion mode through the second machine learning model, and normalizes the reconstruction error into a probability value to obtain the real-time abnormal probability.

[0063] Specifically, the second machine learning model of the present application is trained using a large amount of time series monitoring data of normal infusion processes (i.e. without extravasation), and its learning goal is to minimize the reconstruction error of normal data. In this way, the second machine learning model can learn the typical pattern of how the physiological signals in the puncture area should change under the normal infusion mode. Then, the second machine learning model of the present application receives the dynamic feature vector and groups it into a time series according to a sliding time window where S t is the feature vector at time point t, so as to ensure that the second machine learning model can learn and recognize the abnormal evolution pattern with time dependence.

[0064] More specifically, the present application inputs the sequence S t of the current time window into the trained auto-encoder, which maps it to a feature representation in the latent space, and the decoder tries to reconstruct a sequence from this feature representation, and calculates the reconstruction error between the original input sequence S t and the reconstructed sequence The difference between them is measured by the mean square error, and the reconstruction error Error is obtained, which satisfies the following formula 2:

[0065] (Formula 2).

[0066] The greater the reconstruction error is, the greater the deviation between the current monitored signal pattern and the normal pattern learned by the model is, and the higher the possibility of occurrence of the extravasation anomaly is. Then, the present application maps the continuous reconstruction error value to a standard probability value between 0 and 1 through a normalization function, i.e., a Sigmoid function f:

[0067] (Formula 3).

[0068] In formula 3, P alert That is, the real-time anomaly probability, which is the streaming data constantly updated over time, intuitively reflects the possibility of occurrence of extravasation at the current moment.

[0069] Step S500, fuse the baseline risk probability and the real-time anomaly probability to obtain a comprehensive risk score; dynamically adjust the early warning threshold based on the baseline risk probability, wherein the higher the baseline risk probability is, the lower the early warning threshold used is; when the comprehensive risk score exceeds the adjusted early warning threshold, generate and push an early warning signal.

[0070] Specifically, the formula for calculating the comprehensive risk score satisfies the following formula 4:

[0071] (Formula 4),

[0072] In formula 4, Fused_Risk is the comprehensive risk score; a is a weight hyperparameter between 0 and 1, which determines the relative importance of static risk and dynamic risk in the final decision. It can be optimized and adjusted through a historical verification set. For example, a = 0.7, which means that the decision relies more on static evaluation; for example, a = 0.3, which means that the decision relies more on real-time signals.

[0073] More specifically, the early warning threshold of the present application is not fixed but is personalized down-regulated according to the baseline risk probability of the patient:

[0074] (Formula 5),

[0075] In formula 5, is a basic threshold, for example, 0.8; is an adjustment coefficient, a positive coefficient; is an early warning threshold, P pre is higher, is lower, thereby realizing the reduction of the early warning threshold for high-risk patients.

[0076] More specifically, the present application compares the comprehensive risk score and the dynamically adjusted early warning threshold, makes a final decision, if , generates and pushes an early warning signal; otherwise, , continues to monitor.

[0077] Step S600, based on the early warning signal and its corresponding clinical verification result, the first machine learning model and the second machine learning model are optimized and updated.

[0078] Specifically, the early warning signal of the present application includes trigger time, comprehensive risk score, patient ID and other early warning information. The clinical verification result is the on-site check and final judgment of medical staff on each early warning to constitute a feedback data set, including true positive, i.e. early warning is correct, and indeed exosmosis or early signs occur; false positive, i.e. early warning is wrong, and actual exosmosis does not occur; false negative, i.e. no early warning, but exosmosis occurs subsequently.

[0079] Specifically, the present application monitors the model performance indicators in real time, such as sensitivity, specificity, false positive rate, false negative rate, etc., and collects feedback opinions of clinical users. Each false positive and false negative case is analyzed to locate the cause of model failure, for example, whether the infusion characteristics of a certain new listed chemotherapy drug are not covered by historical data; whether the prediction of patients with special complications is biased; then, the present application performs model optimization experiments in an isolated offline environment or simulation test platform to solve the problems found in the analysis, for example, for new drugs, enhance related training samples based on feedback data; then the present application performs small-scale clinical verification on the optimized model through A / B testing and other methods, compares it with the old model, confirms its performance improvement and does not introduce new problems, and then performs full-scale deployment and update.

[0080] Specifically, the present application adopts the following strategies to iteratively optimize the first machine learning model and the second machine learning model:

[0081] 1) Regular retraining: periodically, such as monthly / quarterly, the newly collected feedback data set Data feedback is combined with the historical data Data old set to form a new training set, and based on the existing model Model old , a new model Model new is retrained, and the formula is as follows:

[0082] (Formula 6).

[0083] ​The application can ensure that the model absorbs the latest clinical practice knowledge, adapts to the data distribution drift caused by the change of new drugs, new therapies or nursing standards, and prevents the model performance from invalidating over time through this periodic retraining.

[0084] 2) Online learning: especially suitable for the second machine learning model, using real-time data flow to fine-tune in a reinforcement learning manner, for example, when it is detected that the infusion model of a certain patient is continuously misjudged as abnormal (false positive), but clinically confirmed as normal, the model can be fine-tuned online with the normal data flow of the patient to quickly use the individual specificity of the patient. In this way, the model can be dynamically adjusted to cope with slowly changing patient population characteristics or individual differences, and the level of personalization of real-time early warning can be improved.

[0085] 3) Bias correction: continuously monitor the performance differences of the model in different patient populations such as different ages and disease types. If it is found that the model has systematic bias in false negative rate or false positive rate in a certain population, bias correction is needed. At this time, the application re-trains the model using techniques such as re-sampling and cost-sensitive learning to ensure the fairness and robustness of the model.

[0086] Specifically, the method of the application further comprises: based on the real-time infusion rate, nurse operation interval historical data and patient pause record, dynamically updating the total remaining time of infusion through a prediction model, and synchronizing the time information to the hospital management information system and the patient mobile terminal application.

[0087] More specifically, the application directly obtains the actual infusion rate of the drug from the intelligent infusion pump, rather than the theoretical rate preset by the medical order, as well as the nurse operation interval historical data and patient temporary pause record (such as the historical behavior pattern of the patient actively pausing infusion due to personal needs such as toilet, eating, etc.), and dynamically calculates and continuously corrects the total remaining time through the first machine learning model and the second machine learning model based on the above real-time and historical data. Then, the application writes the end time generated by dynamic prediction back to the hospital management information system (HIS) in real time to facilitate the situation grasped by the outpatient doctors, pharmacy, bed management center and other collaborative parts. In addition, the application also accesses the hospital official APP, and the patient can view the waiting information in real time through the APP after making an appointment, such as there are X patients in front of you, and the estimated waiting time is about XX minutes; and the application actively reminds the patient to prepare for treatment through the APP or SMS at an appropriate time before treatment (such as 30 minutes in advance), for example, please prepare for treatment, and the treatment will be performed at XX:XX, so as to effectively reduce the invalid waiting of the patient in the hospital and improve the treatment experience. In addition, the application can also be linked with the bed / seat management system of the day chemotherapy center, and automatically trigger the notification to call the next patient when the position is vacated.

[0088] Specifically, the method of the present application executes a preset digital chemotherapy scheme through an intelligent infusion pump, and the digital chemotherapy scheme is automatically synchronized from a hospital management information system and contains infusion sequence, speed and time parameters of multiple drugs.

[0089] More specifically, in the central intelligent management cloud platform, a standardized chemotherapy scheme library is pre-established and maintained by pharmacists and clinical experts, such as FOLFOX, TP scheme, etc. Each scheme is a structured data template, which contains drug types and sequence, infusion parameters, and liquid changing logic. When a doctor writes a chemotherapy order for a patient in HIS, the system automatically synchronizes the selected digital scheme to the cloud platform through standard medical interfaces such as HL7, FHIR, etc., and associates it to the specified intelligent infusion pump. The nurse selects the patient and the synchronized order on the PDA (mobile terminal for medical staff), and the entire sequential infusion plan (including infusion speed, time, sequence of different drugs) can be issued to the OncoSmart intelligent infusion pump at the bedside. After receiving the instruction, the intelligent infusion pump does not execute immediately, and the nurse needs to scan the patient's wristband and drug bar code with the pump body. The system automatically compares the scanning information with the order issued by the cloud, and realizes five checks of "correct drug, correct patient, correct dose, correct speed, and correct time". Only when all the checks are passed, the pump is unlocked and ready to execute. After the check is passed, the intelligent infusion pump starts to automatically execute the scheme, strictly following the preset drug sequence, speed and time for infusion; after the current group of drugs is infused, the pump will automatically pause and prompt the nurse to change the next group of drugs; after the nurse changes, the pump will automatically call the parameters of the next group of drugs and continue infusion, without the need for manual setting again.

[0090] More specifically, the status during the entire execution process of the present application, such as infusion, XX% completed, XX minutes remaining, current drug is XX, etc., is real-time returned to the central intelligent management cloud platform through the hospital network, and the platform pushes the information to the medical staff PDA and the central electronic display board of the nurse station in real time, so that the nursing staff can remotely monitor the real-time treatment progress of the patient. In addition, the present application can also push abnormal alarms in real time, such as blocked pipe, air bubbles, low battery, about to finish infusion, etc., to ensure that problems are handled in time.

[0091] Please refer to Figure 2 The framework diagram of the intelligent infusion management system 200 of the present application is shown in FIG. 1, which includes:

[0092] The acquisition module 210 is configured to acquire static feature data and chemotherapy treatment data of the patient, and acquire time sequence monitoring data of the puncture area of the patient.

[0093] The preprocessing and extraction module 220 is configured to pre-process and extract features of the time sequence monitoring data to obtain a dynamic feature vector for abnormal identification.

[0094] The prospective risk assessment module 230 is configured to input the static feature data and the chemotherapy treatment data into the trained first machine learning model for processing, and output a baseline risk probability representing an individual patient's baseline risk of extravasation;

[0095] The dynamic anomaly detection module 240 is configured to input the dynamic feature vector into the trained second machine learning model for processing, and output a real-time anomaly probability representing the occurrence of extravasation at the current time;

[0096] The fusion decision and early warning module 250 is configured to fuse the baseline risk probability and the real-time anomaly probability to obtain a comprehensive risk score, and dynamically adjust the early warning threshold based on the baseline risk probability, wherein the higher the baseline risk probability, the lower the early warning threshold to be adopted; when the comprehensive risk score exceeds the adjusted early warning threshold, an early warning signal is generated and pushed;

[0097] The optimization module 260 is configured to optimize and update the first machine learning model and the second machine learning model based on the early warning signal and the corresponding clinical verification result.

[0098] It should be noted that the intelligent infusion management system provided in the above embodiments and the intelligent infusion management method provided in the above embodiments belong to the same concept, wherein the specific manner in which each module and unit performs operations has been described in detail in the method embodiments, which will not be repeated here. The intelligent infusion management system provided in the above embodiments can be used in actual applications, and the above functions can be completed by different functional modules according to needs, that is, the internal structure of the system is divided into different functional modules to complete all or part of the functions described above, and this is not limited herein.

[0099] Embodiments of the present application also provide a computer device, comprising: one or more processors; a storage device configured to store one or more programs, when the one or more programs are executed by the one or more processors, the computer device implements the intelligent infusion management method provided in each of the above embodiments.

[0100] Figure 3 The structure of the computer system of the computer device suitable for the embodiments of the present application is shown. It should be noted that, Figure 3 The computer system 300 of the electronic device shown is only an example, and should not limit the functions and use range of the embodiments of the present application.

[0101] As Figure 3As shown, the computer system 300 includes a central processing unit (CPU) 301 which can perform various appropriate actions and processes in accordance with programs stored in a read only memory (ROM) 302 or loaded into a random access memory (RAM) 303 from a storage section 308, such as the methods described in the above embodiments. Various programs and data required for the operation of the system are also stored in the RAM 303. The CPU 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304. The following are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; the storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as necessary. A removable recording medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 310 as necessary, so that a computer program read therefrom is installed into the storage section 308 as necessary.

[0102] In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer tool program. For example, embodiments of the present application include a computer program product including a computer program carried on a computer-readable medium, the computer program containing a computer program for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by the communication section 309, and / or installed from the removable recording medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the system of the present application are performed.

[0103] It should be noted that the computer readable medium shown in the embodiments of the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, be an electronic, magnetic, optical, electromagnetic, infrared or semiconductor system, device or apparatus, or any combination thereof. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, a flash memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present application, the computer readable signal medium can include a data signal carried in a baseband or as a carrier wave in a propagated data signal, in which the computer readable computer program is carried. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium, which can send, propagate or transmit programs for use by or in connection with an instruction execution system, apparatus or device. The computer program contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0104] The flowcharts and block diagrams in the drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In the flowcharts or block diagrams, each block can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order from that shown in the drawings. For example, two blocks represented in succession can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0105] The units described in the embodiments of the present application can be implemented in the form of tools, or can be implemented in the form of hardware, and the described units can also be arranged in a processor. In some cases, the names of these units do not constitute a limitation on the units themselves.

[0106] Another aspect of the present application also provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor of a computer, so that the computer executes the intelligent infusion management method as described above. The computer readable storage medium can be included in the computer device described in the above embodiments, or can exist separately without being assembled into the computer device.

[0107] Another aspect of the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the intelligent infusion management method provided in each of the above embodiments.

[0108] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not used to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes completed by those skilled in the art without departing from the spirit and technical thought disclosed by the present application should be covered by the claims of the present application.

Claims

1. An intelligent infusion management method, characterized in that, Includes the following steps: Step S100: Obtain the patient's static characteristic data and chemotherapy treatment data; obtain the temporal monitoring data of the patient's puncture area; Step S200: Preprocess and extract features from the time-series monitoring data to obtain a dynamic feature vector for anomaly identification; Step S300: Input the static feature data and chemotherapy treatment data into the trained first machine learning model for processing, and output the baseline risk probability characterizing the individual patient's basic risk of extravasation. Step S400: Input the dynamic feature vector into the trained second machine learning model for processing, and output the real-time anomaly probability representing the current time of extravasation. Step S500: The baseline risk probability and the real-time anomaly probability are fused to obtain a comprehensive risk score; The warning threshold is dynamically adjusted based on the baseline risk probability, wherein the higher the baseline risk probability, the lower the warning threshold is; when the comprehensive risk score exceeds the adjusted warning threshold, a warning signal is generated and pushed. Step S600: Based on the warning signal and its corresponding clinical validation results, optimize and update the first machine learning model and the second machine learning model.

2. The method according to claim 1, characterized in that, In step S100, the static feature data and chemotherapy treatment data include the patient's age, gender, BMI, medical history, chemotherapy drug properties and infusion parameters; the time-series monitoring data includes infrared thermal imaging data, tissue bioimpedance data, spectral data and image data of the patient's puncture area.

3. The method according to claim 1, characterized in that, In step S300, the first machine learning model is a logistic regression model.

4. The method according to claim 1, characterized in that, In step S400, the second machine learning model is an autoencoder model based on LSTM, and the second machine learning model is trained using time-series data of a normal infusion process to learn the normal infusion pattern.

5. The method according to claim 4, characterized in that, Step S400 further includes: organizing the dynamic feature vector into a time series according to a time window, calculating the reconstruction error between the time series and the normal infusion mode through the second machine learning model, and normalizing the reconstruction error into a probability value to obtain the real-time abnormal probability.

6. The method according to claim 1, characterized in that, The method further includes: dynamically updating the total remaining infusion time based on real-time infusion rate, historical data of nurse operation intervals and patient pause records through a predictive model, and synchronizing this time information to the hospital management information system and patient mobile terminal applications.

7. The method according to claim 1, characterized in that, The method executes a preset digital chemotherapy protocol using an intelligent infusion pump. The digital chemotherapy protocol is automatically and synchronously loaded from the hospital management information system and includes parameters such as the infusion sequence, speed, and duration of various drugs.

8. An intelligent infusion management system, characterized in that, include: The acquisition module is used to acquire static characteristic data and chemotherapy treatment data of patients; and to acquire temporal monitoring data of the patient's puncture area. The preprocessing and extraction module is used to preprocess and extract features from the time-series monitoring data to obtain dynamic feature vectors for anomaly identification. The prospective risk assessment module is used to input the static feature data and chemotherapy treatment data into the trained first machine learning model for processing, and output a baseline risk probability that characterizes the individual patient's basic risk of extravasation. The dynamic anomaly detection module is used to input the dynamic feature vector into the trained second machine learning model for processing, and output the real-time anomaly probability representing the current time of extravasation. The fusion decision and early warning module is used to fuse the baseline risk probability with the real-time anomaly probability to obtain a comprehensive risk score; The warning threshold is dynamically adjusted based on the baseline risk probability, wherein the higher the baseline risk probability, the lower the warning threshold is; when the comprehensive risk score exceeds the adjusted warning threshold, a warning signal is generated and pushed. An optimization module is used to optimize and update the first machine learning model and the second machine learning model based on the warning signal and its corresponding clinical validation results.

9. A computer-readable storage medium, characterized in that, It stores computer-readable instructions, which, when executed by the computer's processor, cause the computer to perform the intelligent infusion management method as described in any one of claims 1 to 7.

10. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the intelligent infusion management method according to any one of claims 1 to 7.

Images