Picc conduit safety monitoring and alarm system and method based on Internet of Things

Abstract

The invention relates to the technical field of medical equipment monitoring, in particular to a Picc catheter safety monitoring and alarm system and method based on the Internet of Things. The method specifically comprises the steps that multi-modal physiological parameters are collected in real time through an intelligent sensor, multi-level fusion denoising and normalization processing are carried out through an edge calculation module, and an anti-loss seal is generated; the cloud end adopts an adaptive hybrid model to analyze time series data and group patient association features, realizes unsupervised anomaly detection through comparative learning and a variational auto-encoder, and performs lightweight deployment on a complex model in combination with a knowledge distillation technology; dynamically calculating a risk level based on a Bayesian abnormal score and a weighted time sequence index, adopting a grading early warning strategy to match a differentiated alarm notification mode, synchronously generating personalized nursing suggestions, and optimizing monitoring frequency; data, risk prompt and decision support are integrated through a visual interaction interface. According to the invention, accurate identification, real-time early warning and active intervention of the abnormal state of the PICC are realized, and the false alarm rate and the medical risk are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical equipment monitoring, and in particular to a Picc catheter safety monitoring and alarm system and method based on the Internet of Things. Background Art

[0002] PICC (Peripherally Inserted Central Catheter) is a widely used intravenous infusion tool in clinical practice. It can provide long-term intravenous treatment support and is suitable for patients who require continuous infusion, parenteral nutrition or chemotherapy. However, due to the long-term indwelling of PICC catheters in the body, a series of safety hazards may occur during their use, including but not limited to catheter displacement, blockage, infection and venous thrombosis. These risks not only affect the patient's treatment effect, but may also cause serious complications and even endanger life. The existing PICC catheter monitoring method mainly relies on manual regular inspections and ultrasound evaluations. This method has problems such as poor real-time performance, reliance on medical experience, limited detection frequency and inability to provide continuous monitoring data, resulting in some abnormal situations not being discovered and handled in a timely manner.

[0003] The Chinese invention patent application with publication number CN118762838A discloses an image-based PICC intravenous therapy information monitoring system and method. By performing time-series analysis on high-definition images of the patient's puncture site at different time points within the target monitoring period, it is determined whether the patient's puncture site has signs of infection. This can improve the accuracy and timeliness of infection monitoring, help detect signs of infection early, and thus take timely measures to reduce the incidence and severity of infection.

[0004] In recent years, with the development of Internet of Things technology, intelligent medical monitoring systems have been gradually applied to clinical scenarios. By collecting patient physiological data through sensors and combining wireless communication, cloud computing and artificial intelligence technologies for data analysis and anomaly detection, it is expected to improve the safety monitoring capabilities of PICC catheters. However, the current intelligent monitoring system still has deficiencies in sensor stability, data processing accuracy, anomaly prediction capabilities and alarm timeliness, and cannot fully meet clinical needs. Therefore, there is an urgent need for a PICC catheter safety monitoring and alarm system based on the Internet of Things to improve the real-time, accuracy and intelligence level of catheter monitoring, thereby effectively reducing the incidence of PICC catheter-related complications and improving patient safety and medical quality. Summary of the Invention

[0005] The purpose of the present invention is to address the problems existing in the background technology and propose a Picc catheter safety monitoring and alarm system and method based on the Internet of Things.

[0006] The technical solution of the present invention is a Picc catheter safety monitoring and alarm method based on the Internet of Things, which includes the following specific implementation steps:

[0007] S1, regularly collect Picc catheter environmental data and generate physiological parameter data stream;

[0008] S2, based on a multi-level fusion denoising method, performs multi-scale decomposition and adaptive threshold denoising on the original signal through wavelet transform and empirical mode decomposition, combines Kalman filtering to optimize the smoothness of the time series signal, uses Z-score normalization to generate standardized data, and generates anti-loss seals based on hash functions and finite field operations;

[0009] S3: Data reliability is tested based on anti-loss seals. An adaptive hybrid model is constructed, integrating a graph neural network with a Transformer+LSTM model. Patient relationship graphs are used to perform group association analysis and capture long- and short-term temporal features. Contrastive learning is used to construct positive and negative samples to distinguish anomalies. A variational autoencoder is used to reconstruct the error detection anomaly distribution. Knowledge distillation is used to lightweight deploy complex models and output test results.

[0010] S4. Calculate the comprehensive risk value by combining the weighted time series anomaly index, dynamically adjust the risk threshold, determine the warning level, divide the warning level into three levels of low, medium and high, and adopt a graded alarm strategy;

[0011] S5. Patients are divided into low, medium, and high risk levels based on risk grading results, and differentiated nursing measures are formulated based on these levels. Monitoring frequency and nursing plans are dynamically adjusted based on patient historical data. Operation reminders are intelligently pushed, and nursing logs are automatically generated simultaneously to optimize risk strategies.

[0012] S6. Visual display of standardized physiological data, test results, risk levels, corresponding alarm strategies, personalized maintenance recommendations, and optimal care plans.

[0013] Preferably, the implementation process of the multi-level fusion denoising method is as follows:

[0014] S21. Use wavelet transform to perform multi-scale decomposition on the original signal, extract different frequency components, perform threshold denoising on the high-frequency components, and use adaptive soft threshold method to dynamically adjust the threshold according to the noise level:

[0015]

[0016] Where W f (a, b) represents the signal coefficient after wavelet transformation; f(t) represents the original sensor signal at time t; ψ(t) represents the mother wavelet function; a represents the scale parameter; b represents the translation parameter; represents the wavelet coefficient after denoising; λ represents the adaptive threshold; σ represents the noise standard deviation; N represents the data length; sign() represents the sign function;

[0017] S22. Decompose the signal f(t) into intrinsic mode functions. Based on spectrum analysis and IMF energy threshold screening, dynamically remove high-frequency noise and low-frequency trend noise:

[0018] EMD decomposition:

[0019] Calculate the energy ratio of the IMF components:

[0020] Set the energy threshold E th , if Ei <E th , it is considered as a noise component and is removed;

[0021] Where, IMF i (t) represents the i-th natural mode component; r n (t) represents the residual component; n represents the total number of natural mode components;

[0022] S23. Introduce Kalman filtering, model the relationship between signal and noise through state equation, calculate and update the Kalman gain after predicting the state and covariance, realize signal smoothing and abnormality prediction, and optimize the denoising effect.

[0023] Preferably, the optimization process for optimizing the denoising effect is as follows:

[0024] S31, state space modeling: define the state equation as: X t =Ax t-1 +BU t +W t 、Z t =HX t +V t ;

[0025] Where, X t represents the real signal at the current time t; Z t represents the measurement signal; A represents the state transfer matrix; B and U t Represent the control input matrix and control input respectively; W t 、V t represents process noise and measurement noise; H represents the observation matrix;

[0026] S32. Forecast state estimation:

[0027] Forecast covariance matrix:

[0028] in, Indicates the predicted state at the current moment; represents the optimal state estimate at the previous moment; represents the prediction error covariance matrix; Pt-1 represents the error covariance matrix of the previous moment; Q represents the process noise covariance matrix;

[0029] S33. Calculate the Kalman gain:

[0030] Update the state estimate:

[0031] Update the covariance matrix:

[0032] Where K t represents the Kalman gain; R represents the measurement noise covariance matrix; T represents the transpose operation of the matrix; represents the optimal state estimate at the current moment, that is, the final denoised signal; P t represents the updated error covariance matrix; I represents the identity matrix;

[0033] S34. Output optimized denoised data

[0034] Preferably, the process of generating the anti-loss seal is as follows:

[0035] S41, calculate the information packet Inp=H(data);

[0036] Where H is a predefined hash function; data is the binary data form of standardized data;

[0037] S42, select a random number k∈F p , calculate the basic package parameter PB = r + k·Inp mod p;

[0038] Where p is a predefined large prime number that satisfies p=3mod4; F p is a predefined finite field; r is a predefined random p The selected elements in

[0039] S43, calculate the derived package parameter PD = s + k·Inp mod p;

[0040] Where s is a predefined random p The selected elements in

[0041] S44. Calculate the comprehensive packaging parameter PS = k·(2r+τ)+k 2 ·Inp+ρk mod p;

[0042] Among them, τ, ρ are predefined packing factors, τ, ρ∈F p ;

[0043] S45. Generate an anti-loss seal AS = {basic packaging parameters PB, derived packaging parameters PD, comprehensive packaging parameters PS}.

[0044] Preferably, the verification process for data reliability based on the anti-loss seal is as follows:

[0045] S51, calculate the information packet Inp′=H(data′);

[0046] Wherein, data′ is the binary data of the received standardized physiological data data;

[0047] S52. Construct the following equation: Eq1 = (PB) 2 +τ·PD; Eq2=U+PS·Inp′ mod p;

[0048] Eq3=(PD) 2 +ρ·PB; Eq4=V+PS·Inp′ mod p;

[0049] Among them, U and V are predefined deblocking factors, U = r 2 +τs mod p, V=s 2 +ρr mod p;

[0050] S53. If Eq1 = Eq2 and Eq3 = Eq4, it means that there is no packet loss in the standardized data, that is, the reliability of the data is guaranteed.

[0051] Preferably, the detection process of the adaptive hybrid model is as follows:

[0052] S61. Construct a patient relationship graph, treating each patient as a node in the graph and defining patients who use the same type of PICC catheter and have similar treatment plans as edges. The graph neural network (GNN) captures the potential relationships between different patients by training the nodes and edges, learns the feature representation of each patient, and identifies normal and abnormal patterns.

[0053] S62, Transformer processes input data through the self-attention mechanism, captures the long-term dependent features in the time series, focuses on the correlation between each time step in the PICC catheter data, and identifies long-term trends;

[0054] S63,LSTM is used to capture short-term temporal dependencies and extract short-term information from the short-term fluctuations in the monitoring data of the PICC catheter;

[0055] Preferably, the detection process of abnormal distribution detection by reconstruction error of variational autoencoder is as follows:

[0056] S71. Identify abnormal data without labels through contrastive learning, construct positive and negative samples, and learn how to distinguish between the two:

[0057] Positive samples: Select samples from normal PICC catheter monitoring data;

[0058] Negative samples: Negative samples are constructed by artificial generation and abnormal catheter monitoring data;

[0059] S72. Use the variational autoencoder (VAE) to learn the distribution of normal PICC catheter data and determine whether the data is abnormal by calculating the reconstruction error.

[0060] Preferably, the warning level classification process is as follows:

[0061] S81. Calculate the overall risk score using the Bayesian anomaly score combined with the weighted temporal anomaly index:

[0062] Bayesian Anomaly Score:

[0063] Weighted time series anomaly index:

[0064] w t =e -λ(T-t) ;

[0065] Where S BRS represents the Bayesian anomaly score, that is, the probability of anomaly A occurring given the test result X; P(A) represents the anomaly prior probability of historical data statistics; P(X|A) represents the probability of observing the current data X under an abnormal situation; P(X) represents the total probability density of the current data X; S WTAI represents the weighted time series anomaly index; w t Represents the weight coefficient; A t represents the anomaly score at time t; T represents the observation time window; λ represents the decay coefficient;

[0066] S82, based on risk score S BRS , use the adaptive threshold setting method to classify risks and determine the warning level:

[0067] Classification strategy: Use adaptive dynamic classification, based on the mean-variance method of historical abnormal data (adjust the alarm threshold:

[0068]

[0069]

[0070] Where μ Arepresents the mean of the anomaly score, reflecting the average level of the Bayesian anomaly score of each sample in the historical data; N represents the number of historical data samples; S BRS,i represents the Bayesian anomaly score of the i-th sample; σ A represents the standard deviation of the anomaly score, reflecting the fluctuation or dispersion of the Bayesian anomaly score in historical data;

[0071] S83. Set risk level:

[0072] Level 1: Low risk, S BRS <μ A ;

[0073] Level 2: Medium risk, μ A ≤S BRS <μ A +(1+α·S WTAI )σ A ;

[0074] Level 3: High risk, S BRS ≥μ A +(1+α·S WTAI )σ A ;

[0075] Among them, α is the weight factor, α>0.

[0076] The technical solution of the present invention is a Picc catheter safety monitoring and alarm method based on the Internet of Things, which is used to implement the above-mentioned Picc catheter safety monitoring and alarm method based on the Internet of Things, including:

[0077] Smart PICC catheter sensor module for monitoring physiological parameters;

[0078] Edge data processing module, used to pre-process the collected physiological parameter data;

[0079] A cloud-based data analysis module uses big data analysis and deep learning algorithms to intelligently evaluate PICC catheter usage and identify catheter anomalies.

[0080] Intelligent alarm and notification module, used to predict potential risks and send alarm information through smart terminals;

[0081] Intelligent decision support module, which combines historical data with personalized patient information to provide medical advice;

[0082] User interaction module, used to provide data visualization interface.

[0083] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0084] The present invention designs a Picc catheter safety monitoring and alarm system and method based on the Internet of Things. It collects multimodal physiological parameters in real time through flexible microsensors, combines multi-level fusion denoising strategies (wavelet transform, EMD decomposition and Kalman filtering) with an anti-loss seal mechanism, significantly improves signal quality and ensures data integrity, and effectively solves the problems of high false alarm rate and poor real-time performance caused by noise interference in traditional monitoring. The cloud uses an adaptive hybrid model (GNN+Transformer+LSTM) to realize group data association analysis and time series dependency modeling, and combines contrastive learning with unsupervised anomaly detection of variational autoencoders. It breaks through the limitations of single patient data and improves the accuracy of identifying catheter displacement, blockage and infection; the dynamic risk grading strategy based on Bayesian scoring and weighted time series index optimizes the allocation of medical resources through graded alarms (low, medium and high risk) and personalized nursing recommendations, reducing the risk of excessive warnings and delayed intervention; knowledge distillation technology deploys complex models in a lightweight manner, supports real-time reasoning and low-latency response, and the user interaction module integrates data visualization and decision support to form a closed-loop management from monitoring and analysis to intervention, ultimately achieving accurate early warning of abnormal PICC catheter status, active intervention and comprehensive improvement of clinical nursing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is a system architecture diagram of the Internet of Things-based Picc catheter safety monitoring and alarm system proposed by the present invention;

[0086] Figure 2 This is a flow chart of a Picc catheter safety monitoring and alarm method based on the Internet of Things proposed by the present invention. DETAILED DESCRIPTION

[0087] Example 1, as Figure 1 As shown, the Internet of Things-based Picc catheter safety monitoring and alarm system proposed in the present invention includes: an intelligent PICC catheter sensor module, an edge data processing module, a cloud data analysis module, an intelligent alarm and notification module, an intelligent decision support module, and a user interaction module.

[0088] The intelligent PICC catheter sensor module uses a flexible microsensor embedded in the outer wall of the catheter (without affecting catheter operation) to monitor physiological parameters, including but not limited to catheter position, blood flow resistance, temperature, and pH value;

[0089] The edge data processing module pre-processes the collected physiological parameter data, including but not limited to denoising and normalization;

[0090] The cloud-based data analysis module uses big data analysis and deep learning algorithms to intelligently evaluate PICC catheter usage and identify catheter abnormalities, including but not limited to blockage, displacement, and infection.

[0091] The intelligent alarm and notification module uses AI analysis to predict potential risks and send alarm information through smart terminals;

[0092] The intelligent decision support module combines historical data with personalized patient information to provide medical advice;

[0093] The user interaction module provides a data visualization interface.

[0094] Example 2, as Figure 2 As shown, the Picc catheter safety monitoring and alarm method based on the Internet of Things proposed in the present invention is applied to the Picc catheter safety monitoring and alarm system based on the Internet of Things proposed in Example 1. The specific implementation steps are as follows:

[0095] S1. The intelligent PICC catheter sensor module completes device initialization and baseline value calibration through a flexible microsensor array (including but not limited to pressure sensor, flow sensor, temperature sensor, pH sensor);

[0096] After that, the system enters the specific implementation phase. The intelligent PICC catheter sensor module regularly collects PICC catheter environmental data (including but not limited to catheter internal pressure, blood flow rate, temperature, and pH value) to form a physiological parameter data stream.

[0097] Afterwards, the physiological parameter data stream is transmitted to the edge data processing module in real time through the NB-IoT unit;

[0098] S2, the edge data processing module builds a multi-level fusion denoising strategy, removes high-frequency noise through wavelet transform, removes trend noise by combining EMD decomposition, and uses Kalman filtering to smooth the signal. The specific implementation process is as follows:

[0099] S21. Preliminary noise suppression based on wavelet transform extracts signal features at different scales and disperses the noise to high-frequency components. Specifically:

[0100] S2101, Wavelet decomposition: Use wavelet transform to perform multi-scale decomposition on the original signal and extract different frequency components:

[0101] Where W f (a, b) represents the signal coefficients after wavelet transformation; f(t) represents the original sensor signal at time t; ψ(t) represents the mother wavelet function; a represents the scale parameter (controls the frequency resolution); b represents the translation parameter (controls the time position);

[0102] S2102, threshold denoising: Perform threshold denoising on the high-frequency components using an adaptive soft threshold method, specifically:

[0103]

[0104] Where, represents the wavelet coefficient after denoising; λ represents the adaptive threshold; σ represents the noise standard deviation; N represents the data length; sign() represents the sign function;

[0105] Based on this: Adopting the adaptive threshold method, the threshold is dynamically adjusted according to the noise level to improve the adaptability of denoising and reduce information loss;

[0106] S22, Empirical Mode Decomposition (EMD) denoising: Combined with IMF energy threshold screening, it dynamically removes noise components to avoid information loss while enhancing the fidelity of low-frequency signals. Specifically:

[0107] S2201, EMD decomposition: Decompose the signal f(t) into a series of intrinsic mode functions (IMFs):

[0108]

[0109] Where, IMF i (t) represents the i-th intrinsic modal component, i.e., the signal components of different scales; r n (t) represents the residual component, i.e., trend noise; n represents the total number of intrinsic modal components;

[0110] S2202: Remove high-frequency noise and low-frequency trend noise based on spectrum analysis, specifically:

[0111] Calculate the energy ratio of the IMF components:

[0112] Set the energy threshold E th , if Ei <E th , it is considered as a noise component and is removed;

[0113] Based on this: Combined with IMF energy threshold screening, the noise component is dynamically removed to avoid information loss while enhancing the fidelity of low-frequency signals;

[0114] S23. Time series optimization based on Kalman filtering: Wavelet transform and EMD have significantly reduced noise, but there may still be slight signal fluctuations. To further improve the smoothness and predictive ability of the signal, Kalman filtering is introduced to smooth the signal under low noise conditions and provide outlier prediction capabilities, further improving the denoising effect. Specifically:

[0115] S2301, State Space Modeling: Let signal X t is the state variable, the measured value Z t Affected by noise, the state equation is: X t =Axt-1 +BU t +W t 、Z t =HX t +V t ;

[0116] Where, X t represents the real signal at the current time t; Z t represents the measurement signal (including noise); A represents the state transfer matrix (set to the identity matrix); B and U t Represent the control input matrix and control input (negligible in this embodiment); W t 、V t represents process noise and measurement noise; H represents the observation matrix, i.e., the relationship between measurement and state, which is set as the unit matrix H=I in this embodiment;

[0117] S2302, recursive filtering calculation:

[0118] (1) Prediction steps:

[0119] Predicted state estimation:

[0120] in, Represents the predicted state at the current moment, that is, the true signal value at the current moment is inferred based on the information at the previous moment; represents the optimal state estimate at the previous moment;

[0121] Forecast covariance matrix:

[0122] in, represents the prediction error covariance matrix, that is, the uncertainty of the prediction signal; P t-1 represents the error covariance matrix of the previous moment; Q represents the process noise covariance matrix, which describes the degree of random change of the signal itself;

[0123] (2) Update steps:

[0124] Calculate the Kalman gain:

[0125] Where K t represents the Kalman gain, i.e., the weight of the measurement information and the prediction information, which determines the degree of trust in the measurement value during the update; R represents the measurement noise covariance matrix, i.e., the magnitude of the sensor measurement noise; T represents the transpose operation of the matrix;

[0126] Update the state estimate:

[0127] Where, Represents the optimal state estimate at the current moment, that is, the final denoised signal;

[0128] Update the covariance matrix:

[0129] Where, P t represents the updated error covariance matrix, that is, the size of the estimated error at the current moment; I represents the unit matrix to ensure dimension matching;

[0130] Based on this: Kalman filtering can smooth signals in low-noise conditions and provide outlier prediction capabilities, further improving denoising effects;

[0131] S24. Since PICC catheter safety monitoring involves multiple sensor data (including but not limited to catheter internal pressure, blood flow rate, temperature, and pH value), different data have different dimensions and distributions, and therefore need to be normalized for comprehensive analysis. Therefore, the Z-score normalization method is used to normalize the data of different modalities after denoising to obtain standardized physiological data;

[0132] S25. Generate an anti-loss seal for the standardized data. The process of generating the anti-loss seal is as follows:

[0133] S2501, calculate information packet Inp=H(data);

[0134] Where H is a predefined hash function; data is the binary data form of standardized data;

[0135] S2502, select a random number k∈F p , calculate the basic package parameter PB = r + k·Inp mod p;

[0136] Where p is a predefined large prime number that satisfies p=3mod4; F p is a predefined finite field; r is a predefined random p The selected elements in

[0137] S2503, calculate the derived package parameter PD = s + k·Inp mod p;

[0138] Where s is a predefined random p The selected elements in

[0139] S2504. Calculate comprehensive packaging parameters PS = k·(2r+τ)+k 2 ·Inp+ρk mod p;

[0140] Among them, τ, ρ are predefined packing factors, τ, ρ∈F p ;

[0141] S2505. Generate anti-loss seal AS = {basic packaging parameters PB, derived packaging parameters PD, comprehensive packaging parameters PS};

[0142] S26, transmitting {anti-loss seal AS = {basic packaging parameter PB, derived packaging parameter PD, comprehensive packaging parameter PS}, standardized physiological data data} to the cloud data analysis module;

[0143] The S3 cloud-based data analysis module uses an adaptive hybrid model (GNN+Transformer+LSTM), combined with self-supervised learning and knowledge distillation. Through multi-layer feature fusion through deep learning, it achieves accurate analysis, anomaly identification, and risk assessment of PICC catheter monitoring data. The specific implementation process is as follows:

[0144] S31. Receive {anti-loss seal AS = {basic encapsulation parameters PB, derived encapsulation parameters PD, comprehensive encapsulation parameters PS}, standardized physiological data data}, extract the anti-loss seal AS = {basic encapsulation parameters PB, derived encapsulation parameters PD, comprehensive encapsulation parameters PS} and the standardized physiological data data, and perform packet loss detection on the standardized physiological data data. The detection process is as follows:

[0145] S3101, calculate information packet Inp′=H(data′);

[0146] Wherein, data′ is the binary data of the received standardized physiological data data;

[0147] S3102, construct the following equation: Eq1 = (PB) 2 +τ·PD; Eq2=U+PS·Inp′mod p; Eq3=(PD) 2 +ρ·PB; Eq4=V+PS·Inp′ mod p;

[0148] Among them, U and V are predefined deblocking factors, U = r 2 +τs mod p, V=s 2 +ρr mod p;

[0149] S3103. If Eq1 = Eq2 and Eq3 = Eq4, it means that no packet loss occurs during the transmission of the standardized physiological data from the edge data processing module to the cloud data analysis module, which means that the data reliability is guaranteed; otherwise, an alarm is immediately issued;

[0150] S32. Build an adaptive hybrid model that integrates graph neural networks (GNNs) and time series modeling methods (Transformer + LSTM) to achieve a global perspective analysis of single PICC catheter monitoring data and group data (including but not limited to monitoring data of other patients in the hospital). Combine the PICC catheter sensor data with data from other patient groups in the hospital to improve the system's anomaly recognition capabilities. Specifically:

[0151] S3201. Constructing a patient relationship diagram:

[0152] (1) Graph structure construction: Each patient is treated as a node in the graph. Node features include but are not limited to the monitoring data of the PICC catheter sensor and the patient's basic clinical information (including but not limited to treatment history and disease progression);

[0153] (2) Edge definition: If two patients use the same type of PICC catheter and have similar treatment plans, they are connected by an edge;

[0154] (3) The role of GNN: GNN captures the potential relationships between different patients by training nodes and edges, and learns the feature representation of each patient to identify normal and abnormal patterns at the group level;

[0155] S3202, Time Series Modeling: Use Transformer and LSTM to capture the long-term and short-term dependencies of time series data, specifically:

[0156] Transformer: Transformer processes input data through a self-attention mechanism, capturing long-term dependencies in time series. It focuses on the correlation between time steps in PICC catheter data and identifies long-term trends.

[0157] LSTM: LSTM is used to capture short-term temporal dependencies, effectively extract short-term information from short-term fluctuations in PICC catheter monitoring data, and strengthen continuity modeling between local time steps.

[0158] Based on this, the GNN first constructs a patient relationship graph, providing global information at the group level. After receiving this information, the Transformer and LSTM model the time series data of PICC catheters, capturing long-term and short-term dependencies, improving the model's recognition capabilities, generating structured association features between patients, and identifying potential abnormal trends predicted by the time series model.

[0159] S33. Build an anomaly detection model, perform unsupervised anomaly detection on the data through contrastive learning and variational autoencoder (VAE), and learn the anomaly distribution. Specifically:

[0160] S3301. Through contrastive learning, identify abnormal data in the absence of labels, construct positive and negative samples, and learn how to distinguish between the two:

[0161] Positive samples: Select samples from normal PICC catheter monitoring data;

[0162] Negative samples: Negative samples are artificially generated (including but not limited to adding noise) or constructed through abnormal catheter monitoring data;

[0163] S3302, performing reconstruction error detection using a variational autoencoder (VAE): Using the VAE, the distribution of normal PICC catheter data is learned, and reconstruction error is calculated to determine whether the data is abnormal.

[0164] It should be noted that the working principle of VAE is: VAE compresses the input data into a latent space through the encoder, and then reconstructs it back to the original data through the decoder. If the reconstruction error is large, it means that the distribution of the input data is significantly different from that of the training data, which may be abnormal data;

[0165] Based on this, both contrastive learning and VAE are used for unsupervised learning. The former learns similarity by constructing positive and negative samples, while the latter determines anomalies by reconstructing errors, which enhances the model's anomaly detection capability and reduces dependence on labeled data.

[0166] S34: Lightweight the anomaly detection model constructed in step S33 through knowledge distillation to achieve real-time reasoning, specifically:

[0167] S3401, Cloud Model Training: Build an anomaly detection model in the cloud to train abnormal patterns in PICC catheter data;

[0168] S3402, Knowledge Distillation: Use knowledge distillation to compress complex cloud models into lighter models. Use the output of the teacher model (cloud model) to train the student model (lightweight model).

[0169] Distillation process: The output of the teacher model is provided to the student model as soft targets, helping the student model to learn and reason efficiently on edge devices;

[0170] S3403, deploy the distilled student model (lightweight model), the lightweight model performs the inference task, and outputs the detection result {the probability P(X|A) of observing the current data X under abnormal conditions, the abnormal score A at time t t};

[0171] S35, the detection result {the probability P(X|A) of observing the current data X under abnormal conditions, the abnormal score A at time t t}Transmit to the intelligent alarm and notification module;

[0172] S4, the intelligent alarm and notification module builds an intelligent alarm and risk classification warning method to achieve accurate early warning of abnormal PICC catheter status, promptly notify medical staff and patients, and provide reasonable risk level assessment, specifically:

[0173] S41. Calculate the overall risk score using the Bayesian Risk Score (BRS) combined with the Weighted Temporal Anomaly Index (WTAI), specifically:

[0174] Bayesian Anomaly Score:

[0175] Weighted time series anomaly index:

[0176] w t =e -λ(T-t) ;

[0177] Where S BRS represents the Bayesian anomaly score, that is, the probability of anomaly A occurring given the test result X; P(A) represents the anomaly prior probability of historical data statistics; P(X|A) represents the probability of observing the current data X under an abnormal situation; P(X) represents the total probability density of the current data X; S WTAI represents the weighted time series anomaly index, that is, the degree of abnormal accumulation of recent data; w t Represents the weight coefficient, the closer to the current time point, the higher the weight; A t represents the anomaly score at time t; T represents the observation time window; λ represents the attenuation coefficient, which determines that the data farther away from the current time point T has less impact on the total score;

[0178] S42, based on the risk score S calculated in step S41 BRS , use the adaptive threshold setting method to classify risks and determine the warning level, specifically:

[0179] Grading strategy: Adaptive Risk Stratification (ARS) is used to adjust the alarm threshold based on the mean-variance method of historical abnormal data:

[0180]

[0181]

[0182] Where μ Arepresents the mean of the anomaly score, reflecting the average level of the Bayesian anomaly score of each sample in the historical data; N represents the number of historical data samples; S BRS,i represents the Bayesian anomaly score of the i-th sample; σ A represents the standard deviation of the anomaly score, reflecting the fluctuation or dispersion of the Bayesian anomaly score in historical data;

[0183] Risk level setting:

[0184] Level 1: Low risk, S BRS <μ A ;

[0185] Level 2: Medium risk, μ A ≤S BRS <μ A +(1+α·S WTAI )σ A ;

[0186] Level 3: High risk, S BRS ≥μ A +(1+α·S WTAI )σ A ;

[0187] Among them, α is the weight factor, α>0, that is, when the historical anomalies accumulate more, the system will increase the sensitivity of the risk alert;

[0188] S43. Adopt a graded alert strategy based on the risk level to ensure timely and accurate notifications without disrupting the patient's normal life. The alert strategy grading table is shown in Table 1.

[0189] Table 1 Alert strategy classification table

[0190]

[0191] S44, transmitting the risk level and the corresponding alarm strategy to the intelligent decision support module;

[0192] S5. After obtaining the risk level, the intelligent decision support module makes a decision on the current patient's health status and provides targeted treatment suggestions, specifically:

[0193] S51. Based on the risk classification result in step S4, each monitored object will be assigned a risk level L (low risk, medium risk, high risk), corresponding to specific operational recommendations, and L∈{1,2,3} is defined to represent {low, medium, high risk} respectively;

[0194] S52. Give personalized maintenance suggestions based on risk level L:

[0195] (1) Low risk (L=1): Normal use, regular maintenance:

[0196] Recommendation: Follow the routine care plan and check the catheter status daily, including whether there is leakage or signs of infection;

[0197] Medical staff remind: Pay attention to the fixation of the catheter to avoid catheter dislocation or twisting;

[0198] Monitoring recommendations: Maintain normal monitoring frequency, collect data once every 24 hours, and record the skin condition around the catheter;

[0199] (2) Medium risk (L=2): Increase monitoring frequency and focus on nursing care:

[0200] Possible risks include but are not limited to loosening of catheter fixation, local redness and swelling, slight exudation, and tendency to obstruction;

[0201] Recommendation: Increase the monitoring frequency and check every 6 to 12 hours, paying special attention to whether there is further redness, swelling, or increased pain around the catheter.

[0202] Additional nursing measures include but are not limited to: re-fixing the catheter to ensure stability; changing the dressing in case of slight exudation and performing sterile disinfection; appropriately adjusting the patient's arm movements to avoid excessive pulling on the catheter;

[0203] (3) High risk (L=3): Immediate intervention, emergency treatment:

[0204] Possible risks: including but not limited to severe exudation, large area of redness and swelling, suspected infection, catheter blockage, and catheter displacement;

[0205] Treatment measures include but are not limited to: immediately notifying medical staff and assessing whether the catheter needs to be removed or replaced; using thrombolytics to treat catheter blockage; if signs of infection (including but not limited to local redness, swelling, and fever) occur, the patient can be sent for culture and treated with antibiotics; monitoring the patient's vital signs to prevent serious complications;

[0206] S53. Generate a personalized nursing optimization plan, specifically:

[0207] Adjust the nursing plan individually based on the patient's previous nursing records (including but not limited to early intervention of nursing plans for high-risk patients);

[0208] Adjust the frequency of monitoring data collection, and shorten the data collection interval for high-risk patients;

[0209] Intelligent push nursing instructions remind nurses when to perform specific nursing operations, including but not limited to catheter flushing and dressing changes;

[0210] S54. Automatically record all nursing operations and form a nursing log for doctors and nurses to review. Based on historical data, optimize risk stratification strategies, combine machine learning to summarize the treatment methods of similar cases, and promote the optimal nursing plan.

[0211] S6. The user interaction module visually displays standardized physiological data, test results, risk levels, corresponding alarm strategies, personalized maintenance recommendations, and optimal care plans.

[0212] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A Picc catheter safety monitoring and alarm method based on the Internet of Things, characterized in that: The specific implementation steps include the following: S1, regularly collect Picc catheter environmental data and generate physiological parameter data stream; S2, based on a multi-level fusion denoising method, performs multi-scale decomposition and adaptive threshold denoising on the original signal through wavelet transform and empirical mode decomposition, combines Kalman filtering to optimize the smoothness of the time series signal, uses Z-score normalization to generate standardized data, and generates anti-loss seals based on hash functions and finite field operations; S3: Data reliability is tested based on anti-loss seals. An adaptive hybrid model is constructed, integrating a graph neural network with a Transformer+LSTM model. Patient relationship graphs are used to perform group association analysis and capture long- and short-term temporal features. Contrastive learning is used to construct positive and negative samples to distinguish anomalies. A variational autoencoder is used to reconstruct the error detection anomaly distribution. Knowledge distillation is used to lightweight deploy complex models and output test results. S4. Calculate the comprehensive risk value by combining the weighted time series anomaly index, dynamically adjust the risk threshold, determine the warning level, divide the warning level into three levels of low, medium and high, and adopt a graded alarm strategy; S5. Patients are divided into low, medium, and high risk levels based on risk grading results, and differentiated nursing measures are formulated based on these levels. Monitoring frequency and nursing plans are dynamically adjusted based on patient historical data. Operation reminders are intelligently pushed, and nursing logs are automatically generated simultaneously to optimize risk strategies. S6. Visual display of standardized physiological data, test results, risk levels, corresponding alarm strategies, personalized maintenance recommendations, and optimal care plans.

2. The Picc catheter safety monitoring and alarm method based on the Internet of Things according to claim 1 is characterized in that: The implementation process of the multi-level fusion denoising method is as follows: S21. Use wavelet transform to perform multi-scale decomposition on the original signal, extract different frequency components, perform threshold denoising on the high-frequency components, and use adaptive soft threshold method to dynamically adjust the threshold according to the noise level: Where W f (a, b) represents the signal coefficient after wavelet transformation; f(t) represents the original sensor signal at time t; ψ(t) represents the mother wavelet function; a represents the scale parameter; b represents the translation parameter; represents the wavelet coefficient after denoising; λ represents the adaptive threshold; σ represents the noise standard deviation; N represents the data length; sign() represents the sign function; S22. Decompose the signal f(t) into intrinsic mode functions. Based on spectrum analysis and IMF energy threshold screening, dynamically remove high-frequency noise and low-frequency trend noise: EMD decomposition: Calculate the energy ratio of the IMF components: Set the energy threshold E th , if Ei <E th , it is considered as a noise component and is removed; Where, IMF i (t) represents the i-th natural modal component; r n (t) represents the residual component; n represents the total number of natural mode components; S23. Introduce Kalman filtering, model the relationship between signal and noise through state equation, calculate and update the Kalman gain after predicting the state and covariance, realize signal smoothing and abnormality prediction, and optimize the denoising effect.

3. The method for Picc catheter safety monitoring and alarm based on Internet of Things according to claim 2, characterized in that: The optimization process for optimizing denoising effect is as follows: S31, state space modeling: define the state equation as: X t =Ax t-1 +BU t +W t , Z t =HX t +V t ; Where, X t represents the real signal at the current time t; Z t represents the measurement signal; A represents the state transfer matrix; B and U t Represent the control input matrix and control input respectively; W t 、V t represents process noise and measurement noise; H represents the observation matrix; S32. Forecast state estimation: Prediction covariance matrix: P t - =AP t-1 A T +Q; in, Indicates the predicted state at the current moment; represents the optimal state estimate at the previous moment; P t - represents the prediction error covariance matrix; P t-1 represents the error covariance matrix of the previous moment; Q represents the process noise covariance matrix; S33. Calculate Kalman gain: K t =P t - H T (HP t - H T +R) -1 ; Update the state estimate: Update the covariance matrix: P t =(IK t H)P t - ; Where K t represents the Kalman gain; R represents the measurement noise covariance matrix; T represents the transpose operation of the matrix; represents the optimal state estimate at the current moment, that is, the final denoised signal; P t represents the updated error covariance matrix; I represents the identity matrix; S34. Output optimized denoised data 4. The method for Picc catheter safety monitoring and alarm based on Internet of Things according to claim 1, characterized in that: The process of generating the anti-loss seal is as follows: S41, calculate the information packet Inp=H(data); Where H is a predefined hash function; data is the binary data form of standardized data; S42, select a random number k∈F p , calculate the basic package parameter PB = r + k·Inp mod p; Where p is a predefined large prime number that satisfies p=3mod4; F p is a predefined finite field; r is a predefined random p The selected elements in S43, calculate the derived package parameter PD = s + k·Inp mod p; Where s is a predefined random p The selected elements in S44. Calculate the comprehensive packaging parameter PS = k·(2r+τ)+k 2 ·Inp+ρk mod p; Among them, τ, ρ are predefined packing factors, τ, ρ∈F p ; S45. Generate an anti-loss seal AS = {basic packaging parameters PB, derived packaging parameters PD, comprehensive packaging parameters PS}.

5. The method for Picc catheter safety monitoring and alarm based on Internet of Things according to claim 4, characterized in that: The inspection process for the reliability of the data based on the anti-loss seal is as follows: S51, calculate the information packet Inp′=H(data′); Wherein, data′ is the binary data of the received standardized physiological data data; S52. Construct the following equation: Eq1 = (PB) 2 +τ·PD; Eq2=U+PS·Inp′mod p; Eq3=(PD) 2 +ρ·PB;Eq4=V+PS·Inp′mod p; Among them, U and V are predefined deblocking factors, U = r 2 +τs mod p, V=s 2 +ρr mod p; S53. If Eq1 = Eq2 and Eq3 = Eq4, it means that there is no packet loss in the standardized data, that is, the reliability of the data is guaranteed.

6. The Picc catheter safety monitoring and alarm method based on the Internet of Things according to claim 1 is characterized in that: The detection process of the adaptive hybrid model is as follows: S61. Construct a patient relationship graph, treating each patient as a node in the graph and defining patients who use the same type of PICC catheter and have similar treatment plans as edges. The graph neural network (GNN) captures the potential relationships between different patients by training the nodes and edges, learns the feature representation of each patient, and identifies normal and abnormal patterns. S62, Transformer processes input data through the self-attention mechanism, captures the long-term dependent features in the time series, focuses on the correlation between each time step in the PICC catheter data, and identifies long-term trends; S63 and LSTM are used to capture short-term temporal dependencies and extract short-term information from the short-term fluctuations in the monitoring data of the PICC catheter.

7. The method for Picc catheter safety monitoring and alarm based on Internet of Things according to claim 1, characterized in that: The detection process of abnormal distribution detection through variational autoencoder reconstruction error is as follows: S71. Identify abnormal data without labels through contrastive learning, construct positive and negative samples, and learn how to distinguish between the two: Positive samples: select samples from normal PICC catheter monitoring data; Negative samples: Negative samples are constructed by artificial generation and abnormal catheter monitoring data; S72. Use the variational autoencoder (VAE) to learn the distribution of normal PICC catheter data and determine whether the data is abnormal by calculating the reconstruction error.

8. The method for Picc catheter safety monitoring and alarm based on Internet of Things according to claim 1, characterized in that: The process of classifying warning levels is as follows: S81. Calculate the overall risk score using the Bayesian anomaly score combined with the weighted temporal anomaly index: Bayesian Anomaly Score: Weighted time series anomaly index: w t =e -λ(T-t) ; Where S BRS represents the Bayesian anomaly score, that is, the probability of anomaly A occurring given the test result X; P(A) represents the anomaly prior probability of historical data statistics; P(X|A) represents the probability of observing the current data X under an abnormal situation; P(X) represents the total probability density of the current data X; S WTAI represents the weighted time series anomaly index; w t Represents the weight coefficient; A t represents the anomaly score at time t; T represents the observation time window; λ represents the attenuation coefficient; S82, based on risk score S BRS , use the adaptive threshold setting method to classify risks and determine the warning level: Classification strategy: Use adaptive dynamic classification, based on the mean-variance method of historical abnormal data (adjust the alarm threshold: Where μ A represents the mean of the anomaly score, reflecting the average level of the Bayesian anomaly score of each sample in the historical data; N represents the number of historical data samples; S BRS,i represents the Bayesian anomaly score of the i-th sample; σ A represents the standard deviation of the anomaly score, reflecting the fluctuation or dispersion of the Bayesian anomaly score in historical data; S83. Set risk level: Level 1: Low risk, S BRS <μ A ; Level 2: Medium risk, μ A ≤S BRS <μ A +(1+α·S WTAI )σ A ; Level 3: High risk, S BRS ≥μ A +(1+α·S WTAI )σ A ; Where α is the weight factor, α>0.

9. A Picc catheter safety monitoring and alarm method based on the Internet of Things, which is used to implement the Picc catheter safety monitoring and alarm method based on the Internet of Things according to any one of claims 1 to 8, characterized in that: include: Smart PICC catheter sensor module for monitoring physiological parameters; Edge data processing module, used to pre-process the collected physiological parameter data; A cloud-based data analysis module uses big data analysis and deep learning algorithms to intelligently evaluate PICC catheter usage and identify catheter anomalies. Intelligent alarm and notification module, used to predict potential risks and send alarm information through smart terminals; Intelligent decision support module, which combines historical data with personalized patient information to provide medical advice; User interaction module, used to provide data visualization interface.

Images