Artificial liver equipment operation data acquisition and analysis system and method

Abstract

The invention provides an artificial liver device operation data acquisition and analysis system and method, and relates to the field of medical information.The method comprises the steps that firstly, operation parameters of an artificial liver device are captured in real time, and standardized data streams are formed through time sequence alignment and noise filtering; then, deep feature extraction is carried out on flow and pressure time sequences, dynamic correlation of pressure-flow parameters is established by adopting a hidden space sparse constraint mechanism, and nonlinear abnormal coupling features between the flow and the pressure time sequences are identified; and finally, accurate positioning of equipment abnormity is realized through feature analysis so as to break through limitation of traditional static threshold value monitoring, early warning is performed in advance through time sequence dynamic offset features in early stages such as pipeline micro leakage and progressive blockage, multiple abnormal coupling effects in a composite fault are effectively distinguished, and the fault diagnosis accuracy is improved. The timeliness and accuracy of equipment operation monitoring are remarkably improved, and the safety and controllability of the artificial liver treatment process are guaranteed.

Description

Technical Field

[0001] The present application relates to the field of medical information, and more particularly, in an embodiment of the present application, to an artificial liver device operation data collection and analysis system and method. Background Art

[0002] As a core treatment device for replacing liver function in vitro, the operation status of artificial liver equipment is directly related to the reliability of blood purification treatment and the life safety of patients. In clinical applications, the pipeline system involved in the operation of the equipment needs to maintain key parameters such as accurate plasma separation flow and stable transmembrane pressure. Any minor pipeline blockage, leakage or component performance drift may cause treatment interruption or even medical accidents.

[0003] Traditional monitoring systems usually use a fixed threshold alarm mechanism, but this static monitoring method has significant limitations in the face of complex and changing dynamic operating environments: on the one hand, the operating parameters of artificial liver equipment have strong time-series correlation characteristics, and simple point-like threshold judgments cannot capture the dynamic coupling relationship between parameters; on the other hand, early abnormalities in the pipeline system often manifest as implicit time-series offsets of flow-pressure parameters. When such abnormalities have not yet exceeded the preset threshold, it is difficult for traditional systems to identify their potential risk patterns. Especially when dealing with complex faults such as progressive pipeline blockage caused by high-viscosity blood and intermittent efficiency attenuation of circulating pumps, existing technologies lack the ability to deeply interact with the full time-series characteristics of multiple parameters, which can easily lead to early warning delays or misjudgments and cannot meet the stringent requirements of clinical treatment for equipment operation stability.

[0004] Therefore, an optimized artificial liver device operation data acquisition and analysis scheme is desired. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiment of the present application provides an artificial liver device operation data acquisition and analysis system and method, which first captures the artificial liver device operation parameters in real time, and forms a standardized data stream through time alignment and noise filtering. Then, the flow and pressure time series are deeply extracted and the implicit spatial sparse constraint mechanism is used to establish the dynamic correlation of pressure-flow parameters, and the nonlinear abnormal coupling characteristics between the two are identified. Finally, the accurate positioning of equipment abnormalities is achieved through feature analysis to break through the limitations of traditional static threshold monitoring. In the early stages of pipeline micro-leakage and progressive blockage, early warning is given through the dynamic offset characteristics of the time series, which effectively distinguishes the multiple abnormal coupling effects in complex faults, and significantly improves the timeliness and accuracy of equipment operation monitoring, ensuring the safety and controllability of the artificial liver treatment process.

[0006] According to one aspect of the present application, a system for collecting and analyzing operation data of an artificial liver device is provided, comprising: The artificial liver device operation data acquisition and storage module is used to collect the artificial liver device operation data using the flow sensor and pressure sensor deployed in the artificial liver device, and store it in the operation data monitoring database according to the timestamp; A data preprocessing module, used for extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; A fine-grained interaction analysis module is used to interactively analyze the pre-processed flow time queue and the pre-processed pressure time queue to obtain pressure-flow time series fine-grained interaction coding features, wherein the fine-grained interaction analysis module is used to: process the pre-processed flow time queue and the pre-processed pressure time queue in a manner of full-time sparse constraint fine-grained interaction based on time series implicit association to obtain pressure-flow time series fine-grained interaction coding features; The abnormality judgment module is used to determine whether there is an abnormality in the management system of the artificial liver device based on the pressure-flow time series fine-grained interactive coding characteristics.

[0007] According to another aspect of the present application, a method for collecting and analyzing operation data of an artificial liver device is provided, comprising: Using the flow sensor and pressure sensor deployed in the artificial liver device to collect the operation data of the artificial liver device, and storing it in the operation data monitoring database according to the timestamp; Extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; Interactively analyzing the preprocessed flow time queue and the preprocessed pressure time queue to obtain pressure-flow time series fine-grained interactive coding features, including: processing the preprocessed flow time queue and the preprocessed pressure time queue using a full-time sparse constraint fine-grained interactive method based on time series implicit association to obtain pressure-flow time series fine-grained interactive coding features; Based on the pressure-flow time series fine-grained interactive coding features, it is determined whether there is an abnormality in the management system of the artificial liver device.

[0008] Compared with the prior art, the present application provides an artificial liver device operation data acquisition and analysis system and method, which first captures the artificial liver device operation parameters in real time, and forms a standardized data stream through time alignment and noise filtering. Then, deep feature extraction is performed on the flow and pressure time series, and an implicit spatial sparse constraint mechanism is used to establish a dynamic association between pressure-flow parameters, and identify the nonlinear abnormal coupling characteristics between the two. Ultimately, the accurate positioning of equipment abnormalities is achieved through feature analysis to break through the limitations of traditional static threshold monitoring. In the early stages of pipeline micro-leaks and progressive blockages, early warnings are given through the dynamic offset characteristics of the time series, effectively distinguishing the multiple abnormal coupling effects in complex faults, and significantly improving the timeliness and accuracy of equipment operation monitoring, ensuring the safety and controllability of the artificial liver treatment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 4 is a system block diagram of an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application.

[0011] Figure 2 Schematic diagram of data flow of an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application.

[0012] Figure 3 4 is a block diagram of a fine-grained interactive analysis module in an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application.

[0013] Figure 4 The block diagram is a pressure-flow time series fine-grained interaction unit in the artificial liver device operation data acquisition and analysis system according to an embodiment of the present application.

[0014] Figure 5 The block diagram is a pressure-flow time series fine-grained aggregation subunit in the artificial liver device operation data acquisition and analysis system according to an embodiment of the present application.

[0015] Figure 6 The present invention is a flowchart of a method for collecting and analyzing operation data of an artificial liver device according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0017] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0018] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0019] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0020] In clinical applications of artificial liver devices, the pipeline system needs to accurately control key parameters such as plasma separation flow and transmembrane pressure. Any slight blockage, leakage or component performance drift may lead to treatment interruption or medical accidents. Traditional monitoring systems use fixed threshold alarms, but this static method cannot capture the dynamic coupling relationship between equipment operating parameters, especially when facing early hidden anomalies, such as timing deviations of flow and pressure. Traditional systems find it difficult to identify potential risks. Therefore, the existing technology lacks in-depth analysis of multi-parameter timing characteristics, which can easily lead to early warning delays or misjudgments and cannot meet the strict requirements of clinical treatment for equipment stability.

[0021] In response to the above technical problems, the technical concept of this application is to build an intelligent perception system for the operation status of artificial liver equipment through multi-dimensional time series modeling and cross-parameter dynamic correlation analysis. Specifically, firstly, the flow and pressure sensors are used to capture the operation parameters of artificial liver equipment in real time, and a standardized data stream is formed through time series alignment and noise filtering; then the flow and pressure time series are deeply extracted to capture the gradual abnormal pattern implicit in the parameter evolution (such as the slow drop in flow caused by pipeline blockage); then the implicit spatial sparse constraint mechanism is used to establish a dynamic association between pressure-flow parameters in the feature space, and identify the nonlinear abnormal coupling characteristics between the two (such as sudden pressure changes accompanied by abnormal flow fluctuations); finally, the accurate positioning of equipment abnormalities is achieved through feature analysis. This solution breaks through the limitations of traditional static threshold monitoring, and can provide early warnings through time series dynamic offset characteristics in the early stages of pipeline micro-leakage and progressive blockage, while effectively distinguishing the multiple abnormal coupling effects in complex faults, significantly improving the timeliness and accuracy of equipment operation monitoring, and ensuring the safety and controllability of the artificial liver treatment process.

[0022] The present application proposes an artificial liver device operation data acquisition and analysis system. Figure 1 4 is a system block diagram of an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of data flow of an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the artificial liver device operation data acquisition and analysis system 100 includes: an artificial liver device operation data acquisition and storage module 110, which is used to use the flow sensor and pressure sensor deployed in the artificial liver device to collect the artificial liver device operation data, and store it in the operation data monitoring database according to the timestamp; a data preprocessing module 120, which is used to extract the artificial liver device operation data from the operation data monitoring database, and perform time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; a fine-grained interaction analysis module 130, which is used to perform interaction analysis on the preprocessed flow time queue and the preprocessed pressure time queue to obtain a pressure-flow time series fine-grained interaction coding feature; an abnormality judgment module 140, which is used to determine whether there is an abnormality in the management system of the artificial liver device based on the pressure-flow time series fine-grained interaction coding feature.

[0023] In the artificial liver device operation data acquisition and analysis system 100, the artificial liver device operation data acquisition and storage module 110 is used to collect artificial liver device operation data using flow sensors and pressure sensors deployed in the artificial liver device, and store them in the operation data monitoring database according to timestamps. It should be understood that the flow sensors and pressure sensors deployed in the artificial liver device can capture subtle changes during the operation of the device, including but not limited to key indicators such as the precise flow of plasma separation and transmembrane pressure, providing a solid foundation for subsequent data analysis. In addition, since the artificial liver device involves a complex dynamic operating environment, there is a strong temporal correlation characteristic between its operating parameters, and any slight change may indicate a potential problem. Therefore, by accurately recording the flow and pressure values ​​at each time point, the actual operating state of the device can be more accurately reflected. Storing the collected data according to timestamps not only helps to maintain the temporal order consistency of the data, but also facilitates temporal alignment and noise filtering during subsequent processing, thereby forming a standardized data stream. This is particularly critical for identifying abnormal conditions such as early pipeline micro-leakage or progressive blockage. In addition, this timestamp-based data storage method allows the retrospective analysis of historical data to identify abnormal patterns even when faced with problems such as instantaneous measurement deviations caused by vibration of the internal circulation pump of the device, changes in blood composition, or zero drift of the sensor itself. For example, in the case of partial blockage of the pipeline, although the initial pressure change may be very small, this change trend will gradually emerge over time. By comparing data at different time points, such slowly developing faults can be more acutely detected, and timely measures can be taken to avoid more serious consequences. Specifically, flow sensors and pressure sensors are the core components of data acquisition. They are carefully designed and placed in key parts of artificial liver devices. Flow sensors are used to accurately measure the amount of blood flowing through the pipeline system. Their working principle is based on the detection of physical effects caused by liquid flow (such as thermal effects, changes in ultrasonic propagation speed, or electromagnetic induction), and can provide high-precision flow readings. Pressure sensors measure the pressure inside the system by sensing the force of the fluid on its sensitive element, usually using capacitive, piezoresistive or resonant technologies. The combination of the two can fully cover the two most critical parameters during the operation of the artificial liver device, namely flow and pressure, and provide a solid foundation for subsequent data analysis. Once the sensors start working, they will continue to generate a continuous stream of data. These raw data contain information about the working status of the device at every moment, but due to the limitations of the sensor's own characteristics, such as signal noise, zero drift, etc., directly using these unprocessed data for analysis may lead to misjudgment. Therefore, in actual operation, a series of measures need to be taken to ensure that the collected data is both true and reliable.On the one hand, it is necessary to ensure that the sensor is installed in the right position to avoid measurement errors caused by external interference; on the other hand, the sensor needs to be calibrated regularly to compensate for performance deviations caused by environmental factors or long-term operation. Next, when the flow sensor and pressure sensor generate data, they will send the data to the central control system through a specific communication protocol (such as RS-485, Modbus or CAN bus, etc.). In this process, each data packet will be accompanied by an accurate timestamp, which is generated by the clock inside the device to mark when the data was collected. In this way, all data are ensured to be arranged correctly in the order of occurrence, and it is convenient to query and trace the device status within a specific time period in the future. In particular, when an abnormal situation occurs, the root cause of the problem can be located by viewing the data records of the corresponding time period. After arriving at the central control system, these timestamped data will be further processed and stored in the operation data monitoring database. The design of the database must fully consider the characteristics of large data volume and frequent updates. Usually, a high-performance relational database management system (RDBMS), such as MySQL or PostgreSQL, or a non-relational database (NoSQL), such as MongoDB, is selected to facilitate efficient management and retrieval of massive data. Before data is stored in the database, some necessary preprocessing steps need to be performed, such as removing duplicates, filling missing values, and format conversion, to ensure that each record in the database is complete and standardized. In addition, in order to improve query efficiency, the timestamp field is often indexed so that users can quickly locate the data within the date range of interest. It is worth noting that as the artificial liver device continues to operate, the amount of data in the database will continue to grow. This requires the system to have good scalability and maintenance capabilities, not only to be able to cope with the increasing demand for data storage, but also to ensure stability and security under long-term operation. To this end, in addition to adopting an efficient database management strategy, a data archiving mechanism is implemented, that is, migrating older historical data to low-cost storage media, while keeping the latest active data easily accessible. This can save storage costs and maintain the efficient operation of the system. In this way, through precise sensor layout, effective data transmission methods, and scientific data storage solutions, not only real-time monitoring of the device status is achieved, but also the foundation for subsequent in-depth analysis of the device health status is laid.

[0024] In an embodiment of the present application, the data preprocessing module 120 is used to extract the artificial liver device operation data from the operation data monitoring database, and perform time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue, including: extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting on the data according to timestamps to obtain a flow time queue and a pressure time queue; performing data preprocessing on the flow time queue and the pressure time queue to obtain the preprocessed flow time queue and the preprocessed pressure time queue.

[0025] Specifically, the operation data of the artificial liver device is extracted from the operation data monitoring database, and the operation data is time-sequenced according to the timestamp to obtain the flow time queue and the pressure time queue. It should be understood that the flow and pressure data collected by the sensor often have problems such as timestamp dispersion, sampling frequency difference or data packet transmission delay. Due to the dynamic characteristics of the pipeline system during the operation of the device (such as changes in blood viscosity, circulation pump speed adjustment), the timing correlation between parameters has non-uniform lag characteristics. If the original collected unaligned data is directly used for analysis, the phase misalignment of the flow and pressure signals may be caused, thereby concealing the true dynamic coupling relationship between the two. For example, when the pipeline is partially blocked, the response of the pressure sensor may lag behind the flow change. If the timing is not strictly aligned, the model will not be able to accurately capture this cross-parameter causal correlation pattern, resulting in failure of anomaly detection. In addition, occasional data loss or abnormal timestamps during the operation of the device (such as timing confusion caused by sensor clock drift) will also destroy the continuity of the parameter sequence and affect the stability of subsequent timing modeling. Based on this, the present application arranges the artificial liver device operation data in time sequence according to the timestamp to obtain the flow time queue and the pressure time queue. In particular, in a specific example of the present application, after extracting the original operation data from the database, the flow and pressure data are arranged in time sequence according to a unified time reference, including timestamp calibration, missing value interpolation, and equal interval resampling. This step reconstructs the discrete sensor data into a flow time queue and a pressure time queue that are strictly arranged in time order, ensuring that the two parameter sequences are completely aligned in the time dimension, and that the physical events corresponding to each data point have a clear causal time sequence association. For example, for the lag of pressure data caused by data transmission delay, the timestamp compensation mechanism is used to re-match it with the flow data of the same period to eliminate the timing noise introduced by communication jitter.

[0026] Specifically, the flow time queue and the pressure time queue are preprocessed to obtain the preprocessed flow time queue and the preprocessed pressure time queue. It should be understood that the original acquisition data of the flow and pressure sensors often contain a variety of interference factors: high-frequency noise of the signal caused by the vibration of the internal circulation pump of the equipment, instantaneous measurement deviation caused by changes in blood components, and baseline fluctuations caused by the zero drift of the sensor itself. These noises will mask the true trend of parameter changes. For example, the pressure drop caused by micro-leakage in the pipeline may be submerged by random noise, and the abnormal flow fluctuation caused by the sudden change of blood viscosity is easily confused with the pseudo signal generated by sensor drift. In addition, the problem of inconsistent numerical dimensions caused by differences in sampling mechanisms of different sensors (such as pulse counting of flow sensors and analog quantity acquisition of pressure sensors) will affect the effect of subsequent time series encoding due to differences in characteristic scales if they are directly input into the model without processing, thereby reducing the sensitivity of the model to hidden abnormal patterns. Therefore, in order to make the data more accurately reflect the real operating state of the artificial liver device and reduce the impact of data errors on subsequent analysis results, the present application performs data preprocessing on the flow time queue and the pressure time queue to obtain the preprocessed flow time queue and the preprocessed pressure time queue. In a specific embodiment of the present application, a sliding average filtering technique is used to smooth the time series data of flow and pressure to remove high-frequency noise. Specifically, for the data value of each time point in the flow time queue and the pressure time queue, several adjacent data points are taken before and after, and the average value of these data points is calculated as the new value of the time point. In addition, for the interruption of the time series caused by data loss or abnormality, the missing values ​​can be filled by linear interpolation to ensure the continuity of the flow time queue and the pressure time queue. Through such a preprocessing process, a more accurate and stable preprocessed flow time queue and preprocessed pressure time queue are finally obtained, which provides a reliable data basis for subsequent fine-grained interactive analysis.

[0027] In the above-mentioned artificial liver device operation data acquisition and analysis system 100, the fine-grained interaction analysis module 130 is used to interactively analyze the pre-processed flow time queue and the pre-processed pressure time queue to obtain pressure-flow time series fine-grained interaction coding features, wherein the fine-grained interaction analysis module is used to: use the full-time sparse constrained fine-grained interaction method based on time series implicit association to process the pre-processed flow time queue and the pre-processed pressure time queue to obtain pressure-flow time series fine-grained interaction coding features.

[0028] Figure 3 FIG. 4 is a block diagram of a fine-grained interactive analysis module in an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application. Figure 3As shown, in an embodiment of the present application, the fine-grained interaction analysis module 130 includes: a flow pressure sequence encoding unit 131, which is used to perform sequence encoding on the preprocessed flow time queue and the preprocessed pressure time queue to obtain a flow timing implicit pattern feature encoding vector and a pressure timing implicit pattern feature encoding vector; a pressure-flow timing fine-grained interaction unit 132, which is used to perform a full-time implicit fine-grained interaction analysis based on sparse constraints on the flow timing implicit pattern feature encoding vector and the pressure timing implicit pattern feature encoding vector to obtain a pressure-flow timing fine-grained interaction encoding matrix as the pressure-flow timing fine-grained interaction encoding feature.

[0029] More specifically, in an embodiment of the present application, the flow-pressure sequence encoding unit 131 is used to: perform sequence encoding based on a forward LSTM model on the preprocessed flow time queue and the preprocessed pressure time queue to obtain the flow time series implicit pattern feature encoding vector and the pressure time series implicit pattern feature encoding vector.

[0030] It should be understood that, considering that the time series changes of flow and pressure parameters often contain complex dynamic patterns, for example, the initial stage of pipeline blockage may be manifested as a periodic slight attenuation of flow, or a gradual change in the frequency of pressure fluctuations caused by the decline in the efficiency of the circulating pump. Although these abnormal patterns have been freed from noise interference in the preprocessed data, their evolution laws are still implicit in the high-dimensional time series association. Methods based on statistical features or fixed rules are difficult to effectively capture such nonlinear dynamic characteristics. Especially when the equipment is in a sub-healthy state, physiological fluctuations (such as natural changes in the patient's blood viscosity) and pathological deviations (such as local leakage of the pipeline) may exist in the parameter sequence at the same time. The numerical difference between the two at a single time point is small, but the implicit pattern of their time series evolution trend is significantly different, and it is necessary to deeply explore the long-range dependency relationship in the time dimension for accurate distinction. To this end, in the technical solution of the present application, the preprocessed flow time queue and the preprocessed pressure time queue are sequence encoded based on the forward LSTM model to obtain the flow time series implicit pattern feature encoding vector and the pressure time series implicit pattern feature encoding vector. It is worth noting that the forward LSTM model learns the long-term and short-term dependencies in the parameter sequence step by time through the memory gating mechanism, and maps the original time series data into a high-dimensional implicit feature vector. Specifically, for the pre-processed flow time queue, the forward LSTM network analyzes the cumulative change trend of the historical flow values ​​to identify gradual abnormal patterns such as "slow drop in flow-short-term recovery-secondary attenuation"; for the pre-processed pressure time queue, it captures potential fault features such as "gradual expansion of pressure fluctuation amplitude" or "slow baseline shift", so as to better understand the change law of artificial liver equipment operating parameters over time.

[0031] Figure 4FIG. 1 is a block diagram of a pressure-flow time series fine-grained interaction unit in an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application. Figure 4 As shown, in an embodiment of the present application, the pressure-flow time series fine-grained interaction unit 132 is used to perform a full-time implicit fine-grained interaction analysis based on sparse constraints on the flow time series implicit pattern feature coding vector and the pressure time series implicit pattern feature coding vector to obtain a pressure-flow time series fine-grained interaction coding matrix as the pressure-flow time series fine-grained interaction coding feature, including: a feature one-dimensional convolution phase space reconstruction subunit 1321, used to perform a feature one-dimensional convolution phase space reconstruction on the flow time series implicit pattern feature coding vector and the pressure time series implicit pattern feature coding vector to obtain a set of flow local time series implicit pattern feature coding vectors and a set of pressure local time series implicit pattern feature coding vectors; a pressure-flow time series fine-grained aggregation subunit 1322, used to perform an implicit feature fine-grained space dynamic aggregation based on sparse constraints on the set of flow local time series implicit pattern feature coding vectors and the set of pressure local time series implicit pattern feature coding vectors to obtain the pressure-flow time series fine-grained interaction coding matrix.

[0032] It should be understandable that in the operation scenario of artificial liver equipment, complex faults such as progressive pipeline blockage caused by high-viscosity blood and intermittent performance attenuation of circulating pumps frequently occur. This type of fault is not caused by a single parameter abnormality, but the result of the mutual influence and joint action of flow and pressure parameters. For example, micro-leakage in the pipeline will cause periodic slight fluctuations in pressure parameters in the early stage, while the flow parameter will show a slow decline trend after a few minutes. This cross-parameter abnormal correlation not only has nonlinear time delay characteristics (such as pressure fluctuations leading flow changes by 3-5 sampling cycles), but also its interaction mode will show fine-grained differences with the fault type: the performance attenuation of the circulating pump may be manifested as a strong correlation between high-frequency oscillations of pressure and step-by-step decreases in flow, while pipeline blockage presents a weak coupling of pressure baseline elevation and linear attenuation of flow. Traditional methods use simple feature splicing or linear correlation analysis, which cannot capture such interaction details with temporal and spatial heterogeneity, and it is even more difficult to distinguish the differences in coupling paths of different fault modes in the feature space. Therefore, the present application performs a full-time implicit fine-grained interaction analysis based on sparse constraints on the flow time series implicit pattern feature coding vector and the pressure time series implicit pattern feature coding vector to obtain a pressure-flow time series fine-grained interaction coding matrix.

[0033] Specifically, the encoding vector is first reconstructed in phase space through one-dimensional convolution, and the high-dimensional feature vectors of flow and pressure are mapped into a set of codes with local temporal structure to simulate their potential dynamic trajectory in the abstract feature space. Subsequently, a fine-grained dynamic interaction model is established in the implicit space: a fine-grained spatial interaction encoding matrix of pressure and flow characteristics is constructed in a data-driven manner to capture the nonlinear fine-grained interactive correlation pattern between the two in the time dimension, such as the temporal correspondence between the flow rate slow-down trend and the pressure fluctuation amplitude change. In this process, the spatial sparse constraint mechanism applies selective filtering to the fine-grained spatial interaction encoding matrix, suppresses redundant interaction noise, and strengthens the expression weight of key coupling features, such as identifying the abnormal proportional relationship between the pressure decay rate and the flow compensation rate in the pipeline micro-leakage scenario. In this way, the association between the flow and pressure feature encoding vectors at different times can be accurately located. For example, in every second of the equipment operation, it is analyzed how a change in a certain dimension in the flow feature vector triggers the response of the corresponding dimension or other dimensions of the pressure feature vector, thereby establishing a full-time and all-round parameter interaction relationship.

[0034] Specifically, the characteristic one-dimensional convolution phase space reconstruction subunit 1321 is used to perform characteristic one-dimensional convolution phase space reconstruction on the flow time series implicit pattern feature coding vector and the pressure time series implicit pattern feature coding vector to obtain a set of flow local time series implicit pattern feature coding vectors and a set of pressure local time series implicit pattern feature coding vectors, which are expressed as the characteristic one-dimensional convolution phase space reconstruction formula:

[0035] in, is the feature encoding vector of the implicit pattern of traffic time series, is the characteristic encoding vector of the implicit pattern of pressure time series, To reconstruct the characteristic one-dimensional convolution phase space, , , and are the first, second, and third vectors in the set of implicit pattern feature encoding vectors of local traffic time series. and The local temporal implicit pattern feature encoding vector of traffic, is the set of feature encoding vectors of the implicit pattern of local traffic time series, , , and are the first, second, and third vectors in the set of implicit pattern feature encoding vectors of local pressure time series. and The pressure local temporal implicit pattern feature encoding vector, is the set of feature encoding vectors of the implicit pattern of the local pressure time series, and and The length is the same. It should be understood that since the abnormal coupling characteristics of the operating parameters of the artificial liver device often manifest as nonlinear time-series dynamic patterns (such as the time-delayed correlation between the flow rate slow-down and pressure fluctuation caused by pipeline blockage), it is difficult for traditional methods to directly capture such implicit spatiotemporal interaction relationships from static feature vectors. By introducing a one-dimensional convolution operation, the static feature vectors, namely the feature encoding vectors of the implicit pattern of the flow time series and the feature encoding vectors of the implicit pattern of the pressure time series, can be regarded as potential trajectories in the abstract feature space. The local perception mechanism in the form of a sliding window is used to simulate the feature observations under different time delays, thereby reconstructing a set of local time series patterns that can characterize the dynamic evolution of flow and pressure parameters, namely, the set of feature encoding vectors of the implicit pattern of the flow local time series and the set of feature encoding vectors of the implicit pattern of the pressure local time series. This process not only realizes the deep mining of the internal structural information of the original feature encoding vectors of the implicit pattern of the flow time series and the implicit pattern of the pressure time series, but also generates multi-perspective local feature representations under different pattern dimensions through the parallel operation of multiple sets of convolution kernels. In this way, through nonlinear dimensionality increase operations, the scattered isolated feature information can be integrated into a distributed representation with rich semantic information, thereby enhancing the model's perception of the fine-grained interaction between flow and pressure parameters. It is worth noting that the feature redundancy formed during the reconstruction process is not invalid information, but a repeated representation of the original features from different local perspectives. This redundant feature can effectively improve the model's robustness to noise interference and provide a more discriminative interaction pattern foundation for subsequent fine-grained feature aggregation based on sparse constraints.

[0036] Figure 5 FIG. 1 is a block diagram of a pressure-flow time series fine-grained aggregation subunit in an artificial liver device operation data acquisition and analysis system according to an embodiment of the present application. Figure 5As shown, in an embodiment of the present application, the pressure-flow time series fine-grained aggregation subunit 1322 is used to perform sparse constraint-based implicit feature fine-grained spatial dynamic aggregation on the set of the flow local time series implicit pattern feature coding vectors and the set of the pressure local time series implicit pattern feature coding vectors to obtain the pressure-flow time series fine-grained interactive coding matrix, including: flow-pressure implicit feature fine-grained spatial interactive coding secondary subunit 1322-1, used to calculate the set of the flow local time series implicit pattern feature coding vectors and the pressure local time series implicit pattern feature coding vectors. The implicit feature fine-grained spatial interaction coding matrix between each group of corresponding flow local time series implicit pattern feature coding vectors and pressure local time series implicit pattern feature coding vectors in the set of time series implicit pattern feature coding vectors is obtained to obtain a set of flow-pressure implicit feature fine-grained spatial interaction coding matrices; the pressure-flow time series dynamic aggregation secondary subunit 1322-2 is used to dynamically aggregate the set of flow-pressure implicit feature fine-grained spatial interaction coding matrices based on the weighting of the spatial sparse constraint factor to obtain the pressure-flow time series fine-grained interaction coding matrix.

[0037] Specifically, the flow-pressure implicit feature fine-grained spatial interaction coding secondary subunit 1322-1 is used to calculate the implicit feature fine-grained spatial interaction coding matrix between each corresponding group of flow local time series implicit pattern feature coding vectors and pressure local time series implicit pattern feature coding vectors in the set of the flow local time series implicit pattern feature coding vectors and the set of the pressure local time series implicit pattern feature coding vectors to obtain a set of flow-pressure implicit feature fine-grained spatial interaction coding matrices, which is expressed as a flow-pressure implicit feature fine-grained spatial interaction coding formula:

[0038] in, is the transpose operation, and They are and The corresponding weight matrix is, is matrix multiplication, yes and The length of the vector after multiplication, yes and The flow-pressure implicit feature fine-grained spatial interaction coding matrix between them. It should be understood that considering that the traditional method is limited by the expressive power of the preset interaction mode, it is difficult to capture the complex nonlinear time series coupling relationship between the operating parameters of the artificial liver equipment, especially in the face of complex faults such as pipeline micro-leakage and progressive blockage. The abnormal correlation between the parameters often presents fine-grained features such as dynamic delay and dimensional intersection. To this end, by constructing an implicit feature fine-grained space, the interactive learning of flow and pressure features is transformed from explicit rule constraints to data adaptive pattern mining. Specifically, the present application dynamically establishes an association mapping of feature vector pairs (i.e., each group of corresponding flow local time series implicit pattern feature coding vectors and pressure local time series implicit pattern feature coding vectors) in the latent space through query operations, and uses nonlinear transformation to map the original high-dimensional features to the implicit space, so that the small fluctuations of the local time series pattern of the flow and the time series offset of the pressure response form a measurable interactive relationship in the abstract space. The generation of the implicit feature fine-grained spatial interaction coding matrix is ​​essentially to encode the temporal dynamic coupling characteristics of flow and pressure parameters into the numerical strength of the matrix elements. Each element value in the implicit feature fine-grained spatial interaction coding matrix represents the interaction significance of the corresponding feature dimension combination. This design enables the system to automatically learn the most essential interaction mode between parameters, breaking through the a priori assumptions of traditional methods on interaction relationships, and providing a discriminative interactive information foundation for subsequent fine-grained feature aggregation based on sparse constraints, significantly improving the recognition accuracy of early hidden faults and the decoupling ability of complex anomalies.

[0039] In an embodiment of the present application, the pressure-flow time series dynamic aggregation secondary subunit 1322-2 is used to perform dynamic aggregation based on spatial sparse constraint factor weighting on the set of flow-pressure implicit feature fine-grained spatial interaction coding matrices to obtain the pressure-flow time series fine-grained interaction coding matrix, including: performing connection optimization based on phase space projection on each flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of flow-pressure implicit feature fine-grained spatial interaction coding matrices to obtain an optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix. A set of code matrices; performing Frobenius norm-based normalization processing on each optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices to obtain a set of flow-pressure implicit feature fine-grained spatial sparse constraint factors; based on the set of flow-pressure implicit feature fine-grained spatial sparse constraint factors, performing weighted aggregation on the set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices to obtain the pressure-flow time series fine-grained interaction coding matrix.

[0040] Specifically, each flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of flow-pressure implicit feature fine-grained spatial interaction coding matrices is subjected to connection optimization based on phase space projection to obtain a set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices, which is expressed as follows using a connection optimization formula based on phase space projection:

[0041] in, yes The eigenvalues, yes The eigenvalues, The natural constant The exponential function value with base , yes and The flow-pressure implicit characteristic fine-grained relative phase between , , is each eigenvalue in the fine-grained phase matrix of the implicit characteristic of flow-pressure, yes and The flow-pressure implicit characteristic fine-grained phase matrix between yes The inverse matrix of yes The optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix after optimization. It should be understood that since the abnormal coupling characteristics of the artificial liver device pipeline system are often manifested as cross-parameter timing offsets and dimensional cross-effects (such as the time-delayed correlation between pressure mutations and flow rate drops), it is difficult for traditional methods to suppress global redundant noise while retaining local interaction details. By introducing the phase space projection mechanism, the flow-pressure implicit feature fine-grained spatial interaction coding matrix in the implicit feature fine-grained space is mapped to a low-dimensional flat topology, and the dynamic correlation pattern of the feature pairs is reconstructed using the local phase representation analysis. Specifically, by constructing the flow-pressure implicit feature fine-grained phase matrix and combining its inverse matrix to jointly optimize the original flow-pressure implicit feature fine-grained spatial interaction coding matrix, the state space mapping from local nonlinear connection to global flat representation is achieved. This process effectively eliminates redundant nonlinear coupling relationships, so that the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix can significantly improve the discriminability and computational efficiency of feature expression while retaining key abnormal coupling features, providing a high-quality feature interaction foundation for subsequent sparse constraint-based adaptive aggregation, thereby enhancing the early recognition capability of complex faults such as pipeline micro-leakage and progressive blockage.

[0042] Specifically, each optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix is ​​normalized based on the Frobenius norm to obtain a set of flow-pressure implicit feature fine-grained spatial sparse constraint factors, which is expressed as a normalization processing formula based on the Frobenius norm:

[0043] in, To calculate the square of the matrix Frobenius norm, yes The corresponding flow-pressure implicit characteristic fine-grained spatial factor, yes Activation function, yes The corresponding flow-pressure implicit feature fine-grained spatial sparse constraint factor. It should be understood that the distribution characteristics of the elements in the fine-grained spatial interaction coding matrix of the optimized flow-pressure implicit feature are that only a few elements are significant and most of them are close to zero. By introducing the normalization mechanism of the Frobenius norm, the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix can be globally scaled and normalized, so that the element distribution of the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix satisfies the energy conservation constraint and highlights the sparsity characteristics. In this way, a dynamic weight system of feature interaction patterns can be constructed, in which the flow-pressure implicit feature fine-grained spatial sparse constraint factor converts the importance of feature interaction information into a computable numerical indicator by quantifying the sparsity of the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix. In this way, the spatial sparse constraint mechanism is like a feature filter, which can enhance the ability to capture key criteria for complex faults such as pipeline micro-leakage and progressive blockage by suppressing non-critical interaction noise. Moreover, the spatial sparse constraint mechanism, as the control center of adaptive aggregation, can dynamically allocate fusion weights according to the sparsity of each optimized flow-pressure implicit feature fine-grained spatial interaction encoding matrix, so that the final feature representation can adaptively focus on the most discriminative interaction mode. Through the deep integration of matrix norm theory and feature selection mechanism, the refinement of feature interaction modeling is guaranteed, and the transparency of model decision-making is improved through interpretable sparse representation, providing a high-purity feature foundation for the subsequent abnormal mode decoupling based on fine-grained spatial dynamic aggregation.

[0044] Specifically, based on the set of flow-pressure implicit feature fine-grained spatial sparse constraint factors, the set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices is weighted aggregated to obtain the pressure-flow time series fine-grained interaction coding matrix, which is expressed as the pressure-flow time series fine-grained weighted aggregation formula:

[0045] in, It is the pressure-flow time series fine-grained interaction coding matrix. It should be understood that the weighted aggregation mechanism dynamically allocates fusion weights according to the sparsity of each optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix, rather than simply averaging or splicing multiple optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices. Among them, a high sparsity factor indicates that the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix contains more significant abnormal interaction patterns, and therefore is given a higher weight in the aggregation. This dynamic weighting strategy essentially constructs a quality evaluation system for feature interaction information, so that the interaction details of the key criteria dominate the final feature representation. In this way, by suppressing the noise interference of the low-quality optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix, the model's ability to capture key features of complex faults such as pipeline micro-leakage and progressive blockage can be significantly improved. Therefore, based on the dynamic weight allocation mechanism of the flow-pressure implicit feature fine-grained spatial sparse constraint factor, the final pressure-flow time series fine-grained interaction coding matrix has both high discrimination and interpretability, providing a structured feature basis for the precise positioning of abnormal patterns. This technology has made a breakthrough in solving the problem of information imbalance in multi-source heterogeneous feature fusion, enabling the system to perform fine-grained dynamic perception of the equipment operating status in the full time domain.

[0046] In the artificial liver device operation data acquisition and analysis system 100, the abnormality judgment module 140 is used to determine whether the management system of the artificial liver device is abnormal based on the pressure-flow time series fine-grained interactive coding features, including: inputting the pressure-flow time series fine-grained interactive coding matrix into the abnormality judge based on the classifier to determine whether the management system of the artificial liver device is abnormal. It should be understood that the pressure-flow time series fine-grained interactive coding matrix contains rich and detailed interactive information of flow and pressure parameters during the operation of the artificial liver device. However, this information exists in the form of a matrix, and it is difficult to intuitively judge whether the device is abnormal. The abnormality judge based on the classifier has powerful data classification and pattern recognition capabilities, and can perform in-depth analysis on complex data in the matrix and mine hidden abnormal patterns. For example, it is difficult to directly distinguish whether the device is in normal operation or abnormal state by simply observing the matrix values, and the abnormality judge can quickly make accurate judgments by extracting and analyzing the matrix features. In this way, once the abnormality judge identifies an abnormality, the system can immediately issue an early warning signal to remind medical staff and technicians to reduce the risk of treatment interruption or medical accidents caused by equipment failure. Specifically, as a powerful machine learning tool, the classifier can automatically learn from data and identify the difference between normal and abnormal patterns. In the specific implementation process, algorithms such as support vector machine (SVM), random forest or neural network are usually selected to build anomaly detectors. Among them, support vector machine is good at processing high-dimensional data and can still maintain good generalization ability when the number of samples is limited; random forest is known for its good robustness and noise resistance, and can still provide reliable prediction results when facing noisy data sets; neural networks, especially deep learning models, can automatically learn complex nonlinear relationships in data, and have unique advantages in capturing subtle changes in equipment operation. In the classifier-based anomaly detector, the training process requires a large amount of labeled data as a basis, which can be obtained through historical records or generated through simulation experiments. During the training stage, the classifier will continuously adjust the internal parameters according to the input data and its corresponding label (normal or abnormal) to minimize the prediction error. For example, when using a neural network, the weights between the layers are updated through the back propagation algorithm to gradually improve the accuracy of the model in distinguishing different states. As the training progresses, the classifier-based anomaly judge gradually learns to identify abnormal patterns hidden in the data and can make correct judgments when encountering new data in the future. Then, when the pressure-flow time series fine-grained interaction encoding matrix is ​​input into the classifier-based anomaly judge, the anomaly judge will quickly analyze the input data and give a corresponding judgment result, that is, whether the current artificial liver device is in normal working condition.If the judgment result indicates that there is an abnormality, the system will immediately send out an alarm signal to alert medical staff or technicians. This instant feedback mechanism greatly improves the speed of fault response and helps to take timely measures to prevent the problem from getting worse. For example, when early signs of a minor leak in a pipeline are detected, maintenance personnel can be immediately arranged to inspect and repair it, avoiding more serious consequences due to delays.

[0047] In summary, the artificial liver device operation data acquisition and analysis system 100 based on the embodiment of the present application is explained, which first captures the artificial liver device operation parameters in real time, and forms a standardized data stream through time alignment and noise filtering. Then, the flow and pressure time series are deeply extracted and the implicit spatial sparse constraint mechanism is used to establish the dynamic association of pressure-flow parameters, and the nonlinear abnormal coupling characteristics between the two are identified. Finally, the accurate positioning of equipment abnormalities is achieved through feature analysis to break through the limitations of traditional static threshold monitoring. In the early stages of pipeline micro-leakage and progressive blockage, early warning is given through the dynamic offset characteristics of the time series, which effectively distinguishes the multiple abnormal coupling effects in complex faults, and significantly improves the timeliness and accuracy of equipment operation monitoring, ensuring the safety and controllability of the artificial liver treatment process.

[0048] As described above, the artificial liver device operation data acquisition and analysis system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the artificial liver device operation data acquisition and analysis system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the artificial liver device operation data acquisition and analysis system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the artificial liver device operation data acquisition and analysis system 100 can also be one of the many hardware modules of the terminal device.

[0049] Alternatively, in another example, the artificial liver device operation data acquisition and analysis system 100 and the terminal device may also be separate devices, and the artificial liver device operation data acquisition and analysis system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0050] Figure 6 FIG. 1 is a flow chart of a method for collecting and analyzing data of an artificial liver device according to an embodiment of the present application. Figure 6As shown, according to the embodiment of the present application, the artificial liver device operation data collection and analysis method includes: S110, using the flow sensor and pressure sensor deployed in the artificial liver device to collect the artificial liver device operation data, and storing it in the operation data monitoring database according to the timestamp; S120, extracting the artificial liver device operation data from the operation data monitoring database, and performing time series sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; S130, interactively analyzing the preprocessed flow time queue and the preprocessed pressure time queue to obtain a pressure-flow time series fine-grained interactive coding feature, including: using a full-time sparse constraint fine-grained interactive method based on time series implicit association to process the preprocessed flow time queue and the preprocessed pressure time queue to obtain a pressure-flow time series fine-grained interactive coding feature; S140, based on the pressure-flow time series fine-grained interactive coding feature, determining whether the management system of the artificial liver device has an abnormality.

[0051] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned artificial liver device operation data acquisition and analysis method have been described in the above reference. Figures 1 to 5 The artificial liver device operation data acquisition and analysis system has been described in detail, and therefore, its repeated description will be omitted.

[0052] In summary, the method for collecting and analyzing the operation data of the artificial liver device based on the embodiment of the present application is explained, which first captures the operation parameters of the artificial liver device in real time, and forms a standardized data stream through time alignment and noise filtering. Then, the flow and pressure time series are deeply extracted, and the implicit spatial sparse constraint mechanism is used to establish the dynamic association of the pressure-flow parameters, and the nonlinear abnormal coupling characteristics between the two are identified. Finally, the accurate positioning of equipment abnormalities is achieved through feature analysis to break through the limitations of traditional static threshold monitoring. In the early stages of pipeline micro-leakage and progressive blockage, early warning is given through the dynamic offset characteristics of the time series, which effectively distinguishes the multiple abnormal coupling effects in complex faults, and significantly improves the timeliness and accuracy of equipment operation monitoring, ensuring the safety and controllability of the artificial liver treatment process.

[0053] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0054] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0056] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0057] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0058] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

Claims

1. An artificial liver device operation data acquisition and analysis system, characterized in that: include: The artificial liver device operation data acquisition and storage module is used to collect the artificial liver device operation data using the flow sensor and pressure sensor deployed in the artificial liver device, and store it in the operation data monitoring database according to the timestamp; A data preprocessing module, used for extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; A fine-grained interaction analysis module is used to interactively analyze the pre-processed flow time queue and the pre-processed pressure time queue to obtain pressure-flow time series fine-grained interaction coding features, wherein the fine-grained interaction analysis module is used to: process the pre-processed flow time queue and the pre-processed pressure time queue in a manner of full-time sparse constraint fine-grained interaction based on time series implicit association to obtain pressure-flow time series fine-grained interaction coding features; The abnormality judgment module is used to determine whether there is an abnormality in the management system of the artificial liver device based on the pressure-flow time series fine-grained interactive coding characteristics.

2. The artificial liver device operation data acquisition and analysis system according to claim 1, characterized in that: The data preprocessing module is used to: Extracting the artificial liver device operation data from the operation data monitoring database, and arranging the data in time sequence according to timestamps to obtain a flow time queue and a pressure time queue; Data preprocessing is performed on the flow time queue and the pressure time queue to obtain the preprocessed flow time queue and the preprocessed pressure time queue.

3. The artificial liver device operation data acquisition and analysis system according to claim 1, characterized in that: The fine-grained interaction analysis module includes: A flow pressure sequence encoding unit, used for performing sequence encoding on the pre-processed flow time queue and the pre-processed pressure time queue to obtain a flow time series implicit pattern feature encoding vector and a pressure time series implicit pattern feature encoding vector; The pressure-flow time series fine-grained interaction unit is used to perform a full-time implicit fine-grained interaction analysis based on sparse constraints on the flow time series implicit pattern feature coding vector and the pressure time series implicit pattern feature coding vector to obtain a pressure-flow time series fine-grained interaction coding matrix as the pressure-flow time series fine-grained interaction coding feature.

4. The artificial liver device operation data acquisition and analysis system according to claim 3, characterized in that: The flow pressure sequence encoding unit includes: performing sequence encoding based on the forward LSTM model on the pre-processed flow time queue and the pre-processed pressure time queue to obtain the flow time series implicit pattern feature encoding vector and the pressure time series implicit pattern feature encoding vector.

5. The artificial liver device operation data acquisition and analysis system according to claim 4, characterized in that: The pressure-flow time series fine-grained interaction unit includes: A characteristic one-dimensional convolution phase space reconstruction subunit is used to perform characteristic one-dimensional convolution phase space reconstruction on the flow time series implicit pattern characteristic coding vector and the pressure time series implicit pattern characteristic coding vector to obtain a set of flow local time series implicit pattern characteristic coding vectors and a set of pressure local time series implicit pattern characteristic coding vectors; The pressure-flow time series fine-grained aggregation subunit is used to perform sparse constraint-based implicit feature fine-grained spatial dynamic aggregation on the set of the flow local time series implicit pattern feature coding vectors and the set of the pressure local time series implicit pattern feature coding vectors to obtain the pressure-flow time series fine-grained interactive coding matrix.

6. The artificial liver device operation data acquisition and analysis system according to claim 5, characterized in that: The pressure-flow time series fine-grained aggregation subunit includes: The flow-pressure implicit feature fine-grained spatial interaction coding secondary subunit is used to calculate the implicit feature fine-grained spatial interaction coding matrix between each corresponding group of flow local time series implicit pattern feature coding vectors and pressure local time series implicit pattern feature coding vectors in the set of flow local time series implicit pattern feature coding vectors and the set of pressure local time series implicit pattern feature coding vectors, so as to obtain a set of flow-pressure implicit feature fine-grained spatial interaction coding matrices; The pressure-flow time series dynamic aggregation secondary subunit is used to dynamically aggregate the set of the flow-pressure implicit feature fine-grained spatial interaction coding matrices based on the weighting of the spatial sparse constraint factor to obtain the pressure-flow time series fine-grained interaction coding matrix.

7. The artificial liver device operation data acquisition and analysis system according to claim 6, characterized in that: The pressure-flow time series dynamic aggregation secondary subunit is used for: Performing phase space projection-based connection optimization on each flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of flow-pressure implicit feature fine-grained spatial interaction coding matrices to obtain a set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices; Each optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix in the set of the optimized flow-pressure implicit feature fine-grained spatial interaction coding matrix is ​​normalized based on the Frobenius norm to obtain a set of flow-pressure implicit feature fine-grained spatial sparse constraint factors; Based on the set of flow-pressure implicit feature fine-grained spatial sparse constraint factors, the set of optimized flow-pressure implicit feature fine-grained spatial interaction coding matrices is weighted aggregated to obtain the pressure-flow time series fine-grained interaction coding matrix.

8. The artificial liver device operation data acquisition and analysis system according to claim 7, characterized in that: The abnormality judgment module is used to input the pressure-flow time series fine-grained interactive coding matrix into a classifier-based abnormality judgement module to determine whether there is an abnormality in the management system of the artificial liver device.

9. A method for collecting and analyzing operation data of an artificial liver device, characterized in that: include: The flow sensor and pressure sensor deployed in the artificial liver device are used to collect the operation data of the artificial liver device, and the operation data are stored in the operation data monitoring database according to the timestamp; Extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue; Interactively analyzing the preprocessed flow time queue and the preprocessed pressure time queue to obtain pressure-flow time series fine-grained interactive coding features, including: processing the preprocessed flow time queue and the preprocessed pressure time queue using a full-time sparse constraint fine-grained interactive method based on time series implicit association to obtain pressure-flow time series fine-grained interactive coding features; Based on the pressure-flow time series fine-grained interactive coding features, it is determined whether there is an abnormality in the management system of the artificial liver device.

10. The method for collecting and analyzing operation data of an artificial liver device according to claim 9, characterized in that: Extracting the artificial liver device operation data from the operation data monitoring database, and performing time sequence sorting and data preprocessing on the artificial liver device operation data to obtain a preprocessed flow time queue and a preprocessed pressure time queue, including: Extracting the artificial liver device operation data from the operation data monitoring database, and arranging the data in time sequence according to timestamps to obtain a flow time queue and a pressure time queue; Data preprocessing is performed on the flow time queue and the pressure time queue to obtain the preprocessed flow time queue and the preprocessed pressure time queue.

Images