An ai system for monitoring physiological indexes of mouse running in real time

Abstract

The application discloses an AI system for monitoring physiological indexes of mouse running in real time, and belongs to the technical field of animal experiment monitoring. The application solves the problem that the prior art cannot acquire data in real time and is difficult to meet the precise research requirement, deploys an edge base station unit around a mouse activity area, verifies a single parameter threshold value, analyzes multi-parameter correlation, combines behavior-physiology integrated anomaly detection, recognizes abnormal behavior-physiology combination, gives a confidence score according to a deviation index, generates an abnormal data packet, divides a priority level according to the confidence score and potential severity of the abnormality, dynamically adjusts a transmission strategy, processes the abnormal data packet by using a time sequence prediction model, and gives an early warning when a prediction value exceeds a safety threshold value, so that critical and important abnormal data can be quickly and preferentially transmitted, abnormal physiological states of the mouse are discovered and handled in time, and the scientificity and reliability of experiments are improved.

Description

Technical Field

[0001] This invention relates to the field of animal experiment monitoring technology, specifically to an AI system for real-time monitoring of physiological indicators of mice running. Background Technology

[0002] In animal experiments, mice are frequently used in studies of exercise physiology, pharmacology, and disease models. Real-time monitoring of physiological indicators in mice during running is crucial for understanding their athletic ability and physiological state.

[0003] However, existing monitoring methods are mostly offline, which cannot acquire data in real time, and are complex to operate, making it difficult to meet the needs of precise research.

[0004] Therefore, not content with existing needs, this paper proposes an AI system for real-time monitoring of physiological indicators of mice running. Summary of the Invention

[0005] The purpose of this invention is to provide an AI system for real-time monitoring of physiological indicators of mice running. By deploying edge base station units around the mouse's activity area, it identifies abnormal behavior-physiological combinations based on single-parameter threshold verification and multi-parameter correlation analysis, combined with behavior-physiology integrated anomaly detection. It assigns confidence scores based on deviation indices and generates abnormal data packets. Priority levels are assigned based on the anomaly's confidence score and potential severity, and the transmission strategy is dynamically adjusted. A time-series prediction model is used to prioritize the processing of abnormal data packets, and an early warning is issued when the predicted value exceeds a safety threshold. This ensures that critical and important abnormal data can be transmitted quickly and preferentially, enabling timely detection and handling of abnormal physiological states in mice, improving the scientific rigor and reliability of experiments, and solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An AI system for real-time monitoring of physiological indicators in mice during running, comprising:

[0008] The multimodal sensing data unit is configured to acquire mouse multimodal data in real time based on multiple sensors and perform front-end preprocessing on the raw signals;

[0009] The edge base station unit is configured to be deployed around the mouse's activity area to receive real-time data streams from multiple sensors; analyze the real-time data streams, identify data with abnormal patterns, and mark the anomalies.

[0010] Set threshold ranges for each physiological parameter; verify whether each physiological parameter exceeds the preset threshold and identify abnormal patterns; analyze the correlation of multiple parameters through a decision tree model, and identify abnormal behavior-physiological combinations by combining behavior-physiological integrated anomaly detection.

[0011] The abnormal behavior-physiological pattern is compared with the historical normal pattern to calculate the deviation index; a confidence score is assigned based on the deviation index to generate abnormal data packets.

[0012] Prioritization is based on anomaly confidence scores and potential severity, and transmission strategies are dynamically adjusted.

[0013] The AI ​​analysis unit is configured to receive and process multimodal data packets from the base station, and prioritize the processing of critical and important abnormal data packets according to the abnormal priority transmission strategy.

[0014] Based on physiological signal characteristics, behavioral signal characteristics, and historical data, a time-series prediction model is used to predict the trajectory of key physiological indicators in the near future, extract data packets whose predicted values ​​exceed the safety threshold, and analyze abnormal fluctuations in physiological indicators.

[0015] Data packets whose predicted values ​​exceed the safety threshold are prioritized for early warning. When the predicted value exceeds the safety threshold, an early warning is issued in advance. Based on the warning information, mice are examined and intervened, and the results are fed back.

[0016] Further, the edge base station unit includes:

[0017] The anomaly detection module is configured to set a reasonable single-parameter threshold range for each physiological parameter based on the normal physiological changes of mice and experimental requirements; based on the single-parameter threshold, it verifies whether each physiological parameter exceeds the preset threshold range, identifies patterns that do not conform to the normal physiological changes, and records the abnormal conditions, including: abnormal parameter name, abnormal value, and time point when the threshold is exceeded.

[0018] The normal correlation patterns between different physiological parameters are identified by using a decision tree model, and whether the correlation patterns between different physiological parameters are abnormal is analyzed based on multi-parameter correlation.

[0019] Rules for detecting abnormal behavior-physiological integration are set, and normal behavior-physiological combination patterns are defined. Based on the detection of abnormal behavior-physiological integration, the behavior classification results are combined with the physiological response patterns for analysis to identify abnormal behavior-physiological combinations.

[0020] The deviation index is calculated by comparing the abnormal behavioral-physiological patterns with the historical normal patterns.

[0021] Each abnormal behavior-physiological combination is assigned a confidence score based on the deviation index, and an abnormal data packet is generated.

[0022] Based on the confidence score and the potential severity of the anomaly, each anomalous data packet is classified into priority levels, including critical anomalies, important anomalies, suspected anomalies, and reference anomalies.

[0023] Based on different anomaly levels, the edge base station dynamically adjusts the transmission strategy for each abnormal data packet.

[0024] Furthermore, based on different anomaly levels, the edge base station dynamically adjusts the transmission strategy for each abnormal data packet, including:

[0025] For critical and abnormal data packets, they are immediately transmitted through a dedicated high-speed channel, enjoying the highest priority and lowest latency guarantee;

[0026] Important and abnormal data packets are placed in a priority queue to ensure transmission within a short time.

[0027] For suspected abnormal data packets, transmit them in sequence while ensuring the transmission of critical and important data;

[0028] After feature extraction and intelligent compression of reference abnormal data packets and normal data packets, they are transmitted in batches during periods of low network load.

[0029] Furthermore, the edge base station unit also includes:

[0030] The data preprocessing module is configured to perform preliminary processing on the received multimodal data, including data cleaning, noise reduction, and feature extraction.

[0031] The timestamp generation module is configured to automatically add a corresponding timestamp to each data packet when receiving multimodal data, thereby generating multimodal data packets with timestamps.

[0032] Furthermore, the AI ​​analysis unit includes:

[0033] The data parsing module is configured to decrypt the received multimodal data packets, parse the data packet content, and check the integrity and timestamp of the data packets;

[0034] The data processing module is configured to prioritize critical and important abnormal data packets based on an anomaly priority transmission strategy and a time-series prediction model, so as to respond to and handle potential abnormal issues in a timely manner.

[0035] Furthermore, the AI ​​analysis unit also includes:

[0036] The feature analysis module is configured to extract physiological and behavioral signal features from multimodal data based on a deep learning model; align the physiological and behavioral signal features along the time axis; and perform correlation analysis with historical data to automatically generate a rule base for association.

[0037] The anomaly correlation module is configured to extract data packets whose predicted values ​​exceed the safety threshold and analyze the abnormal fluctuations in their physiological indicators.

[0038] By combining historical data and behavioral signals, the causes of abnormal fluctuations can be analyzed.

[0039] The system generates early warning information by comprehensively considering abnormal factors, including abnormal fluctuations in physiological indicators, fluctuation amplitude, fluctuation time points, possible cause analysis, and suggested handling measures.

[0040] Furthermore, the AI ​​analysis unit also includes:

[0041] The warning priority division module is configured to divide the warning priority of each data packet whose predicted value exceeds the safety threshold, and classify the predicted value that far exceeds the safety threshold or shows serious abnormal fluctuations as critical warnings.

[0042] Predicted values ​​that are close to the upper or lower limit of the safety threshold, or that show a moderate degree of abnormal fluctuation, are classified as important warnings.

[0043] Predicted values ​​that slightly exceed the safety threshold or show minor abnormal fluctuations are classified as suspected warnings.

[0044] Predicted values ​​that are within the normal range but exhibit slight fluctuations are classified as reference warnings.

[0045] Furthermore, the AI ​​analysis unit also includes:

[0046] The information feedback module is configured to examine and intervene in mice based on early warning information, and to provide feedback on the processing results;

[0047] Based on feedback results and evaluation data, optimize the time series prediction model and threshold settings, and dynamically adjust the thresholds to reduce false alarms.

[0048] Furthermore, the multimodal sensing data unit includes:

[0049] The front-end processing module is configured to amplify the received sensor signals to improve the signal-to-noise ratio; and to use a low-pass filter to remove high-frequency noise to ensure signal purity.

[0050] The sampling and quantization module is configured to sample and quantize the signals from each sensor according to the set sampling frequency and quantization accuracy.

[0051] Furthermore, the multimodal sensing data unit further includes:

[0052] The device association module is configured to select the appropriate communication protocol based on the experimental site to connect multiple sensors with the edge base station.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] In this invention, edge base station units are deployed around the mouse's activity area. Physiological parameters are verified for abnormality based on single-parameter thresholds. Multi-parameter correlations are analyzed using a decision tree model, and abnormal behavior-physiological combinations are identified through behavior-physiological integrated anomaly detection. Anomaly data packets are generated by assigning confidence scores based on deviation indices. Priority levels are assigned based on the anomaly's confidence score and potential severity, and transmission strategies are dynamically adjusted. A time-series prediction model is used to prioritize the processing of anomaly data packets, predicting the trajectory of key physiological indicators in the near future and providing early warnings when predicted values ​​exceed safety thresholds. This not only ensures rapid and prioritized transmission of critical and important anomaly data, improving the timeliness of anomaly identification and processing, but also optimizes resource utilization efficiency, ensuring efficient operation under different network loads. This enables timely detection and processing of abnormal physiological states in mice, providing strong support for animal experimental research and contributing to improved scientific rigor and reliability. Attached Figure Description

[0055] Figure 1 This is a flowchart of the AI ​​system for real-time monitoring of physiological indicators of mice running, as described in this invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] To address the technical issues that existing monitoring methods are mostly offline, unable to acquire data in real time, and are complex to operate, making them unsuitable for precise research, please refer to [link to relevant documentation]. Figure 1 This embodiment provides the following technical solution:

[0058] An AI system for real-time monitoring of physiological indicators in mice during running, comprising:

[0059] A multimodal sensing data unit is configured to acquire real-time multimodal data of mice based on multiple sensors and to perform front-end preprocessing of the raw signals. Specifically, by integrating a heart rate sensor, temperature sensor, respiration sensor, and motion posture sensor into a lightweight animal vest or collar, it is used to acquire real-time data on heart rate changes, body temperature, respiratory rate and depth, and changes in motion posture and gait during mouse running. The multimodal sensing data unit includes:

[0060] The front-end processing module is configured to amplify the received sensor signals to improve the signal-to-noise ratio; and to use a low-pass filter to remove high-frequency noise to ensure signal purity.

[0061] The sampling and quantization module is configured to sample and quantize the signals from each sensor according to the set sampling frequency and quantization accuracy; and convert the analog signals into digital signals for subsequent data transmission and analysis.

[0062] The device association module is configured to select the appropriate communication protocol based on the experimental site to establish communication connections between multiple sensors and the edge base station. For example, in a laboratory environment, Wi-Fi is selected to facilitate medium-distance data transmission and ensure fast data transfer; in a larger experimental site, LoRa is selected to facilitate long-distance, low-power data transmission.

[0063] An edge base station unit is configured to be deployed in the mouse activity area, such as around a treadmill or open field, to receive real-time data streams from multiple sensors; analyze the real-time data streams, identify data with abnormal patterns, and mark the anomalies; analyze multi-parameter correlations using a decision tree model, and identify abnormal behavior-physiological combinations by combining behavioral-physiological integrated anomaly detection; assign confidence scores based on deviation indices and generate abnormal data packets; prioritize anomalies based on their confidence scores and potential severity, and dynamically adjust the transmission strategy; the edge base station unit includes:

[0064] The anomaly detection module is configured to set reasonable single-parameter threshold ranges for each physiological parameter based on the normal physiological changes of mice and experimental requirements. For example, the normal range for the rate of blood glucose decrease is set to no more than 0.5 mmol / L per minute, and the normal range for the acceleration of heart rate increase is no more than 20 beats / minute². Based on the single-parameter thresholds, the module verifies whether each physiological parameter exceeds the preset threshold range, such as the rate of blood glucose decrease and the acceleration of heart rate increase. It identifies patterns that do not conform to the normal physiological change pattern and records the abnormalities, including: the name of the abnormal parameter, the abnormal value, and the time point when the threshold is exceeded.

[0065] The decision tree model is used to clarify the normal correlation patterns between different physiological parameters. Based on the multi-parameter correlation, the correlation patterns between different physiological parameters are analyzed to determine whether they are abnormal. For example, the decoupling between exercise intensity and heart rate response, or the mismatch between blood glucose level and energy consumption are detected. When it is detected that the heart rate does not rise accordingly when the exercise intensity increases, or when there is a significant deviation in the relationship between blood glucose level and energy consumption, it is considered that the multi-parameter correlation is abnormal.

[0066] Rules for detecting abnormal behavior-physiological integration are established, and normal behavior-physiological combination patterns are defined. Based on the detection of abnormal behavior-physiological integration, behavioral classification results are combined with physiological response patterns for analysis to identify abnormal behavior-physiological combinations. For example, detecting high heart rate phenomena in a resting state or the lack of a normal lactate rise response during strenuous exercise; when mice are at rest, high heart rate phenomena are detected, or a normal lactate rise response is not observed during strenuous exercise, etc., to identify abnormal behavior-physiological combinations. The abnormal behavior-physiological patterns are compared with historical normal patterns, and their deviation index is calculated. This index can be comprehensively calculated based on factors such as the number of abnormal parameters, the degree of abnormality, and the duration, reflecting the degree of difference between the abnormal pattern and the normal pattern.

[0067] Each abnormal behavior-physiological combination is assigned a confidence score based on a deviation index, and an abnormal data packet is generated. Based on the confidence score and the potential severity of the anomaly, each abnormal data packet is prioritized into critical, important, suspected, and reference anomalies. Each abnormal data packet is appended with detailed anomaly description metadata, including: anomaly type, start time, severity, list of involved parameters, and confidence score. The edge base station dynamically adjusts the transmission strategy for each abnormal data packet according to its anomaly level. Critical abnormal data packets are immediately transmitted via a dedicated high-speed channel, enjoying the highest priority and lowest latency. Important abnormal data packets are placed in a priority queue to ensure transmission within a short time. Suspected abnormal data packets are transmitted sequentially while ensuring the transmission of critical and important data. Reference abnormal data packets and normal data packets undergo feature extraction and intelligent compression, and are transmitted in batches during periods of low network load. During transmission, abnormal data packets are encrypted and their integrity is verified to ensure data security and reliability. Simultaneously, the transmission status and timestamp of each abnormal data packet are recorded for subsequent tracking and analysis.

[0068] The data preprocessing module is configured to perform preliminary processing on the received multimodal data, including data cleaning, noise reduction, and feature extraction.

[0069] The timestamp generation module is configured to automatically add a corresponding timestamp to each data packet when receiving multimodal data, generating multimodal data packets with timestamps to ensure data time synchronization.

[0070] The AI ​​analysis unit is configured to receive and process multimodal data packets from the base station. Based on physiological signal characteristics, behavioral signal characteristics, and historical data, it uses a time-series prediction model to predict the trajectory of key physiological indicators in the near future and provides early warnings when the predicted values ​​exceed safety thresholds. The AI ​​analysis unit includes:

[0071] The data parsing module is configured to decrypt the received multimodal data packets, parse the packet content, check the integrity and timestamp of the packets, and ensure the timeliness and accuracy of the data.

[0072] The data processing module is configured to prioritize critical and important abnormal data packets based on an anomaly priority transmission strategy and a time-series prediction model, so as to respond to and handle potential abnormal issues in a timely manner.

[0073] The feature analysis module is configured to extract physiological and behavioral signal features from multimodal data based on a deep learning model. For example, it can distinguish between heart rate increases caused by running and those caused by stress at the signal source. It can also align physiological signal features such as blood glucose and heart rate with behavioral signal features such as steady running, acceleration, turning, and resting on the time axis, and perform correlation analysis with historical data to automatically generate a rule library for association.

[0074] The anomaly correlation module is configured to extract data packets whose predicted values ​​exceed a safety threshold and analyze abnormal fluctuations in their physiological indicators; for example, a rapid increase or decrease in heart rate exceeding a certain range within a short period of time, such as an increase or decrease of more than 50 beats per minute; combining historical data and behavioral signals to analyze the causes of abnormal fluctuations; for example, determining whether the fluctuations are caused by strenuous exercise, stress, or other pathological factors; and generating early warning information by comprehensively considering abnormal factors, including the physiological indicators of abnormal fluctuations, fluctuation amplitude, fluctuation time points, possible cause analysis, and suggested treatment measures.

[0075] The warning priority classification module is configured to classify warning priorities for each data packet whose predicted value exceeds the safety threshold. Predicted values ​​far exceeding the safety threshold or exhibiting severe abnormal fluctuations are classified as critical warnings, which may pose an immediate threat to the health of mice; for example, a heart rate exceeding 650 beats / minute or blood glucose below 1.5 mmol / L requires immediate action. Predicted values ​​close to the upper or lower limit of the safety threshold, or exhibiting moderate abnormal fluctuations, are classified as important warnings; for example, a heart rate close to 600 beats / minute or blood glucose close to 10 mmol / L requires action within a short period. Predicted values ​​slightly exceeding the safety threshold, or exhibiting minor abnormal fluctuations, are classified as suspected warnings; these may be caused by experimental errors or other non-pathological factors and should be handled while ensuring the handling of critical and important warnings. Predicted values ​​within the normal range but exhibiting minor fluctuations are classified as reference warnings, mainly used for recording and subsequent analysis.

[0076] The information feedback module is configured to examine and intervene in mice based on early warning information and provide feedback on the processing results; based on the feedback results and evaluation data, it optimizes the time-series prediction model and threshold settings, and dynamically adjusts the thresholds to reduce false alarms; for example, it adjusts model parameters to improve prediction accuracy and ensures that the system can adapt to new experimental conditions and individual differences in mice.

[0077] The beneficial effects achieved by the above are as follows: By deploying edge base station units around the mouse activity area, physiological parameters are verified for abnormality based on single-parameter thresholds; multi-parameter correlations are analyzed using a decision tree model; abnormal behavior-physiological combinations are identified by combining behavior-physiological integrated anomaly detection; confidence scores are assigned based on deviation index to generate abnormal data packets; priority levels are assigned based on the confidence scores and potential severity of the anomalies, and transmission strategies are dynamically adjusted; a time-series prediction model is used to prioritize the processing of abnormal data packets, predict the trajectory of key physiological indicators in the near future, and provide early warnings when predicted values ​​exceed safety thresholds; this not only ensures that critical and important abnormal data can be transmitted quickly and preferentially, improving the timeliness of anomaly identification and processing, but also optimizes resource utilization efficiency, ensuring efficient operation under different network loads; thus, it enables timely detection and processing of abnormal physiological states in mice, providing strong support for animal experimental research and helping to improve the scientific rigor and reliability of experiments.

[0078] Working principle: Real-time acquisition of various physiological and behavioral data from mice, followed by preprocessing to ensure data accuracy and usability; identification and labeling of abnormal patterns through single-parameter thresholding and multi-parameter correlation analysis, combined with behavioral-physiological integrated anomaly detection; assignment of confidence scores, prioritization based on severity, and transmission to the AI ​​analysis unit according to different strategies; the AI ​​analysis unit prioritizes high-priority abnormal data and generates early warning information; based on the severity of the anomaly, warnings are categorized into four levels: critical, important, suspected, and reference, enabling researchers to take timely measures; through hierarchical processing and dynamic adjustment, real-time monitoring and anomaly warning of mice's physiological state are achieved, improving experimental efficiency and accuracy.

[0079] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or high-voltage switchgear that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or high-voltage switchgear.

[0080] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI system for real-time monitoring of physiological indicators of mouse running, characterized in that, include: The multimodal sensing data unit is configured to acquire mouse multimodal data in real time based on multiple sensors and perform front-end preprocessing on the raw signals; An edge base station unit is configured to be deployed around the mouse's activity area to receive real-time data streams from multiple sensors; Analyze real-time data streams, identify data with abnormal patterns, and mark the anomalies. Set threshold ranges for each physiological parameter; Verify whether each physiological parameter exceeds a preset threshold and identify abnormal patterns; Decision tree model is used to analyze the correlation of multiple parameters, and abnormal behavior-physiological combinations are identified by combining behavior-physiological integration anomaly detection. The abnormal behavior-physiological pattern is compared with the historical normal pattern to calculate the deviation index; a confidence score is assigned based on the deviation index to generate abnormal data packets. Prioritization is based on anomaly confidence scores and potential severity, and transmission strategies are dynamically adjusted. The AI ​​analysis unit is configured to receive and process multimodal data packets from the base station, and prioritize the processing of critical and important abnormal data packets according to the abnormal priority transmission strategy. Based on physiological signal characteristics, behavioral signal characteristics, and historical data, a time-series prediction model is used to predict the trajectory of key physiological indicators in the near future. Extract data packets whose predicted values ​​exceed the safety threshold and analyze abnormal fluctuations in physiological indicators; Data packets whose predicted values ​​exceed the security threshold are prioritized for early warning, and early warnings are issued when the predicted values ​​exceed the security threshold. Based on the early warning information, the mice were examined and intervened, and the results were reported back.

2. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 1, characterized in that, The edge base station unit includes: The anomaly detection module is configured to set a reasonable single-parameter threshold range for each physiological parameter based on the normal physiological changes of mice and experimental requirements; based on the single-parameter threshold, it verifies whether each physiological parameter exceeds the preset threshold range, identifies patterns that do not conform to the normal physiological changes, and records the abnormal conditions, including: abnormal parameter name, abnormal value, and time point when the threshold is exceeded. The normal correlation patterns between different physiological parameters are identified by using a decision tree model, and whether the correlation patterns between different physiological parameters are abnormal is analyzed based on multi-parameter correlation. Rules for detecting abnormal behavior-physiological integration are set, and normal behavior-physiological combination patterns are defined. Based on the detection of abnormal behavior-physiological integration, the behavior classification results are combined with the physiological response patterns for analysis to identify abnormal behavior-physiological combinations. The deviation index is calculated by comparing the abnormal behavioral-physiological patterns with the historical normal patterns. Each abnormal behavior-physiological combination is assigned a confidence score based on the deviation index, and an abnormal data packet is generated. Based on the confidence score and the potential severity of the anomaly, each anomalous data packet is classified into priority levels, including critical anomalies, important anomalies, suspected anomalies, and reference anomalies. Based on different anomaly levels, the edge base station dynamically adjusts the transmission strategy for each abnormal data packet.

3. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 2, characterized in that, Based on different anomaly levels, the edge base station dynamically adjusts the transmission strategy for each abnormal data packet, including: For critical and abnormal data packets, they are immediately transmitted through a dedicated high-speed channel, enjoying the highest priority and lowest latency guarantee; Important and abnormal data packets are placed in a priority queue to ensure transmission within a short time. For suspected abnormal data packets, transmit them in sequence while ensuring the transmission of critical and important data; After feature extraction and intelligent compression of reference abnormal data packets and normal data packets, they are transmitted in batches during periods of low network load.

4. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 2, characterized in that, The edge base station unit further includes: The data preprocessing module is configured to perform preliminary processing on the received multimodal data, including data cleaning, noise reduction, and feature extraction. The timestamp generation module is configured to automatically add a corresponding timestamp to each data packet when receiving multimodal data, thereby generating multimodal data packets with timestamps.

5. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 3, characterized in that, The AI ​​analysis unit includes: The data parsing module is configured to decrypt the received multimodal data packets, parse the data packet content, and check the integrity and timestamp of the data packets; The data processing module is configured to prioritize critical and important abnormal data packets based on an anomaly priority transmission strategy and a time-series prediction model, so as to respond to and handle potential abnormal issues in a timely manner.

6. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 5, characterized in that, The AI ​​analysis unit also includes: The feature analysis module is configured to extract physiological and behavioral signal features from multimodal data based on a deep learning model; align the physiological and behavioral signal features along the time axis; and perform correlation analysis with historical data to automatically generate a rule base for association. The anomaly correlation module is configured to extract data packets whose predicted values ​​exceed the safety threshold and analyze the abnormal fluctuations in their physiological indicators. By combining historical data and behavioral signals, the causes of abnormal fluctuations can be analyzed. The system generates early warning information by comprehensively considering abnormal factors, including abnormal fluctuations in physiological indicators, fluctuation amplitude, fluctuation time points, possible cause analysis, and suggested handling measures.

7. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 6, characterized in that, The AI ​​analysis unit also includes: The warning priority division module is configured to divide the warning priority of each data packet whose predicted value exceeds the safety threshold, and classify the predicted value that far exceeds the safety threshold or shows serious abnormal fluctuations as critical warnings. Predicted values ​​that are close to the upper or lower limit of the safety threshold, or that show a moderate degree of abnormal fluctuation, are classified as important warnings. Predicted values ​​that slightly exceed the safety threshold or show minor abnormal fluctuations are classified as suspected warnings. Predicted values ​​that are within the normal range but exhibit slight fluctuations are classified as reference warnings.

8. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 7, characterized in that, The AI ​​analysis unit also includes: The information feedback module is configured to examine and intervene in mice based on early warning information, and to provide feedback on the processing results; Based on feedback results and evaluation data, optimize the time series prediction model and threshold settings, and dynamically adjust the thresholds to reduce false alarms.

9. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 1, characterized in that, The multimodal sensing data unit includes: The front-end processing module is configured to amplify the received sensor signals to improve the signal-to-noise ratio; and to use a low-pass filter to remove high-frequency noise to ensure signal purity. The sampling and quantization module is configured to sample and quantize the signals from each sensor according to the set sampling frequency and quantization accuracy.

10. The AI ​​system for real-time monitoring of physiological indicators of mouse running according to claim 9, characterized in that, The multimodal sensing data unit further includes: The device association module is configured to select the appropriate communication protocol based on the experimental site to connect multiple sensors with the edge base station.

Images