A method and system for monitoring medical equipment data

Abstract

The present invention belongs to the field of data analysis, and provides a method and system for monitoring information of medical equipment data, which obtains historical data of different types of medical equipment; performs correlation analysis on the historical data of different types of medical equipment to obtain a first related equipment group; obtains current detection data of the first related equipment group, and uses Kalman filtering to cross-validate the detection data of the first related equipment group for each patient; when the detection data of patients greater than a preset proportion fail to cross-validate, it is determined that an abnormal device appears in the first related equipment group. The above scheme can improve the sensitivity of medical equipment data monitoring.

Description

Technical Field

[0001] The present invention belongs to the field of data analysis, and in particular relates to an information monitoring method and system for medical device data. Background Art

[0002] With the development of modern medical technology, various medical devices are widely used in hospitals and home care to monitor patients' physiological parameters and assist doctors in diagnosis and treatment. For example, electrocardiographs (ECG), blood pressure monitors (ABPM), blood oxygen saturation meters (SpO 2 ) and other equipment, which can collect key physiological data such as heart rate, blood pressure, and blood oxygen saturation. These medical devices provide real-time monitoring and data recording of the patient's physiological state, providing an important reference for doctors' clinical judgment and diagnosis and treatment.

[0003] However, during the long-term use of medical equipment, problems such as sensor aging, calibration inaccuracy, and external interference often lead to distortion of measurement data, affecting doctors' diagnosis results. In complex medical environments, doctors and caregivers are faced with a large amount of equipment data, which often needs to be manually monitored and analyzed. If there are slight fluctuations or abnormalities in the equipment data, it may be difficult for medical staff to detect them in time, especially when the patient's own physiological state is already significantly abnormal, and small abnormalities in the equipment are more difficult to detect. This not only increases the workload of doctors, but may also delay the discovery and handling of equipment problems. Summary of the invention

[0004] In order to solve the problems in the prior art, the present invention provides an information monitoring method for medical device data, the method comprising the following steps:

[0005] Obtain historical data for different types of medical devices;

[0006] Performing correlation analysis on historical data of different types of medical equipment to obtain a first related equipment group;

[0007] Acquire current detection data of a first related device group, and use Kalman filtering to cross-validate the detection data of the first related device group for each patient;

[0008] When the detection data of patients greater than a preset proportion fail the cross-validation, it is determined that an abnormal device appears in the first related device group.

[0009] Another aspect of the present invention provides an information monitoring system for medical device data, comprising the following modules:

[0010] Acquisition module, used to obtain historical data of different types of medical equipment;

[0011] An analysis module, used for performing correlation analysis on historical data of different types of medical devices to obtain a first related device group;

[0012] A verification module, used for obtaining current detection data of the first related device group, and cross-verifying the detection data of the first related device group of each patient using Kalman filtering;

[0013] The detection module is used to determine that an abnormal device appears in the first related device group when the detection data of patients greater than a preset proportion fail cross-validation.

[0014] The present invention uses correlation analysis to cross-validate data from different types of medical devices, and can effectively identify inconsistencies in data between devices. The Kalman filter algorithm is used to fuse and correct data from multiple devices, reducing data deviations caused by single device failures or measurement errors, thereby improving the accuracy and reliability of device detection data.

[0015] The present invention monitors the device data in the first related device group in real time and detects the data consistency between the devices based on the cross-validation results. When the test data of multiple patients fail the cross-validation and exceed the preset ratio, the system automatically identifies the abnormal device and issues an alarm. This mechanism can timely detect potential failures or performance degradation of the device, and avoid the device working in an abnormal state for a long time affecting the diagnosis result.

[0016] The present invention is compatible with different types of medical equipment, including electrocardiographs, blood pressure monitors, blood oxygen saturation meters, etc., to achieve data linkage monitoring of multiple devices. This method is not only applicable to data verification of the same type of equipment, but can also establish effective correlation analysis between different types of equipment, thus expanding the application scope of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0019] Below, the invention is preferably described in conjunction with the accompanying drawings and specific implementation methods.

[0020] This embodiment solves the above problem through the following steps:

[0021] In one embodiment, reference Figure 1 The present invention provides an information monitoring method for medical device data, which aims to comprehensively and accurately monitor and analyze various types of data collected by medical devices to ensure the reliability of device operation and the accuracy of patient physiological data. Medical devices include but are not limited to electrocardiographs (ECG), blood pressure monitors (ABPM), blood oxygen saturation meters (SpO 2 ) and other devices are used to obtain key physiological parameters of patients, such as heart rate, blood pressure, and blood oxygen. By monitoring and analyzing the data from these devices, medical staff can understand the patient's physiological status in real time and make corresponding clinical decisions based on these data.

[0022] Information monitoring in the present invention refers to the process of acquiring, analyzing, processing and detecting abnormalities of data generated by medical devices. This monitoring method can not only capture the detection data generated in real time by various medical devices, but also further verify and analyze these data to ensure their accuracy and consistency, and avoid errors caused by equipment failure, sensor failure or data interference. By comprehensively analyzing the data of multiple devices, this method can achieve early warning of equipment abnormalities and improve the accuracy of clinical diagnosis.

[0023] The medical device data of the present invention covers the raw data collected by various medical devices and the processed physiological parameters. There are various types of medical devices, and the data output by them may be in the form of time series, discrete data or other structured information. Therefore, the present invention designs a monitoring method suitable for different types of data to ensure the validity and reliability of the data.

[0024] Get historical data on different types of medical devices.

[0025] In the implementation of the method of the present invention, the recorded data generated by different types of medical devices in a certain period of time in the past are extracted and collected. The medical devices may include but are not limited to electrocardiographs (ECG), blood pressure monitors (ABPM), blood oxygen saturation meters (SpO 2 ), ventilators, injection pumps and other medical monitoring equipment. These devices can monitor different physiological parameters and record the monitoring results in the form of digital or analog signals in the storage unit inside the device.

[0026] Historical data refers to time series data or other types of data information that are continuously generated and stored during the operation of the device, including but not limited to heart rate, blood pressure, blood oxygen saturation, respiratory rate, drug infusion volume, etc. These historical data can be processed high-level data or raw data collected by the device. By obtaining these data, the present invention can analyze the historical performance of the medical device, identify whether the device has abnormalities or deviations during operation, and provide a basic basis for subsequent data verification and anomaly detection.

[0027] The acquisition process includes retrieving these historical data from the local storage of the device, the hospital information management system (HIS), the medical Internet of Things (IoMT) platform, the cloud database or other distributed data storage systems. The data can be acquired through a wired or wireless data transmission protocol, such as extracting data through an interface standard (such as HL7, DICOM or FHIR protocol) to ensure that all device data can be uniformly read and stored.

[0028] Correlation analysis is performed on historical data of different types of medical equipment to obtain a first related equipment group.

[0029] During the implementation of the method of the present invention, by analyzing the statistical relationship between the historical data collected by multiple medical devices, a device combination with a high correlation between these devices, i.e., a first correlated device group, is identified. The analysis process aims to find devices that can verify or corroborate the accuracy of their monitoring data with each other, so as to achieve multi-source data fusion and anomaly detection.

[0030] The correlation analysis mentioned above refers to the mathematical statistical analysis of historical data from different types of medical devices. Specific methods may include but are not limited to Pearson Correlation Coefficient, Spearman's Rank Correlation Coefficient, Mutual Information, Dynamic Time Warping (DTW), etc. Through correlation analysis, the linear or nonlinear relationship between device data can be measured, and the degree of mutual influence between physiological parameters monitored by different devices can be quantified, so as to find out the interrelated device groups.

[0031] Since the data formats, sampling frequencies, units, etc. generated by different medical devices may be different, directly analyzing these data may lead to inaccurate results. Therefore, the present invention implements a data preprocessing step to ensure that the data can be correctly analyzed and compared.

[0032] By removing missing values, duplicate values, and outliers from the data, the integrity and consistency of the data are ensured. For example, if the monitoring data of an electrocardiograph has abnormal fluctuations, the system will identify and process these data through outlier detection algorithms (such as box plot method and Z-Score method).

[0033] Different devices may have different sampling frequencies. For example, the sampling frequency of electrocardiograph data is higher, while that of blood pressure monitor is lower. In order to make these data comparable in time, interpolation technology is used to up-sample low-frequency data or down-sample high-frequency data to ensure that data from all devices can be analyzed synchronously based on the same time axis.

[0034] Different devices may record the same physiological parameter in different units, and these units need to be converted during the data preprocessing stage. For example, heart rate data can be unified into "beats per minute" (bpm) as a comparison standard.

[0035] After completing the data preprocessing, the present invention performs correlation analysis on the historical data of different types of medical devices. Correlation analysis aims to measure the statistical relationship between the data of different devices, quantify their relevance, and identify device combinations that can be cross-validated. Due to the large amount of historical data, the present invention adopts a distributed computing method based on big data to improve the analysis efficiency.

[0036] In order to cope with the high computational complexity of big data, the present invention adopts a distributed computing framework (such as Spark or MapReduce) to distribute the correlation calculation tasks to multiple computing nodes for parallel execution, so that a large amount of medical device data can be processed in a short time.

[0037] Alternatively, the Pearson correlation coefficient can be used to calculate the linear correlation between two sets of data. It is suitable for combinations of devices where the data present a linear relationship. For example, the heart rate and pulse data of an electrocardiograph and a blood pressure monitor usually present a strong linear correlation.

[0038] Alternatively, the Spearman rank correlation coefficient can be used, which is suitable for data with nonlinear relationships, especially when the relationship between the devices is not strictly linear. For example, the correlation between the data of a blood pressure monitor and a blood oxygen saturation meter may be nonlinear, which can be effectively analyzed by the Spearman coefficient.

[0039] Optionally, mutual information can be used to capture complex nonlinear relationships between device data. Mutual information quantifies the amount of shared information between two random variables and can identify deeper interactions between devices.

[0040] Based on the above correlation analysis, the present invention classifies different devices into different device groups according to the strength of the correlation, wherein the device combination with a correlation coefficient higher than a set threshold is called the first related device group. The data of these device groups have significant mutual correlation and can verify each other's accuracy.

[0041] The setting of correlation threshold can be determined by historical experience data or machine learning models. For example, the Pearson correlation coefficient r>0.75 is set as a strong correlation standard, and the threshold of the Spearman coefficient is set as , the mutual information threshold is dynamically adjusted according to the distribution of the data.

[0042] The automatic generation of equipment groups through big data technology can not only significantly reduce manual intervention, but also ensure the accuracy of equipment group division. For example, the correlation coefficient between electrocardiograph and blood pressure monitor is 0.85, while the correlation coefficient between blood oxygen saturation meter and them is 0.55. Then electrocardiograph and blood pressure monitor are automatically classified as the first related equipment group, while blood oxygen saturation meter does not belong to this group.

[0043] Assume that a patient in a hospital has an electrocardiograph (ECG), a blood pressure monitor (ABPM), and a blood oxygen saturation meter (SpO 2 ). Through correlation analysis of a large amount of historical data, it is found that the Pearson correlation coefficient between the heart rate data and pulse data of the electrocardiograph and the blood pressure monitor is 0.88, and through dynamic time warping, it is found that the time series trends of the two are consistent, while the mutual information between the blood oxygen saturation meter and the two is low, and the correlation is insufficient. Therefore, the system classifies the electrocardiograph and the blood pressure monitor into the first related equipment group, while the blood oxygen saturation meter is not included in this group.

[0044] In this way, the present invention can automatically identify device combinations that can be cross-validated based on big data analysis, laying the foundation for subsequent device data consistency verification and anomaly detection.

[0045] Get current detection data of the first related device group.

[0046] In order to obtain the current detection data from the first related device group, the present invention obtains the latest detection data generated by each device in real time through the data interface and communication protocol of each device. The device interface can be wired (such as USB, Ethernet) or wireless communication (such as Wi-Fi, Bluetooth, ZigBee), and follows the medical data communication standard protocol (such as HL7, DICOM, FHIR, etc.) to ensure the reliability and security of data transmission.

[0047] Since the collection time of different devices may be inconsistent, in order to ensure the timeliness and consistency of the data, the present invention adopts a time synchronization mechanism, such as using the Network Time Protocol (NTP), to synchronize the clocks of each device in the first related device group so that its detection data can be extracted within the same time window. This synchronization operation ensures that when comparing data across devices, all data sources are from the same time point, avoiding comparison errors caused by time differences.

[0048] When each device completes data collection, the device will package the detection data into a standardized data packet and transmit it to the data management center through a data stream. The present invention will use distributed data stream technologies such as message queues (such as Kafka, RabbitMQ) to achieve real-time extraction and storage of the detection data of the first related device group. The data stream can be time series data or discrete data, depending on the type of device and the collection method of monitoring data.

[0049] Different types of medical equipment may generate data in different formats. In order to ensure the comparability of data from different devices, the present invention implements the unification and standardization of data formats. For example, an electrocardiograph may generate continuous time series data, while a blood pressure monitor generates discrete periodic data. Therefore, when acquiring the current test data, the system will convert the format and unify the data, including unifying the data units to the same standard, such as unifying the heart rate to beats per minute (bpm) and the blood pressure to millimeters of mercury (mmHg).

[0050] When acquiring the current detection data, the present invention implements a real-time error checking mechanism to ensure the integrity and accuracy of the data during transmission. This step includes checking and error correction of the data packet, such as using a cyclic redundancy check (CRC) or a hash function to verify whether an error occurs during data transmission. At the same time, through data integrity verification, it is ensured that the detection data of each device obtained from the device group is the latest and valid.

[0051] Kalman filtering is used to cross-validate the detection data of the first related device group for each patient.

[0052] In the present invention, each patient's physiological parameters (such as heart rate, blood pressure, blood oxygen saturation, etc.) are regarded as the hidden state of a dynamic system. Each patient always uses the same set of equipment, such as using the same equipment to measure heart rate, blood pressure, blood oxygen saturation, etc. during treatment.

[0053] Assume that the patient's actual physiological state is (such as real heart rate, blood pressure, blood oxygen saturation), the detection data of each medical device at the Kth moment is the observed value , then the state equation and observation equation of the dynamic system can be expressed as:

[0054] Equation of state: ,

[0055] Where A is the state transfer matrix, is the system status at the previous moment, is the process noise, assuming , N represents normal distribution, Q is the process noise covariance matrix.

[0056] Observation equation: ,

[0057] Among them, H is the observation matrix, which defines the measurement model of different devices. is the observation noise, assuming , N represents normal distribution, and R is the observation noise covariance matrix.

[0058] At each moment K, the Kalman filter first estimates the state of the previous moment. and the state equation predicts the current system state and the covariance matrix ,

[0059] Prediction status: ,

[0060] Prediction covariance matrix: ,

[0061] Using the current detection data obtained from the first related equipment group Update, compare the device data with the predicted state through the observation equation, and calculate the updated state estimate and covariance update ,

[0062] Kalman gain: ,

[0063] Status Update: ,

[0064] Covariance update: ,

[0065] Through this step, the Kalman filter can perform weighted fusion on the data of different devices, filter out the observation noise, and obtain the optimal state estimate at the current moment through recursive updating.

[0066] In the present invention, Kalman filtering is used to fuse and verify the detection data of multiple devices (such as electrocardiographs and blood pressure monitors). There is a significant inconsistency between the residuals of the Kalman filter This indicates that the detection data between devices are inconsistent, which may indicate an abnormality or failure of a certain device.

[0067] If the absolute value of the residual If the preset threshold is exceeded, the system will consider that the data in the device group does not meet the cross-validation criteria and thus determine that the device is abnormal.

[0068] When the detection data of patients greater than a preset proportion fail the cross-validation, it is determined that an abnormal device appears in the first related device group.

[0069] Specifically, the present invention cross-validates the test data of each patient. If the data between multiple devices in a related device group are inconsistent and exceed a set threshold, the test data of the patient is deemed to have failed the cross-validation. When multiple patients use the same device group and the proportion of their test data that simultaneously fails the cross-validation is greater than a preset threshold, the system will comprehensively evaluate and determine that one or more devices in the device group may have a fault or abnormality.

[0070] For each patient, the system obtains test data from the first related device group and performs cross-validation through Kalman filtering. Exceeding the preset threshold 𝜖, that is:

[0071] ,

[0072] If the patient's test data fails the cross-validation, the system will record the cross-validation failure result for the patient.

[0073] The preset ratio refers to the proportion of patients who failed cross-validation among a certain number of patients. This ratio can be set based on historical data, device usage environment, and clinical needs. For example, the preset ratio threshold is set to 20%, that is, when more than 20% of the patients fail to pass the cross-validation test data when using the same device group, the system will trigger further abnormal device detection.

[0074] The system will analyze the data of all patient groups using the same first-related device group and count how many patients failed the cross-validation process.

[0075] When the failure ratio exceeds the preset ratio threshold, the system will further analyze the specific devices in the first related device group to identify possible abnormal devices. The specific device abnormality determination can be achieved through the following two methods:

[0076] Analyze each device in the relevant device group one by one, and judge the reliability of each device by the size of the residual in the cross-validation process. For example, calculate the degree to which the data of each device deviates from the system state estimate. If the deviation of a device is continuously large, the device is considered an abnormal device.

[0077] By reviewing the historical test data and maintenance records of the device and combining it with the current cross-validation failure, it is determined whether a device has abnormal conditions in multiple patients. For example, if an electrocardiograph shows abnormal heart rate measurement in multiple patients, the electrocardiograph is determined to be an abnormal device.

[0078] When an abnormal device is identified in the equipment group, the system will automatically trigger an alarm mechanism to notify medical personnel to check, calibrate or replace the device. At the same time, the system will record the identification process and fault details of the abnormal device for subsequent tracking and maintenance.

[0079] On the other hand, the present invention also provides an information monitoring system for medical device data, comprising the following modules:

[0080] Acquisition module, used to obtain historical data of different types of medical equipment;

[0081] An analysis module, used for performing correlation analysis on historical data of different types of medical devices to obtain a first related device group;

[0082] A verification module, used for obtaining current detection data of the first related device group, and cross-verifying the detection data of the first related device group of each patient using Kalman filtering;

[0083] The detection module is used to determine that an abnormal device appears in the first related device group when the detection data of patients greater than a preset proportion fail cross-validation.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0085] The present invention does not particularly specify the structure of some modules, which shall be subject to the contents recorded in the prior art. The prior art mentioned in the above background technology section and the specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters. The scope of protection of the present invention shall be subject to the contents actually recorded in the claims.

Claims

1. A method for monitoring medical device data, characterized in that: The method comprises the following steps: Obtain historical data for different types of medical devices; Performing correlation analysis on historical data of different types of medical equipment to obtain a first related equipment group; Acquire current detection data of a first related device group, and use Kalman filtering to cross-validate the detection data of the first related device group for each patient; When the test data of patients greater than a preset proportion fail the cross-validation, it is determined that an abnormal device appears in the first related device group; The cross-validation of the detection data of the first related equipment group of each patient by using Kalman filtering comprises: Assume the patient's true physiological state is x k , the detection data of each medical device at the kth moment is the observation value z k , then the state equation and observation equation of the system are expressed as: Equation of state: x k =Ax k-1 +w k Among them, A is the state transfer matrix, x k-1 is the system state at the previous moment, w k is the process noise, w k ~N(0,Q), N represents normal distribution, Q is the process noise covariance matrix; observation equation: z k =Hx k +v k Among them, H is the observation matrix, v k is the observation noise, v k ~N(0,R), N represents the normal distribution, R is the observation noise covariance matrix; At each time point k, the Kalman filter first estimates the state of the previous time point. and the state equation predicts the current system state and the covariance matrix Prediction status: Prediction covariance matrix: Using the current detection data z obtained from the first related equipment group k Update, compare the device data with the predicted state through the observation equation, and calculate the updated state estimate and covariance update P k Kalman gain: Status Update: Covariance update: Kalman filtering performs weighted fusion of data from different devices, filters out observation noise, and obtains the optimal state estimate at the current moment through recursive updating; The residual of the Kalman filter is If the absolute value of the residual |e k |If the preset threshold is exceeded, the system will consider that the data in the device group does not meet the cross-validation criteria and thus determine that the device is abnormal.

2. The information monitoring method for medical device data according to claim 1, characterized in that: The correlation analysis includes using: Pearson correlation coefficient or Peelman rank correlation coefficient or mutual information.

3. The information monitoring method for medical device data according to claim 1, characterized in that: When acquiring the current detection data of the first related device group, the network time protocol is used to synchronize the clocks of the devices in the first related device group so that the detection data can be extracted within the same time window.

4. The method for monitoring medical device data according to claim 1, characterized in that: After determining that an abnormal device appears in the first related device group, each device in the related device group is analyzed one by one. During the cross-validation process, the reliability of each device is determined by the size of the residual.

5. An information monitoring system for medical equipment data, characterized in that: The system includes the following modules: Acquisition module, used to obtain historical data of different types of medical equipment; An analysis module, used for performing correlation analysis on historical data of different types of medical devices to obtain a first related device group; A verification module, used for obtaining current detection data of the first related device group, and cross-verifying the detection data of the first related device group of each patient using Kalman filtering; A detection module, configured to determine that an abnormal device appears in the first related device group when the detection data of patients greater than a preset proportion fail cross-validation; The cross-validation of the detection data of the first related equipment group of each patient by using Kalman filtering comprises: Assume the patient's true physiological state is x k , the detection data of each medical device at the kth moment is the observation value z k , then the state equation and observation equation of the system are expressed as: Equation of state: x k =Ax k-1 +w k Among them, A is the state transfer matrix, x k-1 is the system state at the previous moment, w k is the process noise, w k ~N(0,Q), N represents the normal distribution, Q is the process noise covariance matrix; Observation equation: z k =Hx k +v k Among them, H is the observation matrix, v k is the observation noise, v k ~N(0,R), N represents the normal distribution, R is the observation noise covariance matrix; At each time point k, the Kalman filter first estimates the state of the previous time point. and the state equation predicts the current system state and the covariance matrix Prediction status: Prediction covariance matrix: Using the current detection data z obtained from the first related equipment group k Update, compare the device data with the predicted state through the observation equation, and calculate the updated state estimate and covariance update P k Kalman gain: Status Update: Covariance update: Kalman filtering performs weighted fusion of data from different devices, filters out observation noise, and obtains the optimal state estimate at the current moment through recursive updating; The residual of the Kalman filter is If the absolute value of the residual |e k |If the preset threshold is exceeded, the system will consider that the data in the device group does not meet the cross-validation criteria and thus determine that the device is abnormal.

6. The information monitoring system for medical equipment data according to claim 5, characterized in that: The correlation analysis includes using: Pearson correlation coefficient or Peelman rank correlation coefficient or mutual information.

7. The information monitoring system for medical equipment data according to claim 5, characterized in that: When acquiring the current detection data of the first related device group, the network time protocol is used to synchronize the clocks of the devices in the first related device group so that the detection data can be extracted within the same time window.

8. The information monitoring system for medical device data according to claim 5, characterized in that: After determining that an abnormal device appears in the first related device group, each device in the related device group is analyzed one by one. During the cross-validation process, the reliability of each device is determined by the size of the residual.

Images