Multi-parameter cooperative detection method and system for medical equipment

Abstract

The invention discloses a medical equipment multi-parameter cooperative detection method and system, and relates to the field of data processing, and the method comprises the steps: building a three-dimensional parameter group matrix through the dynamic work, static performance and environment coupling parameters of equipment, building a double-layer parameter correlation degree model based on an improved fuzzy analytic hierarchy process, and calculating a cross influence coefficient and a cooperative deviation degree; synchronizing equipment and environment data through a high-precision timestamp, and constructing a space-time marked joint data set; extracting multi-parameter time series data features through a convolutional long-short-term memory network, identifying collaborative anomalies by using a dynamic time warping algorithm, and marking an abnormal time window; and in combination with the cross influence weight distribution table, through an adaptive weighting algorithm optimization model, generating an equipment health score and a collaborative optimization suggestion. The method has the advantages that by integrating dynamic and static parameters, environmental data and the AI technology, accurate monitoring, abnormal recognition and early warning of the equipment operation state are achieved, safety is effectively improved, and the fault risk is reduced.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to a method and system for multi-parameter collaborative detection of medical devices. Background Technology

[0002] With the continuous advancement of medical technology, medical devices are becoming increasingly complex in type and function. Ensuring the accuracy, safety, and reliability of these devices during use has become a crucial issue for the medical industry. Traditional medical device testing methods often employ single devices for individual testing, which presents numerous problems. Therefore, multi-parameter collaborative testing methods for medical devices have emerged.

[0003] Current methods on the market typically rely on single parameters or simple anomaly thresholds for judgment, lacking a comprehensive analysis of the complex relationships between various equipment parameters. These methods often overlook the dynamic changes in equipment operating status and the potential impact of environmental factors on equipment performance, resulting in an inability to effectively capture early signs of potential equipment failures. Furthermore, many existing methods cannot perform deep learning analysis on time-series data, leading to low accuracy when processing multi-parameter collaborative data, and a tendency to miss or misjudge anomalies. Current methods also typically only provide basic fault diagnosis after anomaly identification, lacking in-depth analysis of the causes of the faults and optimization suggestions, making it difficult to provide continuous health management support for the equipment. Summary of the Invention

[0004] To improve existing methods and systems, this paper proposes a multi-parameter collaborative detection method and system for medical devices. This method integrates dynamic and static parameters, environmental data, and AI technology to achieve accurate monitoring, anomaly identification, and early warning of device operating status, effectively optimizing device performance, improving safety, and reducing the risk of failure.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Multi-parameter collaborative detection methods for medical devices include: The dynamic working parameter set, static performance parameter set, and environmental coupling parameter set of the medical equipment are obtained through the equipment operation system, and a three-dimensional parameter set matrix is ​​constructed. A two-level parameter correlation model is constructed based on the improved fuzzy hierarchical analysis method. By introducing a dynamic weight adjustment factor and a parameter sensitivity decay function, the cross-influence coefficient and cooperative deviation of each parameter group are calculated, and a cross-influence weight allocation table is constructed. By synchronously collecting equipment operating status data and environmental sensor data using high-precision timestamp alignment technology, a joint dataset with spatiotemporal tags is established. Historical collaborative datasets are input into a convolutional long short-term memory network for training. By embedding a parameter attention mechanism in the memory unit, the fusion features of multi-parameter time series data are extracted. By comparing the similarity between the joint dataset and normal collaborative feature data using the dynamic time warping algorithm, collaborative anomalies between parameters are identified, and anomalous time windows are marked. The weight allocation of the abnormal parameter group is obtained by matching the collaborative abnormal parameters with the cross-influence weight allocation table, and the weights are adjusted in real time by an adaptive weighting algorithm to optimize the two-layer parameter correlation model. The optimized two-layer parameter correlation model is used to calculate the cooperative deviation between parameters, and a health score is generated by combining the equipment health status benchmark curve.

[0006] Preferably, the step of acquiring the dynamic operating parameter set, static performance parameter set, and environmental coupling parameter set of the medical device through the device operation system, and constructing a three-dimensional parameter set matrix specifically includes: Real-time operating parameters, including voltage, current, temperature, and pressure data, are obtained from the operating system of medical equipment. Obtain the equipment's performance parameters during the initial stage of operation, including the equipment's accuracy, stability, and load capacity; Data related to the device's environment is acquired through sensors, including temperature, humidity, radiation intensity, and air quality. The dimensions of the three-dimensional parameter group matrix include parameter type dimension, time series dimension, and spatial distribution dimension; The data is organized and categorized based on the three dimensions, and the elements of the three-dimensional parameter group matrix are filled in.

[0007] Preferably, the construction of the two-layer parameter correlation model based on the improved fuzzy hierarchical analysis method, by introducing a dynamic weight adjustment factor and a parameter sensitivity decay function, calculates the cross-influence coefficient and cooperative deviation of each parameter group, and constructs a cross-influence weight allocation table, specifically including: Based on the obtained three-dimensional parameter group matrix, an upper-level parameter group inter-correlation matrix and a lower-level intra-group parameter correlation matrix are constructed. The upper-level parameter group inter-correlation matrix represents the relationship between each parameter group, and the lower-level intra-group parameter correlation matrix represents the relationship between each parameter within each parameter group. Based on the correlation matrix between upper-level parameter groups and the correlation matrix within lower-level groups, a two-layer fuzzy judgment matrix is ​​constructed by comparing the degree of influence between different parameters and parameter groups using fuzzy hierarchical analysis. By introducing dynamic weighting factors and attenuation factors for each parameter, their influence weights under different times and operating states can be adjusted. Based on the introduction of a weighted factor two-layer fuzzy judgment matrix, the degree of influence between each pair of parameter groups is obtained. The cross-influence coefficient of each parameter group is calculated by normalizing the elements of the fuzzy judgment matrix, and a cross-influence weight allocation table is constructed. Based on the obtained cross-influence coefficients of each parameter group, the deviation is measured by combining the deviations of each parameter within the parameter group. By introducing a decay factor, the cooperative deviation of each parameter group is obtained.

[0008] Preferably, the step of synchronously collecting device operating status data and environmental sensor data using high-precision timestamp alignment technology to establish a joint dataset with spatiotemporal tags specifically includes: Based on the acquisition time of operating status data and environmental sensor data obtained by sensors of each device, the device data and environmental data are accurately synchronized through a high-precision time synchronization protocol, and a precise timestamp is added to each data point. Based on the synchronized data, spatiotemporal markers are added to each data record, including equipment operation cycle phase markers and environmental disturbance event markers. The equipment operation cycle phase markers are marked according to the specific stages of equipment operation, and the environmental disturbance event markers include changes in the external environment that affect equipment operation. Align device operating status data and environmental sensor data with precise timestamps, and construct a joint dataset by combining spatiotemporal tags.

[0009] Preferably, the step of inputting the historical collaborative dataset into a convolutional long short-term memory network for training, and extracting the fusion features of multi-parameter time-series data by embedding a parameter attention mechanism in the memory units, specifically includes: We acquire historical collaborative datasets, train a model combining convolutional layers and LSTM, extract spatial features from the input data through convolutional layers, and then input the extracted features into the LSTM layer to capture long-term temporal dependencies of the data through memory units. By adding a parameter attention layer to the memory cell of ConvLSTM, the importance of each parameter at different time steps is calculated, attention weights are generated, and used to weight the input time-series data. The model's performance is evaluated using a validation set, and the model's hyperparameters are adjusted based on the validation set results. Based on the trained ConvLSTM model, the fusion features of multi-parameter time series data are extracted to obtain normal collaborative feature data.

[0010] Preferably, the step of comparing the similarity between the joint dataset and normal collaborative feature data using a dynamic time warping algorithm to identify collaborative anomalies between parameters and marking the anomalous time window specifically includes: Synchronize the joint dataset with normal collaborative feature data based on timestamps; For each pair of time series based on the joint dataset and normal collaborative feature data, calculate the DTW distance between them; The shortest matching path between time series is obtained using the DTW algorithm, and the similarity between the two is calculated. Based on the analysis results of historical data, a distance threshold is set. If the DTW distance in a certain period exceeds the threshold, the data in that period is different from the normal collaborative feature data, and it is determined to be an abnormal collaborative parameter. The period in question is recorded as an abnormal time window. For each abnormal period, the start and end times of the abnormality are marked with a timestamp.

[0011] Preferably, the step of matching the collaborative anomaly parameters with the cross-influence weight allocation table to obtain the weight allocation of the anomaly parameter group, and adjusting the weights in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model specifically includes: Based on the obtained collaborative anomaly parameters, they are compared with the data in the cross-influence weight allocation table to obtain the weight allocation data of the collaborative anomaly parameters; The weight values ​​of the collaborative anomaly parameters are dynamically updated using a genetic algorithm, and the updated data is compared with the normal collaborative feature data again to calculate the similarity. Based on the weight data of the collaborative parameters that are judged to be normal, the cross-influence weight allocation table data is updated, and the two-layer parameter correlation model is optimized.

[0012] Preferably, the step of calculating the cooperative deviation between parameters based on the optimized two-layer parameter correlation model and generating a health score by combining it with the equipment health status benchmark curve specifically includes: The cooperative deviation between parameters is calculated based on the optimized two-layer parameter correlation model; Based on the equipment's historical operating data, a health baseline curve is established, including the range of key parameter values ​​for the equipment under various operating conditions. By comparing the device's current operating data with a health baseline curve, the deviation is calculated, and a health score is obtained.

[0013] Preferably, the system also includes generating optimization reports that automatically mark parameters with calibration deviations exceeding 5% as red priority; recommend preventative maintenance plans based on equipment usage time; and generate a visual diagnostic interface containing parameter correlation graphs, highlighting abnormal transmission paths in the graphs.

[0014] Furthermore, a multi-parameter collaborative detection system for medical devices is proposed, including: Data acquisition module: The module collects dynamic operating parameters, static performance parameters and environmental coupling parameters of the medical equipment in real time through the equipment operation system and environmental sensors, and constructs a three-dimensional parameter group matrix; Parameter correlation module: The module is based on the improved fuzzy hierarchical analysis method, constructs a two-level parameter correlation model, calculates the cross-influence coefficient and cooperative deviation of each parameter group, and generates a cross-influence weight allocation table; Feature extraction training module: The module trains the historical collaborative dataset through a convolutional long short-term memory network and embeds a parameter attention mechanism to extract fused features of multi-parameter time series data; Collaborative Anomaly Module: This module uses a dynamic time warping algorithm to compare the similarity between the joint dataset and normal collaborative feature data, and identifies and marks collaborative anomalies and anomalous time windows between parameters; Anomaly parameter matching and optimization module: The module matches collaborative anomaly parameters with a cross-influence weight allocation table, and adjusts the weights in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0015] Compared with the prior art, the advantages of the present invention are: By comprehensively utilizing the dynamic and static operating parameters of the equipment, environmental factors, and advanced artificial intelligence technology, a comprehensive monitoring and early warning system for the operational status of medical equipment has been achieved. By constructing a three-dimensional parameter matrix and introducing an improved fuzzy hierarchical analysis method and dynamic weight adjustment mechanism, the interrelationships and their collaborative deviations between parameters can be accurately captured, enabling timely detection of potential equipment failure risks. Furthermore, by combining high-precision timestamp alignment technology and the deep learning capabilities of convolutional long short-term memory networks, fusion features can be extracted from complex time-series data, further improving the accuracy of data analysis. Through dynamic time warping algorithms for data similarity comparison, collaborative anomalies can be accurately identified and abnormal time windows can be marked, providing strong support for equipment health assessment and optimization. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method proposed in this invention; Figure 2 This is a schematic diagram of the construction of the three-dimensional parameter group matrix proposed in this invention; Figure 3 This is a schematic diagram of the two-layer parameter correlation model proposed in this invention; Figure 4 This is a schematic diagram illustrating the construction of the joint dataset proposed in this invention; Figure 5 This is a schematic diagram of the extraction and fusion features proposed in this invention; Figure 6 This is a schematic diagram of the collaborative anomaly identification parameters proposed in this invention; Figure 7 This is a schematic diagram of the optimized two-layer parameter correlation model proposed in this invention; Figure 8 This is a schematic diagram illustrating the generation of health scores proposed in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] A multi-parameter collaborative detection system for medical devices, including: Data acquisition module: The module collects dynamic operating parameters, static performance parameters and environmental coupling parameters of the medical equipment in real time through the equipment operation system and environmental sensors, and constructs a three-dimensional parameter group matrix; Parameter correlation module: The module is based on the improved fuzzy hierarchical analysis method, constructs a two-level parameter correlation model, calculates the cross-influence coefficient and cooperative deviation of each parameter group, and generates a cross-influence weight allocation table; Feature extraction training module: The module trains the historical collaborative dataset through a convolutional long short-term memory network and embeds a parameter attention mechanism to extract fused features of multi-parameter time series data; Collaborative Anomaly Module: This module uses a dynamic time warping algorithm to compare the similarity between the joint dataset and normal collaborative feature data, and identifies and marks collaborative anomalies and anomalous time windows between parameters; Anomaly parameter matching and optimization module: The module matches collaborative anomaly parameters with a cross-influence weight allocation table, and adjusts the weights in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0019] See Figure 1 As shown, the multi-parameter collaborative detection method for medical devices includes: Step 1: Obtain the dynamic operating parameter set, static performance parameter set, and environmental coupling parameter set of the medical equipment through the equipment operation system, and construct a three-dimensional parameter set matrix; Step 2: Construct a two-layer parameter correlation model based on the improved fuzzy hierarchical analysis method. By introducing a dynamic weight adjustment factor and a parameter sensitivity decay function, calculate the cross-influence coefficient and cooperative deviation of each parameter group, and construct a cross-influence weight allocation table. Step 3: Synchronously collect equipment operating status data and environmental sensor data using high-precision timestamp alignment technology to establish a joint dataset with spatiotemporal tags; Step 4: Input the historical collaborative dataset into the convolutional long short-term memory network for training. By embedding a parameter attention mechanism in the memory unit, the fusion features of multi-parameter time series data are extracted. Step 5: By comparing the similarity between the joint dataset and the normal collaborative feature data using the dynamic time warping algorithm, collaborative anomalies between parameters are identified, and anomalous time windows are marked. Step 6: Match the collaborative anomaly parameters with the cross-influence weight allocation table to obtain the weight allocation of the anomaly parameter group, and adjust the weights in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model; Step 7: Calculate the cooperative deviation between parameters based on the optimized two-layer parameter correlation model, and generate a health score by combining it with the equipment health status benchmark curve.

[0020] See Figure 2 As shown, the dynamic operating parameter set, static performance parameter set, and environmental coupling parameter set of the medical equipment are obtained through the equipment operation system, and a three-dimensional parameter set matrix is ​​constructed, specifically including: Real-time operating parameters, including voltage, current, temperature, and pressure data, are obtained from the operating system of medical equipment. Obtain the equipment's performance parameters during the initial stage of operation, including the equipment's accuracy, stability, and load capacity; Data related to the device's environment is acquired through sensors, including temperature, humidity, radiation intensity, and air quality. The dimensions of the three-dimensional parameter group matrix include parameter type dimension, time series dimension, and spatial distribution dimension; The data is organized and categorized based on the three dimensions, and the elements of the three-dimensional parameter group matrix are filled in.

[0021] See Figure 3 As shown, a two-layer parameter correlation model is constructed based on the improved fuzzy hierarchical analysis method. By introducing a dynamic weight adjustment factor and a parameter sensitivity decay function, the cross-influence coefficient and cooperative deviation of each parameter group are calculated, and a cross-influence weight allocation table is constructed, specifically including: Based on the obtained three-dimensional parameter group matrix, an upper-level parameter group inter-correlation matrix and a lower-level intra-group parameter correlation matrix are constructed. The upper-level parameter group inter-correlation matrix represents the relationship between each parameter group, and the lower-level intra-group parameter correlation matrix represents the relationship between each parameter within each parameter group. Based on the correlation matrix between upper-level parameter groups and the correlation matrix within lower-level groups, a two-layer fuzzy judgment matrix is ​​constructed by comparing the degree of influence between different parameters and parameter groups using fuzzy hierarchical analysis. By introducing dynamic weighting factors and attenuation factors for each parameter, their influence weights under different times and operating states can be adjusted. Based on the introduction of a weighted factor two-layer fuzzy judgment matrix, the degree of influence between each pair of parameter groups is obtained. The cross-influence coefficient of each parameter group is calculated by normalizing the elements of the fuzzy judgment matrix, and a cross-influence weight allocation table is constructed. Based on the obtained cross-influence coefficients of each parameter group, the deviation is measured by combining the deviations of each parameter within the parameter group. By introducing a decay factor, the cooperative deviation of each parameter group is obtained.

[0022] Specifically, based on the acquired three-dimensional parameter group matrix, the upper-level parameter group correlation matrix represents the mutual influence relationship between different parameter groups, such as the dynamic working parameter group, the static performance parameter group, and the environmental coupling parameter group. For each parameter group, an intra-group parameter correlation matrix is ​​constructed to represent the relationship between each parameter in the group, such as voltage, current, and temperature. Based on the upper and lower layer correlation matrices, a two-layer fuzzy judgment matrix is ​​constructed using fuzzy hierarchical analysis to compare the degree of influence between and within parameter groups. The fuzzy judgment matrix is ​​represented by triangular fuzzy numbers. Consistency checks are performed on each layer of the fuzzy judgment matrix, and the consistency ratio is calculated. First, the fuzzy numbers are converted into clear values, and then the consistency index is calculated. Dynamic weighting factors are introduced for each parameter group and parameters within each group to reflect their importance at different times and under different operating conditions. These dynamic weighting factors can be dynamically adjusted according to the equipment's operating status. A sensitivity decay factor is also introduced for each parameter to represent the change in the parameter's sensitivity over time or operating condition. The formula is as follows:

[0023] in, Let be the parameter sensitivity decay function. Let t be the decay rate constant, and t be the current time. This is the initial time. Normalize the upper-level fuzzy judgment matrix, calculate the fuzzy weight of each parameter group, obtain the clear weight by defuzzification (such as taking a weighted average), calculate the cross-influence coefficient of the parameter group based on the upper-level fuzzy judgment matrix, organize the cross-influence coefficient into matrix form, and generate a cross-influence weight allocation table. For each parameter group, the deviation of parameters within the group is calculated. Based on the cross-influence coefficient and the within-group deviation, combined with the attenuation factor, the cooperative deviation of the parameter group is calculated using the following formula:

[0024] in, For the degree of coordination deviation, This is the cross-influence coefficient. This is within-group bias. As the attenuation factor, For parameter groups.

[0025] See Figure 4As shown, the joint dataset with spatiotemporal tags is established by synchronously collecting equipment operating status data and environmental sensor data using high-precision timestamp alignment technology. Specifically, this includes: Based on the acquisition time of operating status data and environmental sensor data obtained by sensors of each device, the device data and environmental data are accurately synchronized through a high-precision time synchronization protocol, and a precise timestamp is added to each data point. Based on the synchronized data, spatiotemporal markers are added to each data record, including equipment operation cycle phase markers and environmental disturbance event markers. The equipment operation cycle phase markers are marked according to the specific stages of equipment operation, and the environmental disturbance event markers include changes in the external environment that affect equipment operation. Align device operating status data and environmental sensor data with precise timestamps, and construct a joint dataset by combining spatiotemporal tags.

[0026] Specifically, the operation of equipment usually goes through multiple stages, such as startup, operation, and shutdown. Each stage has different working characteristics. In order to track the changes in the status of the equipment, a phase marker of the equipment operation cycle is added to each data point according to the specific stage of the equipment operation. Environmental disturbance event markers refer to disturbances caused by external environmental factors during the operation of the equipment. Environmental disturbance event markers are added to each data point to reflect the impact of environmental changes on the equipment.

[0027] See Figure 5 As shown, historical collaborative datasets are input into a convolutional long short-term memory network for training. By embedding a parameter attention mechanism in the memory units, the fusion features of multi-parameter time-series data are extracted, specifically including: We acquire historical collaborative datasets, train a model combining convolutional layers and LSTM, extract spatial features from the input data through convolutional layers, and then input the extracted features into the LSTM layer to capture long-term temporal dependencies of the data through memory units. By adding a parameter attention layer to the memory cell of ConvLSTM, the importance of each parameter at different time steps is calculated, attention weights are generated, and used to weight the input time-series data. The model's performance is evaluated using a validation set, and the model's hyperparameters are adjusted based on the validation set results. Based on the trained ConvLSTM model, the fusion features of multi-parameter time series data are extracted to obtain normal collaborative feature data.

[0028] Specifically, the purpose of convolutional layers is to extract spatial features from data through local receptive fields. For input temporal data, convolution operations are performed through kernel functions. The spatial features extracted by convolution are fed into LSTM layers for temporal modeling. LSTM can capture long-term dependencies in temporal data. To weight the influence of different parameters at different time steps, an attention mechanism is added to the memory unit of ConvLSTM to optimize the model. Specifically, attention weights are embedded in the input or intermediate computation of LSTM to adjust the contribution of different parameters through weighting. The attention mechanism can calculate the attention weights at each time step using the following formula:

[0029] in, Let be the attention weight for the i-th parameter at time t. For calculation parameters The importance score function can typically be calculated using an inner product or a feedforward neural network; The obtained attention weights are used to weight the input temporal data or the hidden state of the LSTM in order to better capture important features. The trained ConvLSTM model is used to extract fusion features from the temporal data. The final LSTM output is obtained as the fusion feature of the multi-parameter temporal data, representing the temporal dependence of the data in multiple dimensions.

[0030] See Figure 6 As shown, the dynamic time warping algorithm is used to compare the similarity between the joint dataset and normal collaborative feature data to identify collaborative anomalies between parameters and to mark the anomalous time windows. Specifically, these anomalies include: Synchronize the joint dataset with normal collaborative feature data based on timestamps; For each pair of time series based on the joint dataset and normal collaborative feature data, calculate the DTW distance between them; The shortest matching path between time series is obtained using the DTW algorithm, and the similarity between the two is calculated. Based on the analysis results of historical data, a distance threshold is set. If the DTW distance in a certain period exceeds the threshold, the data in that period is different from the normal collaborative feature data, and it is determined to be an abnormal collaborative parameter. The period in question is recorded as an abnormal time window. For each abnormal period, the start and end times of the abnormality are marked with a timestamp.

[0031] Specifically, the joint dataset and normal collaborative feature data are synchronized in time. For each pair of time series, the distance between them is calculated using the dynamic time warping algorithm. The DTW algorithm finds the best matching path by calculating the cumulative distance matrix and measures the similarity between the two time series by the shortest matching path. By analyzing historical data, a threshold for DTW distance is set. If the DTW distance of a certain period exceeds the threshold, it indicates that the similarity between the joint dataset and the normal collaborative feature data of that period is low, and there may be an anomaly. The timestamp intervals corresponding to the periods that are determined to be collaborative anomalies are recorded as an anomaly time window.

[0032] See Figure 7 As shown, the weight allocation of the abnormal parameter group is obtained by matching the collaborative anomaly parameters with the cross-influence weight allocation table, and the weights are adjusted in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model. Specifically, this includes: Based on the obtained collaborative anomaly parameters, they are compared with the data in the cross-influence weight allocation table to obtain the weight allocation data of the collaborative anomaly parameters; The weight values ​​of the collaborative anomaly parameters are dynamically updated using a genetic algorithm, and the updated data is compared with the normal collaborative feature data again to calculate the similarity. Based on the weight data of the collaborative parameters that are judged to be normal, the cross-influence weight allocation table data is updated, and the two-layer parameter correlation model is optimized.

[0033] See Figure 8 As shown, the calculation of the cooperative deviation between parameters based on the optimized two-layer parameter correlation model, combined with the equipment health status baseline curve, generates a health score, specifically including: The cooperative deviation between parameters is calculated based on the optimized two-layer parameter correlation model; Based on the equipment's historical operating data, a health baseline curve is established, including the range of key parameter values ​​for the equipment under various operating conditions. By comparing the device's current operating data with a health baseline curve, the deviation is calculated, and a health score is obtained.

[0034] Specifically, the operation data of the equipment is extracted from the historical collaborative dataset, covering the key parameters of the equipment under normal operation. A health baseline curve is established for each key parameter to represent its value range under normal operation. The health baseline curve can be generated by statistical methods. The health baseline curves of all key parameters are integrated to form a multi-parameter health baseline curve set. For each key parameter, the deviation of the current operating data from the health baseline curve is calculated. Combining the collaborative deviation and the deviation of each parameter, the overall health score of the equipment is calculated using a weighted average method. By setting a threshold for health scores, potential equipment malfunctions can be identified. By combining coordination deviation and deviation, key parameters or parameter groups that cause a decline in health scores can be identified, providing a basis for equipment maintenance.

[0035] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0036] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0037] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-parameter collaborative detection method for medical devices, characterized in that, include: The dynamic working parameter set, static performance parameter set, and environmental coupling parameter set of the medical equipment are obtained through the equipment operation system, and a three-dimensional parameter set matrix is ​​constructed. A two-level parameter correlation model is constructed based on the improved fuzzy hierarchical analysis method. By introducing a dynamic weight adjustment factor and a parameter sensitivity decay function, the cross-influence coefficient and cooperative deviation of each parameter group are calculated, and a cross-influence weight allocation table is constructed. By synchronously collecting equipment operating status data and environmental sensor data using high-precision timestamp alignment technology, a joint dataset with spatiotemporal tags is established. Historical collaborative datasets are input into a convolutional long short-term memory network for training. By embedding a parameter attention mechanism in the memory unit, the fusion features of multi-parameter time series data are extracted. By comparing the similarity between the joint dataset and normal collaborative feature data using the dynamic time warping algorithm, collaborative anomalies between parameters are identified, and anomalous time windows are marked. The weight allocation of the abnormal parameter group is obtained by matching the collaborative abnormal parameters with the cross-influence weight allocation table, and the weights are adjusted in real time by an adaptive weighting algorithm to optimize the two-layer parameter correlation model. The optimized two-layer parameter correlation model is used to calculate the cooperative deviation between parameters, and a health score is generated by combining the equipment health status benchmark curve.

2. The multi-parameter collaborative detection method for medical devices according to claim 1, characterized in that, The calculation of the collaborative deviation between parameters based on the optimized two-layer parameter correlation model, combined with the equipment health status baseline curve to generate a health score, specifically includes: The cooperative deviation between parameters is calculated based on the optimized two-layer parameter correlation model; Based on the equipment's historical operating data, a health baseline curve is established, including the range of key parameter values ​​for the equipment under various operating conditions. By comparing the device's current operating data with a health baseline curve, the deviation is calculated, and a health score is obtained.

3. The multi-parameter collaborative detection method for medical devices according to claim 1, characterized in that, It also includes generating optimization reports, such as automatically marking parameters with calibration deviations exceeding 5% as red priority; recommending preventative maintenance plans based on equipment usage time; and generating a visual diagnostic interface containing parameter correlation graphs, highlighting abnormal transmission paths in the graphs.

4. A multi-parameter collaborative detection system for medical devices, used to implement the multi-parameter collaborative detection method for medical devices as described in any one of claims 1-3, characterized in that, include: Data acquisition module: The module collects dynamic operating parameters, static performance parameters and environmental coupling parameters of the medical equipment in real time through the equipment operation system and environmental sensors, and constructs a three-dimensional parameter group matrix; Parameter correlation module: The module is based on the improved fuzzy hierarchical analysis method, constructs a two-level parameter correlation model, calculates the cross-influence coefficient and cooperative deviation of each parameter group, and generates a cross-influence weight allocation table; Feature extraction training module: The module trains the historical collaborative dataset through a convolutional long short-term memory network and embeds a parameter attention mechanism to extract fused features of multi-parameter time series data; Collaborative Anomaly Module: This module uses a dynamic time warping algorithm to compare the similarity between the joint dataset and normal collaborative feature data, and identifies and marks collaborative anomalies and anomalous time windows between parameters; Anomaly parameter matching and optimization module: The module matches collaborative anomaly parameters with a cross-influence weight allocation table, and adjusts the weights in real time through an adaptive weighting algorithm to optimize the two-layer parameter correlation model; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images