Auxiliary diagnosis data feature analysis method, system and device based on deep learning
Through deep learning-based methods, data performance parameters of auxiliary diagnostic data are obtained, complex data indicators are analyzed, data enhancement and segmentation are performed, and parallel processing ports are dynamically selected, which solves the problem of poor flexibility in auxiliary diagnostic data processing in the existing technology, and achieves efficient and accurate data processing.
Patent Information
- Application Number
- CN202510391178.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-31
- Publication Date
- 2025-08-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing auxiliary diagnostic systems rely on fixed algorithms and preset rules, resulting in inefficient data processing and inability to adapt to large-scale, multi-dimensional, and highly complex medical data, and poor flexibility.
Using a deep learning-based method, by obtaining data performance parameters of auxiliary diagnostic data, analyzing data complex indicators, determining data enhancement parameters, and performing initial enhancement processing, dividing them into each data layer for distributed processing, dynamically selecting parallel processing ports, and performing feature analysis.
It realizes accurate identification and processing of different diagnostic data characteristics, improves data processing efficiency and accuracy, optimizes data flow, maximizes parallel processing capabilities, and improves computing resource utilization and processing speed.
Smart Images

Figure CN120544832A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of feature analysis technology, and in particular to a method, system, and device for analyzing feature of auxiliary diagnosis data based on deep learning. Background Art
[0002] With the continuous development of medical and health management, auxiliary diagnosis systems are becoming increasingly important. Existing auxiliary diagnosis data feature analysis systems increase data diversity through fixed data preprocessing processes by transforming, scaling, or adding noise to the original data, and then perform data analysis based on the data diversity.
[0003] For example, the invention patent with publication number CN113545787B discloses a method for analyzing an electrocardiogram signal, comprising: measuring an electrocardiogram signal using a sensor; detecting one or more specified waveforms in the electrocardiogram signal; obtaining multiple electrocardiogram features from each specified waveform; inputting the electrocardiogram features into multiple candidate models to obtain multiple candidate scores; comparing each candidate score with a standard value to obtain an ideal score among the candidate scores; selecting a candidate model corresponding to the ideal score as a risk prediction model, and predicting disease risk using the risk prediction model.
[0004] For example, the invention patent with publication number CN117958835B discloses an electrocardiogram analysis method for postoperative rehabilitation of patients, which includes: collecting the patient's preoperative electrocardiogram data, using the electrocardiogram data to obtain an electrocardiogram scatter plot, obtaining a distribution adjustment factor of the electrocardiogram scatter plot, obtaining a scatter distribution coefficient of the electrocardiogram scatter plot based on the distribution adjustment factor, thereby obtaining a dispersion coefficient of the patient's electrocardiogram data, obtaining an adaptive neighborhood radius of a clustering algorithm, re-collecting the patient's postoperative electrocardiogram data, and calculating the patient's status score during the rehabilitation period in combination with the clustering algorithm.
[0005] However, in the process of implementing the technical solutions of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:
[0006] In the existing technology, auxiliary diagnosis systems rely on fixed algorithms and preset rules for data analysis. However, for auxiliary diagnosis data, the data has the problems of large volume and high complexity. Therefore, relying on fixed algorithms and preset rules for data analysis often cannot fully adapt to the actual characteristics of the data, resulting in low data processing efficiency and inability to perform refined analysis and adjustments based on different data characteristics. When faced with large-scale, multi-dimensional, and highly complex medical data, its processing capabilities and flexibility are severely limited. Therefore, the current data processing of auxiliary diagnosis data has the problem of poor flexibility. Summary of the Invention
[0007] The embodiments of the present application solve the problem of poor flexibility in data processing of auxiliary diagnostic data in the prior art by providing a method, system and device for auxiliary diagnostic data feature analysis based on deep learning, realize personalized data enhancement and distributed processing according to the characteristics of different data layers, and improve the processing efficiency and accuracy of auxiliary diagnostic data.
[0008] An embodiment of the present application provides a method for analyzing auxiliary diagnostic data features based on deep learning, comprising the following steps: when the auxiliary diagnostic device receives an auxiliary diagnostic instruction signal, the auxiliary diagnostic data is acquired for data loading processing to obtain data representation parameters of the auxiliary diagnostic data; based on the data representation parameters of the auxiliary diagnostic data, analysis is performed to obtain a data complexity index of the auxiliary diagnostic data, and data enhancement parameters are determined based on the data complexity index; according to the data enhancement parameters, the auxiliary diagnostic data is subjected to initial enhancement processing by the auxiliary diagnostic device, and after the initial enhancement processing, the comprehensive feature coefficient of the auxiliary diagnostic data is obtained by analysis; based on the comprehensive feature coefficient of the auxiliary diagnostic data, the distributed processing parameters of the auxiliary diagnostic data are determined, and the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic device to obtain each data layer and the comprehensive feature coefficient of each data layer; according to the comprehensive feature coefficient of each data layer, the distributed data parallel processing port of each data layer is determined, and feature analysis processing is performed.
[0009] Furthermore, the data performance parameters based on the auxiliary diagnostic data are analyzed to obtain the data complexity index of the auxiliary diagnostic data. The specific steps include: obtaining the auxiliary diagnostic data for data loading processing to obtain the data performance parameters of the auxiliary diagnostic data, and the data performance parameters of the auxiliary diagnostic data include: data scale, data loading delay, number of data dimensions, data noise ratio, data missing value ratio and data duplication value ratio; obtaining the data performance permission set preset in the database, and the data performance permission set includes the data loading delay permission value, the data missing value ratio permission value and the data duplication value ratio permission value; obtaining the unit performance factor set preset in the data, and the unit performance factor set includes the unit data scale performance factor, the unit data dimension number performance factor, and the unit data noise ratio performance factor; analyzing the data performance permission set, the unit performance factor set and the data performance parameters of the auxiliary diagnostic data to obtain the data complexity index of the auxiliary diagnostic data; the data complexity index of the auxiliary diagnostic data is used to characterize the data processing difficulty of the auxiliary diagnostic data.
[0010] Furthermore, the data complexity index of the auxiliary diagnosis data is obtained by the following method: ;
[0011] Where, Indicates the data complexity index of auxiliary diagnosis data, Indicates the data scale of auxiliary diagnosis data, Indicates the unit data scale performance factor, Indicates the data loading delay of auxiliary diagnosis data, Indicates the allowed value of data loading delay, Indicates the number of data dimensions of auxiliary diagnosis data, Indicates the unit data dimension quantity performance factor, represents the proportion of data noise in auxiliary diagnosis data, Indicates the performance factor of unit data noise ratio, Indicates the proportion of missing values in auxiliary diagnostic data, Indicates the allowable value of the missing value ratio. Indicates the ratio of repeated values of auxiliary diagnostic data. Indicates the allowed value of the data duplication ratio.
[0012] Furthermore, the data enhancement parameters are determined based on the data complexity index, and the specific steps include: obtaining each reference data complexity index interval preset in the database and the reference data enhancement parameters corresponding to each reference data complexity index interval, comparing the data complexity index of the auxiliary diagnostic data with each reference data complexity index interval, and if the data complexity index of the auxiliary diagnostic data is within a certain reference data complexity index interval, then obtaining the reference data enhancement parameters corresponding to the reference data complexity index interval as the data enhancement parameters; the data enhancement parameters include: data enhancement intensity, data enhancement frequency and data enhancement duration window length.
[0013] Furthermore, the auxiliary diagnostic data is initially enhanced by the auxiliary diagnostic device, and the comprehensive characteristic coefficient of the auxiliary diagnostic data is obtained by analysis after the initial enhancement processing. The specific steps include: performing initial enhancement processing on the auxiliary diagnostic data by the auxiliary diagnostic device, and obtaining the data complexity index after the data enhancement processing; based on the data complexity index after the data enhancement processing, the auxiliary diagnostic data enhancement characterization parameters are statistically processed to obtain the comprehensive characteristic coefficient of the auxiliary diagnostic data; the auxiliary diagnostic data enhancement characterization parameters include: the average execution rate of data enhancement and the average processor utilization rate and floating-point operation frequency in each data enhancement process; the comprehensive characteristic coefficient of the auxiliary diagnostic data is used to characterize the comprehensive processing complexity of the auxiliary diagnostic data after the data enhancement processing.
[0014] Furthermore, the distributed processing parameters of the auxiliary diagnostic data are determined, and the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic equipment to obtain each data layer and the comprehensive characteristic coefficient of each data layer. The specific steps include: obtaining each comprehensive characteristic coefficient interval preset in the database and the reference distributed processing parameters corresponding to each comprehensive characteristic coefficient interval, and comparing them with the comprehensive characteristic coefficient of the auxiliary diagnostic data. If the comprehensive characteristic coefficient of the auxiliary diagnostic data is within a certain preset comprehensive characteristic coefficient interval, then obtaining the reference distributed processing parameters corresponding to the comprehensive characteristic coefficient interval as the distributed processing parameters of the auxiliary diagnostic data; according to the distributed processing parameters of the auxiliary diagnostic data, the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic equipment to obtain each data layer and the comprehensive characteristic coefficient of each data layer; the distributed processing parameters of the auxiliary diagnostic data are the number of distributed processing segmentations.
[0015] Furthermore, the distributed data parallel processing port of each data layer is determined based on the comprehensive characteristic coefficient of each data layer, and characteristic analysis processing is performed. The specific steps include: obtaining port performance parameters of each parallel processing port and analyzing to obtain performance evaluation indicators of each parallel processing port; obtaining parallel processing port performance evaluation indicator thresholds corresponding to each reference comprehensive characteristic coefficient interval of the data layer preset in the database, and comparing them with the comprehensive characteristic coefficient of each data layer; if the comprehensive characteristic coefficient of a data layer is within a reference comprehensive characteristic coefficient interval, obtaining the parallel processing port performance evaluation indicator threshold corresponding to the reference comprehensive characteristic coefficient interval as the parallel processing port performance evaluation indicator threshold of the data layer, and marking it as the parallel port performance threshold of the data layer; statistically comparing the parallel port performance threshold of each data layer with the performance evaluation indicators of each parallel processing port; if the performance evaluation indicator of a parallel processing port is greater than and closest to the parallel port performance threshold of a data layer, then using the parallel processing port as the distributed data parallel processing port of the data layer; the port performance parameters of each parallel processing port include the CPU processing speed, resource utilization, historical error rate and historical average response time of each parallel processing port; the performance evaluation indicator of each parallel processing port is used to characterize the performance of each parallel processing port.
[0016] Furthermore, it also includes transmitting each data layer to the corresponding distributed data parallel processing port for deep learning training, and analyzing to obtain the training performance index, thereby dynamically adjusting the learning rate. The specific steps include: obtaining the real-time performance parameters of the training process, and analyzing and processing to obtain the training performance index; obtaining the first threshold of the training performance index and the second threshold of the training performance index preset in the database, and comparing them with the training performance index. If the training performance index is above the first threshold of the training performance index, the learning rate is increased; if the training performance index is less than the first threshold of the training performance index and above the second threshold of the training performance index, the learning rate is not adjusted; if the training performance index is less than the second threshold of the training performance index, the learning rate is reduced; the real-time performance parameters of the training process include: real-time training rounds, real-time convergence speed, real-time gradient change, real-time network delay, real-time network bandwidth and real-time data transmission speed; the training performance index is used to characterize the training performance of the data layer during the training process.
[0017] Furthermore, the auxiliary diagnostic data feature analysis system based on deep learning includes: a data representation parameter acquisition module, a data enhancement parameter determination module, a comprehensive feature coefficient acquisition module, a distributed processing parameter determination module and a parallel processing port determination module; wherein the data representation parameter acquisition module is used to acquire the auxiliary diagnostic data for data loading processing after the auxiliary diagnostic device receives the auxiliary diagnostic instruction signal, and obtain the data representation parameters of the auxiliary diagnostic data; the data enhancement parameter determination module is used to analyze based on the data representation parameters of the auxiliary diagnostic data, obtain the data complexity index of the auxiliary diagnostic data, and determine the data enhancement parameters based on the data complexity index; the comprehensive feature coefficient acquisition module is used to perform initial enhancement processing on the auxiliary diagnostic data through the auxiliary diagnostic device according to the data enhancement parameters, and analyze after the initial enhancement processing to obtain the comprehensive feature coefficient of the auxiliary diagnostic data; the distributed processing parameter determination module is used to determine the distributed processing parameters of the auxiliary diagnostic data based on the comprehensive feature coefficient of the auxiliary diagnostic data, and segment the auxiliary diagnostic data through the auxiliary diagnostic device to obtain each data layer and the comprehensive feature coefficient of each data layer; the parallel processing port determination module is used to determine the distributed data parallel processing port of each data layer according to the comprehensive feature coefficient of each data layer, and perform feature analysis processing.
[0018] An embodiment of the present application provides an auxiliary diagnostic data feature analysis device based on deep learning, including: a processor, and a memory and a network interface connected to the processor: the network interface is connected to a non-volatile memory in a server; when the processor is running, it calls a computer program from the non-volatile memory through the network interface and runs the computer program through the memory.
[0019] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:
[0020] 1. The auxiliary diagnostic data feature analysis method based on deep learning provided by the present invention obtains the data complexity index of the auxiliary diagnostic data by analyzing the data performance parameters of the auxiliary diagnostic data, thereby being able to characterize the characteristics of the auxiliary diagnostic data, and further realizing the accurate identification and processing of different diagnostic data characteristics, effectively solving the problem of poor flexibility in the data processing of auxiliary diagnostic data in the existing technology.
[0021] 2. The present invention obtains the comprehensive feature coefficients of each data layer by obtaining and analyzing them, thereby optimizing the data processing flow in real time, thereby realizing data hierarchical processing and efficient feature extraction, and improving training efficiency.
[0022] 3. By analyzing and determining the distributed data parallel processing ports of each data layer and performing feature analysis and processing, the most suitable processing ports are dynamically selected according to the characteristics of each data layer, thereby maximizing the parallel processing capabilities of each data layer and improving computing resource utilization and processing speed. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 Flowchart of the deep learning-based auxiliary diagnosis data feature analysis method provided in an embodiment of the present application;
[0024] Figure 2 A schematic diagram of the structure of the deep learning-based auxiliary diagnosis data feature analysis system provided in an embodiment of the present application. DETAILED DESCRIPTION
[0025] The embodiments of the present application solve the problem of poor flexibility in data processing of auxiliary diagnostic data in the prior art by providing a method, system and device for analyzing auxiliary diagnostic data features based on deep learning.
[0026] The deep learning-based auxiliary diagnosis data feature analysis method can be applied to the field of electrocardiogram analysis to assist in the diagnosis of electrocardiogram signals. For example, when an electrocardiogram device receives an electrocardiogram signal, it can use this solution to assist in diagnosis. First, the electrocardiogram data is loaded, and data performance parameters such as the number of data dimensions are obtained. After analysis, the data complexity index of the electrocardiogram auxiliary diagnosis data is obtained. This data is then initially enhanced and finally trained to obtain the performance.
[0027] like Figure 1As shown, it is a flow chart of the auxiliary diagnostic data feature analysis method based on deep learning provided by an embodiment of the present application. The method is applied to an auxiliary diagnostic data feature analysis system based on deep learning. The method includes the following steps: when the auxiliary diagnostic device receives the auxiliary diagnostic instruction signal, the auxiliary diagnostic data is obtained for data loading processing to obtain data representation parameters of the auxiliary diagnostic data; based on the data representation parameters of the auxiliary diagnostic data, analysis is performed to obtain data complexity indicators of the auxiliary diagnostic data, and data enhancement parameters are determined based on the data complexity indicators; according to the data enhancement parameters, the auxiliary diagnostic data is initially enhanced by the auxiliary diagnostic device, and after the initial enhancement processing, the comprehensive feature coefficient of the auxiliary diagnostic data is obtained by analysis; based on the comprehensive feature coefficient of the auxiliary diagnostic data, the distributed processing parameters of the auxiliary diagnostic data are determined, and the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic device to obtain each data layer and the comprehensive feature coefficient of each data layer; according to the comprehensive feature coefficient of each data layer, the distributed data parallel processing port of each data layer is determined, and feature analysis processing is performed.
[0028] Furthermore, based on the data performance parameters of the auxiliary diagnostic data, an analysis is performed to obtain a data complexity index of the auxiliary diagnostic data. The specific steps include: obtaining the auxiliary diagnostic data for data loading processing to obtain the data performance parameters of the auxiliary diagnostic data, the data performance parameters of the auxiliary diagnostic data include: data scale, data loading delay, number of data dimensions, data noise ratio, data missing value ratio and data duplication value ratio; obtaining a data performance permission set preset in the database, the data performance permission set includes a data loading delay permission value, a data missing value ratio permission value and a data duplication value ratio permission value; obtaining a unit performance factor set preset in the data, the unit performance factor set includes a unit data scale performance factor, a unit data dimension number performance factor and a unit data noise ratio performance factor; analyzing the data performance permission set, the unit performance factor set and the data performance parameters of the auxiliary diagnostic data to obtain a data complexity index of the auxiliary diagnostic data; the data complexity index of the auxiliary diagnostic data is used to characterize the data processing difficulty of the auxiliary diagnostic data.
[0029] In this embodiment, it should be noted that data scale refers to the size of the data, which can be obtained by querying the file size or the number of data items. Data loading latency refers to the time it takes to load data from the data store into memory, which can be measured by executing a SQL query and recording the start and end times. The number of data dimensions refers to the number of spatial dimensions occupied by each data item in a dataset. An increase in data dimensionality indicates increased data complexity. One-dimensional data refers to data with only one feature or variable. For example, a signal containing a time series relies solely on a time axis to represent each data point. Two-dimensional data refers to data with two features or variables, used for representing planar data, such as image data. Three-dimensional data refers to data with three features or variables, used for representing volumetric data, such as CT scan images. Higher-dimensional data (four-dimensional and above) refers to complex data with more features, which appears in multidimensional representations of time series data, such as video data. The data noise fraction refers to the proportion of noise (erroneous or irrelevant data) in the data, which can be obtained by using statistical methods such as the interquartile range (IQR) method to identify outliers. The proportion of missing values in the data can be obtained by calculating the proportion of missing values in the total data through Python code, and the proportion of duplicate values in the data set can be obtained by detecting exactly the same records in the data set and calculating their proportion.
[0030] The unit performance influence factor refers to the numerical value of the influence degree of the independent variable under unit conditions. For example, the unit data scale performance factor refers to the numerical value of the influence degree of the unit data size, the unit data dimension number performance factor refers to the numerical value of the influence degree of the unit data dimension number, and the unit data noise ratio performance factor refers to the numerical value of the influence degree of the unit data noise ratio.
[0031] By analyzing the data performance parameters of auxiliary diagnostic data, we obtain the data complexity index of auxiliary diagnostic data, thereby achieving an effective assessment of data processing difficulty. Through comprehensive analysis of the data performance parameters of auxiliary diagnostic data, the complexity of the data is reflected. By obtaining the data performance allowable set and unit performance factor set preset in the database, the consistency of data quality assessment is ensured, thereby optimizing data processing efficiency, reducing invalid calculations, and improving data processing response speed.
[0032] Furthermore, the data complexity index of the auxiliary diagnosis data is obtained by: ;
[0033] Where, Indicates the data complexity index of auxiliary diagnosis data, Indicates the data scale of auxiliary diagnosis data, Indicates the unit data scale performance factor, Indicates the data loading delay of auxiliary diagnosis data, Indicates the allowed value of data loading delay, Indicates the number of data dimensions of auxiliary diagnosis data, Indicates the unit data dimension quantity performance factor, represents the proportion of data noise in auxiliary diagnosis data, Indicates the performance factor of unit data noise ratio, Indicates the proportion of missing values in auxiliary diagnostic data, Indicates the allowable value of the missing value ratio. Indicates the ratio of repeated values of auxiliary diagnostic data. Indicates the allowed value of the data duplication ratio.
[0034] In this embodiment, it should be noted that the data complexity index of the auxiliary diagnostic data is obtained by analyzing the data performance parameters of the auxiliary diagnostic data (data scale, data loading delay, number of data dimensions, data noise ratio, data missing value ratio, and data duplication value ratio). This takes into account the mutual influence between these parameters. For example, an increase in data scale will lead to an increase in data loading delay because more data needs to be loaded and processed. In the case of multi-source data or incomplete data collection, an increase in data scale will lead to an increase in the noise ratio and missing value ratio. An increase in the number of data dimensions will increase data complexity, thereby introducing more noise and missing values. An increase in dimensions will result in more redundant data during processing. An increase in the ratio of noise and missing values will introduce more noise.
[0035] Furthermore, data enhancement parameters are determined based on the data complexity index. The specific steps include: obtaining each reference data complexity index interval preset in the database and the reference data enhancement parameters corresponding to each reference data complexity index interval, comparing the data complexity index of the auxiliary diagnostic data with each reference data complexity index interval, and if the data complexity index of the auxiliary diagnostic data is within a certain reference data complexity index interval, obtaining the reference data enhancement parameters corresponding to the reference data complexity index interval as the data enhancement parameters; the data enhancement parameters include: data enhancement intensity, data enhancement frequency and data enhancement duration window length.
[0036] In this embodiment, it should be noted that data augmentation intensity refers to the magnitude of the data augmentation operation when transforming, expanding, or perturbing the original data. Data augmentation frequency refers to the number of times the augmentation operation is applied per unit time during data augmentation, i.e., multiple transformations are performed within a certain time window to increase the diversity of the training data, with the unit being "times / unit time" (e.g., times / second). Data augmentation duration window length refers to the length of time during which the data augmentation operation is performed (e.g., 5 seconds).
[0037] By mapping the data complexity index of the auxiliary diagnostic data to the preset reference data complexity index intervals in the database, data augmentation parameters are determined. This ensures that the data augmentation process matches the data features, facilitating initial augmentation. For example, auxiliary diagnostic data with higher data complexity (such as data with a high noise ratio, a large number of missing values, or high data dimensionality) generally require higher data processing requirements. These data are often more complex and contain more uncertainty. Therefore, during the augmentation process, more refined parameter adjustments are required to ensure that the augmentation process effectively improves data quality rather than introducing additional noise or interference. High-complexity data generally requires stronger data augmentation intensity, higher data augmentation frequency, and longer duration windows to cope with data diversity and variability. For auxiliary diagnostic data with lower data complexity, the augmentation requirements are relatively low. The data itself is relatively clean and simple in structure. Therefore, the augmentation operation can adopt a lower intensity and shorter duration window, thereby avoiding the unnecessary risks of over-enhancement.
[0038] By determining data enhancement parameters based on data complexity indicators, it is possible to dynamically select appropriate enhancement parameters based on the characteristics of the auxiliary diagnosis data, ensuring that the data enhancement process matches the complexity of the data itself. Specifically, by comparing the data complexity indicators of the auxiliary diagnosis data with the preset reference data complexity indicator range, it is possible to select the appropriate enhancement intensity, frequency, and duration window length for data of different complexities, thereby optimizing the data processing process, effectively improving the accuracy and flexibility of data enhancement, and making the processing process more in line with actual application needs.
[0039] Furthermore, the auxiliary diagnostic data is initially enhanced by the auxiliary diagnostic device, and the comprehensive characteristic coefficient of the auxiliary diagnostic data is obtained by analysis after the initial enhancement processing. The specific steps include: performing initial enhancement processing on the auxiliary diagnostic data by the auxiliary diagnostic device, and obtaining the data complexity index after the data enhancement processing; based on the data complexity index after the data enhancement processing, the auxiliary diagnostic data enhancement characterization parameters are statistically processed to obtain the comprehensive characteristic coefficient of the auxiliary diagnostic data; the auxiliary diagnostic data enhancement characterization parameters include: the average execution rate of data enhancement and the average processor utilization rate and floating-point operation frequency in each data enhancement process; the comprehensive characteristic coefficient of the auxiliary diagnostic data is used to characterize the comprehensive processing complexity of the auxiliary diagnostic data after the data enhancement processing.
[0040] In this embodiment, it should be noted that the auxiliary diagnosis data is initially enhanced by the auxiliary diagnosis device, and the initial enhancement processing can be performed by using a machine learning software library (such as OpenCV) or a natural language processing (NLP) toolkit, such as NLTK (Natural Language Toolkit).
[0041] Initial enhancement processing optimizes the feature representation of auxiliary diagnostic data, making it more applicable and accurate during subsequent processing and analysis. Based on a comprehensive analysis of data complexity indicators and enhancement characterization parameters, combined with parameters such as the average execution rate of data enhancement, average processor utilization, and floating-point operation frequency, the effectiveness and efficiency of the data enhancement process were comprehensively evaluated to ensure optimal processing performance and results.
[0042] It should be noted that the steps and principles for obtaining the data complexity index after data enhancement processing are the same as those for obtaining the data complexity index of auxiliary diagnostic data. The data complexity index after data enhancement processing is obtained by analyzing the data enhancement parameters (data enhancement intensity, data enhancement frequency, and data enhancement duration window length). This is based on the influence of these parameters on the data complexity index after data enhancement processing. For example, enhancement intensity refers to the magnitude of change in the original data caused by the enhancement operation. Enhancement operations with higher intensity will cause larger changes to the original data (such as more significant image distortion, noise addition, or data scaling). High enhancement intensity will increase the complexity of the data. Enhancement frequency refers to the frequency or number of enhancement operations performed during the data enhancement process. A high frequency means that the data is enhanced multiple times in a shorter period of time, making the data richer. An enhancement process with a longer duration means that the data will undergo a longer transformation, making the data more complex.
[0043] The comprehensive characteristic coefficient of the auxiliary diagnostic data is obtained by the following specific steps: obtaining the average execution rate reference value, the average processor usage reference value and the floating-point operation frequency reference value preset in the database, and comparing them with the average execution rate of the data enhancement and the average processor usage and floating-point operation frequency in each data enhancement process, to obtain the comprehensive characteristic coefficient of the auxiliary diagnostic data.
[0044] The comprehensive characteristic coefficient of auxiliary diagnosis data is obtained by: ;
[0045] Where, Represents the comprehensive characteristic coefficient of auxiliary diagnosis data, Represents the complexity index of the data after data enhancement processing, represents the average execution rate of data enhancement, represents the average execution rate reference value, i represents the number of data enhancement, , represents the total number of data augmentations, represents the average processor usage during the i-th data enhancement process, Indicates the floating-point operation frequency during the i-th data enhancement process, Indicates the average processor usage reference value. Indicates the floating-point operation frequency reference value.
[0046] It should be noted that the average data augmentation execution rate refers to the number of data augmentation operations completed per unit time and can be obtained through tools such as the TensorFlow Profiler (TensorFlow Performance Analyzer) or the CUDA Profiler (CUDA Performance Analyzer). The average processor utilization rate indicates the percentage of processor (CPU or GPU) resource usage during the data augmentation process, which can be viewed through the Windows Task Manager. The floating-point operation frequency refers to the speed at which the processor executes floating-point operations and can be obtained through the NVIDIA System Management Interface (nvidia-smi).
[0047] It should also be noted that the data complexity index after data enhancement processing and the auxiliary diagnosis data enhancement characterization parameters obtain the comprehensive characteristic coefficient of the auxiliary diagnosis data, which takes into account the mutual influence relationship between these parameters. For example: a higher processor utilization rate means that more computing resources are invested in the data enhancement task, which will increase the execution rate of data enhancement, but too high a utilization rate will lead to resource overload, resulting in a decrease in execution rate or a decrease in floating-point operation frequency. The floating-point operation frequency reflects the computing power of the processor when performing numerical calculations and is positively correlated with the execution rate. If the floating-point operation frequency is too low, it will lead to inefficiency in the data enhancement process, which will affect the overall execution rate.
[0048] Furthermore, the distributed processing parameters of the auxiliary diagnostic data are determined, and the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic equipment to obtain each data layer and the comprehensive characteristic coefficient of each data layer. The specific steps include: obtaining each comprehensive characteristic coefficient interval preset in the database and the reference distributed processing parameters corresponding to each comprehensive characteristic coefficient interval, and comparing them with the comprehensive characteristic coefficient of the auxiliary diagnostic data; if the comprehensive characteristic coefficient of the auxiliary diagnostic data is within a certain preset comprehensive characteristic coefficient interval, then obtaining the reference distributed processing parameters corresponding to the comprehensive characteristic coefficient interval as the distributed processing parameters of the auxiliary diagnostic data; according to the distributed processing parameters of the auxiliary diagnostic data, the auxiliary diagnostic data is segmented and processed by the auxiliary diagnostic equipment to obtain each data layer and the comprehensive characteristic coefficient of each data layer; the distributed processing parameters of the auxiliary diagnostic data are the number of distributed processing segmentations.
[0049] In this embodiment, by obtaining a preset comprehensive characteristic coefficient interval and comparing it with the comprehensive characteristic coefficient of the auxiliary diagnostic data, suitable data distributed processing parameters are determined, thereby reasonably allocating the segmentation processing tasks of the auxiliary diagnostic data. Based on the distributed processing parameters, the auxiliary diagnostic equipment can efficiently perform data segmentation processing and optimize the comprehensive characteristic coefficient analysis of each data layer, effectively improving data processing efficiency and accuracy, while reducing the waste of computing resources, thereby realizing efficient and accurate analysis and processing of large-scale auxiliary diagnostic data.
[0050] By using the distributed processing parameters of auxiliary diagnostic data, the original data is divided into multiple data layers. Each data layer is processed independently and its corresponding characteristic coefficient is calculated. The comprehensive characteristic coefficient of each data layer is then obtained. This comprehensive characteristic coefficient reflects the processing difficulty of the data layer and provides a basis for subsequent enhanced processing, thereby improving the efficiency and accuracy of data processing.
[0051] Furthermore, based on the comprehensive characteristic coefficient of each data layer, the distributed data parallel processing port of each data layer is determined, and characteristic analysis processing is performed. The specific steps include: obtaining the port performance parameters of each parallel processing port, and analyzing to obtain the performance evaluation index of each parallel processing port; obtaining the parallel processing port performance evaluation index threshold corresponding to each reference comprehensive characteristic coefficient interval of the data layer preset in the database, and comparing it with the comprehensive characteristic coefficient of each data layer; if the comprehensive characteristic coefficient of a data layer is within a reference comprehensive characteristic coefficient interval, then obtaining the parallel processing port performance evaluation index threshold corresponding to the reference comprehensive characteristic coefficient interval as the parallel processing port performance evaluation index threshold of the data layer, and marking it as the parallel port performance threshold of the data layer; statistically comparing the parallel port performance threshold of each data layer with the performance evaluation index of each parallel processing port; if the performance evaluation index of a parallel processing port is greater than and closest to the parallel port performance threshold of a data layer, then using the parallel processing port as the distributed data parallel processing port of the data layer; the port performance parameters of each parallel processing port include the CPU processing speed, resource utilization, historical error rate and historical average response time of each parallel processing port; the performance evaluation index of each parallel processing port is used to characterize the performance performance of each parallel processing port.
[0052] In this embodiment, the CPU processing speed and resource utilization of each parallel processing port can be obtained by using a monitoring tool (such as top or htop). The historical error rate and historical average response time can be obtained from the usage log.
[0053] Obtaining a performance evaluation indicator for each parallel processing port includes the following steps: obtaining a preset performance reference set from a database, wherein the performance reference set includes a CPU processing speed reference value, a resource utilization reference value, an error rate reference value, and an average response time reference value; and comparing the performance reference set with the port performance parameters of each parallel processing port to obtain a performance evaluation indicator for each parallel processing port.
[0054] The performance evaluation index of each parallel processing port is obtained by: ;
[0055] Where, Represents the performance evaluation index of each parallel processing port, Indicates the CPU processing speed of the jth parallel processing port, j represents the number of the parallel processing port, , Indicates the total number of parallel processing ports, Indicates the reference value of CPU processing speed. represents the resource occupancy of the jth parallel processing port, Indicates the reference value of resource utilization. represents the historical error rate of the jth parallel processing port, Indicates the error rate reference value, represents the historical average response time of the jth parallel processing port, represents the reference value of the average response time, and e represents a natural constant.
[0056] Performance evaluation metrics for each parallel processing port are derived by analyzing the port performance parameters of each parallel processing port. This takes into account the interplay between these parameters. For example, CPU processing speed directly affects processing efficiency. A higher CPU speed generally means faster data processing capabilities. Higher resource utilization can lead to processing errors or data loss, which in turn increases the error rate. A lower resource utilization indicates better utilization of system resources and a lower probability of errors. The historical error rate reflects historical operational stability. A higher error rate indicates poor historical operational stability, which can affect the accuracy of performance evaluation metrics. Longer historical average response times indicate higher resource utilization and error rates.
[0057] By dynamically selecting the most appropriate parallel processing port based on the comprehensive characteristic coefficients of each data layer, distributed processing at the data layer is ensured to be performed with the most appropriate computing resources. By acquiring and analyzing the performance parameters of each parallel processing port and comparing them with preset performance evaluation index thresholds, the parallel processing port that matches the characteristics of the data layer is selected, improving the efficiency and accuracy of the data processing process. It also effectively utilizes computing resources, avoiding performance bottlenecks caused by resource overload or mismatch, thereby achieving efficient parallel processing at the data layer, and thus improving the processing speed and accuracy of auxiliary diagnostic data.
[0058] Furthermore, it also includes transmitting each data layer to the corresponding distributed data parallel processing port for deep learning training, and analyzing to obtain the training performance index, thereby dynamically adjusting the learning rate. The specific steps include: obtaining the real-time performance parameters of the training process, and analyzing and processing to obtain the training performance index; obtaining the first threshold of the training performance index and the second threshold of the training performance index preset in the database, and comparing them with the training performance index. If the training performance index is above the first threshold of the training performance index, the learning rate is increased; if the training performance index is less than the first threshold of the training performance index and above the second threshold of the training performance index, the learning rate is not adjusted; if the training performance index is less than the second threshold of the training performance index, the learning rate is reduced; the real-time performance parameters of the training process include: real-time training rounds, real-time convergence speed, real-time gradient change, real-time network delay, real-time network bandwidth and real-time data transmission speed; the training performance index is used to characterize the training performance of the data layer during the training process.
[0059] In this embodiment, the training performance index is obtained, and the specific steps include: obtaining a performance parameter reference set preset in a database, the performance parameter reference set including a real-time training round reference value, a real-time convergence speed reference value, a real-time gradient change reference value, a real-time network delay reference value, a real-time network bandwidth reference value and a real-time data transmission speed reference value, and comparing the performance parameter reference set with the real-time performance parameters of the training process to obtain the training performance index.
[0060] Real-time training rounds, real-time convergence rate, and real-time gradient change can be obtained through command calls in TensorFlow. Real-time network latency can be measured using network analysis tools, such as Nsight Systems or nvidia-smi (NVIDIA System Management Interface). Real-time network bandwidth and real-time data transmission speed can be monitored using nvidia-smi (NVIDIA System Management Interface).
[0061] Get the training performance indicator. The specific method is: ;
[0062] Where, represents the training performance indicator, represents the real-time training round, Indicates the real-time training round reference value, Indicates the real-time convergence speed, Indicates the real-time convergence speed reference value, Indicates the real-time gradient change, Indicates the real-time gradient change reference value, Indicates the real-time network delay, Indicates the real-time network delay reference value, Indicates the real-time network bandwidth, Indicates the real-time network bandwidth reference value. Indicates the real-time data transmission speed, Indicates the reference value of real-time data transmission speed.
[0063] The training performance indicators are obtained by analyzing the real-time performance parameters of the training process (real-time training rounds, real-time convergence speed, real-time gradient change, real-time network latency, real-time network bandwidth, and real-time data transmission speed). The interaction between these parameters is taken into account. For example:
[0064] The number of real-time training rounds affects the convergence rate. More training rounds leads to faster model convergence, but this also depends on the magnitude of the gradient change. Large gradient changes indicate unstable updates during training, which slows convergence. Network latency and bandwidth are related to the efficiency of data transmission in distributed training. Excessive latency or insufficient bandwidth can cause data update lags, affecting the efficiency of gradient updates and, in turn, the rate of training convergence. The speed of real-time data transmission determines the progress of data loading and computation tasks. Excessively slow speeds can lead to bottlenecks in the training process and reduce overall performance.
[0065] By dynamically adjusting the learning rate to optimize the performance of the training process, the training performance of the data analysis model can be improved. By obtaining and analyzing the real-time performance parameters during the training process (such as real-time training rounds, real-time convergence speed, real-time gradient change, real-time network latency, real-time network bandwidth, and real-time data transmission speed), and combining them with the preset training performance indicator threshold, the learning rate of the model can be effectively adjusted. When the training performance indicator exceeds the first threshold, the learning rate is automatically increased; when the training performance indicator is between the first threshold and the second threshold, the learning rate is kept unchanged; if the training performance indicator is lower than the second threshold, the learning rate is reduced to avoid overfitting or oscillation during the training process, thereby improving the training efficiency and accuracy of the data analysis model, and further improving the performance of auxiliary diagnostic data analysis.
[0066] When the training performance index is above the first threshold, training is progressing well and nearing ideal performance. Therefore, increasing the learning rate (for example, by 2%) can accelerate training, reaching the optimal solution or a near-optimal solution as quickly as possible and avoiding computational waste caused by overtraining. When the training performance index is below the first threshold but above the second threshold, the training is in a stable state, meaning the training effect is acceptable but there is no significant trend of acceleration or deceleration. Maintaining the current learning rate can prevent drastic changes and ensure the stability of the training process. When the training performance index is below the second threshold, training is poor, potentially indicating overfitting, vanishing gradients, or exploding gradients. Reducing the learning rate (for example, by 10%) can help the model learn more smoothly and avoid jumping out of the local optimum too quickly during training.
[0067] Model training process: Obtain real-time performance parameters during the training process and perform data preprocessing (including data cleaning using the Pandas library and normalization using scikit-learn). Then, select an appropriate deep learning model structure, such as a convolutional neural network (CNN) or recurrent neural network (RNN), based on the data type and analysis objectives. Train the model using the real-time performance parameters during the training process, dynamically adjust the learning rate based on the training performance indicators, and finally save the trained model and prepare to deploy it in the actual environment for real-time auxiliary diagnosis.
[0068] like Figure 2As shown, it is a structural diagram of the auxiliary diagnosis data feature analysis system based on deep learning provided by an embodiment of the present application. The auxiliary diagnosis data feature analysis system based on deep learning provided by an embodiment of the present application includes: a data representation parameter acquisition module, a data enhancement parameter determination module, a comprehensive feature coefficient acquisition module, a distributed processing parameter determination module and a parallel processing port determination module; wherein, the data representation parameter acquisition module is used to obtain the auxiliary diagnosis data for data loading processing after the auxiliary diagnosis device receives the auxiliary diagnosis instruction signal, and obtain the data representation parameters of the auxiliary diagnosis data; the data enhancement parameter determination module is used to analyze the data representation parameters of the auxiliary diagnosis data based on the data representation parameters of the auxiliary diagnosis data, and obtain the data complexity of the auxiliary diagnosis data. Complex indicators, and determine data enhancement parameters based on the data complexity indicators; a comprehensive feature coefficient acquisition module, which is used to perform initial enhancement processing on the auxiliary diagnosis data through the auxiliary diagnosis equipment according to the data enhancement parameters, and obtain the comprehensive feature coefficients of the auxiliary diagnosis data after the initial enhancement processing; a distributed processing parameter determination module, which is used to determine the distributed processing parameters of the auxiliary diagnosis data based on the comprehensive feature coefficients of the auxiliary diagnosis data, and segment the auxiliary diagnosis data through the auxiliary diagnosis equipment to obtain each data layer and the comprehensive feature coefficients of each data layer; a parallel processing port determination module, which is used to determine the distributed data parallel processing port of each data layer according to the comprehensive feature coefficients of each data layer, and perform feature analysis processing.
[0069] An auxiliary diagnostic data feature analysis device based on deep learning is characterized by: including: a processor, and a memory and a network interface connected to the processor: the network interface is connected to the non-volatile memory in the server; when the processor is running, it calls the computer program from the non-volatile memory through the network interface and runs the computer program through the memory to apply the method as described above.
[0070] To sum up, this embodiment obtains the data performance parameters of the auxiliary diagnostic data and performs data complexity index analysis, thereby dynamically determining the data enhancement parameters, distributed processing parameters and parallel processing port performance evaluation indicators, and then realizes data enhancement and distributed processing according to the characteristics of different data layers, thereby improving the processing efficiency and accuracy of the auxiliary diagnostic data, and effectively solving the problem of poor flexibility in the data processing of the auxiliary diagnostic data in the existing technology.
[0071] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0072] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0073] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0075] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0076] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.
Claims
1. A method for analyzing auxiliary diagnostic data features based on deep learning, characterized in that: The following steps are involved: When the auxiliary diagnosis device receives the auxiliary diagnosis instruction signal, it obtains the auxiliary diagnosis data, performs data loading processing, and obtains data performance parameters of the auxiliary diagnosis data; Analyze the data performance parameters of the auxiliary diagnosis data to obtain the data complexity index of the auxiliary diagnosis data, and determine the data enhancement parameters based on the data complexity index; According to the data enhancement parameters, the auxiliary diagnosis data is initially enhanced by the auxiliary diagnosis device, and the comprehensive characteristic coefficient of the auxiliary diagnosis data is obtained by analysis after the initial enhancement process; Based on the comprehensive characteristic coefficients of the auxiliary diagnosis data, the distributed processing parameters of the auxiliary diagnosis data are determined, and the auxiliary diagnosis data are segmented and processed by the auxiliary diagnosis device to obtain each data layer and the comprehensive characteristic coefficients of each data layer; According to the comprehensive characteristic coefficient of each data layer, the distributed data parallel processing port of each data layer is determined, and characteristic analysis processing is performed.
2. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 1, characterized in that: The data complexity index of the auxiliary diagnosis data is obtained by analyzing the data performance parameters of the auxiliary diagnosis data, and the specific steps include: Acquire auxiliary diagnostic data for data loading processing to obtain data performance parameters of the auxiliary diagnostic data, wherein the data performance parameters of the auxiliary diagnostic data include: data size, data loading delay, number of data dimensions, data noise ratio, data missing value ratio, and data duplicate value ratio; Obtaining a data representation permission set preset in a database, wherein the data representation permission set includes a data loading delay permission value, a data missing value ratio permission value, and a data duplicate value ratio permission value; Obtaining a preset unit performance factor set in the data, wherein the unit performance factor set includes a unit data scale performance factor, a unit data dimension number performance factor, and a unit data noise ratio performance factor; The data complexity index of the auxiliary diagnosis data is obtained by analyzing the data performance allowable set, the unit performance factor set and the data performance parameters of the auxiliary diagnosis data; The data complexity index of the auxiliary diagnosis data is used to characterize the difficulty of data processing of the auxiliary diagnosis data.
3. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 2, characterized in that: The specific method for obtaining the data complexity index of the auxiliary diagnosis data is as follows: ; Where, Indicates the data complexity index of auxiliary diagnosis data, Indicates the data scale of auxiliary diagnosis data, Indicates the unit data scale performance factor, Indicates the data loading delay of auxiliary diagnosis data, Indicates the allowed value of data loading delay, Indicates the number of data dimensions of auxiliary diagnosis data, Indicates the unit data dimension quantity performance factor, represents the proportion of data noise in auxiliary diagnosis data, Indicates the performance factor of unit data noise ratio, Indicates the proportion of missing values in auxiliary diagnostic data, Indicates the allowable value of the missing value ratio. Indicates the ratio of repeated values of auxiliary diagnostic data. Indicates the allowed value of the data duplication ratio.
4. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 1, characterized in that: The specific steps of determining the data enhancement parameters based on the data complexity index include: Obtaining each reference data complexity index interval preset in the database and the reference data enhancement parameter corresponding to each reference data complexity index interval, comparing the data complexity index of the auxiliary diagnosis data with each reference data complexity index interval, and if the data complexity index of the auxiliary diagnosis data is within a certain reference data complexity index interval, obtaining the reference data enhancement parameter corresponding to the reference data complexity index interval as the data enhancement parameter; The data enhancement parameters include: data enhancement strength, data enhancement frequency and data enhancement duration window length.
5. The method for analyzing auxiliary diagnostic data features based on deep learning according to claim 4, characterized in that: The auxiliary diagnosis data is subjected to initial enhancement processing by the auxiliary diagnosis device, and the comprehensive characteristic coefficient of the auxiliary diagnosis data is obtained by analysis after the initial enhancement processing. The specific steps include: Performing initial enhancement processing on the auxiliary diagnosis data through auxiliary diagnosis equipment and obtaining the data complexity index after data enhancement processing; Based on the data complexity index after data enhancement processing, the auxiliary diagnosis data enhancement characterization parameters are statistically processed to obtain the comprehensive characteristic coefficient of the auxiliary diagnosis data; The auxiliary diagnosis data enhancement characterization parameters include: the average execution rate of data enhancement and the average processor usage and floating-point operation frequency during each data enhancement process; The comprehensive characteristic coefficient of the auxiliary diagnosis data is used to characterize the comprehensive processing complexity of the auxiliary diagnosis data after data enhancement processing.
6. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 1, characterized in that: The steps of determining the distributed processing parameters of the auxiliary diagnosis data and segmenting the auxiliary diagnosis data using the auxiliary diagnosis device to obtain each data layer and the comprehensive characteristic coefficient of each data layer include: Obtaining each preset comprehensive characteristic coefficient interval in the database and the reference distributed processing parameters corresponding to each comprehensive characteristic coefficient interval, and comparing them with the comprehensive characteristic coefficient of the auxiliary diagnosis data; if the comprehensive characteristic coefficient of the auxiliary diagnosis data is within a preset comprehensive characteristic coefficient interval, obtaining the reference distributed processing parameters corresponding to the comprehensive characteristic coefficient interval as the distributed processing parameters of the auxiliary diagnosis data; According to the distributed processing parameters of the auxiliary diagnosis data, the auxiliary diagnosis data is segmented and processed by the auxiliary diagnosis device to obtain each data layer and the comprehensive characteristic coefficient of each data layer; The distributed processing parameter of the auxiliary diagnosis data is the number of distributed processing divisions.
7. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 1, characterized in that: The steps of determining the distributed data parallel processing port of each data layer based on the comprehensive characteristic coefficient of each data layer and performing characteristic analysis processing include: Obtaining port performance parameters of each parallel processing port and analyzing them to obtain performance evaluation indicators of each parallel processing port; Obtain the parallel processing port performance evaluation index threshold corresponding to each reference comprehensive characteristic coefficient interval of the data layer preset in the database, and compare it with the comprehensive characteristic coefficient of each data layer. If the comprehensive characteristic coefficient of a data layer is within a certain reference comprehensive characteristic coefficient interval, obtain the parallel processing port performance evaluation index threshold corresponding to the reference comprehensive characteristic coefficient interval as the parallel processing port performance evaluation index threshold of the data layer, and mark it as the parallel port performance threshold of the data layer; Comparing the parallel port performance threshold of each data layer with the performance evaluation index of each parallel processing port, if the performance evaluation index of a parallel processing port is greater than and closest to the parallel port performance threshold of a data layer, then the parallel processing port is used as the distributed data parallel processing port of the data layer; The port performance parameters of each parallel processing port include the CPU processing speed, resource utilization, historical error rate and historical average response time of each parallel processing port; The performance evaluation index of each parallel processing port is used to characterize the performance of each parallel processing port.
8. The deep learning-based auxiliary diagnosis data feature analysis method according to claim 1, characterized in that: It also includes transmitting each data layer to the corresponding distributed data parallel processing port for deep learning training, analyzing the training performance indicators, and dynamically adjusting the learning rate based on them. The specific steps include: Obtain real-time performance parameters of the training process and analyze and process them to obtain training performance indicators; Obtaining a first threshold value of a training performance indicator and a second threshold value of a training performance indicator preset in a database, and comparing them with the training performance indicator, if the training performance indicator is above the first threshold value of the training performance indicator, increasing the learning rate; if the training performance indicator is less than the first threshold value of the training performance indicator and above the second threshold value of the training performance indicator, not adjusting the learning rate; if the training performance indicator is less than the second threshold value of the training performance indicator, reducing the learning rate; The real-time performance parameters of the training process include: real-time training rounds, real-time convergence speed, real-time gradient change, real-time network delay, real-time network bandwidth and real-time data transmission speed; The training performance indicator is used to characterize the training performance of the data layer during the training process.
9. A deep learning-based auxiliary diagnosis data feature analysis system, characterized by: include: Data representation parameter acquisition module, data enhancement parameter determination module, comprehensive feature coefficient acquisition module, distributed processing parameter determination module and parallel processing port determination module; The data performance parameter acquisition module is used to acquire the auxiliary diagnosis data and perform data loading processing to obtain the data performance parameters of the auxiliary diagnosis data after the auxiliary diagnosis device receives the auxiliary diagnosis instruction signal; The data enhancement parameter determination module is used to analyze the data performance parameters of the auxiliary diagnosis data to obtain a data complexity index of the auxiliary diagnosis data, and determine the data enhancement parameters based on the data complexity index; The comprehensive characteristic coefficient acquisition module is used to perform initial enhancement processing on the auxiliary diagnosis data through the auxiliary diagnosis device according to the data enhancement parameters, and analyze the auxiliary diagnosis data after the initial enhancement processing to obtain the comprehensive characteristic coefficient; The distributed processing parameter determination module is used to determine the distributed processing parameters of the auxiliary diagnosis data based on the comprehensive characteristic coefficients of the auxiliary diagnosis data, and to segment the auxiliary diagnosis data through the auxiliary diagnosis device to obtain each data layer and the comprehensive characteristic coefficients of each data layer; The parallel processing port determination module is used to determine the distributed data parallel processing port of each data layer according to the comprehensive characteristic coefficient of each data layer, and perform characteristic analysis processing.
10. A device for analyzing auxiliary diagnostic data features based on deep learning, characterized by: include: A processor, and a memory and a network interface connected to the processor: the network interface is connected to the non-volatile memory in the server; When running, the processor retrieves a computer program from the non-volatile memory through the network interface, and runs the computer program through the memory to apply the method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Methods for analyzing electrocardiogram signals
CN113545787B
Electrocardiogram analysis method for postoperative rehabilitation of patients
CN117958835B
Cited By
Traditional Chinese medicine prescription efficacy prediction system and method based on dual-encoder Transformer
CN121662398A
